Structural Determinants for Substrate Binding and Catalysis by the Structure-specific Endonuclease XPG*

XPG belongs to the Fen1 family of structure-specific nucleases and is responsible for the 3 (cid:1) endonucleolytic incision during mammalian nucleotide excision repair. In addition, it has ill-defined roles in the transcription-coupled repair of oxidative DNA damage and likely also in transcription that are independent of its nuclease activity. We have used DNA binding and footprinting assays with various substrates to gain insight into how XPG interacts with DNA. Ethylation interference footprinting revealed that XPG binds to its substrates through interaction with the phosphate backbone on one face of the helix, mainly to the double-stranded DNA. By comparing DNA binding and cleavage activity using single-/double-stranded DNA junction substrates differing in the length of the single-stranded regions, we have found that the 3 (cid:1) but not the 5 (cid:1) single-stranded arm was necessary for DNA binding and incision activity. Furthermore, we show that although a 5 (cid:1) overhang is not required for XPG activity, an overhang containing double-stranded DNA near the junction inhibits the nuclease but not substrate binding activity. Apparently, junction accessibility or flexibility is important for catalysis but not binding of XPG. These results show that XPG has distinct requirements for binding and cleaving DNA substrates. Nucleases with specificity for DNA structural features rather are important many of DNA metabolism,

XPG belongs to the Fen1 family of structure-specific nucleases and is responsible for the 3 endonucleolytic incision during mammalian nucleotide excision repair. In addition, it has ill-defined roles in the transcriptioncoupled repair of oxidative DNA damage and likely also in transcription that are independent of its nuclease activity. We have used DNA binding and footprinting assays with various substrates to gain insight into how XPG interacts with DNA. Ethylation interference footprinting revealed that XPG binds to its substrates through interaction with the phosphate backbone on one face of the helix, mainly to the double-stranded DNA. By comparing DNA binding and cleavage activity using single-/double-stranded DNA junction substrates differing in the length of the single-stranded regions, we have found that the 3 but not the 5 single-stranded arm was necessary for DNA binding and incision activity. Furthermore, we show that although a 5 overhang is not required for XPG activity, an overhang containing double-stranded DNA near the junction inhibits the nuclease but not substrate binding activity. Apparently, junction accessibility or flexibility is important for catalysis but not binding of XPG. These results show that XPG has distinct requirements for binding and cleaving DNA substrates.
Nucleases with specificity for DNA structural features rather than sequence are important for many aspects of DNA metabolism, including replication, recombination, and repair (1,2). One such structure-specific endonuclease, human XPG, is involved in nucleotide excision repair (NER) 1 in a poorly defined pathway of transcription-coupled base excision repair of oxidative lesions and likely also in efficient RNA polymerase II-mediated transcription (3)(4)(5)(6)(7). Two inherited human syndromes are associated with deficiencies in XPG: xeroderma pigmentosum (XP) and Cockayne syndrome (CS). Although the XP phenotype arises from a defect in NER and is characterized by acute sensitivity to UV light and an extremely high incidence of skin cancer (8), defects in transcrip-tion-coupled base excision repair and transcription are likely responsible for certain manifestations of CS such as skeletal, neurological, and developmental abnormalities (9). Whether a defect in XPG causes XP or combined XP/CS is dependent on the position and type of mutation in the XPG gene. Point mutations resulting in a defect in the nuclease activity generally lead to an XP phenotype, whereas mutations resulting in truncation of the protein cause an XP/CS phenotype (10). These observations have led to the model that the nuclease activity of XPG is important for NER, whereas the involvement of XPG in transcription-coupled base excision repair and transcription requires binding of XPG to specific DNA structures but not its nuclease activity.
At the biochemical level, only the involvement of XPG in NER is understood in any detail. In mammals, NER is the major DNA repair pathway responsible for the removal of adducts formed by UV light and environmental mutagens from DNA (11). Two partially overlapping NER subpathways have been discerned: one process called global genome repair, which is able to eliminate NER lesions from all locations in the genome, and transcription-coupled repair, which is responsible for the accelerated removal of lesions from actively transcribed DNA strands (12). Global genome repair has been elucidated in some detail and can be divided in three major stages: (i) damage recognition and DNA opening to form a bubble-shaped structure around the lesion by the XPC-HR23B, TFIIH, XPA, RPA, and XPG factors, (ii) endonucleolytic incision 5Ј and 3Ј of the lesion by ERCC1-XPF and XPG, respectively, and release of an oligonucleotide of 24 -32 residues, and (iii) gap-filling repair synthesis and ligation to restore the original DNA sequence (13)(14)(15). Within NER, XPG also fulfills roles that are independent of its nuclease activity in the opening of the DNA helix around the lesion and the formation of a stable damage recognition complex (16,17). The physical presence but not the catalytic activity of XPG is also required for the subsequent 5Ј incision by ERCC1-XPF (18). Conversely, the nuclease function of XPG may be regulated through interaction with other NER factors, notably TFIIH, RPA, and XPA (19 -21). In transcription-coupled repair, the damage is thought to be recognized by the damage-stalled RNA polymerase. Although the detailed mechanism of transcription-coupled repair has not yet been clarified, it is thought to proceed similarly to global genome repair in steps subsequent to damage recognition.
