TRF2-Mediated Stabilization of hREST4 Is Critical for the Differentiation and Maintenance of Neural Progenitors


  • Patrick Ovando-Roche,

    1. Institute of Reproductive and Developmental Biology, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, Hammersmith Campus, London, United Kingdom
    Search for more papers by this author
  • Jason S.L. Yu,

    1. Institute of Reproductive and Developmental Biology, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, Hammersmith Campus, London, United Kingdom
    Search for more papers by this author
  • Sarah Testori,

    1. Institute of Reproductive and Developmental Biology, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, Hammersmith Campus, London, United Kingdom
    Search for more papers by this author
  • Chloe Ho,

    1. Institute of Reproductive and Developmental Biology, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, Hammersmith Campus, London, United Kingdom
    Search for more papers by this author
  • Wei Cui

    Corresponding author
    1. Institute of Reproductive and Developmental Biology, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, Hammersmith Campus, London, United Kingdom
    • Correspondence: Wei Cui, Ph.D., Imperial College London, IRDB, Hammersmith Campus, Du Cane Road, London, United Kingdom W12 0NN. Telephone: +44-20-7594-2124; Fax: +44-20-7594-2154; e-mail:

    Search for more papers by this author


Telomere repeat binding factor 2 (TRF2) is a component of the shelterin complex that is known to bind and protect telomeric DNA, yet the detection of TRF2 in extra-telomeric regions of chromosomes suggests other roles for TRF2 besides telomere protection. Here, we demonstrate that TRF2 plays a critical role in antagonizing the repressive function of neuron-restrictive silencer factor, also known as repressor element-1 silencing transcription factor (REST), during the neural differentiation of human embryonic stem cells (hESCs) by enhancing the expression of a truncated REST splice isoform we term human REST4 (hREST4) due to its similarity to rodent REST4. We show that TRF2 is specifically upregulated during hESC neural differentiation concordantly with an increase in the expression of hREST4 and that both proteins are highly expressed in NPCs. Overexpression of TRF2 in hESCs increases hREST4 levels and induces their neural differentiation, whereas TRF2 knockdown in hESCs and NPCs reduces hREST4 expression, hindering their ability to differentiate to the neural lineage. Concurrently, we show that TRF2 directly interacts with the C-terminal of hREST4 through its TRF2 core binding motif [F/Y]xL, protecting hREST4 from ubiquitin-mediated proteasomal degradation and consequently furthering neural induction. Thus, the TRF2-mediated counterbalance between hREST4 and REST is vital for both the generation and maintenance of NPCs, suggesting an important role for TRF2 in both neurogenesis and function of the central nervous system. Stem Cells 2014;32:2111–2122


In developing neural regenerative therapies, it is imperative to understand the molecular mechanisms regulating neurogenesis in the central nervous system (CNS), particularly those involved in the production and maintenance of human neural progenitors cells (NPCs) as these cells can potentially replenish the damaged neural cells that contribute to neurodegenerative diseases. Human embryonic stem cells (hESCs) provide a powerful platform for such studies, as they can be expanded indefinitely while retaining the ability to differentiate into all neural lineages. Moreover, the in vitro neural differentiation of hESCs recapitulates the normal development of the CNS in vivo and involves major transcriptomic changes regulated by a plethora of different factors, including chromatin regulators, transcription modulators, signaling mediators, and microRNA [1, 2]. Among them, repressor element-1 silencing transcription factor (REST), a master transcriptional repressor, has been shown to be a key player in this process [3, 4], although the contribution of various REST isoforms to neural induction is still unclear. Furthermore, since most of these studies were conducted in mouse ESCs (mESCs), the function of REST and its isoforms during hESC neural differentiation remains unknown.

Telomere repeat binding factor 2 (TRF2) is a component of the shelterin complex that binds to telomeric DNA, protecting the chromosome ends and hence maintains genome stability [5]. However, recent evidence demonstrating the ability of TRF2 to bind to non-telomeric regions of the chromosomes [6-8] and to function as a protein hub [9-12] suggests that it may have alternative functions independent of telomere maintenance. Despite the importance of TRF2, its expression and involvement in development and tissue homeostasis remain largely unknown. This is because the majority of studies were performed in either tumor cell lines or primary/immortalized fibroblasts which may not be adequate representatives of normal developmental tissues. Furthermore, TRF2 knockout mice exhibit embryonic lethality, restricting interrogation of TRF2 functions during development [13].

Here we investigated the role of TRF2 in hESCs and their differentiated progenies. Given the abundance of TRF2 in the human brain [14], we focused our studies particularly on the neural differentiation and uncovered strikingly dynamic changes in TRF2 expression during this process. Using gain and loss of function approaches, we have identified a critical role for TRF2 in promoting neural differentiation of hESCs and maintaining NPC neuropotency. Furthermore, we have deciphered the molecular mechanisms responsible for this effect, whereby TRF2 alleviates the repression of REST on neural genes by enhancing the expression of human REST4 (hREST4), a splice variant of human REST, similar in structure to rodent REST4 [15, 16], which counteracts the repressive action of REST on neural genes.

