AKT-dependent signaling of extracellular cues through telomeres impact on tumorigenesis

The telomere-bound shelterin complex is essential for chromosome-end protection and genomic stability. Little is known on the regulation of shelterin components by extracellular signals including developmental and environmental cues. Here, we show that human TRF1 is subjected to AKT-dependent regulation. To study the importance of this modification in vivo, we generate knock-in human cell lines carrying non-phosphorylatable mutants of the AKT-dependent TRF1 phosphorylation sites by CRISPR-Cas9. We find that TRF1 mutant cells show decreased TRF1 binding to telomeres and increased global and telomeric DNA damage. Human cells carrying non-phosphorylatable mutant TRF1 alleles show accelerated telomere shortening, demonstrating that AKT-dependent TRF1 phosphorylation regulates telomere maintenance in vivo. TRF1 mutant cells show an impaired response to proliferative extracellular signals as well as a decreased tumorigenesis potential. These findings indicate that telomere protection and telomere length can be regulated by extracellular signals upstream of PI3K/AKT activation, such as growth factors, nutrients or immune regulators, and this has an impact on tumorigenesis potential.


Introduction
Telomeres are specialized structures at the chromosome ends, which are essential for chromosome-end protection and genomic stability [1]. Vertebrate telomeres consist of tandem repeats PLOS Genetics | https://doi.org/10.1371/journal.pgen.1009410 March 9, 2021 1 / 29 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 of the TTAGGG DNA sequence bound by the so-called shelterin complex, which encompasses the TRF1, TRF2, TIN2, POT1, TPP1 and RAP1 proteins [2]. Shelterin ensures telomere protection by preventing end-to-end chromosome fusions, telomere fragility, and activation of DNA repair activities at chromosome ends [3]. Some components of the shelterin complex have also been described to regulate telomere length. In particular, TPP1 and POT1 have been shown to regulate telomerase activity at telomeres [4,5]. In addition, expression of dominantnegative mutants of TRF1 lead to telomere elongation [6], suggesting that TRF1 is a negative regulator of telomere elongation by telomerase. However, the fact that Trf1 knock-out mice and cells show a normal telomere length [7,8], indicates that the potential role for TRF1 in telomere length regulation is still poorly understood. In spite of these essential roles of shelterin in genome stability, telomere length regulation and cellular viability, very little is known on the potential regulation of telomere protection by key cell signaling pathways. We pioneered the understanding of the post-translational regulation of shelterin components by using an unbiased chemical biology approach [9][10][11]. In particular, we showed that mouse TRF1 is post-translationally modulated by a number of key cellular pathways, including the PI3K/AKT and Ras pathways [9][10][11]. In particular, we showed that small molecule chemical inhibitors of these pathways impair proper telomere protection [9][10][11]. Of particular interest is the regulation of TRF1 by the PI3K/AKT signaling pathway. The PI3K/AKT pathway responds to different extracellular signals including immune regulators, growth factors, and hormones [12,13]. In turn, activation of PI3K/AKT pathway regulates a plethora of target proteins to control metabolism, cell proliferation, survival, and cell growth [12,14]. We recently identified that an important target of the PI3K/AKT pathway is the TRF1 shelterin protein. In particular, we found that mouse TRF1 can be phosphorylated at three different residues by AKT and that these modifications are important for telomere protection [9]. However, a potential role for this TRF1 post-transcriptional modification in vivo is still unknown.
Here, we set to address the in vivo relevance of the regulation of human TRF1 by the PI3K/ AKT pathway by generating TRF1 knock-in human cell lines carrying single mutations in the AKT-dependent phosphorylation sites.

Human TRF1 is phosphorylated by AKT at residues T273 and T358
In a previous report, we described that mouse TRF1 can be directly phosphorylated by AKT at three independent residues (T248, T330, and S344), and that these phosphorylation events are necessary for TRF1 stability [9]. Here, we first set to address whether these findings could be extended to the human TRF1 protein. To this end, we carried out an in vitro kinase assay with affinity purified wild-type human histidine-tagged-TRF1 (His-TRF1-WT) incubated with human purified AKT in the presence of γ-32ATP (Materials and Methods). AKT yielded a clear human TRF1 phosphorylation signal (p-His-TRF1 in Fig 1A and 1B and S1 Source Data), which was dependent on the amount of purified GST-AKT used in the assay (GST-AKT in Fig 1A), supporting the specificity of the signal. As a positive control, we also detected a dose-dependent autophosphorylation of human AKT (Fig 1A).
Next, to identify the potential human TRF1 sites homologous to mouse AKT phosphorylation residues, we performed alignment of mouse and human TRF1 protein sequences using the Basic Local Alignment Search Tool (BLAST) (NCBI) (Fig 1C). We identified human TRF1 residues T273, T344 and T358 as the homologous of mouse TRF1 residues T248, T330, and S344, respectively, and as potential targets of AKT phosphorylation. Based on the consensus sequence for AKT phosphorylation, RxRxxS/T [12,14], out of the three identified residues T273 constitutes the most likely candidate. Interestingly, all three potential TRF1 AKT phosphorylation sites were located near the nuclear localization signal domain (Fig 1D). To in vitro validate these potential phosphorylation sites we generated the His-tagged TRF1 phosphomutant alleles T273A, T334A, and T358A (Materials and Methods) (see S1A and S1B Fig).
The affinity purified His-TRF1 wild-type or mutant proteins were incubated with human purified GST-AKT in the presence of γ-32ATP. We found significantly decreased TRF1 phosphorylation levels in T273A and T358A mutants but not in the T344A mutant compared to wildtype TRF1 (Fig 1E and S1 Source Data). To confirm that only T273 and T358 residues are important for AKT-dependent human TRF1 phosphorylation, we also assayed the phosphorylation level of the double T273A/ T358A mutant. The results show that AKT-mediated phosphorylation of the T273A/ T358A variant is fully blocked.
To address whether phosphorylation at residues T273 and T358 is important for TRF1 binding to telomeres, we generated phosphomimetic variants by replacing T273 and T358 residues for aspartic acid residues, i.e. TRF1-T273D and TRF1-T358D mutants. Affinity purified His-TRF1 wild-type or mutant proteins were incubated with radiolabeled double stranded telomeric DNA (an TTAGGG 7 oligo). The appearance of protein-DNA complexes by electrophoretic mobility shift assay (EMSA) showed that these phosphomimetic TRF1 mutant proteins bind more efficiently to telomeric DNA compared to wild-type TRF1 (Fig 1F and S1 Source Data). Together, these findings indicate that human TRF1 is phosphorylated by AKT at residues T273 and T358 and that these phosphorylation events are important for TRF1 binding to telomeric DNA in vitro.