The enzymatic activities of XPG have been studied in vitro, where it shows remarkable structure-specific properties, cleaving artificial DNA structures that contain ss/dsDNA junctions including bubbles, splayed arms, stem-loops, and flap substrates with the same polarity as in the NER reaction (22)(23)(24). The XPG protein (1186 amino acids) is a member of the Fen1 (flap endonuclease 1) family of structure-specific nucleases (2). Members of this family have a conserved nu-clease core composed of the "N region" (N-terminal) and the "I region" (internal), which contain a number of highly conserved acidic residues that are essential for nuclease activity (25)(26)(27). Furthermore, they share some conserved motifs that are implicated in DNA binding (28,29). The distance between the N and I regions varies among the family members and is about 700 amino acids in XPG. This "spacer region" has been shown to be dispensable for nuclease activity in the yeast XPG homolog Rad2 but is required for protein-protein interactions (30).
Although the Fen1 family members Fen1, XPG, and Exo1 (exonuclease 1) share a conserved nuclease domain and bind to ss/dsDNA junctions, there are differences in the exact substrate specificity of these three proteins. Although XPG efficiently processes bubble or stem-loop substrates, Fen1 and Exo1 only act on substrates that have free 5Ј ssDNA ends (31,32). In the absence of high resolution structures of Fen1 family members bound to DNA, indications of how these proteins might interact with their DNA substrates are based on DNA footprinting experiments of Fen1, Exo1, and the 5Ј exonuclease domain of DNA polymerase I and on modeling studies based on the structures of the free proteins (27,(33)(34)(35)(36)(37). Here, we have analyzed the requirements of XPG for binding to ss/dsDNA junctions using band shift and footprinting methods. Furthermore, we have studied the structural elements required for DNA binding and cleavage activities of XPG by using various substrates. Unexpectedly, XPG displays distinct requirements for binding and cleaving its DNA substrates. These observations might help to elucidate the catalytic and structural roles of XPG in various repair processes.

EXPERIMENTAL PROCEDURES
All of the enzymes were purchased from New England Biolabs unless indicated otherwise. [␥-32 P]ATP and [␣-32 P]ddATP were obtained from Amersham Biosciences. All of the other chemicals and reagent were from Sigma or Fluka. Routine molecular biological procedures were performed using standard protocols (38).
Expression and Purification of WT and Mutant XPG Proteins-Se- The lengths of the individual regions of the oligonucleotides are indicated (in nucleotides) for fixed lengths in the drawing; variable lengths are expressed as n, and the lengths are indicated in the list. Asterisks indicate the 5Ј 32 P-labeled end of the substrates. In figure legends, we use the substrate labels 1-9 and show the length of n in subscript. Oligonucleotide sequences are given under "Experimental Procedures." rum-free conditioned Sf9 insect cells were grown in suspension at 27°C under extensive stirring (according to instructions from Invitrogen). For protein expression, 300 ml of Sf9 cells (1.5 ϫ 10 6 cells/ml) were infected with the appropriate baculovirus at an multiplicity of infection of 5 and were grown in an Erlenmayer flask with air-permeable filter caps at 27°C and 200 rpm. The cells were harvested 48 -54 h after infection by low speed centrifugation and resuspended in 40 ml of buffer A (50 mM potassium phosphate, pH 7.6, 10% glycerol, 5 mM ␤-mercaptoethanol, 1 mM PMSF, containing one EDTA-free protease inhibitor mixture tablet from Roche Applied Science). After incubation on ice for 20 min, NaCl was added to a final concentration of 800 mM, and the lysate was sonicated on ice (Bandelin Sonicator, 60% power at 30% interval) by three 1-min intervals, with at least 2 min of cooling on ice between each interval. The lysate was clarified by centrifugation at 40,000 ϫ g for 45 min at 4°C. The supernatant was supplemented with 1 mM PMSF and incubated overnight with 4 ml of nickel-nitrilotriacetic acid-agarose beads (Qiagen) in presence of 1 mM imidazole. The