Materials and Methods

Cell Culture and Differentiation

hESC lines H1 and H7 were obtained from WiCell Research Institute (Madison, WI, and routinely propagated under feeder-free conditions with mouse embryonic fibroblast conditioned media as described previously [17]. Neural differentiation of hESCs and culture and differentiation of NPCs were performed with our established protocols [2, 18, 19]. Fibroblast differentiation of hESCs was induced by replacing hESC culture medium with Dulbecco's modified Eagle's medium containing 10% fetal calf serum (Sigma-Aldrich, St. Louis, MO, and 1% l-glutamine (Life Technologies, Rockville, MD, for more than 3 weeks. Differentiation of hESCs to hepatocyte progenitors and embryoid bodies (EB) formation were performed as previously described [17, 20, 21].

Expression Plasmids

TRF2-overexpressing lentivector was derived from pLVTHM [22] with TRF2 cDNA isolated from pLPC-TRF2 [23]. Both were obtained from Addgene (Cambridge, MA, Briefly, the stop codon of puromycin-resistant cDNA (PURO) in pPUR (Clontech, Mountain View, CA, was mutated by site-directed mutagenesis, and a 2A foot and mouth disease virus sequence [24] was inserted downstream of the PURO. The PURO-2A fragment was then fused with TRF2 cDNA and inserted into pLVTHM. TRF2-knockdown (TRF2-sh) vector was generated by replacing scrambled short hairpin RNA (shRNA) in pLVTHM with a duplex oligonucleotide targeting TRF2 mRNA 5′-CGCGCAGGAGCATGGTTCCTAATAATACT-GCAGTATTATTAGGAACCATGCTCCTGTTTTT-3′ and 5′-GCAAAAACAGGAGCAT-GGTTCCTAATAATACTGCAGTATTATTAGGAACCATGCTCCTG-3′. Both hREST4 and hREST4m expression vectors were generated from pLCMV-Neo-Myc-REST, a kind gift from Dr. S. Bao [25], by replacing REST C-terminal sequence with an oligonucleotide duplex containing either hREST4 exon N coding sequence or a stop codon. hREST4 lentivector was generated by replacing TRF2 in pLVTHM-PURO-2A-TRF2 with hREST4 cDNA. pCS2-HA-TRF2 was generated by cloning TRF2 cDNA into pCS2-HA. HA (human influenza hemagglutinin) and Myc epitopes were added to the N-termini of TRF2 or REST/hREST4 proteins, respectively. All constructs were verified by sequencing.

Lentiviral Production and Transduction

Lentiviruses were produced by transient transfection of HEK293T cells with a lentivector as well as pCMVΔ8.91 helper and pVSV-G envelope plasmids using standard protocols [22] and immediately stored at −80°C until use. 24 hours prior to infection, hESCs were split with accutase and seeded into Matrigel-coated plates in hESC medium containing 10 µM ROCK inhibitor Y-27632 (Reagents Direct, Encinitas, CA, [26]. Cells were selected for puromycin resistance 72 hours postinfection. NPCs were infected similarly to hESCs but under NPC culture conditions.

Immunoblotting, Immunostaining, and Flow Cytometry

Immunoblotting (IB) and immunostaining were performed as previously described [18] with minor modification. For IB, cells were lysed in cold RIPA buffer (150 mM NaCl, 50 mM Tris-HCl pH 8.0, 12 mM sodium deoxycholate, 0.1% SDS, 1% Nonidet P40) containing protease inhibitor cocktail and 0.2 mM phenylmethanesulfonylfluoride (PMSF) (Sigma-Aldrich). 20 µg proteins was resolved in 7% SDS-polyacrylamide gel. Quantification of protein bands was carried out using Quantity One software (Bio-Rad, Hercules, CA, Statistical analyses were performed by Student's t test from data of three independent experiments. For immunostaining, hESCs and NPCs were seeded onto Matrigel-coated Nunc Thermanox (ThermoFisher Scientific, Waltham, MA and poly-l-lysine/laminin-coated glass coverslips, respectively and were stained with antibodies postfixation. Primary antibodies used are listed in Supporting Information Table S1. Flow cytometry was performed as previously described [18].


Cells were lysed in ice-cold lysis buffer (20 mM HEPES pH 7.4, 10 mM KCl, 100 mM NaCl2, 1 mM EDTA, 0.1 mM EGTA, 0.1% Triton X-100, 0.5% Nonidet P40) containing protease inhibitor cocktail and 0.2 mM PMSF. 1 mg of extracts was incubated with 2–3 µg antibodies at 4°C for 1–3 hours with rotation. Immunocomplexes were captured at 4°C with either 20 µl Protein A/G Plus agarose beads (Santa Cruz Biotechnology Inc., Santa Cruz, CA, for 1 hour or 50 µl Protein G Dynabeads (Life Technologies) for 3 hours. Immunoprecipitates were eluted with 2× Laemmli buffer and boiled for 10 minutes, then analyzed by IB.

hREST4 Ubiquitination Assay

Cells cotransfected with pLCMV-Neo-Myc-hREST4 and HA-Ubiquitin plasmids were treated with 20 µM MG132 (Merck, Whitehouse Station, NY, for 5 hours and then lysed in cold lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 1% Nonidet P40) containing protease inhibitor cocktail, 0.2 mM PMSF, and 1 µM ubiquitin aldehyde (Boston Biochem, Cambridge, MA, 1 mg of lysate was subjected to immunoprecipitation (IP) followed by IB as above.