Generation of human TRF1 knock-in cell lines for AKT-dependent phosphorylation sites
In order to address the importance of AKT-dependent TRF1 phosphorylation in vivo, we set to generate human cell lines knock-in for TRF1 alleles harboring inactivating mutations in AKT-dependent TRF1 phosphorylation sites. In particular, we used the CRISPR-Cas9 technology to generate non-phosphorylatable heterozygous and homozygous TRF1 mutant cell lines, namely TRF1 +/T273A , TRF1 T273A/T273A and TRF1 +/T358A , TRF1 T358A/T358A . To this end, singleguide RNAs (sgRNA) showing the lower off-target mutation score were selected for each mutation using the online CRISPR [15]. In addition, to introduce the substitution of the corresponding bases of each amino acid, we designed a 100-nucleotide single-stranded DNA oligo (ssODN) as the homologous repair template. This ssODNs were designed to change the codon "ACT" by "GCT" and for codon-usage optimization as well as to introduce one additional base mutation to avoid re-cleaving and a new restriction site for subsequent clone genotyping (Fig 2A). The plasmid expressing EGFP, Cas9 and sgRNA were co-transfected with each containing 5 μCi [ƴ-32-P] ATP. The reaction mixtures were resolved by SDS-PAGE and subjected to autoradiography. The SDS-PAGE was Coomassie-stained to quantify total TRF1 as loading control (below panel). B. Quantification of phospho-TRF1 (p-TRF1) levels normalized to total TRF1. C. Alignment of mouse and human TRF1 protein sequences using The Basic Local Alignment Search Tool (BLAST) NCBI. Letters in blue indicated the AKT1 kinase consensus sequence and in red indicate the conserved threonine between human and mouse. D. Schematic representation of human TRF1 protein depicting the N-terminal acidic domain, the TRFH domain, the nuclear localization signal (NLS), and the Myb domain. The location of the threonine residues potentially phosphorylated by AKT (T273, T344, and T358) is indicated. E. Quantification of p-TRF1 levels of non-phosphorylatable TRF1 variants normalized to wild-type p-TRF1. Representative image of in vitro phosphorylation assay of purified His-TRF1 (WT), His-TRF1 (T273A), His-TRF1 (T344A), His-TRF1 (T358A) and His-TRF1 (T273A/ T358A) by GST-AKT1 (0.2 μg). The reaction mixtures were resolved by SDS-PAGE and subjected to autoradiography. The SDS-PAGE was Coomassie-stained to quantify total TRF1 as loading control (below panel). F. Quantification of in vitro TRF1 binding to telomeric dsDNA (TTAGGG) 7 levels of His-tagged TRF1 WT and TRF1 phosphomimetic mutant variants (T273D and T358D). Representative image of the electrophoretic mobility shift assay (EMSA). The arrows show the position of the TRF1-DNA complexes and the free probe. A Coomassiestained SDS-PAGE is shown as loading control (below panel). Error bars represent standard deviation. n number of independent experiments. Student's t test was used for statistical analysis, P values are shown. � , p � 0.05; �� , p � 0.01; ��� , p � 0.001; n.s. = not significant. Graphical scheme of CRISPR-Cas9-based strategy followed to substitute the ACT codon for GCT in exon 6 and 9 to generate TRF1 T273A and TRF1 T358A alleles, respectively. The genomic and the template DNA sequence are shown where the nucleotide substitution for amino acid change and for novel restriction site introduction are indicated in red. B-C. The ssODN into Human embryonic kidney HEK 293T (ATCC CRL-3216). EGFP-positive cells were sorted by flow cytometry and amplified. The presence of the mutation in pooled cells was confirmed by PCR and restriction fragment length polymorphism analysis (RFLP). Single cell dilutions were plated onto 96-weel plates for clonal expansion and further RFLP identification of mutant clones [15]. The selected clones were then validated by Sanger DNA sequencing (Fig 2B and 2C). We isolated five independent heterozygous clones for both TRF1 +/T273A and for TRF1 +/T358A and two independent homozygous clones for TRF1 T273A/T273A and for TRF1 T358A/T358A . First, we performed Western blot analysis of nuclear TRF1 protein levels in wild-type and TRF1 heterozygous and homozygous mutant cell lines (TRF1 +/T273A , TRF1 T273A/T273A , TRF1 +/   T358A , TRF1 T358A/T358A ). We found that TRF1 protein levels were significantly lower in the TRF1 knock-in heterozygous cell lines than in wild-type cells, and we found a further decrease in TRF1 protein levels in the TRF1 knock-in homozygous cells (Fig 2D and S1 Source Data).
To exclude clonal variations, we analyzed TRF1 protein levels in independent clonal isolates (S2A and S2B Fig). We did not find any significant differences between the different clones of the same genotype ruling out that reduced TRF1 levels were due to off-targets effects. These findings may suggest that the mutant TRF1 proteins are less stable than the wild-type TRF1 protein, in agreement with a role of AKT-dependent phosphorylation of these residues in TRF1 protein stability. Interestingly, the decrease in TRF1 protein levels induced by the TRF1 mutations was similar than that induced by treatment of TRF1 wild-type cells with an AKT chemical inhibitor (MK22-06) or with a chemical inhibitor against the AKT upstream kinase PI3K (ETP-47037) (Fig 2D and 2E and S1 Source Data), previously demonstrated by us to reduce TRF1 levels in mouse cells [9]. Of interest, treatment of the different TRF1 mutant cell lines with AKT or PI3K inhibitors did not further decrease TRF1 protein levels, demonstrating that the effect of AKT or PI3K inhibitors on TRF1 protein levels is mediated by the TRF1 residues identified in this manuscript (Fig 2D and 2E).
Wild type cells were knocked down for the endogenous TRF1 by using a sh-TRF1 (S2C Fig  and S1 Source Data) and transfected with a vector harboring either wild-type MYC-tagged TRF1, MYC-tagged TRF1 T273D or with MYC-tagged TRF1 T358D alleles. Cells overexpressing wild type and phosphomimetic TRF1 alleles were then treated either with DMSO or with AKT chemical inhibitor (MK22-06) for 24 hours. Western blot analysis of nuclear TRF1 protein mutant TRF1 T273A (A) and TRF1 T358A (B) alleles were validated by RFLP analysis and Sanger sequencing. Representative images of ethidium bromide stained RFLP agarose gels (left panel) and of Sanger sequence spectrum (right panel) in TRF1 T273A/T273A and TRF1 T358A/T358A knock-in cells, respectively. D. Representative western blot images and quantification (right panel) of total nuclear of TRF1 protein levels in wild-type and heterozygous and homozygous mutant cells, treated either with DMSO or the PI3K inhibitor ETP47037 (10 μM). Phospho-AKT and total AKT were analyzed as positive control for the treatment. SMC1 was used as loading control. E. Representative western blot images and quantification (right panel) of total nuclear of TRF1 