DNase I footprinting and enhanced exonuclease function of the bipartite Werner syndrome protein (WRN) bound to partially melted duplex DNA.

Werner syndrome is a premature aging and cancer-prone hereditary disorder caused by deficiency of the WRN protein that harbors 3' -->5' exonuclease and RecQ-type 3' --> 5' helicase activities. To assess the possibility that WRN acts on partially melted DNA intermediates, we constructed a substrate containing a 21-nucleotide noncomplementary region asymmetrically positioned within a duplex DNA fragment. Purified WRN shows an extremely efficient exonuclease activity directed at both blunt ends of this substrate, whereas no activity is observed on a fully duplex substrate. High affinity binding of full-length WRN protects an area surrounding the melted region of the substrate from DNase I digestion. ATP binding stimulates but is not required for WRN binding to this region. Thus, binding of WRN to the melted region underlies the efficient exonuclease activity directed at the nearby ends. In contrast, a WRN deletion mutant containing only the functional exonuclease domain does not detectably bind or degrade this substrate. These experiments indicate a bipartite structure and function for WRN, and we propose a model by which its DNA binding, helicase, and exonuclease activities function coordinately in DNA metabolism. These studies also suggest that partially unwound or noncomplementary regions of DNA could be physiological targets for WRN.

The RecQ helicases are a family of proteins that catalyze the ATP-dependent unwinding of double-stranded nucleic acids (reviewed in Refs. 1 and 2). There are genes coding for at least five RecQ family members (RECQL, BLM, WRN, RECQ4/RTS, and RECQ5) in the human genome that are highly homologous to one another only within seven distinct sequence motifs that comprise a conserved helicase domain. The hereditary diseases known as Bloom, Werner, and Rothmund-Thomson syndromes are the result of the loss of function of BLM, WRN, or RTS, respectively (3)(4)(5). Although genomic instability is common to each of these diseases, the overall phenotype of each disease is different, indicating that the functions of BLM, WRN, and RTS are distinct or at least partially nonoverlapping. Interestingly, WRN is the only human RecQ protein to have an exonuclease domain in addition to the helicase domain (6). The Werner syndrome (WS) 1 phenotype is characterized by early onset of a number of aging characteristics, including graying and loss of hair, increased wrinkling and ulceration of the skin, osteoporosis, atherosclerosis, and an increased frequency of age-related maladies such as cancer, diabetes, and cataracts. WRNdeficient cells from WS patients have elevated levels of genomic instability typified by increased deletions, insertions, and translocations (7,8), as well as an accelerated rate of telomere loss (9,10). Both the WS cellular phenotype and the WRN primary amino acid sequence point to a role in DNA metabolism that, when absent, results in large scale genetic change.
A number of laboratories (including ours) have successfully overexpressed and purified a recombinant WRN protein, and its basic catalytic activities have been characterized. Consistent with the presence of ATPase/helicase amino acid sequence motifs in the central region of the protein, WRN is a DNA-dependent ATPase (11). In conjunction with ATP hydrolysis, WRN unwinds DNA-DNA and DNA-RNA duplexes that contain a single-stranded region 3Ј to the duplex to be unwound, i.e. 3Ј 3 5Ј directionality (11)(12)(13)(14). More recently, WRN has also been shown to disrupt DNA triplexes (15) and certain G-quartet structures (16). WRN helicase can apparently carry out branch migration of Holliday structures as well (17). In general, studies on WRN, BLM, and Sgs1 (the lone RecQ homolog in Saccharomyces cerevisiae) have shown remarkable similarities in helicase function and DNA substrate specificity (16, 18 -24).
WRN has also been shown to be an exonuclease (25)(26)(27), consistent with the presence of RNase D-type nuclease domains in the N-terminal region of the protein (6). This activity is directed to the 3Ј end of a duplex substrate (3Ј 3 5Ј directionality) preferably containing a 5Ј overhang (recessed 3Ј end) and has not been observed on single-stranded DNA or doublestranded DNA with blunt ends or 3Ј overhangs (25,28). On DNA substrates tested thus far, WRN exonuclease activity is dramatically stimulated by ATP hydrolysis (28), suggesting some cooperativity between the ATPase and exonuclease functions of WRN. However, exonuclease activity can be observed in the absence of ATP (28), and mutant WRN proteins lacking either ATPase/helicase activity or the entire helicase domain still retain exonuclease activity (27), suggesting that the exonuclease and helicase domains are functionally independent and, in all likelihood, physically separate. Importantly, the absence of exonuclease activity in the other human RecQ homologs suggests that the unique WS phenotype may be in part the result of loss of the specific exonuclease function of WRN.
In this study, a defined DNA substrate containing an internal region of noncomplementarity was designed to closely approximate the localized melting of DNA duplexes that occurs during many metabolic processes. The DNA binding affinity and catalytic activities of recombinant wild type and mutant WRN proteins were examined on this substrate and compared with other model DNA substrates. We have demonstrated that, when compared with completely complementary substrates, a DNA substrate containing 21 unpaired nucleotides (nt) in both strands has extremely high affinity for the WRN protein. We have further characterized this binding and the effect of nucleotide cofactors by deoxyribonuclease I (DNase I) footprinting. We show that ATP hydrolysis is not required for high affinity binding of WRN to the melted region, and that the exonuclease domain does not bind stably to this substrate. Nevertheless, the WRN exonuclease function is highly active on this bubble substrate, and this activity is both independent of ATP binding or hydrolysis and further unwinding by WRN helicase function.
Our results indicate coordination between the DNA binding and exonuclease domains of WRN and suggest that a noncomplementary (heteroduplex) or melted region of DNA could be the physiological substrate of WRN. Such structures may be indicative of DNA replication, recombination, or repair intermediates formed in vivo that require processing by WRN helicase and/or exonuclease activity.