beads were collected by low speed centrifugation and resuspended in 10 ml of buffer A containing 800 mM NaCl and 1 mM imidazole and packed in column (1 ϫ 5 cm). The column was washed stepwise with 20 ml of buffer A containing 1 mM imidazole with 500 ml of NaCl, 20 ml of 10 mM imidazole with 300 mM NaCl, or 10 ml of 50 mM imidazole with 300 mM NaCl. XPG was eluted with 6 ml of buffer A containing 100 mM imidazole with 300 mM NaCl. The eluate was then applied on a 1-ml HiTrapS column (Amersham Biosciences) after a 1:2 dilution in buffer B (25 mM HEPES, pH 8.0, 10% glycerol, 1 mM EDTA, 10 mM ␤-mercaptoethanol, 1 mM PMSF) at a flow rate of 0.5 ml/min. The protein was eluted with a 25-ml gradient from 100 mM to 1 M NaCl. XPG eluted between 450 and 700 mM NaCl. Fractions containing XPG were pooled and diluted to 150 mM NaCl with buffer C (25 mM HEPES, pH 7.5, 10% glycerol, 10 mM ␤-mercaptoethanol, 1 mM PMSF) and loaded on a 1-ml HiTrapQ column (Amersham Biosciences). XPG eluted in 2.5 ml between 400 and 500 mM NaCl with a 30-ml gradient from 150 mM to 1 M NaCl. The protein concentration was estimated from absorption at 280 nm. Purified protein was flash frozen in aliquots and stored at Ϫ80°C. Typically, 0.3-1.2 mg of pure XPG protein at a concentration of 0.2-0.5 mg/ml was obtained.
Endonuclease Assay-2.5 nM of 5Ј 32 P end-labeled DNA substrate was incubated with 1.25-5 nM XPG in nuclease buffer (25 mM Tris-Cl, pH 6.8, 10% glycerol, 2.5 mM ␤-mercaptoethanol, 0.5 mg/ml bovine serum albumin, 30 mM KCl) in the presence of 2 mM MgCl 2 or 0.5 mM MnCl 2 in a reaction volume of 20 l for 1-2 h. The reactions were stopped by the addition of an equal volume of formamide loading buffer (90% (w/v) formamide, 0.5ϫ TBE, 0.1% bromphenol blue, and xylene cyanol FF) and heating for 5 min at 95°C. The samples were loaded onto a 15 or 20% (cross-linking ratio 19:1) denaturing polyacrylamide gel containing 1ϫ TBE and 7 M urea and run for 1-2 h at 2000 V. The bands were visualized by autoradiography, and quantification was carried out using a Molecular Dynamics (Sunnyvale, CA) Storm860 Phos-phorImager and Molecular Dynamics ImageQuant version 1.2 software.
Electrophoretic Mobility Shift Assay-5Ј 32 P end-labeled substrate (1 nM) with or without a 10 -500-fold excess of unlabeled competitor DNA was incubated with various amounts of XPG in EMSA buffer (50 mM Tris pH 6.8, 20 mM KCl, 15% glycerol, 0.5 mM ␤-mercaptoethanol, 50 mg/ml bovine serum albumin) in a reaction volume of 20 l. Unless indicated otherwise, no metal cofactor was used. After equilibration at room temperature for 10 -30 min, the samples were loaded on a 5% (37.5:1) native polyacrylamide gel containing 0.5ϫ TBE and run in precooled buffer at 90 V for 2 h at room temperature. The gels were dried, and the bands were visualized by autoradiography.
Exonuclease III Footprinting-The XPG mutant E791A (12-30 nM) and 5Ј 32 P end-labeled substrate (1) 20 (2.5 nM) were incubated at room temperature under EMSA conditions (see above) in the presence of 2 mM MgCl 2 and 100-fold excess of cold dsDNA competitor (substrate 9) for 20 min. ExoIII digestion was initiated at 15°C by the addition of 2 units (1 l) of ExoIII, allowed to proceed for 10 -20 min, and stopped by the addition of EDTA to 50 mM and formamide loading buffer and heating for 10 min at 75°C. The reactions were analyzed on a 15% sequencing gel, and the bands were visualized by autoradiography.
Ethylation Interference DNA Footprinting-5 pmol of 5Ј 32 P endlabeled oligonucleotides A 20 or B 19 was suspended in 50 l of 100 mM calcium cacodylate (Sigma) and added to a volume of 150 l of freshly prepared ethylnitrosourea (170 mg/ml in ethanol). Ethylation was performed in Parafilm-sealed safe lock Eppendorf tubes for 75 min at 50°C under gentle agitation. Excess ethylnitrosourea was then extracted 6 times with a 1-ml portion of water-saturated diethyl ether. The remaining traces of ether were evaporated for 5 min using vacuum centrifugation. The dried ethylated oligonucleotides were redissolved in TE buffer (25 l) and annealed in annealing buffer (see above) with a 5-fold excess of the complementary strand by heating for 5 min at 90°C and cooling to room temperature.
