Conserved Residues of Human XPG Protein Important for Nuclease Activity and Function in Nucleotide Excision Repair*

The human XPG endonuclease cuts on the 3′ side of a DNA lesion during nucleotide excision repair. Mutations in XPG can lead to the disorders xeroderma pigmentosum (XP) and Cockayne syndrome. XPG shares sequence similarities in two regions with a family of structure-specific nucleases and exonucleases. To begin defining its catalytic mechanism, we changed highly conserved residues and determined the effects on the endonuclease activity of isolated XPG, its function in open complex formation and dual incision reconstituted with purified proteins, and its ability to restore cellular resistance to UV light. The substitution A792V present in two XP complementation group G (XP-G) individuals reduced but did not abolish endonuclease activity, explaining their mild clinical phenotype. Isolated XPG proteins with Asp-77 or Glu-791 substitutions did not cleave DNA. In the reconstituted repair system, alanine substitutions at these positions permitted open complex formation but were inactive for 3′ cleavage, whereas D77E and E791D proteins retained considerable activity. The function of each mutant protein in the reconstituted system was mirrored by its ability to restore UV resistance to XP-G cell lines. Hydrodynamic measurements indicated that XPG exists as a monomer in high salt conditions, but immunoprecipitation of intact and truncated XPG proteins showed that XPG polypeptides can interact with each other, suggesting dimerization as an element of XPG function. The mutation results define critical residues in the catalytic center of XPG and strongly suggest that key features of the strand cleavage mechanism and active site structure are shared by members of the nuclease family.

The XPG protein is a DNA endonuclease with remarkable structure-specific properties, cleaving near the junctions between duplex and single-stranded DNA with a defined polarity.
In human cells, XPG functions to cleave on the 3Ј side of a damaged site in DNA during NER, the process that removes injuries induced by UV light and many chemical agents. Individuals with mutations in XPG are almost equally divided between those with the inherited syndrome xeroderma pigmentosum (XP) and those with a combination of both XP and Cockayne syndrome (CS). The XP phenotype includes acute sun sensitivity with a high incidence of cancers and, in some individuals, progressive neurological degeneration. The even more severe CS is associated with a characteristic constellation of skeletal, neurological, and developmental disorders. Whether an individual is affected with XP only or with both XP and CS is correlated with the severity of the mutation in the XPG gene and its effect on protein function (8 -10).
During the NER process, lesions are recognized and removed as part of an oligonucleotide fragment, and the resulting gap is subsequently filled and ligated (11)(12)(13). XPA, XPC, and the single-stranded binding protein RPA preferentially bind to damaged DNA. XPB and XPD proteins are the two DNA helicase subunits of TFIIH, which has a dual function both as a core NER factor and as a basal transcription factor for RNA polymerase II. Its helicase activities and the XPC-hHR23B heterodimer participate in the formation of a key open intermediate required for dual incision (14,15). The incisions made on each side of a lesion are made by two different DNA endonucleases. The ERCC1-XPF heterodimer makes the 5Ј incision (16,17), whereas XPG protein cuts 3Ј to the lesion (6,18). The ERCC1-XPF complex is probably recruited through interaction with XPA and RPA proteins (6,18), whereas XPG co-purifies with TFIIH under some conditions (19) and has been reported to interact with several TFIIH subunits (20). Both nucleases display structure specificity because they cut bubble, splayed arm, or flap structures near the border of single-stranded and duplex DNA. ERCC1-XPF (and its S. cerevisiae counterpart Rad10-Rad1) cuts on the strand that leads off from the junction in the 5Ј 3 3Ј direction (21), whereas XPG (and its S. cerevisiae homolog Rad2) makes endonucleolytic incision close to the 3Ј junction, where single-stranded DNA meets duplex DNA (6,7).