EXPERIMENTAL PROCEDURES
WRN Purification-Recombinant wild type and mutant WRN proteins were overexpressed and purified essentially as described previously (29). The WRN-K577M mutant contains a lysine to methionine point mutation at amino acid residue 577 in motif I of the conserved helicase domain; this protein lacks both ATPase and helicase activities (11,12). The WRN-E84A mutant contains a glutamate to alanine change at residue 84 in the conserved exonuclease domain that abolishes exonuclease activity (25). The WRN⌬369 -1432 contains only the N-terminal 368 amino acids of the full-length protein; however, this truncated protein possesses the entire conserved exonuclease domains and retains exonuclease activity (27). The activities of wild type and mutant WRN proteins used in this study are summarized in Table I. All constructs contained an N-terminal hexahistidine tag to facilitate purification by metal (Ni 2ϩ ) affinity chromatography. Briefly, Spodoptera frugiperda insect cells (Sf9 strain) were infected with baculovirus containing WRN cDNA sequences at an multiplicity of infection of 10 and harvested 72-96 h after infection and stored at Ϫ80°C. The wild type, WRN-K577M, and WRN-E84A proteins were purified by sequential liquid chromatographic steps using DEAE-Sepharose, Q-Sepharose, and nickel-nitrilotriacetic acid-agarose resins (29). The purification of WRN⌬369 -1432 was modified slightly. After cell lysis, WRN⌬369 -1432 protein was bound to DEAE-Sepharose in 150 mM Tris-HCl, pH 8.0, 10 mM NaCl, 0.5% Nonidet P-40, 10% glycerol, 5 mM ␤-mercaptoethanol, 1 mM phenylmethanesulfonyl fluoride, and 2 g/ml each of pepstatin A, leupeptin, chymostatin, and aprotinin, washed extensively with the same buffer minus Nonidet P-40, and eluted with the same buffer (minus Nonidet P-40) containing 60 mM NaCl. The DEAE eluate was then loaded directly onto and subsequently recovered from nickelnitrilotriacetic acid resin, using washing and elution conditions identical to those described for full-length WRN proteins. Mock-infected insect cell lysates were purified in parallel to control for potential contaminating activities. All protein preparations were stored with 100 g/ml bovine serum albumin (BSA) at Ϫ80°C until use.
DNA Substrates-Oligomers were purchased from Integrated DNA Technologies (Coralville, IA) and Operon (Alameda, CA). In 5Ј to 3Ј orientation, sequences of the oligomers used in this study are G80 (AGCTCCTAGGGTTACAAGCTTCACTAGGGTTGTCCTTAGGGTTA-GGGTTAGGGTTACCTACACATGTAGGGTTGATCAGC), G80bub-ble21 (AGCTCCTAGGGTTACAAGCTTCACTAGGGTTGTCCAGTCA-CAGTCAGAGTCACAGTCCTACACATGTAGGGTTGATCAGC), C80 (GCTGATCAACCCTACATGTGTAGGTAACCCTAACCCTAACCCTA-AGGACAACCCTAGTGAAGCTTGTAACCCTAGGAGCT), 51-mer (GG-TACCGGGCCCCCCCTCGAGGTCGACGGTATCGATAAGCTTGATA-TCGAA), and 57-mer (CAGGAATTCGATATCAAGCTTATCGATACC-GTCGACCTCGAGGGGGGGCCCGGTACC). Phosphorothioate linkages between the 5Ј end and 5Ј penultimate nucleotides of the G80, G80bubble21, and C80 oligomers were included to inhibit the activity (5Ј 3 3Ј nuclease) of a minor contaminant present in WRN⌬369 -1432 preparations. The 5Ј ends of these oligomers were labeled using [␥-32 P]ATP (60 Ci, 3000 Ci/mmol, PerkinElmer Life Sciences) and T4 polynucleotide kinase (New England Biolabs, 10 units) using standard methods. After removal of nonincorporated radionucleotide, the labeled oligomers were annealed with a 2-fold excess of their fully or partially complementary partners by heating to 90°C for 5 min and slow cooling to room temperature (25°C); C80 was annealed to either G80 or G80bubble21 to yield fully duplex or 21-nt bubble-containing substrate, respectively (Fig. 1). 5Ј end-labeled 51-mer was annealed to unlabeled 57-mer to create a standard 5Ј overhang substrate for WRN exonuclease activity. Partially and fully duplex substrates were separated from nonannealed and excess single-stranded DNAs by nondenaturing polyacrylamide gel electrophoresis (12%), eluted from gel slices using a gel extraction kit (Qiagen, Valencia, CA), and kept in 10 mM Tris-HCl, pH 8.0, at 4°C. The single-stranded character of the noncomplementary region of the bubble substrate was confirmed by localized incision of that region by S1 nuclease and by its ability to anneal with a fully complementary 21-nt oligomer (data not shown).
DNA Unwinding Assay-DNA substrates (21-nt bubble or fully duplex, 32 P-labeled on the C80 oligomer, 0.5 fmol each) were incubated for 15 min at 37°C with WRN-E84A (30 -240 fmol) in WRN reaction buffer containing 40 mM Tris-HCl (pH 8.0), 4 mM MgCl 2 , 0.1 mg/ml BSA, and 5 mM dithiothreitol. ATP (1 mM) was added unless otherwise indicated. The reactions were terminated by addition of one-sixth volume of helicase stop dye (30% glycerol, 50 mM EDTA, 0.9% SDS, 0.25% bromphenol blue, 0.25% xylene cyanol). Samples were loaded onto a nondenaturing acrylamide (8%) gel that was run in 1ϫ TBE (90 mM Tris borate, pH 8.0, 2 mM EDTA) at 120 V at 25°C. After drying of the gels, visualization and quantitation were accomplished using a Storm 860 phosphorimaging system and ImageQuant software (Molecular Dynamics, Sunnyvale, CA). The band intensities (minus background) for the double-stranded and single-stranded DNA species were determined for WRN-treated and untreated reactions. Percentage of unwinding is calculated as the ratio (ϫ100) of single-stranded species to the total amount of substrate in WRN-treated reactions, including subtraction of the low percentage of single-stranded species in untreated controls.
Exonuclease Assay-In WRN reaction buffer either without ATP or including ATP or ATP␥S (1 mM) where indicated, 32 P-labeled DNA substrates (ϳ0.5 fmol each) were incubated for 1 h at 37°C with wild type or mutant WRN protein (2-600 fmol). The reactions were stopped by addition of an equal volume of formamide loading buffer (95% formamide, 20 mM EDTA, 0.1% bromphenol blue, and 0.1% xylene cyanol), and the DNA products were denatured at 90°C for 5 min and separated by denaturing polyacrylamide (14%) gel electrophoresis in 1ϫ TBE (1.5 h at 40 watts). Size markers were obtained by individual restriction enzyme (AluI, AvrII, BclI, BfaI, HindIII, or NspI) digestion of the appropriately radiolabeled bubble substrate (for restriction sites, see Fig. 1A) at 37°C for 1 (partial) or 3 h. Individual digests were mixed in various combinations, heated to denature DNA products as above, and analyzed in parallel with exonuclease reactions. The gels were dried and subjected to phosphorimaging analysis as above. Exonuclease activity was assessed by comparing the band intensity of the undigested substrate remaining in treated samples with that of the untreated control.