For the DNA binding reaction, 25 nM of substrate 1 (A 20 :B 19 ) were incubated under EMSA conditions described above in presence of a 100-fold excess of cold dsDNA competitor (substrate 9) with 150 nM of WT XPG in a volume of 20 l for 5-10 min at room temperature, resulting in about 40 -50% complex formation. The mixture was run on a 5% native polyacrylamide gel for 90 min at 90 V. The bands of bound and unbound fractions were visualized by autoradiography, excised, and electroeluted (1ϫ TBE buffer) into 10 M NH 4 OAc buffer for 4 h at 90 V. The volume of the electroeluted fractions was reduced on Microcon YM-10 filters (Millipore) and washed twice with TE buffer before precipitation with 100% ethanol in presence of 10 mg/ml tRNA followed by washing twice with 80% ethanol. DNA strand cleavage at ethylated phosphate sites was achieved by incubation in 100 l of 140 mM NaOH at 90°C for 30 min. After neutralization with 15 l of 1 M AcOH, the samples were ethanol-precipitated and resuspended in gel loading buffer. Following quantification of the samples in a scintillation counter, the samples were loaded onto a 15% 7 M urea denaturing gel along with T, C, and G/A sequencing ladders prepared from the same target DNA (40,41). Briefly, the T-specific modification was prepared by incubation of 0.5 pmol of denatured labeled oligonucleotide in 1.27 mM potassium permanganate for 20 min, for C in 3 M hydroxylamine hydrochloride, pH 6.0, and for G/A in 80% formic acid in a volume of 20 l. Modified oligonucleotides were ethanol-precipitated and cleaved at modified sites by incubation for 15 min at 90°C in 1.1 M pyrrolidine and evaporated to dryness. The gels were visualized and quantified by PhosphorImager.

Activity of Purified XPG Protein on ss/dsDNA Junctions-
Wild type and E791A mutant human histidine-tagged XPG endonuclease was purified from Sf9 insect cells as described under "Experimental Procedures" (Fig. 1A). We compared the structure-specific nuclease activity of XPG on splayed arm, bubble, and stem-loop DNA substrates (Fig. 1B). Incubation of XPG in the presence of either MgCl 2 (lanes 2, 5, and 8) or MnCl 2 (lanes 3, 6, and 9) resulted in efficient DNA cleavage of the splayed arm (22 and 91%, respectively) and bubble substrates (70 and 96%), whereas only weak activity was found on the stem-loop substrate (3 and 4%). The cleavage positions were mapped 1 nt into the duplex region of the junction for the splayed arm substrate, whereas the bubble and stem-loop structures were cleaved 1 nt into the ssDNA just 5Ј of the ss/dsDNA junction (data not shown). Although a bubble-shaped structure might formally resemble a reaction intermediate of NER more closely than a splayed arm structure, we decided to perform our further analysis of substrate binding and cleavage by XPG with the splayed arm substrate, because binding of the protein at two equivalent positions of a bubble would complicate such an analysis. We designate a 5Ј ssDNA overhang of a splayed arm substrate as an unpaired overhang with a 5Ј end, whereas a 3Ј ssDNA overhang refers to the unpaired overhang with a 3Ј ssDNA end.
Structure-specific Binding of XPG to ss/dsDNA Junctions-We used EMSA to define conditions that allow specific binding of the XPG endonuclease to ss/dsDNA junctions. Incubation of XPG with a splayed arm substrate (substrate 1 in Table I) in the absence of metal cofactor resulted in the formation of a band indicative of a protein-DNA complex (Fig. 2A, lane 2). Optimal binding was observed over a broad range of pH (6 -8.5) in the presence of bovine serum albumin and glycerol and at a KCl concentration of Ͻ25 mM. Stable DNA binding was observed if the concentration of substrate was Ն1 nM and that of XPG was Ն6 nM. At high XPG concentrations (Ն10 nM), we observed the appearance of a second band, which may be due to the formation of a dimeric XPG-DNA complex.