To gain insight into the important residues for the NER function of XPG, we have introduced site-directed mutations into the conserved N and I regions. A combination of biochemical and cellular tests was then used to measure the function of the mutated proteins. Their intrinsic endonuclease activity was measured on a model bubble substrate, and their repair function was assayed by their ability to form an open preincision complex and to complete dual incision of DNA containing a single cisplatin adduct in the presence of purified NER components. We further tested the mutated XPG proteins for their ability to restore UV resistance to cell lines from two different XP-G individuals. Finally, we used two approaches to examine the possibility that XPG is capable of multimerization. The results emphasize that XPG normally works in the context of a repair complex with other proteins and not as an isolated peptide. We discuss the implications of the results for the XPG incision mechanism and XP-G clinical phenotypes.

EXPERIMENTAL PROCEDURES
Human Cells, Extracts, and Repair Proteins-HeLa cells and fibroblasts from XP-G/CS patient XPCS1RO transformed with an origin of replication deficient SV40 (22) were cultured at 37°C in the presence of 5% CO 2, in regular minimum Eagle's medium (Seromed) supplemented with 5% fetal calf serum, 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. The LBL463 lymphoblastoid cell line was established by Epstein-Barr virus immortalization of a frozen polymorphonuclear lymphocyte blood sample from XP-G patient XP3BR (23). This cell line, the XPG83 lymphoblastoid cell line from XP-G patient XP125LO (24), and the Burkitt lymphoma Raji cell line were grown in suspension in RPMI 1640 medium (Seromed) with 10% fetal calf serum, L-glutamine, and antibiotics. HeLa whole cell extract was prepared as described (25). Purified recombinant XPA, RPA, and ERCC1-XPF from Escherichia coli, XPG from baculovirus, and XPC-hHR23B and TFIIH from HeLa cells were as previously noted (15).
Construction of XPG Mutants-Mutations were created in plasmid pBS-XPG, containing a full-length XPG cDNA with 67 base pairs of natural 5Ј leader (26), using the Sculptor in vitro mutagenesis system (Amersham Pharmacia Biotech). Table I lists the oligonucleotides used to create the mutants. All were confirmed with a deaza-G/A T7 sequencing kit (Amersham Pharmacia Biotech). Fragments containing a mutation were excised by cleavage at nearby restriction sites and used to replace the corresponding fragment in pBS-XPG.
For use with the vaccinia virus-based expression system, the coding and 3Ј untranslated regions of XPG were placed behind the U1A leader (27) in the pGEM-3Zf(ϩ) vector to yield clone pGEM-XPG. For E791A, E791D, A792V, C794A, and C794S, the mutated sequences were subcloned into pGEM-XPG using BspE1 and BsmI sites. A BstBI-BstBI fragment was used for the D77A and D77E substitutions. To generate the ⌬N136 truncation, XPG cDNA was cut with BstBI, filled in with Klenow enzyme and then cut with HindIII, and the resulting 3298-base pair fragment was subcloned between the NcoI (filled-in) and HindIII sites of pGEM-XPG.
Expression of Mutated XPG Proteins-84-mm-diameter Petri dishes containing monolayers of HeLa cells or transformed XPCS1RO fibroblasts were infected with 5 plaque-forming units/cell of vTF7-3, a recombinant vaccinia virus expressing T7 RNA polymerase (28). At 1 h postinfection, the medium was replaced with a transfection mix comprising 3 ml of minimum Eagle's medium, 5 g of plasmid DNA (two portions of 2.5 g for co-expression), and 15 l of TransfectACE (29). After 3 h at 33°C, an additional 7 ml of minimum Eagle's medium supplemented with 5% fetal calf serum was added, and incubation was continued at 33°C. Cells were harvested 24 h later in 1 ml of phosphate-buffered saline, centrifuged at 300 ϫ g, and resuspended in one pellet volume of 50 mM Tris-HCl (pH 8.0), 300 mM NaCl, 0.1% Nonidet P-40 (Buffer A) complemented with 10% glycerol, 1 mM EDTA, 1 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride, 2 g/ml leupeptin, and 1 g/ml pepstatin. The suspension was incubated on ice for 30 min, and then nuclei were removed from the cytosolic extracts by centrifugation at 14,000 ϫ g for 20 min. For metabolic labeling, HeLa cells were infected-transfected as described above. 20 h later, cells were washed and incubated for 1 h with minimum Eagle's medium lacking methionine and cysteine before incubation for 4 h with 100 Ci/ml EasyTag Express Protein labeling Mix (NEN Life Science Products) containing L-[ 35 S]methionine and L-[ 35 S]cysteine. Cells were lysed in 25 mM HEPES-KOH (pH 6.8), 10% glycerol, 2 mM MgCl 2 , and 1 mM dithiothreitol (Buffer B) containing 0.15 M KCl, 0.4% Nonidet P-40, 0.1 mM phenylmethylsulfonyl fluoride, 2 g/ml leupeptin, and 1 g/ml pepstatin. Extracts were cleared as above at 14,000 ϫ g for 20 min.