DNase I Footprinting-32 P-Labeled DNA substrates (21-nt bubble or completely complementary, 0.5 fmol each) were incubated with or without WRN wild type or mutant protein (WRNp, WRN-E84A, WRN-K577M, or WRN⌬369 -1432) for 5-30 min at 4°C or 37°C in WRN reaction buffer (10-l final volume). Where indicated, reactions also included 1 mM ATP or ATP␥S. DNase I (1-l volume, 0.015 units/ reaction at 4°C or 0.00015 units/reaction at 37°C) was then added to the binding reactions followed by incubation at 0°C or 37°C for 10 min. The reactions were terminated by addition of an equal volume of formamide loading buffer. The digestion products of these reactions were resolved by denaturing polyacrylamide (14%) gel electrophoresis in 1ϫ TBE at 40 watts for 2 h and, as above, visualized using a phosphorimager after drying of gels. Size markers were generated as described above (see "Exonuclease Assay").

RESULTS
Most metabolic processes involving DNA (notably including replication, transcription, and nucleotide excision repair) take advantage of the ability of duplex DNA to be locally melted or unwound to allow access of additional DNA metabolizing enzymes. Although WRN alone does not appear to be capable of initiating the melting of fully duplex DNA, we felt it reasonable that a partially unwound region in the midst of fully duplex DNA might be a physiological substrate for WRN. For example, WRN helicase might assist in unwinding at a site that is already partially unwound and/or use an unwound region as an anchor to initiate degradation using its exonuclease function. Consistent with this rationale, various experimental findings have pointed to a role for WRN in DNA replication, recombination, repair, or transcription pathways (17, 30 -34) that involve localized melting of duplex DNA. To examine this possibility, we designed a duplex DNA substrate with an internal noncomplementary (bubble) region of 21 unpaired nucleotides on each strand that would somewhat mimic a melted DNA intermediate (Fig. 1A). The noncomplementary region is asymmetrically positioned between duplex arms of 35 and 24 base pairs (bp). For the sake of reference in certain experiments, the individual strands will be identified as the G-rich and C-rich strands (Fig. 1A). This substrate was then subjected to unwinding, exonuclease, and DNA binding studies with wild type and mutant WRN proteins, using a fully duplex substrate (Fig. 1B) for comparison where appropriate. Because of the existence of both unwinding and exonuclease activities in wild type WRN, mutant proteins were used and/or conditions were manipulated such that the individual catalytic activities could be examined in isolation. For convenience, a comparison of the catalytic properties of the individual WRN proteins used in this study is presented in Table I.
Unwinding of Bubble DNA-Enzymatic studies have demonstrated that WRN helicase requires a 3Ј single-stranded region to the duplex region to be unwound and cannot act on fully duplex, blunt-ended substrates (13,24). We reasoned that providing WRN with an internal noncomplementary (bubble) region in the middle of a DNA substrate might satisfy the single-stranded DNA requirement of WRN helicase and might permit unwinding of an otherwise blunt-ended substrate to occur. We tested the ability of WRN to unwind our substrate containing an internal 21-nt bubble in comparison to fully complementary, blunt-ended double-stranded DNA. Because of the exonuclease activity of wild type WRN protein on these substrates (see below), helicase assays were carried out with WRN-E84A mutant protein, which completely lacks exonuclease activity but retains unwinding capability. The WRN-E84A mutant was able to unwind the 80-mer substrate containing a 21-nt bubble at WRN concentrations of 60 fmol and above ( Fig.  2A). Upon overexposure of gels, small amounts of slowly migrating forked structures were detected, corresponding to unwinding of a single duplex arm of the bubble substrate (data not shown). In contrast, WRN-E84A could not detectably unwind a completely complementary DNA substrate with blunt ends (Fig. 2C), in agreement with earlier reports (13,24). The amount of helicase activity on the bubble substrate increased in a roughly proportional manner with increasing WRN concentration, whereas the fully duplex substrate was not unwound over the range of WRN concentration tested (Fig. 2D). The unwinding activity on the bubble-containing substrate was inherent in the recombinant WRN-E84A, as mock-purified protein did not detectably unwind this substrate (Fig. 2B). As expected, WRN-catalyzed unwinding of the bubble substrate a WRN helicase and exonuclease activities are detectable from 25 to 37°C, but not at 4°C.
b Based on comparison of WRN-K477M DNA binding and exonuclease assays carried out in the presence and absence of ATP, although actual ATP binding studies for this protein have not been done.
c This protein lacks the consensus nucleotide binding/hydrolysis domains present in RecQ homologs and presumably cannot bind or hydrolyze ATP. requires ATP hydrolysis, as single-stranded products were not detectable in reactions lacking ATP (Fig. 2B) or containing ATP␥S, a nonhydrolyzable ATP analog (data not shown). Furthermore, no unwinding activity was observed when reactions were incubated at 4°C. Thus, WRN could unwind our bubblecontaining substrate to some degree, but, by manipulating reaction conditions, unwinding could also be prevented in order that binding and exonuclease activities on the intact substrate can be examined unambiguously.