Telomere Fluorescent In Situ Hybridization and Immuno-FISH

Metaphase Spreads Preparation

Cells were harvested in 0.8% sodium citrate after treatment with 0.5 µg/ml colcemid (Life Technologies) at 37°C for 4 hours and incubated at 37°C for 15 minutes. Cells were then fixed in 3:1 methanol-acetic acid and dropped onto clean glass slides. Fluorescent in situ hybridization (FISH) was performed using Telomere PNA FISH kit (DAKO, Glostrup, Denmark, following manufacturer's instructions. For immuno-FISH, cells were first subjected to immunostaining as above, which were followed by fixation in 3.7% formaldehyde and dehydration in cold ethanol series. The cells were then incubated with PNA probe 5 minutes at 85°C followed by overnight at 37°C, cells were washed in 70% formamide and 2× SSC for 5 minutes each. Finally, cells were dried and mounted with VECTASHIELD (DAKO) solution containing 4,5-diamidino-2-phelylindole.

Telomere Length Assay

Telomere lengths were measured by telomere restriction fragment (TRF) Southern blot using a TeloTAGGG telomere length assay kit (Roche Diagnostics, Basel, Switzerland,

Quantitative Reverse Transcriptase Polymerase Chain Reaction

RNA isolation and cDNA synthesis were performed as described previously [18]. Polymerase chain reaction (PCR) reactions were carried out in an Opticon thermal cycler (Bio-Rad) using the primers listed in Supporting Information Table S2.


TRF2 Protein Is Upregulated During the Neural Differentiation of hESCs and Maintained at High Levels in Neural Progenitors

In order to explore the role of TRF1 and TRF2 in hESCs and their neural progenies, we first examined their mRNA expression in both H1 and H7 hESCs together with hESC-derived NPCs. TRF1 mRNA was highly expressed in hESCs but was radically reduced following neural differentiation and maintained at low levels throughout the NPC culture, whereas the expression of TRF2 transcripts did not exhibit an obvious pattern of change (Fig. 1A). These results are consistent with our previous genome-wide RNA sequencing results [19] and published data [27, 28]. However, in contrast to the mRNA expression, TRF1 protein did not show a considerable reduction upon neural differentiation. More surprisingly, in both H1 and H7 hESC lines, TRF2 protein levels were low in hESCs but dramatically increased following neural differentiation and were maintained at these elevated levels throughout extended culture (Fig. 1B, 1C; Supporting Information Fig. S1A). The discrepancy between mRNA and protein levels of TRF1 and TRF2 indicate that both proteins are regulated at the post-transcriptional level, in line with previous findings [29, 30]. Analysis of additional protein samples from independent differentiation experiments further confirmed the significant upregulation of TRF2 protein in H1 and H7 upon neural differentiation (Fig. 1B, right panels). Furthermore, although TRF2 proteins were predominantly located in cell nuclei of NPCs as they are in hESCs (Fig. 1C, 1D), they were not solely restricted to the telomeres (Fig. 1E). Interestingly, TRF2 protein was found at lower levels in mESCs than in hESCs and did not exhibit any upregulation upon neural differentiation (Supporting Information Fig. S1B). Therefore, TRF2 levels are considerably lower in mouse NPCs than in human NPCs, highlighting yet another difference between human and mouse biology.

Figure 1.

Upregulation of TRF2 protein after neural differentiation of hESCs. (A): Expression of TRF1 and TRF2 mRNAs by quantitative reverse transcriptase polymerase chain reaction during the neural differentiation of H1 and H7 hESCs and subsequent culture of their progenitors. Passage numbers of postdifferentiation are indicated. Data are presented as mean ± SD of six to nine PCRs from three independent experiments. (B): Immunoblotting showing TRF1 and TRF2 protein levels during the neural differentiation of hESCs. Left panels are representative images; right panels are the densitometric analysis as mean ± SD of three experiments. **, p < .005. (C): Immunostaining showing considerably higher TRF2 in NPCs but not in hESCs as indicated by Oct4 and Sox1 antibody staining, respectively. Scale bar = 50 µm. (D): Higher magnification images of TRF2 staining in hESCs and NPCs. Scale bar = 5 µm. (E): Double staining of TRF2 by TRF2 antibody and telomeres by telo-FISH in NPCs. Only a proportion of TRF2 signals show colocalization with telomeres (arrowheads). Scale bar = 5 µm. Abbreviations: hESC, human embryonic stem cell; NPC, neural progenitor cell; TRF, telomere repeat binding factor.

Overall our results indicate that while the expression of TRF1 is relatively similar between hESC and NPCs at various stages of the differentiation, there is contrastingly a considerable upregulation in the expression of TRF2 protein upon neural differentiation. Additionally, TRF2 is abundantly distributed throughout the nuclei of NPCs rather than being confined solely to the telomeres, which suggests an extra-telomeric role for TRF2 during neural differentiation.

