In order to increase the cleavage activity for human telomeric TTAGGG sequences, we constructed several types of chimeric endonuclease versions of TRAS1EN (EN), which cuts the silkworm TTAGG repeats by fusing wildtype TRAS1EN with various parts of TRF1 (T) either at the N- or C-terminal ends of TRAS1EN (Figure 1a). TRF1 has three domains: the homodimerization region (TRFH), a nuclear localization signal (NLS) and a telomeric DNA binding motif (Myb). Thus, we made chimeric EN constructs with a full-length sequence of TRF1 (T-EN, EN-T): only the TRF1-Myb component (M-EN, EN-M) and with the NLS to Myb region of TRF1 (EN-NM). As a negative control, we constructed a TRAS1EN mutant (ENmut, H258A) which has a mutation in the catalytic center that inactivates the endonuclease activity 9, and made respective chimeric constructs with TRF1 (T-ENmut, ENmut-T). We also used the nonspecific cleavage domain of a restriction enzyme FokI (FN) which is often used for chimeric endonuclease design 18 and made FN constructs with a full length of TRF1 (T-FN, FN-T) (Figure 1a). We performed three-dimensional protein modelling in order to estimate each linker length of the chimeric constructs, based on the crystal protein structures. Figure 1b shows an example of the linker design (yellow line) in TRF1-Myb (M; green) and TRAS1EN (EN; cyan) of the chimeric endonuclease M-EN. The predicted structure of M-EN showed that the distance between M and EN was around 2-4 nm, indicating that the length of flexible linker between M and EN should be longer than 10 amino acids (assuming a peptide unit length of 3.8 Å), so that the EN domain can access target DNA effectively 19. Similarly, we estimated and designed the linker for each chimera construct (Additional file 1). We expressed chimeric proteins with Rosetta2 Escherichia coli strain overexpressing tRNAs for rare codons and chaperons to enhance protein expression (Additional file 2a), analysed their qualities and quantities (Additional file 2b) and purified the proteins using the His tag.

In order to assess the cleavage property of each chimeric endonuclease in vitro, the purified protein was added to a reaction mixture including the linearized plasmid DNA containing human (TTAGGG)₈₀ telomeric repeats (pTR80) or nontelomeric sequence (pNTR; see Methods). When a chimeric endonuclease has the ability to cleave human telomeric repeats, pTR80 (3.5 kbp), which has (TTAGGG)₈₀ repeats in the centre, is expected to be digested into approximately 1.8 kb and 1.2 kb fragments (Figure 2a). We found that our chimeric endonucleases (T-EN, M-EN, EN-T, EN-NM, EN-M, T-FN and FN-T) cut the pTR80 plasmid and produced additional 1.8 kb and 1.2 kb bands (Figure 2b, lanes 4-8, 13 and 14, respectively), although the nonfused proteins (T, M and EN; lanes 2, 3 and 9, respectively), a mutant TRAS1EN protein (ENmut; lane 12) and its chimeric proteins with TRF1 (T-ENmut, ENmut-T; lanes 10, 11) did not show the additional bands for pTR80. Although it has been reported that TRAS1EN cleaves vertebrate telomeric repeats 11, TRAS1EN itself (EN, lane 9) did not cleave pTR80 effectively in this study because of a low concentration of purified TRAS1EN. None of the fused and nonfused proteins showed any extra bands for the nontelomeric plasmid pNTR (Figure 2b, lanes 16-30). Thus, these chimeric EN endonucleases and chimeric FN endonucleases fused with the human telomere binding protein TRF1 or with its Myb domain, specifically and effectively cleaved the human (TTAGGG)_n telomeric repeats. As far as we know, this is the first report of endonucleases which cleave (TTAGGG)_n more effectively than other sequences.