Exonuclease Activity of WRN on Bubble DNA-In addition to its nucleic acid unwinding activity, WRN has also been demonstrated to possess an intrinsic 3Ј 3 5Ј exonuclease activity (25)(26)(27). This activity is associated with the N-terminal region of the protein, and, in fact, deletion mutants containing only the N-terminal 368 amino acids retain exonuclease activity (26,27). WRN exonuclease digests duplex substrates from a recessed 3Ј end, but has little or no activity on fully complementary substrates with blunt ends (25,28). However, a singlestranded region inside the duplex might overcome the need for a 5Ј single-stranded overhang (recessed 3Ј end) to achieve WRN exonuclease function. To test this possibility, we compared WRN exonuclease activity on our noncomplementary substrate containing a 21-nt bubble in the midst of a blunt-ended duplex ( Fig. 1A) with a fully duplex substrate containing identical blunt ends (Fig. 1B). To prevent potential interference caused by WRN-catalyzed substrate unwinding, initial experiments were carried out in the absence of ATP. On the fully complementary DNA substrate, there was little or no degradation from the 3Ј end of the labeled strand (Fig. 3C), in agreement with previous reports (25,28). In contrast, a robust 3Ј 3 5Ј exonuclease activity was evident from the end 24 bp from the beginning of the bubble on the noncomplementary substrate, even in the absence of ATP (Fig. 3A). A similarly high level of activity was also observed when the other strand of the substrate was labeled, where the detectable degradation started 35 bp from the edge of the bubble (Fig. 3B). By comparison, WRN exonuclease activity on a duplex substrate with a recessed 3Ј end but without a bubble was barely detectable only at concentrations of ϳ150 fmol and increased progressively from that point over the testable range of WRN concentration (Fig. 3D). When WRN-E84A protein containing a point mutation in the conserved exonuclease domain was incubated with the bubble substrate, no degradation was observed (Fig. 3E), confirming that the avid exonuclease activity associated with the wild type protein is inherent to WRN. The susceptibility of the bubblecontaining and recessed end substrates to WRN exonuclease was assessed by comparing the percentages of undigested substrate remaining (compared with an untreated control) after incubation with various concentrations of wild type WRN. These measurements indicate that, in the absence of ATP, the "preferred" duplex substrate containing a 3Ј recessed end (Fig.  3D) required about 100 times more WRN enzyme than the bubble-containing substrate (Fig. 3, A and B) to achieve an ϳ50% reduction of the amount of original, undigested sub-strate. Thus, exonuclease activity was initiated much more readily on the bubble-containing substrate when compared with the 3Ј recessed end substrate. Another obvious difference in comparing the degradation patterns of the 3Ј recessed end substrate with that of the bubble substrate is the extent of inward digestion. Only a few nucleotides were removed from the 3Ј end of the recessed end substrate (Fig. 3D), suggesting a distributive mode of exonuclease activity. In contrast, WRN  Fig. 1B) control substrate (0.5 fmol, 32 P-labeled on C80 oligomer). D, WRN exonuclease treatment of 3Ј recessed end substrate (51-mer, 32 P-labeled, annealed to 57-mer, ϳ0.5 fmol). E, WRN-E84A treatment of the bubble substrate, as in B. All reactions were incubated with WRN (wild type, 2.25-300 fmol, or WRN-E84A, 120 -240 fmol, as indicated) for 1 h at 37°C in the absence of ATP. After strand separation was achieved by heating at 90°C, the DNA products were resolved by denaturing polyacrylamide (14%) gel electrophoresis and then radioactive species were visualized by phosphorimaging. Arrows denote the point of attack for the 3Ј 3 5Ј exonuclease activity of WRN. The positions of nucleotide markers and the bubble region (curved brackets) were obtained by running denatured restriction digests (see Fig. 1A for sites) of G-rich strand-labeled (panel A; HindIII, 16 nt; AluI-partial, 18 nt; BfaI-partial, 24 nt; NspI, 66 nt; BclI, 73 nt) or C-rich strand-labeled (panel B; NspI, 18 nt; AluI, 62 and 23 nt; BfaI, 54 nt; HindIII, 60 nt; AvrII, 71 nt) substrate alongside exonuclease digestions. The location of the 62-nt AluI marker has been omitted to avoid crowding. exonuclease activity on the bubble substrate proceeded deeply into the substrate on both strands (Fig. 3, A and B), pausing only at positions well removed from the original 3Ј end. On each strand, some digestion occurred into the bubble region and occasionally beyond. These results suggest that, once initiated, WRN degradation of the bubble substrate from a 3Ј end occurs in a processive manner until probably slowed by sequence, structural, or steric considerations. This observation might indicate that individual WRN molecules remain associated with the same molecule of the bubble substrate during the course of digestion.
The heightened exonuclease activity on the substrate containing the noncomplementary region (in comparison with the fully duplex and 3Ј recessed end substrates) could be attributable to properties of the exonuclease domain alone, or to the coordination of the exonuclease domain with an affinity of some other part of the WRN protein to the melted region of the duplex. This question was addressed by examining the activity of WRN⌬369 -1432, a mutant WRN protein containing only the N-terminal exonuclease domain, on the bubble-containing substrate. This truncated form of WRN is an active 3Ј 3 5Ј exonuclease on DNA substrates with 3Ј recessed ends (Ref. 27 and Fig. 4). Notably, this level of WRN⌬369 -1432 exonuclease activity on the 3Ј recessed end substrate is comparable with what is observed using wild type WRN (compare Figs. 3D and 4). However, when WRN⌬369 -1432 was incubated with either the bubble containing or the fully duplex substrate, no exonuclease activity was detected (Fig. 4). These results suggest that another part of WRN outside of the N-terminal 368 amino acids is required to achieve the elevated exonuclease activity on the 3Ј blunt ends of the bubble substrate. Thus, it is likely that regions within the helicase domain and/or the C-terminal portion of WRN have significant affinity for a melted region of duplex (and/or the associated junctions between single-and double-stranded DNA), which mediates the potent 3Ј 3 5Ј exonuclease activity at distances up to 35 bp away.
WRN exonuclease activity on duplex DNA substrates containing recessed 3Ј ends has been shown to be stimulated dramatically by the presence of ATP (28). We wanted to examine whether the potent exonuclease activity observed on the 3Ј ends of the bubble-containing substrate was likewise affected by binding or hydrolysis of an ATP cofactor. To this end, we incubated the bubble substrate with wild type WRN in reaction mixtures containing ATP, nonhydrolyzable analog ATP␥S, or lacking nucleotide altogether. Surprisingly, the substrate was digested extensively by the 3Ј 3 5Ј exonuclease activity of WRN under all three conditions (Fig. 5A). The amount of intact substrate remaining was similar (ϳ4%) in the presence or absence of ATP, and slightly higher (ϳ10%) in the presence of ATP␥S. In addition, differences in digestion pattern were no-ticeable when ATP, ATP␥S, and minus ATP reactions were compared. Exonuclease activity in the presence of ATP increased along with WRN concentration, while remaining relatively constant in the presence of ATP␥S or in the absence of ATP over the same 5-fold range of WRN concentration (Fig.  5A). These results were supported by similar experiments using the WRN-K577M protein, which lacks ATP hydrolysis and DNA unwinding activity. In the absence of ATP or in the presence of ATP or ATP␥S, the exonuclease activities and patterns of WRN-K577M on the bubble substrate are extremely Size markers from C-rich strand-labeled (panel A) and G-rich strandlabeled (panel B) substrates were obtained as described in Fig. 3, and their positions as well as that of the bubble region (curved brackets) are denoted.
