Telomere-binding Protein TRF2 Binds to and Stimulates the Werner and Bloom Syndrome Helicases*

Werner syndrome is a human premature aging disorder displaying cellular defects associated with telomere maintenance including genomic instability, premature senescence, and accelerated telomere erosion. The yeast homologue of the Werner protein (WRN), Sgs1, is required for recombination-mediated lengthening of telomeres in telomerase-deficient cells. In human cells, we report that WRN co-localizes and physically interacts with the critical telomere maintenance protein TRF2. This interaction is mediated by the RecQ conserved C-terminal region of WRN. In vitro, TRF2 demonstrates high affinity for WRN and for another RecQ family member, the Bloom syndrome protein (BLM). TRF2 interaction with either WRN or BLM results in a notable stimulation of their helicase activities. Furthermore, the WRN and BLM helicases, partnered with replication protein A, actively unwind long telomeric duplex regions that are pre-bound by TRF2. These results suggest that TRF2 functions with WRN, and possibly BLM, in a common pathway at telomeric ends.

Defects in members of the RecQ family of DNA helicases are responsible for three distinct human disorders: Werner syndrome (WS) 1 (1), Bloom syndrome (2), and Rothmund-Thomson syndrome (3). All three disorders exhibit a predisposition to cancer and increased genomic instability (4). WS is of particular interest in aging research, because WS patients prematurely display aging features including gray hair, cataracts, osteoporosis, atherosclerosis, and diabetes mellitus (type II) (5). Cells from WS patients show elevated levels of DNA deletions, translocations, and chromosomal breaks (6,7) and display replicative defects including an extended S-phase and premature senescence (8,9). The gene defective in WS (10) encodes a 167-kDa protein (WRN) that has ATPase, 3Ј 3 5Ј helicase, and 3Ј 3 5Ј exonuclease activities (11)(12)(13). The status of the WRN protein in the cell influences not only genome integrity but also the cellular life span.
The state of telomeres in the cell is also an important determinant of cellular life span. Telomeres cap and protect chromosome ends, and the consequences of telomere dysfunction include replicative senescence, apoptosis, and/or genomic instability (reviewed in Ref. 14). Telomere dysfunction results from direct damage, defects in telomere maintenance proteins, and a progressive decline in telomere lengths that occurs with each cell division (reviewed in Ref. 15). Telomere-associated senescence can be bypassed in some cell lines by the expression of telomerase, which extends telomeres (14). Evidence suggests that proper telomere function in mammalian cells is maintained by a nucleoprotein complex, together with structural features of the telomeric DNA. Telomere repeat binding factors TRF1 and TRF2 regulate telomere length, and TRF2 defects induce growth arrest and telomere end fusions (16,17). The TRF proteins specifically bind to duplex telomeric DNA, but not to single stranded DNA, and aid in the formation and stabilization of telomeres in a secondary structure that sequesters and protects the telomeric end (18,19).
Recent evidence suggests that the RecQ helicases may also function in pathways at telomeric ends. Accelerated rates of telomere shortening and defective repair at telomeres were observed in WS fibroblasts (20,21). Altered telomere length dynamics were reported in WS B-lymphoblastoid cell lines (22). Furthermore, the expression of telomerase in WS fibroblasts extended the cellular life span, rescued the premature senescent phenotype (23), and partially reversed the hypersensitivity to 4NQO (24). The molecular basis for the apparent connection between WRN and telomerase is unknown. In yeast Saccharomyces cerevisiae the RecQ homolog, Sgs1, participates in a telomerase-independent pathway for telomere lengthening (25)(26)(27). This mechanism for telomere maintenance, termed alternative lengthening of telomeres (ALT), was also detected in telomerase-negative immortalized mammalian cells (28). ALT cells are characterized by long heterogeneous telomeres and distinct nuclear foci referred to as ALT-associated promyelocytic leukemia (PML) bodies, which contain the PML protein telomeric DNA, recombination proteins RAD52, RAD51, RPA, and telomere-binding proteins TRF1 and TRF2 (29). Recently, WRN was found to co-localize with these ALT-associated PML nuclear bodies (25), and there is evidence for localization of the Bloom syndrome protein (BLM) to these bodies, as well (30). However, it is not known whether RecQ helicases participate in a telomere metabolic pathway in mammalian cells.
We report here molecular evidence for the participation of human RecQ helicases, WRN and BLM, in a protein complex that functions specifically in DNA metabolic processes at telomeric ends. We observed that telomere-binding protein TRF2 physically associates with the WRN protein in vivo and directly binds to both WRN and BLM with high affinity in vitro. In addition, TRF2 stimulates the helicase activity of both WRN and BLM via physical protein interactions. We propose specific roles for the WRN and BLM helicase activities, together with TRF2, in telomeric DNA metabolism.