We further evaluated the cleavage activity of each chimeric endonuclease as described below. Linearized pTR80 or pNTR was added to a dilution series of protein and incubated at 25°C for 30 min. Then the ratio of cleaved bands to uncleaved linear plasmid was calculated using agarose gel electrophoresis (Figure 3a-i). The EC50, defined as the concentration of protein that cuts half of the plasmid, for pTR80 of nonfused TRAS1EN (EN) was 3.9 μM (Figure 3a). In contrast, the EC50s for pTR80 of chimeric EN endonucleases (Figure 3c, M-EN; Figure 3d, T-EN; Figure 3e, EN-M; Figure 3f, EN-NM; Figure 3g, EN-T) were much lower (0.09-0.19 μM) (Figure 3, Additional file 3), suggesting that purified TRAS1EN has only 2%-5% of the specific cleavage activity of the purified chimeric endonucleases. In order to compare the cleavage activity of each chimeric endonuclease, the relative activities for pTR80 and pNTR were calculated from reciprocals of each EC50 (Additional file 3). The relative activities of five chimeric EN endonucleases (M-EN, T-EN, EN-NM, EN-M and EN-T) for pTR80 were 21-45-fold higher than that of TRAS1EN alone (Figure 3j, Additional file 3). The EC50s for pNTR of EN-T, T-FN and FN-T could not be determined, because the purified protein concentration was too low to induce nonspecific cleavage. FN-T showed the highest in vitro cleavage activity for pTR80 because it was 364 times higher than that of TRAS1EN and it cleaved 20 times more telomeric repeats than nontelomeric sequences. EC50s for pTR80 and pNTR of TRAS1EN were the same (Figure 3a). This may appear to be conflicting with our previous report of TRAS1EN being the telomere cleavage endonuclease 11. This could be explained by the different substrates used in the two studies. While in this study we used the pTR80 plasmid with telomeric repeats accounting for 14% of the substrate DNA, in the previous report (TTAGGG)_n-oligonucleotides served as substrates requiring a lower enzymatic specificity 11. Moreover, pTR80 contains human telomeric repeats, whereas the cleavage activity of TRAS1EN is maximum when silkworm telomeric repeat (TTAGG)_n is the target sequence.

In order to understand how our chimeric endonucleases cut the telomeric repeats, we analysed cleavage sites in end-labelled oligo DNA TR5 containing (TTAGGG)₅ repeats, digested with chimeric endonucleases (Additional file 4). We found that chimeric TRAS1EN endonucleases cleaved the T-A junction of the TTAGGG strand and the C-T junction of the CCCTAA strand, which is consistent with the cleavage sites for TRAS1EN 11. This demonstrates that the specificity of these chimeric endonucleases was not changed even though TRAS1EN was fused with TRF1 domains. Interestingly, the cleavage sites of the two chimeric FN endonucleases (cleavage site (C|CCTAA) in T-FN and (CCCTA|A) in FN-T) on the CCCTAA strand were different from each other. The linker lengths of T-FN and FN-T are quite different, which might affect their respective cleavage sites on the CCCTAA strand 20.

Next, we performed electrophoretic mobility shift assays (EMSAs) to test the binding capacities of chimeric endonucleases to the telomeric repeats. Like TRF1 (T) and its Myb domain (M) 16, all EN and FN chimeric endonucleases bound to the telomeric oligonucleotide TR5 (Additional file 5a), but not to the nontelomeric sequence (NTR) (Additional file 5b), suggesting that chimeric endonucleases recognize telomeric repeats based on the binding specificity of TRF1.

In order to test whether these endonucleases also cleave the telomeric repeats in vivo, we introduced endonuclease constructs into two telomerase-negative cancer cell lines, U2OS and CCF-STTG1, and a normal fibroblast cell line HFL-1 using adenovirus vector. Genomic DNA was collected 3 days after virus infection and the telomere length was measured by Southern hybridization (see Methods). In the case of U2OS, CCF-STTG1 and HFL-1 cells, which have long telomeric repeats (over 10 kb), the band intensity of telomeric repeats decreased drastically when EN-T adenovirus was infected, but did not change when GFP-TRF1 (G-T) and FN-T viruses were infected, in comparison with no virus infection (Figure 4a-d). The human L1 sequence and centromeric alpha satellite were used as a nontelomeric internal control for equal loading, and for the comparison of non-specific cleavage, and its banding patterns were not changed in any treatment (Figure 4a-d). These data demonstrate that EN-T was able to cut telomeric repeats specifically and effectively not only in vitro but also in human cell lines in vivo. Unexpectedly, FN-T, which showed the highest activity in vitro (Figure 3j), did not cut the telomeric repeats in human cells. Using U2OS cells showing the most evident telomere digestion in Figure 4a-d, we tested the telomere cleavage by various adenovirus constructs with MOI 100, 10 and 1 (Figure 4e). Two constructs, T-EN and EN-T cleaved telomeric repeats clearly but other constructs, including the TRAS1EN mutant fusion proteins (T-ENmut and ENmut-T), did not alter band intensity. We confirmed that each protein was normally expressed and increased proportionally to the virus titre (Figure 4f).