protein levels in wild-type and homozygous mutant cells, treated either with DMSO, AKT inhibitor MK22-06 (10 μM) or PI3K inhibitor ETP47037 (10 μM). Phospho-AKT and total AKT were analyzed as positive control for the treatments. SMC1 was used as loading control. F. Representative images and quantification (right) of TRF1 immunofluorescence in wild-type and mutant cells. Scale bars, 10 μm. G. Representative western blot images and quantification (right panel) of total nuclear of TRF1 protein levels in wild-type cells knocked down for endogenous TRF1 and overexpressing either wild-type MYC-tagged TRF1, MYC-tagged TRF1 T273D or with MYC-tagged TRF1 T358D alleles, treated either with DMSO or with AKT inhibitor MK22-06 (10 μM). Phospho-AKT and total AKT were analyzed as positive control for the treatment. SMC1 was used as loading control. Error bars represent standard deviation. n number of independent experiments. Student's t test was used for statistical analysis, P values are shown. � , p � 0.05; �� , p � 0.01; ��� , p � 0.001; n.s. = not significant.
https://doi.org/10.1371/journal.pgen.1009410.g002 PLOS GENETICS hTRF1 is regulated by PI3K/AKT pathway levels were performed using anti-TRF1 or anti-MYC antibodies. The results showed that the levels of wild type MYC-TRF1 decreased to 70% upon AKTi treatment while the levels of both MYC-TRF1 T273D and MYC-TRF1 T358D remained largely unaffected (Fig 2G and S1 Source Data). These results further demonstrate that residues T273 and T358 are key for AKT-dependent TRF1 stability.
Human TRF1 knock-in cell lines for AKT-dependent phosphorylation sites show increased proteasome-mediated TRF1 protein degradation Next, we set to determine the molecular mechanisms underlying decreased TRF1 protein levels in the TRF1 mutant cells. Of relevance, we could not attribute the changes in TRF1 protein expression to changes in TRF1 mRNA expression. In particular, although we detected significantly altered TRF1 mRNA levels in cells carrying heterozygous TRF1 alleles, this effect was not detected in cells carrying the TRF1 homozygous alleles (Fig 3A and S1 Source Data).
To directly study TRF1 protein stability, we performed a cycloheximide chase assay of protein degradation (Materials and Methods). Cells carrying the mutant TRF1 alleles showed significantly lower TRF1 protein levels at different times after cycloheximide treatment compared to wild-type cells (Fig 3B and S1 Source Data), suggesting a decreased TRF1 mutant protein stability. Interestingly, treatment with the proteasome inhibitor bortezomib rescued TRF1 protein levels in the different TRF1 mutant cell lines, indicating that lower protein stability of the TRF1 mutants is caused by proteasome-mediated degradation (Fig 3B). As a negative control, we also analyzed the stability of a different shelterin component, RAP1 in both wild-type and TRF1 mutant cell lines (Fig 3C and S1 Source Data). We did not find differences in RAP1 stability between wild-type and TRF1 mutant cell lines (Fig 3C), indicating that the AKT-dependent TRF1-dependent regulation does not affect RAP1 protein levels. Interestingly, it is known that the inhibition of protein synthesis by cycloheximide induces phosphorylation/activation of AKT [16]. Thus, we set to address AKT phosphorylation after cycloheximide treatment in the different Trf1 wild-type and mutant cell lines. We found increased AKT phosphorylation at Ser473 both in TRF1 will-type and mutant cells (Fig 3D  and S1 Source Data). We confirmed these results by measuring phosphorylation of an AKT downstream substrate, the ribosomal protein S6, and again found increased levels of p-S6 in both TRF1 will-type and mutants cell lines, in agreement with increased phospho-AKT ( Fig  3E and S1 Source Data). These findings indicate that the cycloheximide-dependent AKT phosphorylation stabilizes the wild-type TRF1 protein, but this does not occur in the TRF1 mutant cell lines, demonstrating the importance of T273 and T358 TRF1 residues in AKTmediated TRF1 post-translational regulation.
Human TRF1 knock-in cell lines for AKT-dependent phosphorylation sites show increased global and telomeric DNA damage and increased telomere fragility TRF1 has an essential role in protecting the chromosome ends from eliciting a persistent DNA damage response (DDR), as well as in preventing telomere fragility as detected by the occurrence of multiple telomere signals per chromosome end (MTS) [7,8]. Thus, we first set to address the effects of mutant TRF1 proteins in both global and telomeric DNA damage. To this end, we performed either immunofluorescence with γH2AX to detect global DNA damage or double immunofluorescence of γH2AX and the telomeric protein RAP1 to detect DNA damage specifically located at telomeres (the so-called telomere induced foci or TIFs) [17]. We found that homozygous TRF1 mutant cell lines showed significantly increased levels of both global and telomeric DNA damage while heterozygous TRF1 mutant cell lines showed normal global and telomeric DNA damage levels (Fig 4A and 4B and S1 Source Data), indicating that AKT-dependent TRF1 phosphorylation is essential for preventing telomeres from eliciting a DNA damage response. We checked the cell cycle status of wild-type and mutant cells and found no significant differences among the genotypes, ruling out cell cycle-dependent variations in the DDR blot (S3A Fig and S1 Source Data). TRF1 inhibition has been previously shown to induce the so-called multitelomeric signals (MTS), which are associated to increased telomere fragility and increased telomere damage [7,8]. Thus, we next tested whether the non-phosphorylatable TRF1 mutant cells lines showed an increase in MTSs. To this end, we performed telomere quantitative fluorescence in situ hybridization (Q-FISH) on metaphase spreads to visualize telomeres (Materials and Methods) in TRF1 wild-type, as well as TRF1 heterozygous and homozygous mutant cell lines (Fig 4C and S1 Source Data). We found that homozygous TRF1 mutant cells showed significantly elevated number of MTSs compared to the TRF1 wild-type controls. Of interest, one of the heterozygous mutants, TRF1 +/T358A , also showed significantly increased MTSs in this assay, suggesting haploinsufficiency for the MTS phenotype.
TIF and MTS induction after TRF1 depletion is accompanied by activation of the ATM/ ATR signaling pathways and the phosphorylation of their downstream checkpoint kinases CHK2 and CHK1 [7,8]. To address the impact of TRF1-T273A and TRF1-T358A mutant variants on DDR activation, we analyzed phospho-CHK2 and phospho-CHK1 levels by Western blot (S3B and S3C Fig and S1 Source Data). The results show that mutant cells present increased phospho-CHK2 and phospho-CHK1 levels compared to wild-type cells, a DDR characteristic of replicative damage.