FIG. 4. Exonuclease activity of WRN⌬369 -1432 deletion mutant. The 21-nt bubble, fully duplex, or 3Ј recessed end substrate (as described in Fig. 3 (B-D), 0.5 fmol each) were incubated with WRN⌬369 -1432 (150 -600 fmol) for 1 h at 37°C. DNA products of these reactions were analyzed as described in Fig. 3. similar to one another (Fig. 5B). Thus, two lines of evidence demonstrate that ATP hydrolysis is not required for and does not greatly affect the ability of WRN exonuclease to both initiate and continue degradation on the 3Ј ends of our substrate containing a 21-nt bubble. Our results are in contrast with an earlier report (28) that found ATP-dependent stimulation of WRN exonuclease activity on duplex substrates with recessed 3Ј ends, but perhaps more significantly also differ from those of a recent report (35) that indicated an ATP hydrolysis requirement in the 3Ј exonucleolytic degradation by WRN of a variety of substrates containing smaller bubbles or loops. Nevertheless, our data clearly show that ATP binding and hydrolysis are not required for exonucleolytic degradation of the 3Ј ends of a DNA substrate containing a 21-nt noncomplementary region in the midst of duplex DNA.
High Affinity Binding of WRN to Bubble DNA-Studies of WRN helicase and exonuclease functions on various types of DNA substrates suggest that WRN may act preferably on alternate (nonduplex) DNA structures (16,24,35). It is likely that this preference is manifested through differences in binding of WRN to specific structural features of DNA substrates. We used the EMSA to determine whether WRN could form a stable complex with a DNA substrate containing a noncomplementary bubble. To prevent WRN helicase and exonuclease activities on DNA substrates, both the incubation and electrophoretic steps of this assay were carried out at 4°C. By this analysis, WRN formed stable complexes with the bubble-containing substrate in the presence of ATP␥S (Fig. 6) or to a slightly lesser extent in the presence of ATP or without nucleotide cofactors (data not shown). By contrast, no protein-DNA complexes were detected when WRN was incubated with fully duplex DNA or when mock protein preparations (lacking recombinant WRN) were incubated with either DNA substrate ( Fig. 6 and data not shown). Thus, WRN formed stable and specific complexes with partially duplex DNA containing a 21-nt noncomplementary region.
DNase I Footprint of WRN on Bubble DNA-Although the above analysis identifies stable complexes between WRN and DNA containing a noncomplementary bubble, a more precise examination of these complexes requires a more sensitive method. DNase I footprinting has been widely used to pinpoint the binding sites of proteins on specific DNA sequences and structures. Notably, this method is also an equilibrium binding technique, and thus more sensitive than nonequilibrium tech-niques such as EMSA and filter binding assays. We employed this method to determine whether WRN binds stably to a DNA substrate containing a 21-nucleotide noncomplementary region, and to map the regions of DNA covered by bound protein (as measured by inhibition of DNase I endonuclease activity). As for the earlier EMSA experiments, WRN binding and DNase I endonuclease incubations were typically carried out at 4°C in ATP or ATP␥S to prevent WRN unwinding and exonuclease activities and thus maintain integrity of the DNA substrate structure. WRN unwinding activity was undetectable at 4°C (data not shown), and the lack of exonuclease activity by WRN at 4°C was demonstrated by the incubation of WRN and DNA substrate without subsequent DNase I treatment (Fig. 7, A-D,  second lane from left). DNase I digestion of the substrate containing a 21-nucleotide bubble surrounded by 35-bp (long arm) and 24-bp (short arm) duplex regions is shown in Fig. 7 (B and D, third lane from right). In general, the DNase I digestion was weak in the noncomplementary region, probably because of its single-stranded character. Nevertheless, increasing concentrations of WRN clearly inhibited DNase I incision in the noncomplementary region of the substrate (Fig. 7, B and D, lanes [3][4][5][6]. WRN binding to this substrate was complete by 5 min (data not shown) and protected a region surrounding the noncomplementary region, being particularly evident to the short (24 bp) arm of the substrate on the C-rich strand (Fig. 7B). Using relevant restriction sites present in the substrate (Fig.  1A), we estimate that, on the C-rich strand, WRN protects a region that includes the entire bubble, 10 nt of the short arm, and 5-6 nt of the long arm. In addition, the boundary between the region bound by WRN and unbound areas of the substrate could be inferred by the appearance of a significant DNase I-hypersensitive site (indicated by a thick arrow) at ϳ7 nt as well as a minor hypersensitive site (indicated by a thin arrow) at 13 nt from the edge of the bubble toward the 3Ј end of the C-rich strand on the long arm (Fig. 7B). Although the pattern is not as clear because of greater inhibition of DNase I incision on the G-rich strand in the noncomplementary region, WRN binding again inhibited DNase I incision in the vicinity of the bubble (Fig. 7D), covering the entire noncomplementary region plus ϳ8 and 6 nt on the 35-and 24-bp arms, respectively, and creating DNase I-hypersensitive sites on the short arm (toward the 3Ј end of the G-strand). Thus, the binding of WRN to substrate appears to encompass the noncomplementary region with a slight asymmetry toward the 5Ј end of each strand. The   FIG. 6. High affinity binding of WRN to bubble-containing DNA by EMSA. WRN (wild type, 0.5-3.0 l ϭ 12-72 fmol) or mock-purified protein preparation (volume normalized, 1.5 or 3.0 l) was incubated for 30 min at 4°C with either bubble-containing or fully duplex substrate (0.5 fmol each, both 32 P-labeled on the C80 oligomer). Protein-DNA complexes were resolved from free DNA by nondenaturing polyacrylamide (4%) gel electrophoresis.
binding affinity of WRN for this substrate was calculated from the concentration at which ϳ50% of the intensity of a specific band (indicated by asterisk) had disappeared to give a K a of 4 ϫ 10 8 M Ϫ1 . By comparison, no indication of DNase I inhibition was observed on either strand of the fully duplex control substrate, even using WRN concentrations that were greater than 4-fold higher than necessary to yield a "complete" footprint on the bubble substrate (Fig. 7, A and C). On the bubble-containing substrate, similar footprinting patterns were observed when an exonuclease-deficient/helicase-proficient mutant was used instead of wild type protein. This is the case even when DNase I digestion on the bubble-containing substrate was carried out using WRN-E84A in the presence of ATP␥S at 37°C (Fig. 8), conditions that do not permit either helicase or exonuclease activity. Although there were minor differences in the DNase I digestion pattern when the assay is carried out at 37°C, it is clear that WRN-E84A bound to the DNA in the vicinity of the bubble. These findings clearly demonstrate that WRN bound with high affinity to a noncomplementary region of DNA under noncatalytic (4°C) and catalytic (37°C) conditions.