As the TRF1 protein is known to accumulate around telomeres 21, we analysed in order to see if our chimeric endonucleases also localize at telomeres. The subcellular localizations of chimeric endonucleases expressed with HA tags were detected with an anti-HA antibody and telomeres were detected with an antibody against TRF2, another component of the shelterin complex 16. We found that all chimeric endonucleases, except for T-FN, localized at telomeres, as shown by yellow signals in the merged image seen in Figure 5a. The T-FN chimeric endonuclease seemed not to migrate into the nucleus (Figure 5a), which may explain why T-FN could not cut telomeric repeats in cells (Figure 4e).

Indirect immunofluorescence of γH2AX, which is a double-strand-break marker 22, visualized telomere cleavage by chimeric endonucleases in U2OS cells (Figure 5b). Only after infection with the T-EN or EN-T adenovirus did we observe many co-localized spots of chimeric endonuclease and γH2AX (yellow signals in the merged and enlarged images). There were some background γH2AX signals in PML bodies 2223 without virus infection. Using NIH Image J software, we quantified colocalization of HA and γH2AX signals (Figure 5c) and found that 34% of T-EN and 27% of EN-T foci localized at double-strand breaks. These observations confirmed that both T-EN and EN-T localize at telomeres and cut telomeric DNA.

We examined whether the telomere shortening induced by chimeric EN endonucleases could repress the cell proliferation directly. A low concentration of recombinant adenoviruses was added to U2OS and HFL-1 human cells at each passage, every 3-4 days. Twenty-five days after viral infection, the growth of U2OS tumour cells infected with EN-T-expressing adenovirus was inhibited at the higher dose (Figure 6b). In contrast, U2OS cells infected with adenoviruses expressing chimeric proteins, G-T, FN-T, ENmut-T and EN showed normal growth similar to the cells expressing only GFP (G). We further analysed HFL-1 fibroblasts, which have no telomerase activity to elongate shortened telomere. The growth of HFL-1 cells infected with EN-T virus was more clearly inhibited than that of U2OS with the same virus (Figure 6c, d). HFL-1 cell cultures infected with EN-T adenovirus titres of 8.8 pfu/ml were dead by day 17 after infection.

Then we explored the senescence status of U2OS tumour cells expressing chimeric EN endonuclease. Consistent with the results presented in Figure 6b, EN-T expression suppressed proliferation of U2OS cells 2 weeks after infection, while ENmut-T did not exhibit such suppression effect (Figure 7a). We measured telomere length at day 3, 7 and 13 after infection to analyse telomere digestion by EN-T and other expression products (Figure 7b). Telomere digestion was detected from day 7 with the EN-T virus, while not with other viruses including ENmut-T. We also compared the gene expression levels of p53 and p16INK4a (p16) after viral infection (Figure 7c). The p53 and p16 proteins are known as key factors in the telomere DNA damage response. In the experiment of telomere loop structure disruption by overexpression of dominant negative TRF2 (TRF2^ΔBΔM), p53 was induced within a few days and p16 was induced after 8-10 days 24. The p53 and p16 proteins induced cellular senescence independently by inhibiting their downstream target cyclin-dependent kinase 25. In this experiment, the expression of p53 was induced with EN-T virus 7 days after viral infection, but the level of p16 remained stable (Figure 7c). ENmut-T expression also activated p53 slightly, this could be because TRF1-overexpression affected cell cycle progression 26. Nevertheless, EN-T expression clearly has the strongest effect on p53 activation. It was confirmed that the amount of EN-T protein was almost equal to that of ENmut-T (Figure 7c, HA). We also performed a senescence-associated (SA) β-galactosidase assay in order to detect cellular senescence. EN-T-expressing cells showed senescent phenotype, an enlarged and flattened appearance and increased SA β-galactosidase activity (Figure 7d). We found that 70% of cells infected with EN-T virus were positive for SA β-galactosidase staining, while only 14% with the ENmut-T virus (Figure 7e). These results indicate that EN-T virus induces cellular senescence through telomere shortening and represses cell proliferation.