TRF1 knock-in cell lines for AKT-dependent phosphorylation show reduced proliferation ability
The PI3K/AKT pathway is known to promote cell proliferation and survival throughout modulation of different targets [14,18]. As we have shown here that TRF1 is a bona fide target of this pathway, we set to address whether the human TRF1 knock-in cell lines expressing nonphosphorylatable TRF1 proteins in the AKT phosphorylation sites, showed any cell proliferation defects. Interestingly, both heterozygous and homozygous TRF1 mutant cell lines showed reduced cell proliferation as compared to the wild-type TRF1 controls (S4 Fig and S1 Source Data). The TRF1 +/T273A heterozygous cell line showed a severe phenotype which was comparable to that of the homozygous cell lines (S4 Fig). These results suggest that post-translational modifications in the shelterin complex at chromosome ends can modulate the ability of cells to proliferate.

TRF1 knock-in cell lines for AKT-dependent phosphorylation show telomere length defects
Telomere length regulation is a key process in cancer and aging. On one hand, telomere shortening associated to cell division and tissue repair throughout the organismal lifespan is T273A and TRF1 T358AT/358A cell lines by cycloheximide (CHX) pulse chase. Cells were treated with 100 μg/ml of CHX in the absence or presence of 5 μM bortezomib. At the indicated times, the cells were harvested to prepare nuclear extracts, followed by Western blot analysis against TRF1 (B) or RAP1 (C) and SMC1 (as loading control) (lower panels). D-E. Phospho-AKT (D) and phospho-S6 (E) levels were determined in wild type, TRF1 T273A/T273A and TRF1 T358AT/358A cell lines at the indicated times during CHX pulse chase. Total AKT (D) and S6 (E) was used as loading control. The graph represents the mean value for each time point and corresponding standard deviation. n number of independent experiments. Student's t test was used for statistical analysis, P values are shown. � , p � 0.05; �� , p � 0.01; n.s. = not significant.  proposed to be one of the molecular causes underlying aging and age-related diseases [19]. On the other hand, cancer cells need maintenance of telomeric repeats in order to divide indefinitely and they achieve that by activating telomere maintenance mechanisms such as telomerase activity in the majority of tumors, or less frequently, activating alternative mechanisms of telomere lengthening based on homologous recombination between telomeric sequences (Alternative Lengthening of Telomeres or ALT) [20]. As some reports suggested a role for TRF1 as a negative regulator of telomere length [6] while other reports did not show any telomere length defects upon TRF1 genetic deletion in mice [7,8], here we set to investigate the impact of AKT-dependent phosphorylation of TRF1 in telomere length regulation in vivo. To this end, we first studied telomere length by using telomere restriction fragment analysis (TRF) based on Southern blotting [4,21](Materials and Methods). We found significantly decreased telomere length both in heterozygous and homozygous TRF1 mutant cell lines compared to wild-type cells (Fig 5A). These defects were not aggravated when treating with ETP47034, an PI3Kα chemical inhibitor known to inhibit AKT phosphorylation (Fig 5A), again suggesting that the effects of the inhibitor are mediated by the AKT-dependent phosphorylation of TRF1. We confirmed shorter telomeres in cells carrying TRF1 homozygous mutant alleles by an independent technique based on telomere fluorescence known as quantitative telomere FISH or Q-FISH (Materials and Methods), which measures individual telomere signals (Fig 5B and S1 Source Data) [21,22]. In agreement with telomere maintenance defects in the TRF1 mutant alleles, we found that homozygous TRF1 mutants showed significantly shorter telomeres than TRF1 wild-type cells as well as a dramatic increase in the percentage of short telomeres (defined as the % of telomeres showing a lower fluorescence intensity than the 15% percentile of control wild-type cells) (Fig 5B). Furthermore, we also found and increase in the number of undetectable telomere signals (referred here as "signalfree" ends) per metaphase in TRF1 homozygous mutant cells (Fig 5B), again indicative of impaired telomere maintenance. The shorter telomere phenotype was confirmed in independent clones, ruling out potential clonal variations and demonstrating that TRF1 mutant variants are causative of telomere length defects (S5 Fig).
In agreement with defective telomere maintenance of AKT non-phosphorylatable TRF1 mutants, we observed that the shorter telomere phenotype was aggravated with increasing cell passage in both homozygous mutant cell lines compared to wild-type cells (Fig 5C). Importantly, the short telomere phenotype as well as the presence of short telomeres and of "signalfree ends" was rescued upon over-expression of wild-type flag-tagged TRF1, but only in TRF1 T273A/T273A cells and not in TRF1 T358A/T358A cells (Figs S6A and 5B and 5C,). These findings indicate that telomere shortening phenotype induced by the TRF1 T273A/T273A mutant protein is the direct effect of the mutation as this effect can be rescued by over-expression of the wild-type TRF1 protein. However, telomere shortening induced by the TRF1 T358A/T358A mutant, is likely to be indirect as it cannot be rescued by over-expression of the wild-type TRF1 protein.
It is possible that the TRF1 T358A/T358A mutation may also be modifying the structure of the telomere capping complex, thus impeding proper access of the wild-type TRF1 protein.
To understand the consequences of TRF1 abrogation, we downregulated TRF1 protein levels by using and shRNA against TRF1 (Materials and Methods). To this end, wild-type and T273A , TRF1 +/T358A and TRF1 T358A/T358A cell lines. Scale bars, 10 μm. B. Representative images (above panel) and quantification (below panel) of percentage of cells presenting three or more γH2AX and RAP1 co-localizing foci (TIFs) (white arrowhead). Scale bars, 10 μm. C. Representative images (above panel) and quantification (below panel) of multitelomeric signals (MTS) (white arrow heads). Insets represent high magnification images. Scale bars, 5μm. Error bars represent standard deviation. n number of independent experiments (A,B) or number of metaphases (C). Student's t test was used for statistical analysis, P values are shown. � , p � 0.05; �� , p � 0.01; ��� , p � 0.001.
https://doi.org/10.1371/journal.pgen.1009410.g004 PLOS GENETICS hTRF1 is regulated by PI3K/AKT pathway We confirmed decreased TRF1 mRNA and protein levels in the cell lines transduced with the shTRF1 (S6B and S6D Fig and S1 Source Data). We next measured telomere length by telomere restriction analysis at increasingly cell passages (Fig 5D). In agreement with previous reports [6], TRF1 knock-down induced a significant progressive telomere lengthening in both wild-type and in TRF1 mutant cells confirming a TRF1 role as a negative regulator of telomere elongation. We confirmed these results by measuring telomere length by Q-FISH in metaphase spreads from TRF1 +/+ , TRF1 T273A/T273A and TRF1 T358A/T358A cells expressing shTRF1 at passage 14 (S6D Fig). These results also indicate that telomere length defects shown by TRF1 T273A/T273A and TRF1 T358A/T358A mutant cells are not due to the lower protein stability and consequent reduction in TRF1 protein levels in mutant cells (Figs 2D, 2F, and 3B) but rather to a de novo gain of function of these mutations (Fig 5D).
The progressive telomere elongation observed upon TRF1 downregulation in wild-type, TRF1 T273A/T273A and TRF1 T358A/T358A cells indicates a telomerase-dependent telomere lengthening and suggests that depleted TRF1 levels may allow telomerase to access telomeres. In order to address whether telomerase overexpression was able to bypass defective telomere maintenance in the mutant TRF1 cell lines, we transfect wild-type, TRF1 T273A/T273A and TRF1 T358A/T358A cells with a vector containing the human TERT [23]. We confirmed telomerase over expression by qPCR (S6E Fig and S1 Source Data). We next measured telomere length by telomere restriction analysis at increasingly cell passages (Fig 5E). The results show that telomerase overexpression induces a significant and progressive telomere lengthening in both wild-type and in TRF1 mutant cells.
We and others previously showed that induction of pluripotency by transduction of the Yamanaka factors in mouse and human induced pluripotent stem (iPS) cells leads to de novo telomere elongation mediated by telomerase [24,25]. Thus, we next set to address telomere length dynamics upon induction of pluripotency in TRF1 +/+ cells, as well as in the TRF1 T273A/ T273A and TRF1 T358A/T358A mutants. To this end, wild-type and mutant cell lines were transfected with OCT4, SOX2, KLF4 and c-MYC containing vectors to induce reprogramming. First, in agreement with induction of pluripotency, we observed increased Nanog mRNA levels in all the cell lines upon reprogramming with the Yamanaka factors (S7A Fig and S1 Source Data). Of interest, TRF1 transcriptional levels did not change upon induction of reprogramming (S7B Fig and S1 Source Data), in contrast with previous reports in which TRF1 was found to be over-expressed upon reprogramming of mouse embryonic fibroblasts [26]. This could be due to the fact that the HEK293T cell line used here is embryo-derived and thus have already elevated TRF1 levels. We next measured in parallel telomere length both by telomere restriction analysis and by Q-FISH in metaphase spreads from reprogrammed TRF1 +/+ , TRF1 T273A/T273A and TRF1 T358A/T358A cells at passage 14 (S7C and S7D Fig and S1 Source Data). Although TRF1 T273A/T273A and TRF1 T358A/T358A cells showed shorter telomeres than wild-type controls before reprogramming, induction of pluripotency resulted in significant telomere elongation both in TRF1 wild-type and TRF1 mutant cells compared to the nonreprogrammed counterparts (S7C and S7D Fig). As the levels of TRF1 did not significantly magnification images. Scale bars, 5μm. Quantification of the mean telomere length, the percentage of short telomeres and the number of signalfree ends per metaphase by Q-FISH analysis. The percentage of short telomeres is defined as fluorescence intensity less that the 15 th