Upregulation of TRF2 Is Specific to Human Neural Progenitors

Since TRF2 exhibited a more distinctive change during the neural differentiation of hESCs, we focused our subsequent studies on TRF2. We asked whether TRF2 upregulation is a general phenomenon in hESC differentiation to all lineages or is specifically restricted to NPCs. To address this, we differentiated hESCs via EB formation to the three germ layers. Following differentiation, TRF2 was mainly detected in neural progenitors that were positive for Sox2 but negative for Oct4 and HNF4α (Hepatocyte Nuclear Factor 4 alpha, Supporting Information Fig. S2A). Similarly, TRF2 levels were found to be low in several non-neural cell lines (Supporting Information Fig. S2B). To further corroborate these findings, we differentiated hESCs specifically to neural, hepatic, or fibroblast lineages in which majority of the cells displayed their lineage-specific morphology and corresponding gene expression profiles (Fig. 2A, 2B). TRF2 was upregulated only in NPCs but not in hepatocyte progenitors or fibroblasts (Fig. 2C). Furthermore, when NPCs were further differentiated into postmitotic neurons or glia, TRF2 levels were again reduced (Fig. 2D–2E). Our data demonstrate that TRF2 expression is dynamically regulated during neural differentiation and neuron/glia formation with higher levels being restricted to NPCs.

Figure 2.

Comparison of TRF2 expression in hESC-derived NPCs and hESC-derived other cell types. (A): Phase-contrast images of hESCs and their derived progenies: NPC, FC, HPC. (B): Expression of lineage-specific marker genes by quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) in the indicated cells. Data are presented as mean ± SD of six RT-PCRs from three independent experiments. (C): TRF2 protein expression by immunoblotting in the above cells. (D, E): NPCs from early (D) and late (E) neural differentiation were cultured in the presence or absence of GF for 14 days and then analyzed by either immunostaining of the cells with indicated antibodies (upper panels) or immunoblotting of their cell extracts with TRF2 antibody (lower panels). Scale bar = 50 µm in all microscopic images. Abbreviations: FC, fibroblasts; GF, growth factor; GFAP, glial fibrillary acidic protein; HPC, hepatocyte progenitor cell; hESC, human embryonic stem cell; NPC, neural progenitor cell; TRF, telomere repeat binding factor.

Differential Expression of TRF2 Neither Correlates with Telomere Length Nor Telomerase Expression

Since the protective and regulatory role of TRF2 with regards to telomeres has been well-characterized [31, 32], we investigated whether the dramatic changes of TRF2 expression during hESC neural differentiation correlated with changes in either telomere length or telomerase expression. In hESCs, telomeres are long and stable due to high telomerase expression. Upon neural differentiation, expression of telomerase was considerably downregulated and telomeres shortened with subsequent cell proliferation. Following extended culture of NPCs, telomerase was reactivated in latter passages, which led to subsequent lengthening of the telomeres (Fig. 3A, 3B). These dynamic changes of telomere length were notably different to that of TRF2 expression (Fig. 3B vs. 1B). H1 hESCs had similar telomere lengths to the p4 NPCs (Fig. 3B), but they expressed much lower levels of TRF2 (Fig. 1B), whereas although telomeres were significantly longer in p4 than p27 NPCs, both passages expressed similar levels of TRF2. Additionally, in situ immunostaining of chromosomal spreads also showed a lack of correlation between telomere length and TRF2 expression (Fig. 3C). Overall, these studies therefore demonstrate the independence of TRF2 expression from both telomere length and telomerase expression.

Figure 3.

No correlation between TRF2 expression and telomere length or telomerase expression in NPCs. (A): Expression of telomerase catalytic subunit hTERT mRNA in hESCs and their neural progenitors by quantitative reverse transcriptase polymerase chain reaction. Data are presented as mean ± SD of three experiments. (B): Dynamic changes of telomere length in hESCs and their NPCs by TRF Southern blotting. (C): Comparison of telomere length (upper panel, by telomere PNA-FISH) and TRF2 expression (lower panel, by immunostaining) in hESCs and their late passage NPCs. Inserts are the same images but with a higher magnification. Abbreviations: hESC, human embryonic stem cell; NPC, neural progenitor cell; TRF, telomere repeat binding factor.

Ectopic Expression of TRF2 in hESCs Promotes Neural Differentiation

Since TRF2 is significantly upregulated during neural differentiation of hESCs, we speculated that it might also play a critical role in this process. We applied gain and loss of function approaches to address this, whereby TRF2 was ectopically expressed in H1 hESCs by lentiviral transduction in the first instance (Fig. 4A). Following selection, several colonies in the TRF2-overexpressing (TRF2-ov) plates emerged with neural rosette-like structures (Fig. 4B), and this differentiation became more evident upon continuous culture with the appearance of bipolar neural progenitors (Supporting Information Fig. S3A). In contrast, no such phenotype was observed in the controls. These morphological changes were reproducible in repeated experiments. Gene expression analysis revealed a downregulation of pluripotent markers Oct4 and Nanog, coupled with an upregulation of neural markers Sox1, Mash1, and Snap25 in TRF2-ov cells (Fig. 4C, 4D). Flow cytometry and immunostaining also showed that a proportion of TRF2-ov hESCs expressed neural lineage markers despite being cultured under hESC culture conditions (Fig. 4E–4G). Furthermore, when TRF2-ov hESCs were transferred into neural differentiation conditions, they differentiated faster when compared to control cells. Although a high proportion of control cells still expressed high levels of Oct4 by day 12 of the differentiation, very few TRF2-ov cells were positive for Oct4 and the majority of them expressed the neural progenitor marker Nestin (Supporting Information Fig. S3B, S3C). Given that the neural differentiation emerged before passage 4 in TRF2-ov hESCs, and that telomere attrition was detected after passage 5 (Fig. 4H), it is unlikely that this phenotype is a consequence of telomere shortening. Furthermore, we did not detect any increase of γH2AX in TRF2-ov hESCs (Fig. 4D), indicating that the differentiation is unlikely to be attributed to DNA damage. These results therefore demonstrate that TRF2 plays a positive role in promoting neural differentiation of hESCs, which is independent from its role in regulating telomere length and protection of the chromosomes.