In contrast, the WRN⌬369 -1432 deletion mutant protein containing only the exonuclease domain did not create a detectable footprint (even at significantly higher molar ratios of protein to DNA) in the vicinity of the bubble (Fig. 9). The absence of a footprint near the bubble with WRN⌬369 -1432 indicates that this region of the protein is not sufficient (and probably not involved) for the binding observed with the fulllength protein. Notably, no inhibition of DNase I incision was observed near either end of the bubble-containing substrate using wild type WRN, WRN-K477M, WRN-E84A, or WRN⌬369 -1432 when the assay was carried out at 4°C (Figs.  7 (B and D), 9, and 10B). A similar result was obtained when the DNase I assay was done with WRN-E84A at 37°C (Fig. 8). These results suggest that, although the exonuclease activity was directed at both ends of the substrate (see Fig. 3, A and B), the interaction of the WRN exonuclease domain with the DNA ends must be such that DNase I incision in those regions was not affected.
We also tested the effect of nucleotide cofactors on the binding of WRN to the bubble containing substrate. As ATP hydrolysis is prevented by carrying out the binding and footprinting incubations at 4°C, no difference was observed when WRN was incubated with the bubble substrate in the presence of ATP or ATP␥S (data not shown). In contrast, when nucleotide cofactors were omitted during the binding incubation, the footprint was much less distinct at comparable WRN concentrations (Fig.  10A). At a 2-4-fold higher WRN concentration in the absence of ATP, the pattern of DNase I digestion was similar if not identical to the pattern observed in the presence of ATP␥S (Fig.  10A). The similarity between these footprints indicates that the noncomplementary region is bound by WRN in the same conformation whether nucleotide cofactor is associated with the protein or not. To further test this hypothesis, we compared the binding of the wild type protein with that of WRN-K577M containing a point mutation in a conserved nucleotide binding/ hydrolysis motif that eliminates ATP hydrolysis and helicase activity (11,12). As expected, WRN-K577M had no unwinding activity on this bubble substrate (data not shown). When WRN-K577M was incubated with the bubble substrate in the presence of ATP␥S, it protected an area surrounding the bubble on the C-rich strand WRN from DNase I digestion (Fig. 10B)  Markers from bubble substrate with 32 P radiolabel on the G-rich strand were created as described in Fig. 3. Marker III is a mixture of BfaI, NspI, and HindIII digestions, whereas marker IV is a mixture of BclI and AluI digestions. pattern similar to that for wild type WRN. However, when compared with wild type protein in the presence of ATP␥S (Fig.  7B), a significantly higher concentration of WRN-K577M was necessary to achieve the same level of protection, as was the case with the wild type WRN footprint in the absence of ATP (Fig. 10A). Taken together, our results indicate that the binding of WRN to the bubble-containing substrate does not require a nucleotide cofactor. However, nucleotide binding significantly enhances the binding of WRN to the noncomplementary region of the substrate. By extension, it also appears that mutation of the conserved lysine to methionine in motif I of the RecQ helicase domain may eliminate not only the ATP hydrolysis activity but also the ATP binding capability of WRN. Another possibility is that this mutation lowers the DNA binding affinity of the WRN-K577M protein. However, the manner in which full-length WRN binds to the bubble structure appears to be similar, regardless of the presence or absence of nucleotide cofactors. Most importantly, we can conclude that the highly specific binding of WRN to the melted region of our substrate (that is observed with or without ATP) underlies the potent 3Ј 3 5Ј exonuclease activity that occurs on the nearby blunt ends. DISCUSSION Genomic instability in the premature aging and cancerprone hereditary disease, WS, is thought to be caused by DNA metabolic defects that result from a lack of WRN protein. This hypothesis has been strengthened by the identification and characterization of 3Ј 3 5Ј helicase and 3Ј 3 5Ј exonuclease activities of WRN. In this study, a substrate containing a 21-nt noncomplementary region (bubble) in the midst of duplex DNA (Fig. 1A) has been shown to be bound with high affinity by WRN using the EMSA and DNase I protection (footprinting) assays. The footprinting technique was further exploited to determine the precise region of DNA bound by WRN and the participation of the exonuclease and nucleotide binding domains of WRN in binding to this structure. We also demonstrate that this high affinity binding underlies an extremely efficient exonuclease activity directed at the nearby 3Ј ends of this substrate. In addition, a significant percentage of this substrate can be completely unwound by WRN helicase activity, although only at substantially higher WRN concentrations than necessary to achieve optimal exonuclease activity. We hypothesize that WRN binding to a melted region of duplex, as demonstrated by DNase I protection, serves as an anchor that permits the exonuclease domain of WRN to access nearby 3Ј ends at a much higher frequency than would occur by constant cycling of WRN on and off of a substrate. This notion would explain the vastly increased amounts of initiation and inward degradation of WRN exonuclease activity on this substrate containing a 21-nt noncomplementary region, in comparison to fully duplex and 3Ј recessed end substrates.