Incubation of the substrate 1 with XPG in the presence of Mg 2ϩ or Mn 2ϩ resulted in cleavage ( Fig. 2A, lanes 3 and 4), indicating that the XPG-DNA complex observed in EMSA is in an active conformation. The XPG mutant E791A, which was shown to have severely reduced endonuclease activity (26), formed a more stable complex with substrate 1 in the presence of Mg 2ϩ and Mn 2ϩ , although the substrate was partially cleaved under the conditions used ( Fig. 2A, lanes 8 and 9). Through small variations in pH or enzyme concentration, cleavage could be prevented entirely without compromising DNA binding (data not shown). The presence of Ca 2ϩ , which does not support nuclease activity in most enzymes, resulted in the formation of stable protein-DNA complex with both WT and the E791A mutant ( Fig. 2A, lanes 5 and 10). Apparently, the presence of divalent metal ions is not required for DNA binding, and the identity of the metal has no significant influence on the binding affinity.
To determine whether the interaction of XPG with the substrate was specific for a ss/dsDNA junction, we tested the effect of various DNA competitors on the binding reaction. Oligonucleotides containing only the ssDNA (Fig. 2B, lanes 3-5) or dsDNA (lanes 6 -8) part of the splayed arm substrate had no effect on the binding reaction (a Ͼ500-fold excess was needed to detect weak competition), whereas the addition of a 40-fold excess of specific competitors, either flap substrate 3 (lanes 9 -11) or splayed arm substrate 1 (lanes 12-14), abolished the binding reactions under the same conditions. These observations demonstrate that XPG shows a strong binding preference for DNA molecules containing ss/dsDNA junctions.
Exonuclease III Footprinting of XPG-DNA Complex-To gain more insight into the way XPG is bound to the ss/dsDNA junctions, we used various footprinting assays. We first used ExoIII footprinting. ExoIII is a 3Ј to 5Ј exonuclease that digests ssDNA or dsDNA in a progressive fashion. If an additional protein is bound to the DNA, ExoIII stalls on the 3Ј side of the protein-binding site, leading to preferential incision at that position. Because the exonuclease requires the presence of Mg 2ϩ for activity, footprinting was performed with the XPG active site mutant E791A under conditions allowing only Յ5% of incision to occur (data not shown). To detect the 3Ј borders of the junction binding, XPG was incubated with the 5Ј endlabeled substrate and the unprotected DNA degraded by ExoIII following equilibration (Fig. 3, lanes 6, 7, 12, and 13).
On the incised strand, where XPG WT protein cleaves 1 nt into the duplex region (lane 4), strong bands could be detected 5-7 bases in the duplex region away from the junction. Bands of higher mobility, 3-5 nt in the single-stranded 5Ј overhang, were detectable. The meaning of these bands is unclear, because our further footprinting experiments (see below) did not reveal any DNA interactions upstream of this region. Furthermore, very weak bands were observed 14 -15 nt in the duplex region, which were more pronounced using shorter digestion times or lower ExoIII concentrations (data not shown). However, these milder digestion conditions failed to reveal the prominent boundary of 5-7 nt into the duplex region. Apparently, ExoIII is able to displace XPG from the ss/dsDNA junction under certain conditions. On the complementary strand, a clear band was obvious on the first nt on either side of the junction, indicating that the binding site of XPG starts around the junction in the dsDNA region. The faster migrating fragments are most likely products of nonspecific degradation, because the control without XPG (lane 11) shows the same pat-tern. A second XPG-specific band was mapped close to the end of the 3Ј overhang. Again, our studies were not able to explain the nature of this specific band. The pattern of ExoIII footprint suggests that XPG contacts the incision as well as the complementary strand of its substrate, mainly in the dsDNA region.
Phosphate Ethylation Interference DNA Footprinting-To gain more detailed insight of how XPG binds to junctions, we performed ethylation interference DNA footprinting (42), which directly detects contacts of the protein with the phosphates in the substrate DNA. We treated either the incised or complementary 5Ј end-labeled strand in single-stranded form with ethylnitrosourea to yield Ͻ1 phosphotriester bond/strand, annealed it to its partner strand and incubated the resulting splayed arm substrate 1 with XPG. Bound and unbound fractions were separated by EMSA, the corresponding bands were excised, and the DNA was isolated by electroelution. The oligonucleotides were cleaved at the ethylated positions with sodium hydroxide, and the resulting fragments were resolved on a sequencing gel. The pattern of bands obtained revealed three main regions in which XPG contacted the phosphodiester bonds of the DNA on the incised strand (Fig. 4A, left panel): close to the junction in the 5Ј single-stranded overhang (4 to 1 nt away from the junction), directly 3Ј to the cleavage site (2-5 nt into the dsDNA region), and further upstream in the duplex region (10 -13 nt into the dsDNA region). Significant contacts to the DNA could also be observed on the complementary strand in one continuous stretch 1-14 nt away from the junction in the duplex region (Fig. 4A, right panel). Three independent experi- ments for each strand were performed and quantified (Fig. 4B).