Immunoblotting, Immunopurification, and Immunoprecipitation-Immunoblotting was performed on cell extracts (10 -50 g of protein), or 0 -40 ng of purified XPG plus 20 g of bovine serum albumin (BSA) with the mouse anti-XPG monoclonal antibody 8H7 (14) essentially as described previously (10). XPG proteins were immunopurified by mixing 4 l of 8H7 antibody with HeLa extracts completed to 100 l with Buffer A. After incubation for 1 h at 4°C, the XPG-8H7 antibody complex was coupled with 10 l of protein A-Sepharose beads (CL-4B, Amersham Pharmacia Biotech) for 1.5 h at 4°C. Beads were subsequently washed twice in 500 l of Buffer A and then twice in 500 l of Buffer B and finally resuspended in 20 l of Buffer B for use in the endonuclease assay.
For wild-type XPG/⌬N136 XPG co-immunoprecipitation, cells were labeled for 4 h with L-[ 35 S]methionine and L-[ 35 S]cysteine and then lysed as described above. Approximately 200 g of HeLa extracts were immunoprecipitated with 3 l of crude rabbit polyclonal anti-N terminus XPG antibody (30) in Buffer B containing 0.15 M KCl and 0.4% Nonidet P-40. After extensive washing with the same buffer, the beads were resuspended in SDS gel loading buffer, boiled for 5 min, and electrophoresed on 5% polyacrylamide gels.
Endonuclease Assay-The substrate was a "bubble" formed by annealing two 90-mer oligonucleotides with a central 30-nucleotide unpaired region (6). One strand was previously 5Ј-labeled with polynucleotide kinase and [␥-32 P]ATP. The bubble substrate (0.2-0.5 ng) was incubated with the immunopurified and immobilized XPG proteins for 3 h at 37°C. Reactions were stopped by addition of 30 mM EDTA, 0.6% SDS, and 0.4 mg/ml proteinase K. After 30 min at 37°C followed by phenol extraction and ethanol precipitation, samples were resuspended in formamide dye mix, heated at 95°C, and loaded onto denaturing 12% polyacrylamide gels. Maxam-Gilbert reactions (AϩG and CϩT) were performed on the 5Ј-labeled oligonucleotide to localize the site of XPG cleavage.
In Vitro DNA Repair Assay-Covalently closed circular DNA containing a site-specific 1,3-intrastrand d(GpTpG)-cisplatin cross-link has been described in detail (31,32). Repair was reconstituted in 7.5-l reactions with purified proteins (per reaction, 0.36 pmol of XPA, 0.04 pmol of XPC-hHR23B, 0.46 pmol of RPA, 1.5 l of TFIIH heparin-5PW fraction, 0.06 pmol of ERCC1-XPF, and 0.35 pmol of XPG or vaccinia virus-overexpressed wild-type or mutant XPG as indicated) in repair buffer containing 45 mM HEPES-KOH (pH 7.8), 70 mM KCl, 7.5 mM MgCl 2 , 0.9 mM dithiothreitol, 0.4 mM EDTA, 2 mM ATP, 22 mM phosphocreatine (di-Tris salt), 0.375 g of creatine kinase, 3.4% glycerol, 0.02% Nonidet P-40, and 2.5 g of BSA. Repair proteins were preincubated for 10 min at 30°C, 50 ng of DNA containing the single cisplatin adduct were added, and incubation was continued for a further 90 min. As a positive control, a reaction with HeLa whole cell extract incubated for 30 min was usually included. Oligonucleotides excised during repair were detected by end labeling with Sequenase as described (32) and quantified using ImageQuant software (Molecular Dynamics).