Figure 4.

Induction of neural differentiation in TRF2-ov human embryonic stem cells (hESCs). (A): Schematic map of transgene expression in a lentiviral vector. (B): Representative phase-contrast images of control and TRF2-ov hESCs four passages post-selection. Scale bar = 100 µm. (C): TRF2 overexpression in hESCs increased the expression of neural marker genes but reduced the expression of pluripotent genes as detected by quantitative reverse transcriptase polymerase chain reaction in cells five passages after selection. Data are presented as mean ± SD of six repeats of two independent experiments. (D): Immunoblot showing indicated protein expression in control and TRF2-ov hESCs. (E): Flow cytometry analysis of Tra-1–81 and Pax6 in control and TRF2-ov cells (passage 11) shows decreased Tra-1–81 and increased Pax6 positive population in TRF2-ov cells. (F, G): Immunostaining with indicated antibodies depicting areas of neural differentiation in TRF2-ov hESCs (p5–6). Scale bar = 50 µm. (H): Comparison of telomere length in control and TRF2-ov hESCs at two indicated passages by Southern blotting shows that telomere length shortening occurred after five passages in TRF2-ov hESCs. Abbreviations: CAG, CAG promoter; PuroR, puromycin resistant gene; 2A, 2A sequence; TRF2, TRF2 cDNA; TRF2, telomere repeat binding factor 2.

TRF2 Knockdown Hinders Neural Differentiation of hESCs and Neural Competence of NPCs

Complementary to the overexpression experiments, we also examined the effect of TRF2 deficiency in neural differentiation by knocking down TRF2 with shRNA (TRF2-sh) in hESCs and hESC-derived NPCs (Fig. 5A, 5B). Both TRF2-sh hESCs and NPCs did not display clear morphological changes under normal growth conditions (Fig. 5C, 5I, upper panels). The TRF2-sh hESCs exhibited similar levels of pluripotent markers, Oct4, Nanog, and Sox2 (Fig. 5D). However, when differentiated via EB, TRF2-sh hESCs generated much less Tuj1 positive neurons than the controls, even though no clear difference was observed in the staining of endoderm markers, HNF4α and AFP (Alpha-fetoprotein, Fig 5E). Furthermore, direct neural differentiation of TRF2-sh hESCs revealed a reduction in the generation of neural progenitors (Fig. 5C, lower panel; Fig. 5F), which led to decreased number of neurons upon growth factor withdrawal (Supporting Information Fig. S4A). These results indicate that TRF2 is important for the neural differentiation of hESCs.

Figure 5.

Knockdown of TRF2 in hESCs and NPCs hinders neural differentiation and impedes the formation of neurons and glia. (A): Decrease of TRF2 expression in TRF2-knockdown (TRF2-sh) hESCs and NPCs by immunoblotting. (B): Reduction of TRF2 staining in TRF2-sh NPCs. Scale bar = 50 µm. (C): Phase-contrast images of control and TRF2-sh hESCs in hESC (upper) and neural differentiation (lower) culture conditions. (D–F): Knockdown TRF2 in hESCs. (D): Expression of pluripotent genes by quantitative reverse transcriptase polymerase chain (qRT-PCR) in cells under hESC culture conditions. (E): Immunostaining with indicated antibodies in cells after EB differentiation. (F): Expression of neural progenitor markers by qRT-PCR after neural differentiation. (G–J): Knockdown TRF2 in NPCs. Expression of neural makers by qRT-PCR (G) and immunostaining (H). (I): Phase-contrast images of NPCs grown in culture with (upper, +GF) or without growth factors (lower, −GF). (J) NPCs cultured for 2 weeks without GF were stained with Tuj1 and GFAP antibodies. Error bars represent SD from six to eight repeats of two independent experiments in (D, F, and G). Scale bar = 50 µm in (B, E, H, and J) and 100 µm in (C and I). Abbreviations: AFP, alpha-fetoprotein; GF, growth factor; GFAP, glial fibrillary acidic protein; hESCs, human embryonic stem cells; NPC, neural progenitor cell; TRF2, telomere repeat binding factor 2.

In line with our observation in the neural differentiation of TRF2-sh hESCs, knockdown of TRF2 in NPCs also revealed clear reduction in the expression of neural progenitor and neuronal markers, Sox2, Mash1, and Snap25 (Fig. 5G, 5H), despite no clear morphological changes (Fig. 5I, upper panel). There was no apparent cell death in these cells but cell proliferation was slightly reduced (Supporting Information Fig. S4B, S4D). Upon further differentiation to neurons and glia, control cells developed long and multiple processes of interconnecting outgrowths, whereas the TRF2-sh cells appeared flatter, with fewer outgrowths (Fig. 5I, lower panel). Correspondingly, TRF2-sh cells displayed a clear reduction in Tuj1 and GFAP staining (Fig. 5J). Interestingly, knockdown of TRF2 in both hESCs and NPCs did not affect telomere length (Supporting Information Fig. S4E, S4F) and no obvious chromosome end joining was observed (Supporting Information Fig. S4G). These results demonstrate that a deficiency of TRF2 decreases neural gene expression and diminishes the capability of NPCs to further differentiate into neurons and glia without any evident effect on the telomeres. Therefore, TRF2 appears to play an important role not only in the differentiation but also in the maintenance of NPCs.