EXPERIMENTAL PROCEDURES
Proteins-Recombinant histidine-tagged WRN protein was purified using a baculovirus/insect cell expression system as described previously (31). Recombinant histidine-tagged BLM was overexpressed in S. cerevisiae and purified as described previously (32). Recombinant human TRF2 protein, containing an N-terminal histidine tag, was purified using a baculovirus/insect cell expression system. The baculovirus construct for TRF2 expression was generously provided by Dr. Titia de Lange (Rockefeller University, New York, NY). TRF2 protein was purified from expression cultures using a 1-ml HiTrap chelating column charged with Ni 2ϩ ions (Amersham Biosciences), according to the manufacturer's protocol. Eluted TRF2 protein was dialyzed against Buffer D (20 mM Hepes pH 7.9, 100 mM KCl, 3 mM MgCl 2 , 1 mM dithiothreitol, 20% glycerol, and 0.5 mM phenylmethylsulfonyl fluoride) (33), and the final concentration was determined by the Bio-Rad assay using BSA as a standard. The protein was analyzed by SDS-PAGE followed by Coomassie Blue staining. The identity of TRF2 was confirmed by Western blot analysis using anti-TRF2 (sc-8528; Santa Cruz Biotechnology, Santa Cruz, CA).
Coimmunoprecipitation Assay-HeLa nuclear extracts, prepared as described previously (34), were precleared with protein A-Sepharose beads (Amersham Biosciences) for 1 h at 4°C. Extracts (400 l each) were incubated with either 5 g of rabbit anti-TRF2 (H-300; Santa Cruz Biotechnology) or 5 g of rabbit IgG (Santa Cruz Biotechnology) as a negative control, overnight at 4°C. Protein A-Sepharose (40 l) was added to each sample, followed by incubation at 4°C for 1 h. Samples were centrifuged and washed four times with 500 l of buffer (20 mM Hepes (pH 7.1), 100 mM KCl, 10% glycerol, 0.25 mM EDTA, 0.05% Tween 20). Bound proteins were eluted by boiling in sample buffer for 5 min and were analyzed by SDS-PAGE and Western blot using goat anti-TRF2 (1:100; Santa Cruz Biotechnology) and mouse anti-WRN (1:250; Transduction Laboratories) antibodies followed by ECL-Plus detection (Amersham Biosciences).
Transfection and Immunofluorescence-HeLa and U-2 OS cells were grown in Dulbecco's modified Eagle's medium (Invitrogen) and 10% fetal bovine serum on coverslips. The cells were transfected with the pEGFPC3-WRN expression vector as described previously (35). Exponentially growing cells were fixed and permeabilized with 3.7% formaldehyde (Sigma) and 0.4% Triton X-100 in PBS for 15 min at 25°C. Following washes with PBS, the coverslips were incubated in blocking buffer (0.1% Tween 20, 3% BSA in 1ϫ PBS) overnight at 4°C. Coverslips were incubated with mouse monoclonal anti-TRF2 antibodies (1: 250 dilution; Imgenex) for 4 h at 25°C. Following washing, coverslips were incubated with anti-mouse Texas Red (1:200 dilution; Vector Laboratories). After washing, the coverslips were mounted with Vectashield (Vector Laboratories) and viewed using a laser-scanning confocal microscope (Zeiss 410) in two channels (green, 488 nm; red, 568 nm). Images were analyzed with Metamorph imaging system 4.1 software (Universal Imaging Corp.).
GST-WRN-Sepharose Pull Down Assay-The gene encoding TRF2 in the pBacPA8 vector was cloned into the BamHI and EcoRI sites of vector pRSETA (Invitrogen). This plasmid was used to generate [ 35 S]methionine-labeled TRF2 protein by in vitro transcription and translation using a rabbit reticulocyte lysate kit (Promega) and [ 35 S]methionine (Amersham Biosciences) according to the manufacturer's protocol. Pull down assays with various GST-tagged WRN protein fragments with either HeLa NE or 35 S-TRF2 were performed essentially as described previously (34,35). Briefly, GST-WRN fragments were bound to glutathione beads and then incubated with either HeLa NE (400 l) or 35 S-TRF2. After washing, total protein was eluted in sample buffer by boiling and was analyzed by SDS-PAGE. Equal loading of the WRN fragments was verified by either Amido Black or Coomassie staining, and bound 35 S-TRF2 was detected by autoradiography. Bound TRF2 from HeLa NE were detected by Western blot with mouse monoclonal anti-TRF2 antibodies (1:250 dilution; Imgenex).