Open Intermediate Detection-Intermediates with single-stranded character were detected by their sensitivity to KMnO 4 as described previously (14). Covalently closed circular double-stranded M13 DNA was cut at an AvaII site 140 base pairs 3Ј to the platinum lesion, radiolabeled at the 3Ј end using Klenow fragment and [ 32 P]dTTP, and purified over a Sephadex G50 column. The opening reactions (30 l) contained 200 ng of DNA or 60 g of Hela whole cell extract, 60 g of mM phosphocreatine (di-Tris salt), 1.5 g of creatine kinase, 3.4% glycerol, and 10 g of BSA. Repair proteins were preincubated for 10 min at 30°C. DNA was added, and after a further 15 min of incubation at 30°C, KMnO 4 (120 mM stock) was added to a final concentration of 6 mM. After 1 min, the oxidation was stopped by adding ␤-mercaptoethanol to 1 M. Purification of the DNA and analysis on a 6% sequencing gel was performed as described previously (14). UV Survival Assays-Wild-type and mutant XPG cDNAs were cloned into the NotI site of EBO-pLPP (33) or of EBS-PL, a similar episomal vector containing the strong SR␣ promoter (34). 10 7 XP-G or Raji cells were transfected by electroporation with a Gene Pulser (Bio-Rad) at 250 V, 960 microfarads in RPMI 1640 medium containing 20 g/ml recombinant DNA and 400 g/ml E. coli tRNA. Transfectants were selected with increasing doses of hygromycin B (Calbiochem) up to 200 g/ml. Selected cells were UV-irradiated at 254 nm in Hanks' buffer (without phenol red) at a dose rate of 0.12 J/m 2 /s for various times.
For the fluorescence assay, duplicate 200-l samples were incubated at 37°C in 24-well plates in 1.4 ml of RPMI medium containing 10% fetal calf serum. 48 h later, duplicate 200-l aliquots were incubated with 10 l of Alamar blue (Accumed), a fluorometric growth indicator, for 4 h at 37°C in 96-well plates. Fluorescence was measured in a Cytofluor (Millipore) with excitation and emission wavelengths of 530 and 590 nm, respectively (35). For the thymidine uptake assay, quadruplet 200-l samples containing 10 5 cells were incubated for 48 h at 37°C in 96-well plates. 1 Ci of [ 3 H]thymidine was then added to 25 l medium in each well, and incubation was continued for a further 2 h. The plates were frozen and thawed, and the lysed cells were recovered on glass microfiber filters. After thorough washing, the filters were assayed for 3 H content by liquid scintillation counting (36).
Hydrodynamic Measurements-Velocity sedimentation was used to determine an S value for the XPG protein. 3 g of purified XPG (15) or 425 g of crude protein extract from infected-transfected HeLa cells expressing wild-type XPG were layered onto 4.4 ml of 10 -30% glycerol gradients in 25 mM HEPES-KOH (pH 7.4), 0.1 mM EDTA, and 1 mM dithiothreitol (Buffer C) containing either 0.4 or 1 M KCl. After centrifugation in a SW60 rotor at 47,000 rpm at 4°C for 14 h, fractions (0.175 ml) were collected and assayed for the presence of XPG by immunoblotting using the 8H7 monoclonal antibody. For determination of the Stokes radius, 15 g of pure XPG were applied to a fast protein liquid chromatography Superose 6 column (Amersham Pharmacia Biotech). The column was washed at 4°C with Buffer C containing 1 M KCl and 20% glycerol. The resulting 0.3-ml fractions were desalted and concentrated using Microcon filtration units (Amicon), and XPG-containing fractions were identified by gel electrophoresis and Coomassie Blue staining. Fractions were assayed for XPG structure-specific nuclease cleavage of a splayed arm oligonucleotide substrate as described (15).