TRF2 Expression Affects hREST4 But Not REST Expression in hESCs and NPCs

In line with the observed impact upon neural differentiation, TRF2 has also been previously reported to inhibit neuronal differentiation of human embryonic carcinoma NTera2 cells by interacting with REST and stabilizing its expression [33]. Given the importance of REST to neural differentiation, we examined REST expression in hESCs and their neural progenies to see whether this correlates with TRF2 expression as a potential mechanism to explain our observations. Surprisingly, we did not detect any significant differences in REST expression between hESCs and their differentiated NPCs even though TRF2 was clearly upregulated in NPCs (Fig. 6A, 6B). Accordingly, there is no clear correlation between TRF2 and REST expression during hESC neural differentiation. This is further supported by the fact that REST levels showed no change in TRF2-ov hESCs or HEK293T cells and in TRF2-knockdown NPCs (Fig. 6D, 6E; Supporting Information Fig. S5A). Of note, expression of TRF2 and REST was shown to be considerably higher in NTera2 cells than in hESCs, while the level of TRF2 in NTera2 cells was similar to that of hESC-derived NPCs (Supporting Information Fig. S5B). High levels of TRF2 in NTera2 cells, together with the fact that NTera2 cells are more easily differentiated into neural lineages via retinoic acid treatment than hESCs, further support our notion of TRF2 promoting neural differentiation.

Figure 6.

TRF2 affects NPCs via upregulating hREST4 expression. (A): Immunoblotting of cell extracts from H1 and H7 hESCs and their early and late NPCs with indicated antibodies, showing upregulation of TRF2 and hREST4 in NPCs. (B): Quantitative analysis of (A) was performed by densitometry of the blots and presented as histogram. (C): Schematic map of human REST gene, REST/hREST4 transcripts and proteins, and mutated REST4 protein. Open and colored boxes represent exons and coding regions, respectively. hREST4-specific amino acids are shown in red. (D, E): Immunoblotting of cell extracts from TRF2-ov hESCs and TRF2-sh NPCs as well as their corresponding controls with indicated antibodies. Histograms are densitometric analysis of the blot shown on left. (F–H): Ectopic expression of hREST4 in TRF2-sh NPCs as shown by immunoblotting (F) increases expression of neural progenitor markers (G) and formation of neurons and astrocytes (H) in compare to TRF2-sh NPCs. Scale bar = 50 µm. Abbreviations: GFAP, glial fibrillary acidic protein; hESC, human embryonic stem cell; hREST4, human REST4; hRESTm, mutant REST; NPC, neural progenitor cell; NPC-E, early NPCs; NPC-L, Late NPCs; REST, repressor element-1 silencing transcription factor; TRF2, telomere repeat binding factor 2.

Interestingly, the relatively stable REST expression we have observed during neural differentiation of hESCs is different from that reported during mESC neural differentiation, which showed a downregulation of REST. This reduction was thought to alleviate REST repression on neural genes, hence promoting neural differentiation [34]. This has raised the question as to how REST-mediated repression on neural genes is mitigated during hESC neural differentiation. We noticed that the REST antibody, raised against the N-terminus of human REST, consistently detected a 40-kDa band in NPCs, which is lower in molecular weight than that of REST (Fig. 6A, 6F). Although the antibody also detected a couple of other bands, these did not appear when using other batches of the antibody and were therefore considered to be nonspecific (Fig. 6A, 6F; Supporting Information Fig. S6C). Interestingly, this 40 kDa protein exhibited a pattern of upregulation similar to that of TRF2 during the neural differentiation (Fig. 6B) and is of a similar molecular weight to the predicated size of REST-N62, a human REST isoform that is orthologous to the rodent neural-specific REST isoform REST4 [15, 16, 35]. Correspondingly, IB with the REST antibody on HEK293T cell extracts overexpressing REST and REST-N62 revealed the presence of only two bands of 160–200 and 40 kDa, respectively (Supporting Information Fig. S5C). Human REST-N62 is generated by the alternative splicing of a neural-specific exon (exon N) located between exon IV and V to produce a 62 nucleotide insertion in the REST mRNA, containing an in-frame stop codon that causes premature translation termination and synthesis of truncated REST protein (Fig. 6C) [16]. Although no function has been ascribed to REST-N62, the rodent protein orthologous REST4 has been shown to counteract the repressive role of REST in the regulation of its target genes, particularly during neurogenesis [16, 35-38]. While the protein sequence encoded by exon N differs between REST-N62 and rodent REST4 (Supporting Information Fig. S5D), both retain the REST N-terminal domain and five of the nine zinc-finger DNA binding domains, but lack the C-terminal repression domain (Fig. 6C). Hence, we have appropriately termed this REST-N62 variant as human REST4 (hREST4) and given the similarity in protein structure to the rodent REST4, posit its involvement in regulating neural differentiation through a similar mechanism.