The DNase I assays reveal several interesting aspects regarding WRN protein binding to the bubble-containing DNA substrate. ATP is not required for stable binding of WRN to the bubble region, although ATP binding apparently increases the affinity of WRN for substrate significantly. In contrast, a completely duplex substrate of equal length and similar sequence was not detectably bound by WRN, even at substantially higher protein concentrations. From these experiments, we calculate the K a for WRN binding in the presence of ATP (or ATP␥S) to the bubble substrate to be 4 ϫ 10 8 M Ϫ1 . By comparison, the K a for a long single-stranded DNA substrate was estimated to be 10 8 to 10 9 M Ϫ1 and at least 5-fold higher than for double-stranded DNA (29). Thus, WRN binds to the bubble region (or the junctions between double-and single-stranded DNA contained therein) with affinity comparable with that for FIG. 8. Binding of WRN-E84A to the bubble substrate at physiological temperature. The bubble-containing substrate (0.5 fmol, C-rich strand-labeled) was incubated in WRN buffer containing ATP␥S (1 mM) for 30 min at 37°C with WRN-E84A (10 -90 fmol), then treated with DNase I (0.00015 unit) for 10 min at 37°C. DNA products were analyzed as described under "Experimental Procedures" and Fig. 3. The position of the noncomplementary region and the area of protection are indicated at right by curved and right-angle brackets, respectively. Size markers from bubble substrate with the 32 P label on the C-rich strand were generated as described in Fig. 3. Marker II is a mixture of BfaI and AluI digestions; marker V is a mixture of NspI, HindIII, and AvrII digestions. single-stranded DNA and certainly much better than a linear double-stranded DNA substrate.
WRN binding to the bubble-containing substrate protects an area from DNase I digestion that includes the entire melted region of 21 unpaired nucleotides on each strand as well as limited sequence (7-10 bp) within the duplex regions on either side. Therefore, WRN protects an area of DNA on each strand ϳ35-40 nt in length with a slight asymmetry toward the 5Ј side of the bubble. Interestingly, there is no detectable area of protection in the vicinity of either end of this substrate, suggesting that 3Ј end binding events that occur during exonucleolytic processing are transient, and potentially pointing toward a distributive mechanism for WRN exonuclease function. However, we cannot rule out the possibility that stable binding of the exonuclease domain to the 3Ј end does occur under catalytic conditions (using exonuclease-proficient WRN at 37°C) that cannot be tested because of WRN-associated digestion of DNA.
Our bubble-containing substrate is an excellent substrate for the 3Ј35Ј exonuclease activity of WRN. WRN exonuclease can act at either blunt end, degrading from the 3Ј end back toward and, with some frequency, through and beyond the noncomplementary region. By comparison, the comparable blunt-ended fully duplex substrate and a substrate with a 3Ј recessed end are poor substrates for WRN exonuclease activity. This indicates that high affinity binding to the bubble region is responsible for the enhanced exonuclease activity initiated at the blunt ends 35 and 24 bp away. The WRN⌬369 -1432 mutant containing only the intact exonuclease domain cannot bind to or digest the bubble-containing substrate, indicating that the elevated exonuclease activity of wild type WRN requires por-tions of the protein outside of the exonuclease domain to bind to the melted region.
Earlier studies have demonstrated that the presence of an ATP cofactor can stimulate the exonuclease activity of WRN on duplex substrates with 3Ј recessed ends (28). Our results using substrate containing a 21-nt bubble show that the level of WRN exonuclease activity is very high regardless of the presence or absence of nucleotide cofactor. The exonuclease activity of the ATPase/helicase-deficient WRN-K577M protein on this substrate is strikingly similar to that of wild type WRN in the absence of ATP, consistent with the idea that ATP binding and hydrolysis are not necessary for optimal degradation of this particular bubble substrate. The different patterns of WRN exonuclease digestion observed with or without ATP cofactors could be attributed either to ATP-influenced association and dissociation dynamics of WRN to the substrate, or to alteration, denaturation, or ATP-mediated unwinding of the DNA substrate as digestion proceeds. Although ATP binding and hydrolysis are dispensable for the heightened exonuclease activity of WRN on the ends of our bubble substrate, certainly ATP binding and hydrolysis play a role in substrate binding and the dynamics of exonuclease activity on several types of DNA structures (Refs. 28 and 35, and this study).
A discussion of the results of this study would not be complete without comparison to a recently published report (35) that examined WRN exonuclease activity on several types of DNA structures including small bubbles and extrahelical single-stranded loops. Both studies show that an internal bubble permits WRN 3Ј 3 5Ј exonuclease activity on a nearby blunt end duplex, a structure that is resistant to digestion in the context of fully complementary DNA. However, WRN exonuclease activity on a 46-bp substrate with an 8-nt bubble appears to require ATP hydrolysis and is strongly inhibited at 5 nt from the junction of the bubble (or loop) with duplex DNA (35). In contrast, our results using an 80-bp substrate with an internal bubble of 21 nt indicate that (i) WRN exonuclease activity occurs optimally even without ATP or helicase activity and (ii) digestion occurs up to, through, and to some extent beyond the bubble structure without substantial inhibition. The most likely explanation for these differences could stem from the larger size of our substrate and bubble (80 bp, 21-nt bubble) in comparison to theirs (46 bp, 8-nt bubble). Their results suggest that an 8-nt bubble may have to be further unwound by WRN helicase (with the concomitant requirement for ATPase activity) to create a stable binding site, whereas our results demonstrate that a 21-nt bubble constitutes a stable binding site for WRN without additional unwinding. The inhi-bition of exonuclease activity at 5 nt from the edge of the bubble could also be a result of the size of their substrate and its ATPase/helicase requirement. The unwinding activity that occurs in the presence of ATP combined with exonuclease activity at both ends would conceivably cause the short substrate to fall apart as digestion proceeds. As single-stranded DNA is not digested by WRN exonuclease, the denatured substrate is resistant to further degradation. As our substrate is significantly longer and subject to exonuclease degradation in the absence of unwinding activity, it probably retains double-stranded character even after significant degradation from both ends, thus facilitating a higher extent of digestion up to and through the bubble region. Another possibility would be that, under certain circumstances, binding of the WRN helicase domain to the bubble may simply physically inhibit the extent of degradation by the exonuclease domain.