Modeling the XPG Active Site-Three nucleotide cleaving enzymes with significant sequence similarity to XPG, the structures of which have been solved by x-ray crystallography, were superimposed: Thermus aquaticus DNA polymerase (37), T5 D15 exonuclease (3), and T4 RNase H (1). The superposition was entirely automatic and not biased to any particular region; it was performed using MULTISUP, a program based on a pairwise superposition algorithm (38). A total of 103 equivalenced C␣ positions were superimposed to under 1.0 Å root mean squared deviation. Visual inspection of the superposition (see Fig. 8A) indicated that all 103 equivalenced residues lie in one domain of each enzyme. This domain is the catalytic domain and consists of the conserved N and I regions ( Fig. 1), which come together to form the active site. The XPG sequence was manually aligned in this region by maintaining the conserved hydrophobic and expected charged residues of the catalytic center. A secondary structure prediction (39) of the XPG sequence agreed with the superposition and was also used as a guide to the alignment. The XPG sequence has ϳ16% sequence identity within this region; this drops to ϳ11% for the complete alignment. Hence, it was decided to model only the active site region, as this covered all of the mutations made and also the most conserved region. Only small differences were found between models based on each of the three crystal structures or the average of their superimposed coordinates. Because the positions of two Mg 2ϩ metals have been assigned in the active site region of T4 RNase H (1), it was decided to model directly from the RNase H structure. The ligands to both Mg 2ϩ sites are conserved in all sequences. Side chains were replaced between RNase H and XPG using 3D-JIGSAW (40), a program that optimizes the inter-action between all side chains by considering all possible side chain conformer interactions.

Description and Expression of Mutated XPG Proteins-
XP125LO and XP124LO are two sibling XP-G individuals with a very mild XP clinical presentation (41). Their paternal XPG allele contains a premature stop codon at position 960, whereas their maternal allele codes for valine instead of alanine at position 792 (8,10). Immunoblots of cell extracts from these patients have revealed the presence of full-length XPG carrying the A792V substitution but not the truncated XPG protein, suggesting that the latter or its message is unstable (10). Ala-792 is located in a conserved amino acid sequence core within the I region and is surrounded by glutamate and cysteine residues that have been found in the active site of various DNA nucleases (Ref. 26 and Fig. 1). To test the potential functional significance of these residues, site-directed mutagenesis was used to convert Glu-791 to either Ala or Asp, and Cys-794 to either Ala or Ser. Similarly, the highly conserved Asp-77 in the N region was changed to either Ala or Glu. The mutations were confirmed by sequencing, and appropriate sequenced restriction fragments were subcloned into an otherwise wild-type XPG cDNA background to ensure the absence of any additional undesired mutations. The cDNAs were preceded by the 5Ј leader for the human U1 snRNA-associated A protein (27) to improve translational efficiency and were cloned in a plasmid vector behind a promoter for T7 RNA polymerase.
The corresponding XPG proteins were overexpressed in HeLa cells using a vaccinia virus-based expression system (28). Cells were infected with a recombinant vaccinia virus expressing T7 RNA polymerase and then were transfected with the various mutated cDNAs. After overnight incubation, cytosolic extracts were prepared and were examined by immunoblotting to ensure that the wild-type and mutated XPG proteins were equally expressed. All mutated proteins were recognized by the anti-XPG mouse monoclonal antibody 8H7 (14) and migrated on gels at the same position as wild-type XPG ( Fig. 2A). All recombinant proteins were strongly overexpressed compared with the endogenous HeLa XPG ( Fig. 2A, compare the right lanes with lane CSB).