To validate this hypothesis, hREST4 was ectopically expressed (hREST4-ov) in hESCs. These cells exhibited compromised self-renewal and increased cell differentiation, particularly toward the neural lineage even under hESC culture conditions (Supporting Information Fig. S6A–S6D). In addition, when subjected to neural differentiation, hREST4-ov hESCs, like TRF2-ov hESCs, showed a quicker and more efficient conversion (Supporting Information Fig. S6A, S6E). These results indicate that hREST4 undermines the pluripotency of hESCs, possibly through affecting the repression of neural genes by REST. In exploring this relationship between TRF2 and hREST4 further, we showed that overexpressing TRF2 increased hREST4 protein levels in hESCs whereas knockdown of TRF2 reduced hREST4 in NPCs (Fig. 6D, 6E; Supporting Information Fig. S5E). Furthermore, we were able to rescue the defects in neural gene expression and formation of neurons and astrocytes observed in TRF2-sh NPCs through hREST4 overexpression without altering TRF2 levels (Fig. 6F–6H). Overall, our data suggest that TRF2 may act upstream of hREST4 to enhance its expression, subsequently affecting the neural differentiation of hESCs and the preservation of NPC neuropotency.

TRF2 Interacts with hREST4 to Promote Neural Differentiation of hESCs

Although we observed a positive correlation between TRF2 and hREST4 expression, it remained unclear how this occurs. Since overexpression of TRF2 did not increase REST in hESCs (Fig. 6D; Supporting Information Fig. S5E) and hREST4 in HEK-293T cells (Supporting Information Fig. S5A), it is unlikely that TRF2 enhances hREST4 expression through direct regulation of REST transcription or splicing. Given that TRF2 acts as a protein hub and therefore interacts with many proteins [10, 11], it is conceivable that TRF2 may affect hREST4 through protein-protein interaction. Indeed, TRF2 co-IP in NPCs showed that endogenous hREST4 but not REST could be detected in TRF2 immunoprecipitates (Fig. 7A). Furthermore, transient expression of myc-tagged REST or hREST4 and HA-tagged TRF2 in HEK293T cells showed that HA-TRF2 was only detectable in the myc-hREST4 immunoprecipitates but not in myc-REST (Fig. 7B); conversely, only myc-tagged hREST4 could be detected in the HA-TRF2 immunoprecipitates (Fig. 7C).

Figure 7.

TRF2 interacts with hREST4 to inhibit its ubiquitination. (A): Immunoblot of endogenous TRF2 immunoprecipitates or whole cell lysate from NPCs with REST and Rap1 antibodies. (B, C): HEK293 cells were transiently expressed with HA-TRF2, Myc-REST, and Myc-hREST4 individually or combined as indicated. Myc (B) or HA (C) immunoprecipitates were analyzed by IB with Myc and HA antibodies. (D): HA immunoprecipitates from HEK293 cells coexpressing HA-TRF2 with either myc-hREST4 or myc-hREST4m were analyzed by IB with Myc and HA antibodies. (E): TRF2 inhibits hREST4 ubiquitination. Myc-hREST4 and HA-Ub were transiently expressed in HEK293T cells that ectopically express TRF2 or control vectors. Myc immunoprecipitates were immunoblotted with indicated antibodies. (F): A schematic model for the role of TRF2, REST, and hREST4 in regulating neural differentiation of hESCs and functional NPCs. Abbreviations: hESCs, human embryonic stem cells; hREST4, human REST4; HA, human influenza hemagglutinin; IB, immunoblotting; IP, immunoprecipitation; hNPCs, human nuclear progenitor cells; REST, repressor element-1 silencing transcription factor; TRF2, Telomere repeat binding factor 2; Ub, ubiquitin.

Although hREST4 is largely considered as a truncated REST, the 13 amino acids (AAs) encoded by exon N are unique to hREST4 and contain the TRF2 core binding motif [Y/F]xL (Fig. 6C) [9, 10]. To address whether these 13 AAs are critical for hREST4 binding to TRF2, we generated a mutant hREST4 (hREST4m) lacking these 13 AAs (Fig. 6C). When HA-TRF2 with either wild-type myc-hREST4 or mutant myc-hREST4m were transiently transfected into HEK293T cells, only myc-hREST4 was detected in HA-TRF2 immunoprecipitates, indicating that the 13 AAs at the C-terminus of hREST4 are essential for the TRF2-hREST4 interaction (Fig. 7D). As such, the uniqueness of these residues to hREST4 and their absence in full-length REST confer binding specificity in favor of hREST4 to TRF2. We also observed that high levels of TRF2 reduced the ubiquitination of hREST4 (Fig. 7E), which could prevent its proteasome-mediated degradation. Although the identity of the hREST4-specific E3 ubiquitin ligase remains unknown, hREST4 appears to be nonetheless regulated by a proteasomal-mediated degradation mechanism (Supporting Information Fig. S5F). Therefore, high levels of TRF2 stabilize and increase the resistance of hREST4 to degradation, resulting in higher levels of hREST4, which promote the neural differentiation of hESCs.