To determine whether the absence of the metal cofactor had any influence on the incision pattern, we performed the same experiment with the E791A mutant of XPG in the presence of Mg 2ϩ . Within the limitations of our resolution, we were not able to detect any significant differences in the footprints of WT XPG without metal and of E791A with metal (data not shown). In summary, the ethylation interference footprinting shows that XPG endonuclease makes numerous phosphate contacts on both strands around the junction, mainly in the dsDNA region.
Minimal Size of the Single-stranded Overhang Required for Substrate Cleavage and Binding-Because our EMSA and footprinting data revealed specific binding to the double-stranded region at the ss/dsDNA junction but not extensive DNA contacts to the single-stranded regions of the substrate, we decided to determine the minimal lengths of the 5Ј and 3Ј singlestranded overhangs needed for DNA binding and cleavage activity. We therefore tested the incision activity of XPG on splayed arm substrates with various lengths of the ssDNA region of the incised strand, the complementary strand, or both. Shortening the length of the 5Ј overhang from 20 to 2, 1, or 0 nt resulted in a 1.5-fold (2 nt) or 2-fold (1 and 0 nt) reduction of incision activity (Fig. 5, A and C). The position of the incision remained unchanged, yielding 5Ј end-labeled cleavage products of 21, 3, 2, and 1 nt in length, respectively, and a product of 29 nt in length in all cases if the incised strand was labeled at the 3Ј end (data not shown). Truncation of the 3Ј ssDNA overhang from 19 to 3, 2, 1, and 0 nt had a more pronounced influence leading to 2-and 4-fold reduction (3-and 2-nt overhangs) and a trace or no activity (1 and 0 nt overhangs), respectively (Fig. 5, B and C). The efficiency of cleavage of substrates with truncations of the complementary strand was partially restored by substituting Mn 2ϩ for Mg 2ϩ , although this led to the formation of products with different incision sites in some cases (Fig. 5B, lanes 15-20). Apparently, the two catalytic active divalent metal ions can influence the conformation of the DNA substrates, which can affect protein activity under certain conditions. We then examined the effect of simultaneous truncation of both the 5Ј and 3Ј ssDNA overhangs of the splayed arm substrate (Fig. 5D). We used substrates with 2-and 0-nt 5Ј overhangs together with a 3Ј overhang of 20, 3, 2, or 0 nt (Fig. 5C). We observed a combined decrease in incision efficiency by shortening both ssDNA regions. The smallest substrate that was cleaved, albeit at very low efficiency, consisted of a 5Ј overhang of 0 nt and a 3Ј overhang of 2 nt (Fig. 5C, lane 29).
We wondered whether the reduction of cleavage efficiency upon truncation of the 3Ј ssDNA overhang was caused by a reduction in the DNA binding or phosphodiester bond hydrolysis step. Therefore, we performed EMSA assays to test DNA binding activity on substrates with truncated 3Ј overhangs. Indeed, DNA binding decreased monotonically with shortening of the 3Ј ssDNA overhang from 20 to 3, 2, 1, and 0 nt (Fig. 6), indicating that the 3Ј ssDNA overhang is required for binding of XPG to ss/dsDNA junctions. In contrast, truncation of the 5Ј overhang did not show a significant decrease in DNA binding activity (data not shown).  (Table I, substrates 1 and 2). The length of the 5Ј overhang (n) is 2 (lanes 21-25) or 0 nt (lanes 26 -30), and the 3Ј overhang (m) is 19 (lanes 21, 22, 26, and 27), 3 (lanes 23 and 28), 2 (lanes  24 and 29), or 0 nt (lanes 25 and 30).  (lanes 1-4), 2 3 (lanes 5-8), 2 2 (lanes 9 -12), 2 1 (lanes [13][14][15][16], and 2 0 (lanes [17][18][19][20] were incubated with XPG at the following concentrations: lanes 1,5,9,13, and 17, 0 nM; lanes 2, 6, 10, 14, and 18, 3 nM; lanes 3, 7, 11, 15, and 19, 6 nM; lanes 4, 8, 12, 16, and 20, 7.5 nM. Separation of DNA Binding and Cleavage Activity Using Different Junction Structures-Based on our results demonstrating that the minimal requirement for processing a ss/ dsDNA junction is a 2-nt 3Ј overhang, we became interested in determining how replacing the ssDNA overhangs partially or entirely with dsDNA would affect XPG activity (Fig. 7). We first investigated flap substrates, in which either the 5Ј or 3Ј ssDNA overhang was replaced with dsDNA. Substrates in which the 3Ј overhang on the nonincised strand was double-stranded (substrate 