A Pull-down Endonuclease Assay-Previous studies showed that purified XPG is capable of cleaving a DNA substrate containing a centrally unpaired bubble of 30 nucleotides flanked by two duplex regions of 30 base pairs, specifically at the 3Ј end of the unpaired region (6,14). Relatively pure XPG preparations are needed for this type of assay because other nuclease activities in cruder fractions degrade the bubble substrate. To circumvent this problem, we made use of the 8H7 monoclonal antibody to first immunopurify the overexpressed XPG proteins on protein A-Sepharose beads; the immobilized proteins were then mixed with the bubble substrate that had been 5Ј-labeled on one strand. This antibody does not inhibit NER reactions in vitro (14) or the structure-specific cleavage of the bubble substrate by purified XPG in solution (data not shown). Similarly, binding of wild-type XPG to the antibodyprotein A-Sepharose complex did not affect its endonuclease activity or change its substrate specificity: XPG still cut close to the 3Ј junction between single-and double-stranded DNA, generating a labeled fragment of ϳ61 nucleotides (Fig. 2B, lanes  4 -6). The nuclease activity of the overexpressed wild-type XPG protein was very much higher than any endogenous HeLa XPG background (Fig. 2B, compare lanes 4 -6 with lanes 1-3).
The Alanine and Acidic Amino Acid Substitutions Abolish XPG Intrinsic Endonuclease Activity-Because all mutated XPG proteins were immunoprecipitated by the monoclonal antibody with equal efficiency, we were able to test their intrinsic endonuclease activity using this pull-down assay. Conversion of the highly conserved Asp-77 and Glu-791 residues to Ala abolished endonuclease activity (Fig. 2B, lanes 7-9 and 13-15). Changing each of these residues to the other acidic amino acid (namely D77E and E791D) also abolished activity in this assay (Fig. 2B, lanes 10 -12 and 16 -18). In contrast, amino acid changes at Cys-794 only partially reduced endonuclease activity, with C794S being slightly less active than C794A (Fig. 2B, lanes 22-24 and 25-27). The natural A792V substitution present in XP125LO and XP124LO cells also appeared to abolish intrinsic endonuclease activity (Fig. 2B, lanes 19 -21). Because wild-type XPG can utilize either Mg 2ϩ or Mn 2ϩ as cofactor, but not Co 2ϩ (30), we assayed the mutated proteins with Mn 2ϩ as divalent cation. This stimulated endonuclease activity of the overexpressed wild-type XPG protein (Fig. 2C, lanes 1-4). Proteins carrying the Asp-77 or Glu-791 substitutions again showed no detectable activity in this assay (data not shown), but interestingly, weak but measurable activity was found for the A792V protein in the presence of Mn 2ϩ (Fig. 2C, lanes 5  and 6).
The Acidic Amino Acid Substitutions, but Not the Alanine Substitutions, Permit Dual Incision Activity in Reconstituted NER in Vitro-To test the activity of the various mutant XPG proteins in NER, we used a dual incision assay with DNA containing a single cisplatin adduct, purified repair proteins (XPA, XPC-hHR23B, RPA, ERCC1-XPF, and TFIIH), and recombinant XPG produced in the vaccinia virus system. Pilot studies showed that although production of XPG mutant pro-teins in HeLa cells gave the highest yields, production in XPGdefective XPCS1RO fibroblasts was preferable as it eliminated any residual interference from endogenous XPG activity.
The amount of vaccinia virus-expressed wild-type XPG in cytosolic extracts was determined by comparison to known amounts of purified XPG by immunoblotting. To determine the appropriate amounts of protein to use in repair reactions, the wild-type vaccinia-produced XPG was titrated in the reconstituted repair system (data not shown). Based on this titration, we compared the activity of 1.5 ng of vaccinia-produced wildtype and mutant XPG proteins.