Three-dimensional modelling of chimeric proteins was performed with PyMol v0.99-rc6 http://www.pymol.org. We started from the co-crystal structure of the Myb domain of TRF1 and DNA (PDB ID: 1W0T), then overlaid EN (PDB ID: 1WDU) or FN (PDB ID: 1FOK). Although the co-crystal structure of EN and DNA has not yet been clarified, the interaction between EN and DNA was surmized via homology modelling using the crystal structure of DNaseI (PDB ID: 1DNK) 11. The TRF1, TRF1 Myb domain, and TRF1 NLS-Myb domain were amplified by polymerase chain reaction (PCR) with Pfu Turbo DNA polymerase (Stratagene, CA, USA) using primers TRF1-f-Nco-Nhe, TRF1M-f-Nco-Nhe, TRF1NM-f-Nco-Nhe and TRF1-b-Sal-Not (Additional file 6) from HEK293 cell cDNA. EN and EN(H258A) were amplified using primers T1EN-f-Sal-Nhe and T1EN-b-Nco-Not (Additional file 6) from plasmid pHisT1EN or pHisT1EN(H258A) 9 and FN was amplified using primers FN-f-Sal-Nhe and FN-b-Nco-Not (Additional file 6) from plasmid pML109RM119-1 (ATCC biological resource centre). All primers were designed to contain restriction endonuclease sites and PCR products were digested with the corresponding enzymes (Figure 1). The products were cloned into the pET21b expression vector (Novagen), which is an expression vector for E. coli and contains a His tag in the C-terminal end. To produce chimeric constructs, in the case of pET21b-EN-T vector, TRF1 was subcloned into pET21b digested with NheI and NotI, and then TRAS1EN was subcloned into pET21b-TRF1 digested with NheI and NcoI. Other vectors were constructed similarly. The insert sequence of a candidate clone was confirmed. Adenovirus expression vectors were constructed from pAxCAwtit (TAKARA), which contains the CAG (chicken beta-actin promoter and cytomegalovirus immediate-early enhancer) promoter. Insert sequences were amplified from each pET21b vector using 5'-phosphorylated primers Kozak-HA-rbs and His tag (Additional file 6) and were ligated into pAxCAwtit digested with SwaI. Recombinant adenoviruses were generated according to the manufacturer's procedures. Recombinant pAxCAwtit vectors were digested with NspV and the linearized adenovirus DNAs were transfected into human 293 cells. Recombinant adenoviruses were generated in 293 cells. The map of adenovirus vector is shown in Additional file 7.

The plasmids containing chimeric endonucleases were transformed into the E. coli strain Rosetta2(DE3)/pLysS (Novagen) plus pG-KJE8 (TAKARA). We used the Rosetta2 E. coli strain because this strain carried a plasmid containing tRNA genes to decode rare E. coli codons, and the strain was designed to enhance the expression of eukaryotic proteins. The plasmid pG-KJE8, which contains chaperons to facilitate protein folding and enhance production of active proteins, was co-transformed. Transformants were grown in 10 ml LB containing 50 μg/ml ampicillin, 20 μg/ml chloramphenicol, 100 μg/ml kanamycin, 5 ng/ml tetracycline, and 0.5 mg/ml L-arabinose culture at 37°C until the optical density at 600 nm (OD600) reached 0.6. Isopropyl-β-D-thiogalactopyranoside (IPTG) was then added to a final concentration of 1 mM, and incubation was continued for 12 h at 25°C. Protein purification was carried out according to the manufacturer's protocol for Ni-NTA agarose (Qiagen, CA, USA). Glycerol up to a concentration of 5% was added to the eluted proteins which were stored at -80°C. Eluates were run on SDS-PAGE and the target proteins were detected as bands of predicted molecular masses. We calculated the percentage of the target protein in the total eluted proteins from Additional file 2b by quantification with ImageJ v. 1.34s (National Institutes of Health [NIH], MD, USA). The concentrations of the total eluted proteins were determined by Bradford assay using Proteostain Kit (Dojindo, MD, USA).