ELISA Detection of Protein Interactions-ELISA was conducted as described previously (36), with some modification. Wells were coated with 50 l of purified WRN or BLM protein (diluted to 1.5 ng/l in carbonate buffer) or with 50 l of BSA as a background control by incubation for 2 h at 37°C. Wells were washed and incubated with 300 l of blocking buffer (PBS, 3% BSA, 0.1% Tween 20) for 1 h at 37°C. After blocking, various concentrations of TRF2 protein were added (50 l) and incubated for 2 h at 37°C. Following washing, primary antibody (1:1000; rabbit IgG against TRF2 (H-300); Santa Cruz Biotechnology) was added (50 l) and incubated for 1 h at 37°C. Wells were washed, and secondary antibody (1:10,000; anti-rabbit IgG-horseradish peroxidase; Vector Laboratories) was added (50 l) and incubated for 1 h at 37°C. Wells were washed, and bound TRF2 was detected with o-phenylenediamine dihydrochloride followed by termination with 3 M H 2 SO 4 . Absorbance was read at 490 nm, and values were corrected for background in the BSA alone control. For the DNase I-treated control reactions, DNase I (Calbiochem) was included at a concentration of 5 g/ml. To determine the dissociation constant (K d ) for the WRN⅐TRF2 and BLM⅐TRF2 complexes the fraction of immobilized WRN or BLM bound by TRF2 was calculated, and the data were analyzed by Hill plot and Scatchard binding theory, as described previously (36).
Electrophoretic Mobility Shift Assay-Binding reactions (20 l) were conducted in standard reaction buffer (40 mM Tris-HCl (pH 8.0), 4 mM MgCl 2 , 5 mM dithiothreitol, 2 mM ATP, and 0.1 mg/ml BSA). Protein and DNA substrate concentrations were as indicated in the figure legends. Reactions were incubated for 20 min at 4°C and stopped by addition of 3ϫ dye (0.125% bromphenol blue and 40% glycerol) to a 1ϫ final concentration. Samples were loaded on 4% polyacrylamide (29:1) gels and electrophoresed overnight at 30 V, 4°C in 1ϫ TBE (89 mM Tris base, 89 mM boric acid, 2 mM EDTA). Products were visualized using a PhosphorImager and quantitated using ImageQuant software (Molecular Dynamics). The percent of bound DNA substrate was quantitated using the following formula: % bound ϭ 100 ϫ (amount of bound DNA)/(total DNA). Background correction was carried out using control reactions that excluded the enzyme.
Helicase Reactions-Reactions (20 l) were performed in standard reaction buffer at 37°C for 15 min. The DNA substrate concentration was 0.5 nM, and protein concentrations were as indicated in the figure legends. For reactions with the (TTAGGG) 2 forked duplex, the proteins were pre-incubated at 25°C for 5 min prior to addition of the DNA substrate. For reactions with the (TTAGGG) 4 forked duplex, TRF2 was pre-incubated with the DNA substrate for 5 min prior to the addition of the helicase with or without RPA as indicated. Reactions were terminated by the addition of 3ϫ stop dye (50 mM EDTA, 40% glycerol, 0.9% SDS, 0.1% bromphenol blue, and 0.1% xylene cyanol) to a 1ϫ final concentration, along with a 10ϫ molar excess of unlabeled competitor oligonucleotide to prevent reannealing of the unwound single stranded DNA products. Reactions containing the (TTAGGG) 4 forked duplex were incubated with proteinase K (8.3 ng/l final concentration) following termination, as described previously (37). This process disrupts TRF2 and RPA complexes with the DNA substrate and unwound product, respectively, that fail to migrate into the native gel. Products were run on 12% native polyacrylamide gels and analyzed using a Phospho-rImager (Molecular Dynamics). The percent of unwound product (% displacement) was calculated as described previously (37).
Exonuclease Reactions-Reactions (10 l) were performed under standard reaction conditions, and protein concentrations were as indicated in the figure legends. For reactions containing the (TTAGGG) 4 forked duplex (0.5 nM), the substrate was pre-incubated with TRF2 prior to the addition of WRN and then incubated at 37°C for 15 min. For reactions with the duplex substrate containing the 3Ј recessed end, WRN was pre-incubated with TRF2 for 5 min prior to the addition of substrate (3 fmol) and then incubated at 37°C for 1 h. The reactions were terminated by addition of equal volume formamide stop dye (37). Heat-denatured products were run on 14% denaturing polyacrylamide gels and were visualized using a PhosphorImager.