WRN is unique among the human RecQ members in that it FIG. 11. Coordinated action of the bipartite and bifunctional WRN protein. Full-length WRN protein (shaded) is depicted as containing linked but physically separate helicase (hel, circular) and exonuclease (exo, oval) domains. Proposed directions of movement of individual domains are indicated with dashed arrows. A, uncoupled helicase and exonuclease activities of WRN on 3Ј overhang and 3Ј recessed ends, respectively, of a duplex substrate. B, coordinated activity of WRN on partially unwound duplex DNA substrates. Top, as demonstrated in this study, binding of WRN to unwound or noncomplementary region of DNA (probably mediated by the helicase domain) facilitates the activity of the 3Ј 3 5Ј exonuclease domain at a nearby blunt end, leading to degradation of one strand toward the melted region. Bottom, similarly, the binding of WRN to unwound duplex can direct the exonuclease activity to a nick, as shown for DNA containing an extrahelical loop by Shen and Loeb (35), leading to 3Ј 3 5Ј degradation of one strand toward the melted region from that discontinuity in double-stranded DNA. C, dynamic models for coordinated action of the helicase and exonuclease domains of WRN. Left, high affinity binding to an unwound region of DNA facilitates exonuclease activity at a more distant end or nick, accomplished either by bending or spooling of flexible DNA, or by helicase-mediated movement of the bubble along DNA toward the discontinuity. Right, also possible is the concerted degradation of long stretches of one strand by helicase-mediated movement of the bubble ahead of the tethered exonuclease domain. contains 3Ј 3 5Ј exonuclease activity as well as the requisite 3Ј 3 5Ј helicase activity. In this report, we have demonstrated that a helicase-deficient WRN protein is active as an exonuclease and vice versa, demonstrating that the exonuclease and helicase activities can occur independently of one another. Moreover, the WRN⌬369 -1432 mutant containing only the exonuclease domain is catalytically active (Ref. 27 and this study), indicating that the N-terminal part of WRN folds into a functional domain. This evidence suggests that wild type WRN is bipartite as well as bifunctional. Early experiments (13,25,28) implied that the domains have opposing specificities; helicase function requires a single-stranded region 3Ј to the duplex to be unwound, and exonuclease function requires a singlestranded region 5Ј to the duplex to be degraded. This concept is clarified in Fig. 11A, which depicts the individual catalytic activities of WRN acting at the opposite ends of a DNA duplex with flanking 3Ј and 5Ј single-stranded regions. Our experiments suggest another scenario; WRN binds to a partially melted region of DNA and can then degrade one strand of the DNA from certainly a nearby end or nick (Fig. 11B) and possibly a more distant end or nick through a DNA bending or threading mechanism (Fig. 11C, left). WRN helicase activity could increase the original size of the unwound region to create a more stable binding site and/or move the bubble along DNA either toward or away from the end or nick being acted on by the exonuclease domain. Coordinated movement of the WRN helicase and exonuclease domains along one strand of a duplex in a 3Ј 3 5Ј direction could potentially allow the degradation of long stretches of that same strand (Fig. 11C, right). This kind of mechanism might be very useful in a number of DNA metabolic pathways.
The extreme affinity and activity of WRN helicase/exonuclease on DNA substrates containing noncomplementary (or unwound) regions is likely to reflect the physiological function of WRN in DNA metabolism. WRN can unwind DNA structures provided the existence of a 3Ј single-stranded region, but has not yet been shown to be able to unwind DNA that is in a fully duplex form. However, several situations can be envisioned by which WRN might function to bind and at least partially unwind regions within essentially continuous (chromosomal) DNA: (i) binding/unwinding by WRN at specific sequences or sites, (ii) participation of WRN in a complex that initiates or assists in duplex melting, (iii) attraction of WRN to a "bubble" structure already partially unwound by activities of other proteins/complexes, and (iv) specific binding to noncomplementary or unwound regions of DNA that occur during or as a result of DNA metabolic processes. Evidence implicates WRN function at regions destined to become or already unwound. FFA-1, the WRN ortholog in Xenopus laevis, appears to participate with RPA at replication foci (30). In agreement with a possible role for WRN at replication origins and/or forks, WS cells have reduced rates of replication initiation and elongation (31,32), likely resulting in the observed extension of S phase (33). Both WRN and its S. cerevisiae homolog, Sgs1, have been proposed to have positive transcriptional roles (34,36), perhaps mediated through assistance in creation, expansion, or movement of transcription bubbles.
In contrast to the other human RecQ homologs, any proposed role for WRN should consider its combined helicase and exonuclease activities. Thus, WRN might create an optimum binding site through either its unwinding activity or its ability to bind to already melted regions of duplex DNA, then act coordinately as an exonuclease at a nearby 3Ј end, as demonstrated by our in vitro experiments. A significant body of evidence suggests that the RecQ family proteins may be involved in recombination or anti-recombinogenic pathways (1). One pos-sibility suggested by our experiments is that WRN helicase and exonuclease could act in concert to degrade heteroduplex DNA formed during nonhomologous (illegitimate) recombination. Alternatively, as strand invasion into duplex DNA is an initiating step in recombination, WRN or its RecQ cousins may be involved in assisting the strand invasion step or conversely, disrupting the D-loop structure created during strand invasion. Recently, BLM has been shown to bind and melt D-loops (37). Through its analogous helicase activity, WRN might also disrupt strand invasion intermediates and perhaps even degrade the invading 3Ј strand via its associated exonuclease activity. In support of a potential role in recombination or an antirecombinogenic pathway, WRN has been shown to migrate Holliday junctions in vitro (17). It has also been suggested that through an interaction with the Ku heterodimer, WRN may be involved in end processing during nonhomologous end joining pathway of DNA double-strand break repair (38 -40). Perhaps in coordination with Ku, a complex that may also have some DNA unwinding capability, bubbles or forked intermediates are created at damaged ends that are subsequently acted on by WRN exonuclease. Importantly, when provided a small extrahelical loop in the midst of duplex DNA, WRN exonuclease has been shown to degrade at a nearby nick 3Ј to the bubble (35). Such an activity is reminiscent of exonucleases that act from nicks created on the newly replicated strand during the mismatch repair process. A defect in this exonucleolytic processing might be expected to increase the persistence of these nicks and perhaps lead to double-strand breaks and a concomitant increase in deletion, insertion, and translocation mutations that would fit with some aspects of the WS genomic instability phenotype. A mismatch repair defect has been reported in WS cells (41), but further investigations in this area are needed. Another possible role for WRN could be the processing of Okazaki fragments that occur as a result of lagging strand replication, although at this time no direct evidence implicates WRN in such a pathway. Examination of WRN activities on DNA structures that more closely reflect true physiological substrates and a rigorous analysis of the DNA metabolic deficiencies of WS cells should lead to elucidation of the exact function of the WRN protein.