The telomeric repeat plasmid pTR80 was constructed from a PCR product of an extended primer dimer of (TTAGGG)₅ and (CCCTAA)₅ primers. A non-telomeric repeats (NTR) sequence was randomly selected from human mRNA, and 2022-2672 bp of Tat-SF1 (GenBank accession number U76992) was amplified with primers dT18 and NTR-f (Additional file 6) and cloned into the pGEM T-Easy vector (Promega, WI, USA). For quantitative chimeric endonuclease activity assays, plasmid pTR80 and pNTR were digested with the restriction endonuclease ScaI to make a linear substrate with telomeric repeats positioned in the centre of the DNA strand. Then, DNA was purified using the PCR clean-up kit (Sigma, NY, USA). DNA concentration was determined spectrophotometrically at 260 nm. Chimeric endonuclease was added to 10 μl of reaction buffer containing 1.6 ng/μl substrate DNA. Samples were incubated for 1 h at 25°C, and the reaction products were analysed by agarose gel electrophoresis. Ethidium bromide-stained gels were digitized with PrintGraph (Atto, Tokyo, Japan). The intensity of the product bands was quantitated using ImageJ v. 1.34s (NIH). Assays for oligonucleotide endonucleolytic activities were performed as described previously 9. Oligonucleotides were labelled at the 5' end by the enzymatically catalyzed transfer of ³²P from [γ-³²P] adenosine triphosphate (ATP) with T4 polynucleotide kinase. The oligonucleotide concentration was fixed at 100 nM with adding nonlabelled oligonucleotides. The reaction buffer for chimeric endonucleases contained 5 mM HEPES-KOH (pH 7.9), 10 mM NaCl, 2 mM MgCl₂, 100 μg/ml BSA and 500 nM non-specific oligonucleotide, NTR to suppress non-specific binding of misfolded chimeric endonuclease to non-telomeric sequence. This mixture was treated with a purified chimeric endonuclease in 10 μl of reaction buffer for 1 h at 25°C. The reaction mixture was denatured in a loading buffer containing 75% formamide for 5 min at 95°C, immediately chilled on ice and run on a 28% Long Ranger polyacrylamide denaturing sequencing gel (Lonza, Basel, Switzerland). Quantitation of the reaction products was carried out with a BAS 5000 imaging analyser system (Fujifilm).

Various-sized telomeric repeat oligonucleotides were end-labelled with T4 polynucleotide kinase and [γ-³²P] ATP and were used as molecular size markers: dG8, dG14, dG20, dG26, dC9, dC15, dC21 and dC27 (Additional file 6).

DNA probes were 5'-end-labelled as described earlier 9, and were gel-purified. About 15 fmol of DNA was labelled in a 10 μl reaction and the final probe concentration in binding assays was <10 nM. Binding reactions were carried out by incubating the indicated amounts of protein (see figure legends for details) in 10 μl of a reaction mix of 20 mM Hepes (pH 7.9), 150 mM KCl, 0.1 mM EDTA, 1 mM DTT, 4% Ficoll, 5% glycerol, 0.1 mg/ml of BSA and 3 ng/μl salmon sperm DNA. Samples were incubated at 4°C for 60 min and then run on native 5% polyacrylamide gels as described 33.

Two hundred and ninety-three U2OS, CCF-STTG1 and HFL-1 cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum 34. All cell lines were provided from the RIKEN BioResource Center (Ibaraki, Japan). Cells were routinely passaged at 80% confluence. Adenoviruses were used as freeze-thaw lysates, and their titres were estimated following the manufacturer's procedure of the Adenovirus Expression Vector Kit (TAKARA, WI, USA) for each virus (n = 16). Transfections were performed with FuGene HD (Roche) according to the manufacturer's procedure. Four separate microscope images were used for counting the number of cells with CellProfiler 35.