WRN Interacts with TRF2 in Human Cell
Lines-Based on the reported co-localization of WRN with telomere-binding proteins in ALT-associated PML nuclear bodies (25), we tested for a physical association between these proteins. The C-terminal domain of WRN mediates the binding to several known WRNinteracting proteins (39). Therefore, we used a pull-down assay to test for the association of a GST-WRN C-terminal fragment (amino acids 949 -1432) with telomere-binding proteins in HeLa NE. TRF2 co-precipitated with the GST-WRN fragment 949 -1432 but not with the GST protein control as shown by Western blot with antibodies against TRF2 and WRN (Fig. 1A). Amido Black staining of the same membrane shows that extracts were incubated with higher amounts of the GST control, relative to the GST-WRN fragment 949 -1432 (Fig. 1B), attesting to the specificity of the TRF2 interaction with GST-WRN. Comparison with the input indicates that ϳ25% of the TRF2 protein was precipitated. To test for association of endogenous TRF2 and WRN in vivo we performed co-immunoprecipitation experiments using the HeLa NE. A rabbit antibody against the N terminus of human TRF2 precipitated endogenous WRN protein (Fig. 1C, lane 2), along with endogenous TRF2 (Fig. 1D, lane 2). Comparison with the input (Fig. 1C, lane 1) indicates that ϳ15% of the total WRN protein was precipitated. Rabbit IgG alone failed to precipitate either TRF2 or WRN (Fig. 1, C and D, respectively; see lanes 3). Although TRF2 from HeLa NE successfully co-precipitated with a WRN C-terminal fragment (Fig. 1A), we were unable to co-precipitate TRF2 with a WRN antibody, probably because of epitope masking, or because the majority of WRN localizes to non-telomeric regions. Nonetheless, these results demonstrate that a fraction of TRF2 and WRN reside in the same complex in vivo.
Next we tested for WRN and TRF2 co-localization in HeLa cells. In previous studies, the limited sensitivity of the immunofluorescence assay may have prevented the detection of WRN and TRF2 co-localization in telomerase-positive cell lines, including HeLa (25). To increase the assay sensitivity, we transfected human cell lines with a construct to express WRN fused to the enhanced green fluorescent protein (EGFP-WRN). As reported previously for endogenous WRN (25), the EGFP-WRN localized primarily to nuclear foci in the telomerasenegative ALT cell line U-2 OS (Fig. 2, A-C). The majority of these EGFP-WRN foci co-localized with endogenous TRF2 (Fig.  2C), similar to endogenous WRN (25). However, in HeLa cells EGFP-WRN localized primarily to the nucleolus, consistent with previous reports for endogenous WRN (40). Given that TRF2 stains outside the nucleolus, co-localization with EGFP-WRN was not observed in these cells (data not shown). However, in a minority of HeLa cells (about 20% as reported previously (35)), EGFP-WRN was either detected both in the nucleolus and in nuclear foci (Fig. 2E) or exclusively in nuclear foci (Fig. 2I). In these cells, ϳ30 -50% of the EGFP-WRN nuclear foci co-localized with distinct TRF2 foci, although the number of the EGFP-WRN foci was small (Fig. 2, G and K, white arrows). WRN has been observed to relocate into nuclear foci in response to some DNA damaging agents and during S-phase arrest (41,42). Therefore, WRN and TRF2 may associate under specific cellular conditions, during which the enzymes participate in a common pathway.