PLOS GENETICS
hTRF1 is regulated by PI3K/AKT pathway change during reprogramming (S7B Fig), the observed telomere elongation suggests that telomerase upregulation upon induction of pluripotency [24] was able to bypass defective telomere maintenance in the mutant TRF1 cell lines.

Shelterin complex formation in TRF1 knock-in cell lines for AKTdependent phosphorylation
To address whether the expression of different proteins of the shelterin complex was normal in the TRF1 T273A/T273A and TRF1 T358A/T358A knock-in cell lines, we first analyzed the abundance of different shelterins by performing Western blotting of whole cell lysates of wild-type, TRF1 T273A/T273A and TRF1 T358A/T358A cells, as well as the same cells overexpressing wild-type FLAG-TRF1 (Fig 6A and 6B and S1 Source Data). As expected, we found a decrease of TRF1 protein levels in the TRF1 T273A/T273A and TRF1 T358A/T358A cells and this was rescued when expressing the wild-type FLAG-TRF1 in both mutants (Fig 6A and 6B). In spite of decreased TRF1 levels, we observed normal levels of the rest of the shelterin components in TRF1 T273A/ T273A and TRF1 T358A/T358A cells (Fig 6A and 6B), suggesting that the mutant TRF1 proteins do not interfere with the stability of shelterin components.
To address whether the non-phosphorylatable TRF1-T273A and TRF1-T358A mutants were affected in binding to known TRF1-interacting shelterin components, namely POT1 and TIN2 [27][28][29], we transfected wild-type cells with either Flag-TRF1, Flag-TRF1-T273A or Flag-TRF1-T358A in HEK293T cells and performed co-immunoprecipitation with anti-Flag antibody (Fig 6C and 6D and S1 Source Data). As control, expression of the different TRF1 mutant proteins was indistinguishable from that of wild-type TRF1 (Fig 6C). Interestingly, similar amounts of the POT1 and TIN2 shelterins were pulled down in cells expressing either Flag-TRF1, Flag-TRF1-T273A or Flag-TRF1-T358A, indicating that these mutations do not interfere with TRF1 interactions with the shelterin components (Fig 6D and S1 Source Data). Finally, by performing chromatin immunoprecipitation experiments in these transfected wildtype cells, we observed a 60% and 40% reduced TRF1-T273A and TRF1-T358A binding to telomeric DNA in vivo, respectively (Fig 6E and S1 Source Data). Interestingly, binding of TRF2 to the telomere was unaffected in spite of the decreased TRF1 levels bound to telomeres (Fig 6F and S1 Source Data). These findings are in agreement with normal TRF2 foci formation in the cells genetically depleted for TRF1 [8], suggesting that the AKT-dependent TRF1 phosphorylation specifically regulates TRF1 binding to telomeres and that this is sufficient to affect telomere protection and telomere length even in the presence of normal TRF2 levels.