3) were bound and cleaved with similar efficiency as the splayed arm substrate 2 (Fig. 7, compare A, lanes 6 -10 and 1-5,  with B, lanes 3 and 2). Substrates containing dsDNA on the 5Ј overhang (substrate 4) were bound efficiently (Fig. 7A, lanes 11-15) but were not processed under the same conditions (Fig.  7B, lane 4). Apparently, a 5Ј dsDNA overhang interferes with incision but not with binding by XPG. A related substrate in which the dsDNA 5Ј overhang was recessed by 3 nt on the side of the junction (substrate 5) yielding 3 nt of ssDNA adjacent to the junction was bound (Fig. 7A, lanes 16 -20) and processed by XPG (Fig. 7B, lane 5), indicating that ssDNA character of the 5Ј overhang was only required near the junction but not at the 5Ј end. We then investigated substrates in which both the 5Ј and 3Ј overhangs were made of dsDNA (substrate 6), and three-way junctions with (substrate 8) or without mismatches (substrate 7) at the site of the junction. These substrates were bound preferentially compared with the splayed arm, flap, or duplex 5Ј overhang structures (Fig. 7A, compare lanes 21-35 and  1-20), but displayed minimal (gap structure 6; Fig. 7B, lane 6) or no incision (three-way junctions with or without mismatches; Fig. 7B, lanes 7 and 8). Increasing XPG concentration by a factor of 30 and labeling of the complementary strands did not result in cleavage (data not shown). Apparently, the nature of the overhangs in the junctions is not crucial for DNA binding, whereas the nuclease activity of XPG requires the presence of a single-stranded or no 5Ј overhang and is blocked by dsDNA immediately adjacent to the junction. DISCUSSION As part of our studies on the structure-specific endonucleases involved in NER (43), we were interested in characterizing the determinants of XPG interaction with its DNA substrates using binding, cleavage, and footprinting techniques. XPG is a member of the Fen1 family of structure-specific endonucleases, which also includes Fen1 and Exo1. Consistent with their contribution to different repair pathways, these two proteins and XPG show distinct but overlapping substrate specificities and cleave ss/dsDNA junctions in the dsDNA region adjacent to a 5Ј ssDNA overhang (2,31,32).
Junction-specific DNA Contacts of XPG-We established conditions that allowed the stable binding of XPG to its substrates by preventing incision either by omitting the metal cofactor or by using the E791A mutant of XPG, which displays greatly reduced nuclease activity (26). Identical interference patterns were observed in our footprinting studies using WT XPG in the absence (Fig. 4) or the E791A mutant in the presence of the metal cofactor (data not shown), indicating that the metal ion does not significantly influence substrate binding prior to incision. We used several protecting footprinting techniques with XPG-DNA complexes, including exonuclease III,  Table I. A, EMSA with 1 nM substrate in presence of a 150-fold excess (150 nM) of unlabeled dsDNA 9. Substrates 1 20 (lanes 1-5), 3 (lanes 6 -10), 4 (lanes 11-15), 5 (lanes 16 -20), 6 (lanes 21-25), 7 (lanes 26 -30), and 8 (lanes [31][32][33][34][35] were incubated with XPG at the following concentrations: 0 nM (lanes 1, 6, 11, 16, 21, 26, and 31), 0.8 nM (lanes 2, 7, 12, 17, 22, 27, and 32), 1.6 nM (lanes 3, 8,13,18,23,28, and 33), 3.2 nM (lanes 4,9,14,19,24,29, and 34), 6.4 nM (lanes 5, 10, 15, 20, 25, 30, and 35). B, incision assay with 2.5 nM substrates 1 20  DNase I, hydroxyl radical, or copper phenanthroline footprinting. In these techniques, the region of the DNA bound by the protein is protected from degradation, and a fundamentally kinetic parameter, the effect of the protein bound to DNA on the rate of attack of the footprinting reagent, is observed. Except for the studies with exonuclease III (Fig. 3), we failed to observe a clear footprint with any of these methods (data not shown), which might be due to fast on and off rates of XPG binding to DNA. We also used ethylation interference footprinting, a technique that measures the effect of ethylation of individual phosphodiester bonds in terms of DNA binding. In contrast to protection footprinting techniques, with interference methods the effect of DNA modifications on the equilibrium binding strength of a protein is observed. Interference footprinting is thus a thermodynamically rather than kinetically controlled experiment (44).