TRF2 Interacts with the C Terminus of WRN Primarily via the RQC Domain-A series of WRN protein fragments (35) was used to map the site(s) of protein interaction with TRF2. Recombinant protein fragments that spanned the full-length WRN protein were expressed as GST fusion proteins. The fragments were then mixed with radiolabeled TRF2 ( 35 S-TRF2), and the bound 35 S-TRF2 was detected by SDS-PAGE and autoradiography (Fig. 3). 35 S-TRF2 did not interact with GST alone (Fig. 3B, lane 2). However, 35 S-TRF2 showed weak binding to the WRN fragment 114 -240 (lane 4), which contains a portion of the WRN exonuclease domain. Incubation of WRN fragments 500 -946 and 949 -1432 with 35 S-TRF2 resulted in a moderate (lane 6) and intense (lane 7) 35 S-TRF2 band, respectively. These results indicated that TRF2 interacted most strongly with the WRN C terminus but also bound a fragment containing the WRN helicase and conserved RecQ C-terminal (RQC) domains. Next we determined the domain in the WRN C terminus (amino acids 949 -1432) responsible for the strong interaction with 35 S-TRF2. As shown in Fig. 3C, 35   TRF2 Binds to WRN and BLM with High Affinity-To confirm the direct interaction between full-length WRN and TRF2 and to determine the strength of the protein association we performed binding studies in vitro using purified recombinant proteins (Fig. 4A). Because TRF2 bound to the highly conserved RQC domain of WRN (Fig. 3), we also tested another RecQ family member, the BLM protein, for binding to TRF2. BLM encodes a 159-kDa protein with ATPase and helicase activity and shares similar substrate specificities with the WRN helicase (43). To test for binding, ELISA assays were conducted by coating the wells of microtiter dishes with BSA (control), WRN, or BLM followed by incubation with TRF2 protein. Bound TRF2 was detected using an antibody against TRF2 and colorimetric analysis. The addition of TRF2 protein to wells coated with WRN and BLM resulted in an increase in the colorimetric signal, similar to the positive control (wells coated with TRF2) (Fig. 4B). This signal did not decrease in the presence of DNa- , and the nuclear localization sequence (NLS) are shown. B, WRN domains that interact with TRF2. A series of recombinant GST-WRN fragment fusion proteins, shown schematically, and GST alone were bound to glutathione beads and incubated with in vitro-translated TRF2 ( 35 S-TRF2). Total protein was eluted, and equal amounts were run on 10% SDS gels. Bound 35 S-TRF2 was detected by autoradiography. Ten percent of the input was loaded. C, fine mapping of the TRF2 interaction sites in the WRN C terminus. seI, indicating that the observed WRN-TRF2 and BLM-TRF2 interactions were not mediated by DNA. In contrast, no signal was observed when TRF2 was added to BSA-coated wells, which attests to the specificity of the TRF2 interaction with WRN and BLM. Furthermore, the colorimetric signal increased as a function of increasing TRF2 concentration for both WRN (Fig. 4C) and BLM (Fig. 4D) and began to plateau at approximately equal molar TRF2 and WRN and equal molar TRF2 and BLM concentrations. To determine the strength of the WRN⅐ TRF2 and BLM⅐TRF2 physical interaction, the data in Fig. 4, C and D were analyzed by Hill plots, and the apparent dissociation constants (K d ) were determined using the Scatchard binding theory as described previously (36). The apparent dissociation constants for the WRN⅐TRF2 and BLM⅐TRF2 interactions were 2.3 and 2.5 nM, respectively. These values are similar to that reported for the BLM and RPA interaction (K d ϭ 1.3 nM) (36). These analyses revealed a high affinity interaction between TRF2 and the human RecQ helicases, WRN and BLM.
TRF2 Binds to Forked Duplex Molecules Containing Four Telomeric Repeats-To test whether the TRF2 interaction with WRN and BLM modulates their activities, we prepared forked duplex molecules that mimic replication and repair intermediates at telomeric ends. Forked structures were shown previously to be unwound by the WRN and BLM helicases (37,43). The (TTAGGG) 4 forked substrate was designed to bind TRF2 and contained four tandem telomeric repeats within the 34-bp double stranded region. Because TRF2 binds DNA as a homodimer via two Myb domains, two copies of the Myb binding sequence (5Ј-YTAGGGTR-3Ј) are required for binding (18,44). As controls, two other substrates were designed that should not bind TRF2. First, the sequence of the (TTAGGG) 4 fork was scrambled to create a non-telomeric 34-bp forked duplex. In addition, a truncated (TTAGGG) 2 forked duplex was constructed to contain only two telomeric repeats and thus, a single Myb site.
First, we tested for TRF2 binding to the various substrates using a standard gel shift assay. In control experiments, radiolabeled (TTAGGG) 13 fragments were incubated either in the presence or absence of TRF2 and were run on native polyacrylamide gels. The TRF2⅐DNA complexes failed to migrate into the gel, even under various reaction and electrophoresis conditions (data not shown). This was consistent with previous reports, which suggested that the basic N terminus of TRF2 interferes with detection of TRF2⅐DNA complexes (18). However, the (TTAGGG) 13 fragment migrated freely in the absence of TRF2 (Fig. 5A, lane 1) but was retained in the wells upon incubation with TRF2 (lanes 2-3), even in the presence of competitor DNA. Thus, retention of DNA substrates in the wells served as an indicator of TRF2⅐DNA complex formation. Next we tested the ability of TRF2 to bind the forked duplex substrates under identical buffer conditions to those in the WRN and BLM helicase assay. Incubation of TRF2 with the (TTAGGG) 4 forked duplex resulted in a dose-dependent retention of the DNA substrate. In reactions containing 0, 20, and 50 nM TRF2 the percent of DNA substrate that was bound and retained in the wells was 0, 41, and 71%, respectively (Fig. 5B). In contrast, the percentage of substrate that was retained and bound by TRF2 upon incubation of TRF2 (0 -50 nM) with the non-telomeric or the truncated (TTAGGG) 2 forked duplexes was negligible (0 -4.3%; see Fig. 5, C and D). These data indicate that the purified TRF2 protein binds stably to the (TTAGGG) 4 forked duplex but not to the non-telomeric or the truncated (TTAGGG) 2 forked duplexes.