Extracellular cues are transduced to telomeres through PI3K/AKT signaling pathway
The PI3K/AKT pathway responds to different extracellular signals to regulate cell proliferation and growth [12,14]. TRF1 is required for proper telomeric DNA replication [7,8]. However, a direct cause-effect connection between extracellular signaling and changes in telomere regulation have not been demonstrated to date. Since we found here that human TRF1 is regulated by PI3K/AKT pathway, we next set to address the potential in vivo role of AKT-dependent TRF1 phosphorylation in transducing extracellular signaling to telomeres. To this end, we analyzed TRF1 protein levels under normal fed conditions in DMEM+FBS, starved for 24 hours in DMEM without FBS, and either refed for 5 hours with FBS or with FBS plus PI3K inhibitor (ETP47037) in TRF1 wild-type and homozygous knock-in cells for non-phosphorylateble TRF1 alleles (TRF1 T358A and TRF1 T273A ). Interestingly, we find that in wild-type cells, TRF1 levels decrease by 50% under starvation conditions and that upon FBS administration TRF1 levels increase to initial levels. The increase in TRF1 levels upon re-feeding was fully blocked  (Fig 7A and S1 Source Data), indicating that is dependent on the PI3K/AKT pathway. To further demonstrate this, we repeated the experiment in the different TRF1 mutant cells impaired for AKT-dependent TRF1 phosphorylation. Interestingly, in the mutant cell lines, the levels of TRF1-T273A and TRF1-T358A proteins remained unaltered upon changes in the extracellular nutrient availability. As control for PI3K activation/inactivation we assayed phospho-AKT (S473) levels ( Fig 7A).
As the PI3K/AKT pathway is known to be a major signaling pathway downstream of insulin signaling, we next addressed whether addition of insulin to starved cells (DMEM without FBS) affected TRF1 levels (Fig 7B and S1 Source Data). Addition of insulin to starved wild-type cells induced a 80% increase in TRF1 levels, which was impaired in the presence of either PI3K (ETP47037) or AKT chemical inhibitors (MK22-06) (Fig 7B). Again, when we performed the same experiment in different TRF1 mutant cells impaired for AKT-dependent TRF1 phosphorylation, the levels of non-phosphorylatable TRF1 variants, TRF1-T273A and TRF1-T358A, remained unchanged upon addition of either insulin or insulin plus AKTi/ PI3Ki. To address whether the levels of other shelterin components did also change upon extracellular insulin signals, we checked TIN2, TRF2 and RAP1 protein levels in these conditions. TIN2, TRF2 and RAP1 levels remained unaltered upon addition of either insulin or insulin plus AKTi/PI3Ki (Fig 7B). As control for PI3K activation/inactivation we assayed phospho-AKT (S473) levels (Fig 7B).
In summary, these results demonstrate the importance of AKT-dependent TRF1 phosphorylation in residues T273 and T358, for the proper transduction of extracellular cues to telomeres.

AKT-dependent TRF1 phosphorylation impacts on the tumorigenic capability of cancer cells
Both the PI3K/AKT and the telomere maintenance pathways are frequently mutated in cancer [20,30]. Here, we set to address whether human cancer cells mutant for the AKT-dependent TRF1 phosphorylation sites identified here retained their tumorigenic ability. To this end, HEK 293T TRF1 wild-type and homozygous knock-in cells for non-phosphorylatable TRF1 alleles (TRF1 T358A and TRF1 T273A ), were subcutaneously injected into nude mice (Material and Methods). Tumor growth was measured by caliper every second day. Xenografts derived from TRF1 T358A/T358A and TRF1 T273A/T273A cells showed a delayed tumor onset and a significant slower tumor growth as compared to TRF1 +/+ cells (Fig 8A and S1 Source Data). All the mice were sacrificed 15 days post-injection when wild-type tumors reached large volume (1500mm 3 ). Postmortem tumor analysis of mutant xenografts (TRF1 T358A/T358A and TRF1 T273A/T273A ) revealed a significant reduction in tumor size and weight as compared to wild-type tumors (Fig 8A and 8B and S1 Source Data). Tumors derived from TRF1 T358A/ T358A and TRF1 T273A/T273A cells showed a significant decrease in the proliferation marker Ki67 and an increase in the DNA damage marker γH2AX (Fig 8C, 8D, and 8E and S1 Source TRF1 +/+ , TRF1 T358A/T358A and TRF1 T273A/T273A cells. B. Representative western blot images (left) and quantification (right) of TRF1, TIN2, POT1, TRF2 and RAP1 protein levels in TRF1 +/+ , TRF1 T358A/T358A and TRF1 T273A/T273A cells expressing FLAG-TRF1-WT. C. Total TRF1 levels in wild type HEK293T cells expressing either Flag-TRF1, Flag-TRF1-T273A or Flag-TRF1-T358A by western blot. D. Flag-TRF1, Flag-TRF1-T273A and Flag-TRF1-T358A co-immunoprecipitate with endogenous TIN2 and POT1. Co-immunoprecipitation was performed from lysates of HEK293T cells expressing either the empty vector or the indicated Flag-TRF1 alleles. Quantification of TIN2 and POT1 pulled down with anti-Flag. E-F. Representative images of chromatin immunoprecipitation (ChIP) of telomeric DNA and of ALU sequences with anti-Flag (E) and with anti-TRF2 (F) of HEK293T cells expressing the indicated Flag-TRF1 alleles. DNA input signal is also shown. Quantification of telomeric DNA pulled down with anti-Flag (E) or with anti-TRF2 (F) is shown to the right. ChIP values are normalized by the input of each individual sample. Bars and error bars represent mean values ± SE, n = number of independent experiments. Student's ttest was used for the statistical analysis; � p < 0.05, �� p < 0.01, ��� p < 0.001 ns, no significant. Data). TRF1 T358A/T358A and TRF1 T273A/T273A mutant tumors also presented lower cellularity and larger necrotic areas as compared to wild-type tumors (Fig 8C). Interestingly, we did not observed significant differences in apoptosis between wild-type and mutant tumors as indicated by the percentage of AC3 positive cells (Fig 8C and 8F and S1 Source Data). As expected, TRF1 immunofluorescence analysis showed a sharp reduction in TRF1 levels of 65% and 82% in TRF1 T273A/T273A and TRF1 T358A/T358A mutant tumors, respectively (Fig 8C and 8G and S1 Source Data). Finally, telomere Q-FISH analysis showed that TRF1 T273A/T273A and TRF1 T358A/T358A tumor cells showed a 25-33% decrease in mean nuclear telomere fluorescence compared to wild-type tumors (Fig 8C and 8H and S1 Source Data). Together, these results show that non-phophorylatable by AKT TRF1 mutant cells show a decreased tumorigenic potential.

Discussion
Telomeres, the protective structures at chromosome ends, are essential for chromosome stability and cell viability [2]. Telomeres can also control gene-expression programs important for metabolism and pluripotency [31][32][33][34]. These important functions of telomeres are exerted by the binding of the protective shelterin complex to telomeric DNA repeats. In addition, shelterin can also regulate telomerase access to chromosome ends, thereby also regulating telomere length [4,35]. In spite of these key roles of shelterin, very little is known of the regulation of shelterin function by extracellular signals and environmental or developmental cues.
Here, we show that AKT dependent TRF1 phosphorylation has important roles in vivo. We generated non-phosphorylatable human cell lines for the AKT-dependent TRF1 phosphorylation sites by using the CRISPR-Cas9 technology. We demonstrate that AKT-dependent TRF1 phosphorylation is essential for TRF1 protein stability and TRF1 binding to telomeres, and thus, for preventing a persistent telomere DNA damage signaling and occurrence of telomere aberrations. These telomere defects induced by the TRF1 mutants occurred in the absence of changes in the protein levels of the rest of the shelterin components. Binding of the TRF2 shelterin to telomeres was also normal, thus supporting the notion that abolishing AKT-dependent TRF1 phosphorylation is sufficient to disrupt shelterin function even in the presence of the rest of the shelterin components. Importantly, phosphomimetic TRF1 variants show a significant higher affinity binding to telomeric DNA "in vitro", thus supporting a role for these residues in TRF1 binding to telomeres.