Genomic DNA was purified with PUREGENE (Gentra, WA, USA) and digested with HinfI. Digests were separated on a native 0.8% agarose gel, blotted in 0.4 M NaOH onto a Hybond N+ membrane (Amersham, Uppsala, Sweden) and hybridized with the RI-labelled (CCCTAA)₅ and 1490-1467 bp of L1 (Additional file 6; GenBank accession number X52235.1) probes in hybridization buffer (6× saline-sodium citrate [SSC], 1% sodium dodecyl sulphate [SDS], 5× Denhardt's solution, 2.5 μg/ml salmon sperm DNA) at 37°C for 12 h. The membrane was washed several times with wash buffer (4× SSC, 0.1% SDS) for 10 min at 37°C. Quantitation of the reaction products was carried out with a BAS 5000 imaging analyser system (Fujifilm).

Cells were trypsinized, washed once with PBS and subsequently lysed in 200 mM Tris-HCl pH 6.8, 4% SDS, 20% glycerol, 200 μM DTT and 0.2% bromophenol blue. Lysates were separated on SDS polyacrylamide gels (29:1 acrylamide: bisacrylamide, 8% gel) and transferred onto BioTrace PVDF (Pall) for 70 min at 14 V. Membranes were preincubated in TBST (0.1% Tween20 in 1× TBS) containing 5% BSA for 30 min and then incubated with primary antibodies in Can Get Signal Solution 1 (TOYOBO, Osaka, Japan) for 1 h, followed by three 10 min washes with TBST. The following antibodies were used: TRF2, IMGENEX #IMG-124A; γ-H2AX, LPBIO #AR-0149; HA, BETHYL #A190-108A; p53, EXBIO #11-114-C100; p16INK4a, Thermo #MS-383-P0; α-Tubulin, CEDARLANE #CLT9002 and β-Actin, Abcam #ab8226. Membranes were incubated for 1 h with horseradish peroxidase conjugated anti-mouse or anti-rabbit antibodies (Amersham) in Can Get Signal Solution 2 (TOYOBO) and were developed using the ECL PLUS system (Amersham). All samples were loaded in the same gel and the blotted membrane was re-stained with different antibody after stripping in 2% SDS, 100 mM β-mercaptoethanol and 62.5 mM Tris-HCl pH 7 at 50°C for 30 min.

The cellular localization of chimeric endonucleases was visualized through indirect immunofluorescence. Cells were grown on glass-bottomed dishes, washed once with PBS, fixed with methanol, and permeabilized in a Triton X-100 solution (0.5% Triton X-100 in phosphate-buffered saline) four times for 10 min. The cells were subsequently blocked for 1 h at 25°C in 10% sheep serum, stained with primary or fluorescence-conjugated secondary antibodies for 1 h at 37°C, and then visualized under an Olympus IX51 fluorescence microscope. Anti-TRF2 antibody (IMGENEX, CA, USA; #IMG-124A) was used to detect telomere, because TRF2 is another telomere binding protein being independent of TRF1, and anti-HA antibody (BETHYL, TX, USA; #A190-108A) was used to detect chimeric endonucleases. Secondary antibodies were Alexa Fluor 488 or 555-conjugated goat anti-mouse or anti-rabbit antibodies (Molecular Probes, CA, USA).

Senescence-associated beta-galactosidase (SA β-gal) activity at pH 6.0 is a commonly used cellular senescence marker 36. Positive cells were detected by the modified method of Dimri et al. 36. The monolayers of cells were washed with PBS and then fixed with 1% glutaraldehyde for 5 min. The cells were washed again twice with PBS, then staining solution was added [1 mg/ml 5-bromo-4-chloro-3-inolyl-b-D-galactoside (X-gal), 40 mM sodium phosphate, pH 6.0, 5 mM potassium ferrocyanide, 5 mM potassium ferricyanide, 150 mM NaCl, 2 mM MgCl₂]. After the cells were further incubated at 37°C for 24 h, the percentage of stained cells was counted. The number of blue structures was counted in four fields (from a total of average 3.76 × 10³ cells).