Association of TRF2 with WRN and BLM Stimulates Their
Helicase Activity-To address the ability of TRF2 to modulate WRN and BLM activity via protein-protein interactions, we used the (TTAGGG) 2 forked duplex, because it can be unwound by the WRN helicase (37) but is not bound stably by TRF2 (Fig.  5D). Therefore, any effects observed on the helicase activity could be attributed primarily to TRF2 interactions with the helicase enzymes rather than to TRF2 interactions with the DNA. We used suboptimal amounts of helicase enzymes so that either stimulation or inhibition of unwinding activity could be detected. The helicase proteins were pre-incubated with TRF2 prior to the addition of substrate, and the unwound products were visualized by native gel electrophoresis. WRN (0.5 nM; monomer) alone catalyzed unwinding of 39 Ϯ 10% of the duplex molecules (Fig. 6A). Incubation of WRN, together with low concentrations of TRF2 (0.25 to 1 nM; monomer), did not alter the amount of substrate unwound (Fig. 6A). However, incubation of WRN with 2 and 4 nM TRF2 (monomer) resulted in a 1.6and 2.2-fold increase in the percent of unwound substrates, respectively. As a control, when TRF2 (4 nM; monomer) was heat-denatured (Fig. 6A, OEE) prior to incubation with WRN the helicase stimulation was not observed. Maximal stimulation was achieved at an 8:1 molar ratio of TRF2 (monomer) to WRN (monomer); further stimulation was not achieved at higher TRF2 to WRN ratios (data not shown). A similar stimulatory affect was observed (about 2.2-fold) when a non-telomeric 22-bp substrate was used, in which the (TTAGGG) 2 sequence was scrambled. Reactions containing WRN (0.5 nM) alone yielded 29 Ϯ 8% displacement, and pre-incubation with 8 nM TRF2 resulted in 63 Ϯ 14% displacement. In addition, when the experiment was repeated with the (TTAGGG) 2 forked duplex and Ku (which also binds to WRN (45)) in place of TRF2, no stimulation of WRN helicase activity was observed (data not shown). Furthermore, when a bacterial helicase, 0.5 nM UvrD, was incubated with increasing concentrations of TRF2 (0.25-4 nM), no effect on unwinding activity was observed (data not shown). These data indicate that pre-incubation of 4 to 8-fold molar excess (monomer) TRF2, relative to WRN, results in a notable stimulation of the WRN helicase activity on 22-bp forked duplex substrates.
The TRF2 functional interaction with WRN may be conserved among RecQ helicases, as similar results were obtained with BLM. Reactions containing BLM (0.5 nM; monomer) and the (TTAGGG) 2 forked duplex resulted in 20 Ϯ 5% strand displacement (Fig. 6B). Similar to WRN, incubation of BLM (0.5 nM; monomer) with 2 and 4 nM TRF2 (monomer) resulted in a 2-and 3-fold increase, respectively, in the percent of substrate unwound by BLM (Fig. 6B). As for WRN, maximal stimulation was observed at an 8:1 molar ratio of TRF2 to BLM (monomer); further stimulation was not observed at higher TRF2 amounts, relative to BLM (data not shown). In summary, these results indicate that TRF2 complex formation with the RecQ helicases, WRN and BLM, stimulates their unwinding activity.
To assess the ability of TRF2 to modulate the WRN exonuclease activity, we used a duplex DNA substrate with a recessed 3Ј end (38). The digestion products were visualized on a 14% denaturing gel. As shown in Fig. 6C, the WRN exonuclease initiated digestion of the labeled strand at the recessed 3Ј end and produced a ladder of shortened products. Pre-incubation of WRN with increasing concentrations of TRF2, prior to substrate addition, did not alter the pattern of exonuclease products (Fig. 6C). Thus, TRF2 interaction with WRN does not affect the WRN exonuclease.