PLOS GENETICS
hTRF1 is regulated by PI3K/AKT pathway

PLOS GENETICS
hTRF1 is regulated by PI3K/AKT pathway over-expression of wild-type TRF1 in the homozygous TRF1 T273A/T273A mutant cells but not in the TRF1 T358A/T358A cells, suggesting that there are different mechanisms responsible for telomere shortening associated to the AKT non-phosphorylatable TRF1 mutants. In the case of TRF1 T358A/T358A cells, the fact that TRF1 over-expression cannot rescue the short telomere phenotype suggest that this mutation could have a dominant negative effect. These findings are further supported by the fact that, in contrast to mutant TRF1 proteins in the AKT phosphorylation sites which show short telomeres, a complete abrogation of TRF1 proteins levels by shRNA results in telomere elongation as previously reported [6], suggesting that telomere length defects shown by TRF1 T273A/T273A and TRF1 T358A/T358A mutant cells are not due to the lower protein stability and consequent reduction in TRF1 protein levels in mutant cells but rather to a de novo gain of function of these mutations exerting a dominant negative effect. The fact that TRF1 phosphorylation-defective mutants do not impair telomere lengthening by telomerase and given the functional TRF1 role in telomeric DNA replication [7,8], suggest that the effects of T273A and T358A mutations in telomere shortening are due to a defect of these variants in assisting telomere replication either by impeding replication of telomeres or by enhancing the degree to which replication forks stall and telomere breakage occurs. In support of this, we observed elevated MTS and higher levels of phospho CHK1 in TRF1 T273A/T273A and TRF1 T358A/T358A mutant cells, indicative of telomeric DNA replicative damage. Future research is needed to study the effect on telomere maintenance of AKT-dependent regulation of TRF1 in ALT cells.
These findings also suggest that TRF1 may be regulated in response to extracellular signals via AKT-dependent TRF1 phosphorylation. Indeed, we show here that TRF1 levels are regulated by extra-cellular signals mediated by the PI3K/AKT pathway such as starvation or exposure to insulin and that TRF1 mutant cells in the AKT-dependent phosphorylation sites show an impaired response to these proliferative extracellular signals. The response to extracellular cues at telomeres seems to be specific for TRF1 since other shelterin component remain unaffected to milieu changes.
Furthermore, in line with this, we also show a decreased tumorigenesis potential of AKT non-phosphorylatable TRF1 mutants, which is coincidental with shorter telomeres and increased DNA damage in the TRF1 mutant tumors. As PI3K mutations and AKT hyper-activation frequently occurs in cancers, our results suggest that these alterations may also favor TRF1 stability, telomere protection and telomere maintenance, a hallmark of cancer cells [20].

Generation of TRF1 knock-in cell lines by CRISPR-Cas9 system
The sequence for each sgRNA were obtained using an online CRISPR Design Tool [15]. The sgRNA were cloned into the plasmid pSpCas9(BB)-2A-GFP (Addgene,#48138) using a protocol described in [15]. A homologous 100-nt single-stranded DNA oligonucleotides (ssODN) harboring the desired nucleotide substitutions was used as template for homologous recombination repair of the CRISPR-Cas9 cleavage site. The ssODN was designed to change the Cas9 cleavage site to avoid re-cutting and to introduce a new restriction site for genotyping without altering the encoded amino acid sequence [43]. Sequences of sgRNA and ssODN can be found in S1 Table.

Cells and culture conditions
Human embryonic kidney (HEK) 293T cells were purchased from ATCC and grown in Dul-becco´s modified Eagle´s medium (DMEM) with 10% of fetal bovine serum (FBS) and 1% (vol/vol) penicillin-streptomycin (Gibco). The cells were grown at 37˚C with 5% CO 2 . Cell lines were regularly tested for mycoplasma using the Myco Alert Mycoplasma Detection Kit (Lonza). Cells were seeded on 6-well plates at 100,000 cells per ml. At 60% confluence, HEK293T cells were transfected with 2μg Cas9 plasmid and 1μg ssODN using Fugene HD transfection (Promega) following the manufacturers' protocol.
To generate TRF1 knock-in HEK293T cells were transiently co-transfected with plasmid pSpCas9(BB)-2A-GFP and specific ssODN. Two days after cells were trypsinized, washed with PBS and re-suspended in FACS solution (0.1% BSA, 3 mM EDTA in PBS). Single cells of the 10% GFP brightest ones were sorted using the FACS ARIA IIU (Becton Dickinson) and plated onto 96-wells plates for monoclonal cell expansion. Genomic DNA was obtained using a DNeasy kit (69504, QIAGEN) and different TRF1 locus fragments were amplified using Phusion High-Fidelity DNA Polymerase (F53OL, Thermo Fisher) and primers positioned outside of the corresponding mutated sequence (S1 Table). Restriction-fragment length polymorphism (RFLP) analysis using PvuI (R0150, New England Biolabs) for T273A mutation or PstI (R0140, New England Biolabs) for T358A mutation was performed for clone identification. Selected clones were further validated by Sanger sequencing. Homozygous and heterozygous monoclonal cell lines were selected.

Western blot analysis
Nuclear and cytoplasmic protein extracts were obtained using a Nuclear/Cytosolic Fractionation Kit (K266-100, Biovision) and protein concentration was determined using a Bradford Reagent (B6916, Sigma Aldrich). Twenty micrograms of nuclear extracts and forty micrograms of cytoplasmic were separated in 4-12% SDS-PAGE gels (NuPAGE Invitrogen) and transferred to nitrocellulose membranes (Amersham Protan). Blots were incubated with the indicated antibodies. Antibody binding was detected after incubation with a secondary antibody coupled to horseradish peroxidase using a chemiluminescence with ECL detection kit (GE Healthcare). In the protein stability assay, cells were treated with cycloheximide (100 mg/ ml) (Sigma-Aldrich) and collected at indicated time points. Relative protein levels in the immunoblots were quantified using ImageJ image analysis software. The protein levels were normalized to SMC1 and the relative levels at time 0 defined as 1. The

Telomeres Q-FISH and chromosomal aberrations on metaphase
HEK293T wild type and mutant cells were incubated with 0.1 μg/ml during 4 h. After hypotonic swelling in 0.03 M sodium citrate for 30 min at 37˚C, cells were fixed in methanol:acetic acid (3:1). Quantitative telomere fluorescence in situ hybridization (Q-FISH) was performed as described [45]. Images were captured using microscope Leica DM6B using a 100x oil objective. Telomere length was analyzed using Leica Application Suite X Software. The images were analyzed blindly.