In this study, we have demonstrated that TRF2 plays a critical role in the neural differentiation of hESCs and the maintenance of NPCs, independent of its role in regulating telomere length and protecting telomeres. TRF2 protein is dramatically upregulated upon neural differentiation of hESCs and maintained at high levels during the extended culture of NPCs. This upregulation appears to be specific to hESC neural differentiation as neither the differentiation of hESCs to other lineages nor differentiation of mESCs to neural progenitors revealed similar increases. These results are consistent with in vivo data, which show that TRF2 is expressed at a much higher level in the human brain than in other tissues [14], indicating that TRF2 may have an important function within neural tissues. Indeed, in this study we have demonstrated that ectopic expression of TRF2 in hESCs promotes neural differentiation, while knockdown of TRF2 impedes neural progenitor generation and maintenance, ultimately hindering the conversion of NPCs to neurons and glia. Our results suggest that TRF2 is a vital factor for the differentiation to and maintenance of NPCs, which may consequently have important implications on brain function.

In deciphering the molecular mechanisms underlying the role of TRF2 in neural differentiation, we discovered that the differential expression of TRF2 during neural differentiation correlates with the expression of hREST4 protein, a truncated REST generated by the alternative splicing of exon N from the REST gene. Although three transcripts, N4, N50, and N62, can be generated by alternative splicing of exon N at different sites and all lead to premature termination of the REST protein, only the latter two produce a truncated REST protein with same additional 13 AAs at the C-terminus (Supporting Information Fig. S5D), necessary for interaction with TRF2 [16, 39]. Since the N62 transcript was mainly expressed in neural tissues/cells as rodent REST4 [16], we named this protein as hREST4. Rodent REST4 has been widely reported to counteract the repression of REST on target genes, promoting neural gene expression and neurogenesis [35-38] as it lacks the C-terminal repression domain of REST and is unable to interact with CoREST [40, 41]. We show now that hREST4 is also able to promote the neural differentiation of hESCs, possibly through affecting the repression on neural genes by REST. Although overexpression of “hREST4” in murine neuroblastoma cells was found not to reduce the repression of REST on human synapsin promoter-driven luciferase expression [42], this was performed using only a single promoter in a tumour cell line, which may not be the case across all cell lines. Moreover, we demonstrated that TRF2 interacts with hREST4 to protect hREST4 from ubiquitin-mediated degradation by the proteasome, hence positively regulating neural progenitor formation and maintenance.

TRF2 has been reported to be capable of regulating neural differentiation in human embryonic carcinoma NTera2 cells by stabilizing REST expression [33]. Here we have taken this study into a more physiological context and demonstrated that TRF2 is also involved in regulating neural differentiation process of hESCs. However, we show that TRF2 enhances hREST4 expression during this process, which counteracts the repression of REST on neural genes to promote neural progenitor formation. The discrepancy of the two studies could be attributed to the two different cell types used where TRF2, REST, and hREST4 are differentially expressed. We have shown that TRF2 and REST are expressed at considerably higher levels in NTera2 cells than in hESCs, which may not reflect their physiological levels in normal pluripotent cells. This could affect their interaction and function due to the intricacies of protein complex formation. Nonetheless, the fact that high-TRF2 expressing NTera2 cells are more readily differentiated into neural lineages than hESCs further supports the idea of TRF2 promoting neural differentiation.

Reducing REST repression on its target genes plays an important role in controlling neural gene expression and neural differentiation. During neural differentiation of mESCs, repression of REST is alleviated by a considerable reduction of REST protein in which REST undergoes proteasomal degradation through β-TrCP-mediated ubiquitination [3]. However, in both H1 and H7 hESCs, we observed no such reduction of REST during this differentiation, probably due to the expression of HAUSP deubiquitylase and low levels of β-TrCP [25]. Instead, the hESCs exhibit a clear upregulation of hREST4 following neural differentiation, which is regulated by an increase of TRF2. Therefore, we propose an alternative mechanism that regulates this process. High levels of TRF2 stabilize hREST4, enhancing its accumulation, which subsequently counteracts the repression of REST on neural target genes, leading to the induction of neural differentiation (Fig. 7F). Our findings are further supported by the notion that alternative splice variants play a major role in the regulation of gene expression, which is particularly prominent in the CNS [43, 44]. Overall, our results suggest that repression of REST on its target genes is reduced in the neural differentiation of both hESCs and mESCs but is regulated through different mechanisms. The balance of hREST4 and REST is important for human neural differentiation and NPC maintenance for which high levels of TRF2 are crucial.


In summary, we have identified a novel extra-telomeric function for TRF2, a function that is of critical importance to the differentiation and maintenance of human NPCs. These progenitor cells are essential to the proper development and function of the CNS which underscores their therapeutic value in the potential treatment of neurodegenerative disorders. Given the high expression of TRF2 in the human brain, our findings will benefit further exploration of its role in normal brain function and aid in the improved procurement of human NPCs for use in future regenerative therapies.


We thank I. Jaeger and M. Li of Cardiff University for the mESCs, mEpiSCs (mouse post-implantation epiblast-derived stem cells), and mNPCs cell extracts and M. Sheldon for his technical help. We also thank M. Parker and V. Azuara for their comments. We are grateful to Dr. Shideng Bao from the Lerner Research Institute for the REST plasmid. This work was supported by funding from the BBSRC, MRC, and Genesis Research Trust.

Author Contributions

P.O-R.: conception and design, collection and assembly of data, data analysis and interpretation, and manuscript writing; J.S.L.Y.: collection and assembly of data and manuscript writing; S.T. and C.H.: collection, assembly, and analysis of data; W.C.: conception and design, data analysis and interpretation; manuscript writing, and final approval of manuscript.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.