WRN⅐RPA and BLM⅐RPA Retain Activity on Telomeric Substrates Bound by TRF2-Next we determined whether WRN and BLM were able to unwind telomeric duplex molecules that were pre-bound by TRF2. We chose a molar ratio of TRF2 protein to substrate that demonstrated binding to the majority of the (TTAGGG) 4 substrate molecules (71%; see Fig. 5B, lane  3). In earlier studies with the (TTAGGG) 4 substrate, we reported that the WRN helicase did not unwind the full-length duplex region but that the WRN exonuclease was active at the blunt end (37). However, the presence of the human single stranded DNA-binding protein, RPA, allows WRN helicase to unwind long duplexes that are not unwound by WRN alone (37,46). Therefore, we compared the activity of both WRN and the WRN⅐RPA complex on substrates that were either free or prebound by TRF2 (50 nM; monomer), using helicase amounts necessary for maximal unwinding (Fig. 7A). Incubation of WRN (7.5 nM) with either free substrate (lane 3) or TRF2-bound substrate (lane 5) resulted in the loss of the duplex and the appearance of shortened products because of WRN exonuclease activity. TRF2 binding to the telomeric substrate did not block the WRN exonuclease progression (lane 5). This was better visualized when the products were run on a denaturing gel as shown in Fig. 7B. In addition, TRF2 did not stimulate WRN to unwind a long duplex substrate that WRN cannot unwind alone. However, WRN (7.5 nM), complexed with RPA (8.5 nM), maximally unwound 86 Ϯ 6% of the free full-length duplex (Fig. 7A, lane 6) and 90 Ϯ 3% of the TRF2 pre-bound substrate (lane 7). Thus, TRF2 interaction with the substrate did not block the progression of the WRN⅐RPA helicase complex or the WRN exonuclease activity.
Similar to WRN, the BLM protein did not efficiently unwind the full-length 34-bp duplex region of the (TTAGGG) 4 substrate (Fig. 7C, lane 3). Minor displacement (7.3 Ϯ 2.3%; see lane 3) was observed with BLM alone; however, this was similar to that achieved by RPA (11 Ϯ 3%; see lane 4), which can passively promote duplex melting by binding single stranded regions. Unwinding of the 34-bp duplex by BLM is minimally affected by the presence of TRF2 (lane 5), demonstrating that TRF2 is unable to promote BLM unwinding of long duplexes as RPA does. In addition, the presence of TRF2 on the substrate did not affect RPA activity (lane 9). Binding of TRF2 to the DNA appeared to decrease this low level of displacement (lane 5) perhaps by increasing the thermal stability of the duplex. As observed for WRN, the presence of RPA (8.5 nM) allowed the BLM helicase to unwind the (TTAGGG) 4  TRF2-bound substrate did not hinder the progression of the helicase complex (84 Ϯ 5% displacement; see lane 7). In summary, both WRN and BLM partnered with RPA retain helicase activity on substrates bound by TRF2. DISCUSSION In this study, we observed that the Werner syndrome protein partially co-localized and interacted physically with the telomere-binding protein TRF2 in vivo. The interaction was mediated mainly by the highly conserved RQC domain of WRN, which is also present in other RecQ helicases, including the Bloom syndrome protein. Consistent with this, we observed that purified TRF2 has high affinity for WRN and BLM proteins in vitro. We found that complex formation of WRN and BLM with TRF2 was favorable for their helicase activity and resulted in a 2-to 3-fold stimulation of unwinding. Furthermore, WRN⅐RPA and BLM⅐RPA helicase complexes actively unwound long telomeric duplex regions that were pre-bound by TRF2.
The TRF2 stimulation of WRN helicase activity appears to be specific for TRF2. We have tested several proteins that also interact with WRN, including Ku (45), p53 (38), FEN1 (34), BLM (35), and PCNA, 2 and have failed to detect a helicase stimulation. Of these interacting proteins, FEN1 and BLM also bind to the WRN RQC domain (34,35). The mechanism by which TRF2 stimulates WRN and BLM helicase activity clearly differs from RPA-mediated stimulation of unwinding. RPA interaction with WRN and BLM promotes unwinding of long duplexes that are not efficiently unwound by either helicase alone (36,46), whereas TRF2 does not (Fig. 7). TRF2 does not bind to single stranded DNA (18) and thus cannot coat partially unwound strands, as RPA does, to prevent strand re-annealing. Rather, interaction of TRF2 with WRN may stabilize the enzyme in an active form or may enhance the enzyme's interaction with DNA.