RNA and qPCR
Total RNA from cells was extracted with the RNeasy kit (74106, QIAGEN) and reverse transcribed was using the iSCRIPT cDNA synthesis kit (1708891, BIO-RAD) according to manufacturer's protocol. Quantitative real-time PCR was performed with the QuantStudio 6 Flex (Applied Biosystens, Life Technologies) using Go-Taq Green Master Mix (M7123, Promega) according to the manufacturer's protocol. All values were obtained in triplicates. Primers used were RT-TRF1-F and RT-TRF1-R. Alpha Tubulin was used as the housekeeping gene using primers RT-Tubulin-F and RT-Tubulin-R. Sequences can be found in S1 Table. We determined the relative expression of TRF1 in each sample by calculating the 2ΔCT value. For each sample, 2ΔCT was normalized to control 2ΔCT mean.

Telomere Restriction Fragment analysis (TRF)
Cells were isolated and embedded in agarose plugs and digested with MboI (R0147, New England Biolabs). DNA was separated by gel electrophoresis in 0.5X TBE maintained at 14˚C, using a CHEF DR-II pulsed-field apparatus (Bio-Rad) for 12 h at 5 V/cm at a constant pulse time of 5 s. The gel was transfer to nylon membrane (Hybond-XL, GE Healthcare) and probed with a 32 P-labeled telomeric probe (TTAGGG) n (a gift from T. de Lange). Mean TRF length were determined using ImageQuant TL.

Co-immunoprecipitation
For immunoprecipitation, cells were washed with PBS and were lysed in 1X cell lysis buffer (9803, Cell Signaling Technology) supplemented with Complete protease inhibitors (Roche). After incubation on ice for 20 min with occasional mixing, the cell lysates were centrifuged at 14,000 rpm for 10 min at 4˚C and supernatants were collected for immunoprecipitation. We prepared lysates from one 10-cm dish and mixed them with 50 μl of a 50% slurry of Flag M2 agarose beads (A2220, Sigma-Aldrich) overnight, immunoprecipitates were washed four times whit lysis buffer and eluted with Laemmli loading buffer.

ChIP assay and telomere dot-blots
For ChIP analysis, 3 x 10 6 of HEK293T cells stably expressed Flag-TRF1, Flag-TRF1-T273A and Flag-TRF1-T358A cells were used per condition. Formaldehyde was added directly to culture medium to a final concentration of 1% and incubated for 15 min at room temperature (RT) on a shaking platform. Cross-linking was then stopped by addition of glycine to a final concentration of 0.125 M for 5 min at RT. Cross-linking cells were washed twice with cold PBS containing 1 μM PMSF and protease inhibitors and then pelleted. Cells were lysed in lysis buffer (1% SDS, 10 mM EDTA and 50 mM Tris-HCl pH 8.0) containing protease inhibitors at 2 × 107 cells/ml for 20 min at 4˚C. Lysates were sonicated to obtain chromatin fragments <1 kb and centrifuged for 15 min in a microfuge at room temperature. Chromatin was diluted 1:10 with dilution buffer (1.1% Triton X-100, 2 mM EDTA pH 8.0, 150 mM NaCl, 20 mM Tris-HCl pH 8.0) contained protease inhibitors and precleared with 50 μl of protein A/G Plus-Agarose beads (sc-2003, Santa Cruz Biotechnology). After centrifugation, chromatin fragments were incubated with one of the following at 4˚C overnight on a rotating platform: anti-FLAG M2 (F1804, Sigma-Aldrich), anti TRF2 (NB110-57130, Novus Biologics) or normal rabbit IgG (sc-2025, Santa Cruz Biotechnology). Samples were then immunoprecipitated with 50 μl of protein A/G Plus-Agarose beads were added and incubated for 2 h. The immunoprecipitated pellets were washed once with IP Wash A (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl pH 8.0), and IP Wash B (150 mM NaCl, and then with 0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl pH 8.0), and IP Wash C (500 mM NaCl, and next with 0.25 M LiCl, 1% Nonidet P-40, 1% sodium deoxycholate, 1 mM EDTA, and 10 mM Tris-HCl pH 8.0) and finally with TE (10 mM Tris-HCl pH 8.0 and 1 mM EDTA) two times. The chromatin was eluted from the beads twice by incubation with 250 μl 1% SDS and 0.05 M NaHCO 3 during 15 min at RT with rotation. After adding 20 μl of 5 M NaCl, the crosslink was reversed overnight at 65˚C. Samples were supplemented with 20 μl of 1 M Tris-HCl pH 6.5, 10 μl of 0.5 M EDTA, 20 μg of RNase A, and 40 μg of proteinase K, and were incubated for 1 h at 45˚C. DNA was recovered by phenol-chloroform extraction and ethanol precipitation, denatured (0.3 M NaOH) at 50˚C for 1 h, neutralized (1 M Ammonium acetate), and transferred to a Hybond-N+ membrane (Amersham) on a dot blot, and hybridized with a telomeric probe obtained from a plasmid containing 1.6 kb of TTAGGG repeats (gift from T. de Lange, Rockefeller University) or with ALU sequence probe. The signal was quantified with the ImageJ software. The amount of telomeric DNA immunoprecipitated was calculated in each ChIP based on the signal relative to the corresponding total telomeric DNA signal.

Xenografts experiments
For xenografts experiments, 6-week athymic nude females were obtained from Harlan (Foxn1 nu/nu ). Mice were maintained at the Spanish National Cancer Centre (CNIO) in accordance with the recommendations of the Federation of European Laboratory Animal Science Associations (FELASA) under specific pathogen-free conditions. All animal experiments were performed in accordance with the guidelines stated in the International Guiding Principles of Biomedical Research Involving Animals, developed by the Council for International Organizations of Medical Sciences (CIOMS), and were approved by Ethical Committee (CEIyBA) (IACUC.003-2015, CBA-032015). Along with those guidelines, mice were monitored in a daily or weekly basis and they sacrificed in CO 2 chambers when the human end-point was considered.
Wild type and mutants HEK293T cells were dissociated and re-suspended in DMEM and Matrigel in a 1:1 ratio. Foxn1 nu/nu mice were subcutaneously injected with 4 x 10 6 cells in 100 ul of DMEM:Matrigel. Mice were weighted and tumors were measured every 2-4 days. Tumor volume was determined by the following equation: V = a � b 2 , were a and b are tumor length and width respectively.
Supporting information S1  tively. Those clones used throughout the manuscript are labeled in red. The fold change in TRF1 levels with regards to wild type cells is indicated below the images. SMC1 was used as a loading control. C. Quantification of TRF1 transcriptional levels by q-PCR and TRF1 protein levels in TRF1 +/+ cells transfected with an sh-TRF1. SMC1 was used as a loading control. A representative western blot images of total nuclear TRF1 protein levels is shown in right panel. Error bars represent standard deviation. n number of independent experiments. Student's t test was used for statistical analysis, P values are shown. ��� , p � 0.001.