The interface of XPG binding to DNA revealed by these studies is summarized in Fig. 8. XPG makes phosphate contacts mainly in the double-stranded region of the DNA on both strands spanning about 12 nucleotides. The strongest phosphate contacts mapped to one face of the B-DNA helix, suggesting that XPG interacts preferentially with the splayed arm substrate on the side where the 5Ј single-stranded overhang protrudes from the junction.
XPG makes three additional phosphate contacts in the 5Ј ssDNA overhang adjacent to the junction on the same face as the contacts to the dsDNA, which have however only a minor influence on binding affinity. Footprinting studies of exonuclease I and DNA polymerase I revealed similar modes of binding, with main contacts being made on one face of the dsDNA and less extensive contacts on the ssDNA overhangs. The orientation of binding of exonuclease I and DNA polymerase I appears to be different, because it makes contacts on the 3Ј ssDNA overhang and that face of the helix (34,35). Overall it appears that XPG and the other Fen1 family nucleases interact in a similar way with the double-stranded portion of a junction containing substrate through multiple nonspecific phosphodiester backbone contacts.
An intriguing observation from the footprinting studies is that although XPG is highly specific for binding to ss/dsDNA junctions, not many contacts to the ssDNA overhangs were observed. These observations were further emphasized when we investigated how far the ssDNA overhangs could be truncated without loosing cleavage and binding activity of XPG. Although deletion of the 5Ј overhang only reduced XPG incision activity by 50%, at least two nucleotides of the 3Ј overhang were required to retain substantial XPG binding and cleavage activity. Similar observations were made for Fen1 and the exonuclease domain of DNA polymerase I, where shortening of the 5Ј overhang only had a minor effect on the activity of the two enzymes (35,45). Because we failed to observe phosphate contacts to the 3Ј overhang in our footprinting assays, it will be of interest to determine what contacts are responsible for the specific binding to the ss/dsDNA junctions at the molecular level. XPG might recognize distortion or bending of the DNA bases at the junction in the 3Ј single-stranded overhang.
Distinct Requirements for Binding and Cleaving Substrates by XPG-After showing that only small parts of the ssDNA overhangs were required for junction specific cleavage by XPG, we thought it might be insightful to test to what extent dsDNA overhangs support XPG activity. Our results indicate that 3Ј overhangs can be single-or double-stranded without significantly influencing substrate binding and catalysis by XPG. The nature of the 5Ј overhang is more critical; 5Ј dsDNA overhangs strongly interfere with XPG cleavage activity, regardless of the nature of the 3Ј overhang. The inhibition of cleavage of a 5Ј dsDNA overhang by XPG is in agreement with a study of cleavage activity of XPG on model substrates of transcriptioninduced immunoglobulin switch regions, in which XPG cleaved substrates with a ssDNA overhang, but not dsDNA 5Ј overhangs (46). Substrates in which the 5Ј dsDNA overhang is recessed to yield 3 nt of ssDNA adjacent to the junction are efficiently processed by XPG, indicating that the substrates containing 5Ј dsDNA overhangs likely inhibit XPG activity by restricting accessibility and/or flexibility of the substrate at the junction. Because 5Ј dsDNA overhangs do not interfere with substrate binding by XPG, this accessibility/flexibility appears to be only important for catalysis not binding. Particularly instructive in this respect are three-way junction substrates, which are bound severalfold more efficiently than splayed arm substrates by XPG but are completely resistant to cleavage. It was shown that cationic surfaces of proteins can induce substantial DNA bending by neutralizing phosphates on one face of the helix (47). Because we have found that also XPG binds to the phosphodiester backbone on one face of the helix, it is possible that substrates like three-way junctions that already contain a bent backbone favor the XPG-substrate interaction. XPG and other structure-specific nucleases might recognize ss/dsDNA junctions by sensing the bending of the junction. The flexibility/accessibility of the junction could be a necessary second step for rendering the enzyme catalytically competent. DNA binding of XPG to specific DNA structures may be important for the additional roles of the protein in transcriptioncoupled base excision repair and transcription that do not require its endonuclease activity (4 -7). Therefore, the binding of XPG to three-way junctions raises the question about the potential physiological significance of this activity. Three-way junctions are possible intermediates in homologous recombination and replication (48). Although XPG has not yet been directly implicated in these processes, it is tempting to speculate that the severe phenotype of certain XPG patients and XPG knockout mice might partially stem from a special role of XPG in binding distorted DNA structures that are intermediates in transcription, replication, or recombination (49).
Further studies are required to determine in detail how XPG and other Fen1 family members recognize ss/dsDNA junctions. A mechanistically particularly intriguing question is why certain substrates are bound but not cleaved by XPG. Understanding this difference will be crucial for the elucidating the structural and catalytic roles of XPG.