What role might WRN, and possibly BLM, play at the telomeres? Recently, several enzymes involved in double strand break repair including the Ku heterodimer with the catalytic subunit of the DNA-dependent protein kinase (DNA-PKcs) (47) and the Rad50⅐MRE11⅐NBS1 complex (48) have been observed to localize to telomeres. There is evidence that defects in Ku, DNA-PKcs, and NBS1 result in dysfunctional telomeres in mammalian cells (49,50). Similarly, a role for WRN in double strand break repair is supported by a physical and functional interaction with Ku and the increased sensitivity of WS cells to ionizing radiation (45,51). Similar to WRN, both Ku and NBS1 also associate with TRF2 (48,52). How these repair enzymes act in telomere maintenance is not known, but they likely 2 V. Bohr, unpublished information.  4 forked duplex (0.5 nM) was pre-incubated with or without TRF2 for 5 min at 25°C. Reactions were initiated by adding either the helicase alone or together with RPA and were incubated for 15 min at 37°C. OEs, heatdenatured substrate. A, analysis of WRN helicase activity on TRF2-bound substrates. Where indicated, the reactions contained 50 nM TRF2 (monomer), 7.5 nM WRN (monomer), and 8.5 nM RPA (heterotrimer) (lanes 1-8). The products were analyzed on a 12% native denaturing gel. The percent displacement (%D) values represent the mean and standard deviation from three independent reactions. B, analysis of WRN exonuclease activity on TRF2-bound substrates. The reactions contained 50 nM TRF2, 7.5 nM WRN, and 8.5 nM RPA. The products were analyzed on a 14% denaturing gel. C, analysis of BLM helicase activity on TRF2-bound substrates. The reactions contained 50 nM TRF2, 7.5 nM BLM (monomer), and 8.5 nM RPA, where indicated. Products were analyzed on a 12% native gel. The %D values represent the mean and standard deviation from three independent experiments. function in the cellular response to dysfunctional telomeres. Recent reports suggest that cells may respond to the state or structure of telomeres rather than to the telomeric length (15). Blackburn (15) has proposed a model where telomeres exist in (and switch between) two states, an open accessible form and a closed protected form. An example of a closed form that blocks telomerase access was observed with mammalian telomeres bound by TRF2 (19). Here the telomeric 3Ј single stranded tail invades the duplex region resulting in a large t-loop that is stabilized by a D-loop at the site of invasion (19). The enzymes involved in double strand break repair and recombination, including WRN, may function in the cellular response to persistent unprotected telomeric structures, which are sensed as dysfunctional.
The telomere state model proposes that shortened telomeres have a higher probability of persisting in the open unprotected form and that the action of telomerase or recombination pathways, ALT, can switch the telomere to the closed protected form (15). Although WS cells display an accelerated rate of telomere shortening, the mean telomere lengths of the senescent WS cells were longer than those of the control fibroblasts (21). Therefore the premature senescence observed in WS cells is most likely not triggered by abnormally short telomeres but may still be related to telomere dysfunction. Consistent with this, the expression of telomerase was shown to prevent senescence in WS cells (23), as demonstrated for normal somatic cell lines. This suggests that telomeres in WS may have a higher probability of persisting in the unprotected form even when they are sufficiently long. Therefore, expression of telomerase may complement the need for WRN to act in a response pathway at unprotected open telomeres. One possibility is that WRN functions in the recombination based pathway for lengthening telomeres, or ALT pathway, as demonstrated for Sgs1 in yeast (25).
Alternatively, WRN and BLM may act to resolve structures at telomeres. The existence of secondary structures at telomeric ends necessitates a molecular mechanism for switching to an open accessible form, to allow for the progression of replication forks and the action of repair and telomere lengthening machineries. Our observation that WRN and BLM unwind telomeric substrates pre-bound by TRF2 suggests that these enzymes may act to resolve TRF2-mediated structures, such as t-loops. We propose that WRN helicase activity at telomeres, and potentially BLM, is tightly regulated at the level of recruitment rather than activity. Consistent with this, WRN localizes primarily to the nucleolus in HeLa and primary cell lines but has been observed to exit the nucleolus and co-localize with RPA and/or Rad51 in response to replication fork blocks and some damaging agents (41,42). The physical interaction we observed between TRF2 and WRN suggests that TRF2 has the potential to recruit WRN to the telomeres when WRN has exited the nucleolus. BLM is expressed primarily during the S/G 2 -phase (5) and thus may be recruited by TRF2 to specifically participate in the resolution of telomeric structures during DNA replication. We have demonstrated that WRN⅐RPA and BLM⅐RPA helicase complexes have the potential to unwind long telomeric D-loops in vivo, because they have been shown to unwind duplexes up to 250 bp long (36,46), and their progression is not blocked by TRF2 (Fig. 7) or Ku (37). In contrast, TRF1 interaction with telomeric DNA substrates was reported to block the progression of DNA polymerase ␣ (44). Therefore the RecQ helicases may also function to relieve the TRF1-and TRF2-mediated blocks to some DNA polymerases at telomeric ends. The precise role of WRN, and potentially BLM, at telomeric ends remains to be determined.