Oligonucleotides excised during repair were detected by end labeling as described (32). Vaccinia-expressed wild-type XPG (Fig. 3, lane 2) showed the same pattern of excision products as found with baculovirus-expressed XPG (lane 11) or HeLa whole cell extract (lane 12). No incision products were observed when XPG was omitted (lane 1) or when an extract containing overexpressed CSB protein was used (lane 3). The mutant XPG proteins differed in their repair activity. The nonconservative D77A and E791A changes completely inactivated repair function (Fig. 3, lanes 4 and 6), but the acidic substitution mutants D77E and E791D retained considerable activity in the reconstituted system. In this particular experiment, they exhibited 37 and 97% of normal (lanes 5 and 7), but in other experiments, both mutants were close to 100% of normal. The A792V mutant XPG gave a very weak repair signal (ϳ8% of normal), which could be increased by using more mutant protein (data not shown). The C794A and C794S XPG proteins were intermedi- ate, displaying 63-73% of normal repair activity in this experiment (Fig. 3, lanes 9 and 10).
Activity of Mutant XPG Proteins in Open Intermediate Formation-The mutant XPG proteins were tested for their ability to sustain formation of a key preincision open intermediate in NER. Open intermediates with single-stranded character were detected by their sensitivity to a 1-min pulse of KMnO 4 following a 15-min incubation of damaged DNA with repair proteins, as described (14).
Because of the distortion caused by the cisplatin adduct, there is an intrinsic permanganate sensitivity of the T residues at positions Ϫ5, Ϫ4, Ϫ2, and 0, and at G(ϩ1) as revealed by the BSA-only sample (Fig. 4A, lane 14). NER-related open inter-mediates produced during the 1-min period of KMnO 4 modification are detectable by intensification of sensitivity at these positions, and the appearance of new diagnostic bands at positions Ϫ7, Ϫ9, and Ϫ10, and ϩ5. Opening to positions Ϫ7, Ϫ9, and Ϫ10 can take place in the absence of XPG protein (Fig. 4,  lanes 3 and 6). This XPG-independent opening is probably equivalent to the preincision complex that has been designated PIC2 (42).
When XPG protein with nuclease activity was included in reaction mixtures, 3Ј incisions were formed and accumulated throughout the 15-min incubation period (Fig. 4A, lane 4). To ensure the assignment of these bands, reactions were done without permanganate treatment so that incision products could be visualized separately. The 3Ј incision bands were absent in a reaction mixture containing HeLa cell extract instead of purified incision proteins (Fig. 4B, lane 1), because the extract can carry out DNA synthesis and ligation after incision.

FIG. 2. Intrinsic endonuclease activity of mutant XPG proteins in the pull-down assay.
A, immunoblot analysis of overexpressed wild-type and mutated XPG proteins. Extracts were prepared from vTF7-3-infected HeLa cells transfected with cDNAs for CSB as a negative control, wild-type XPG, or various mutated XPG proteins as indicated. Lysates were electrophoresed through a 5% polyacrylamide gel, blotted, and probed with the mouse anti-XPG monoclonal antibody 8H7. Note that the amount of endogenous HeLa XPG (lane CSB) is negligible compared with the overexpressed proteins and that XPG, which has a predicted molecular mass of 133 kDa, migrates very close to the 205-kDa myosin marker (lane M). B, endonuclease activity of mutated XPG proteins. The bubble substrate was 5Ј-labeled on one strand, gel purified, and incubated for 3 h with increasing amounts of immunopurified XPG. The autograph in A was first scanned to evaluate the amount of XPG relative to the total protein concentration in each HeLa extract. Equivalent 1:2:3 ratios of XPG were immunopurified from appropriate amounts of the extracts. These varied from 25-75 g to 64 -192 g of total protein, so a range of 25, 70, and 195 g was chosen for the control CSB lysate. The products of the endonuclease reactions were analyzed on a denaturing 12% polyacrylamide gel. Maxam-Gilbert sequencing ladders of the labeled strand are displayed as markers. C, endonuclease activity of wild-type and A792V XPG proteins in the presence of Mn 2ϩ . Equivalent amounts of XPG proteins were immunopurified from 80 -100-g samples of total protein. The reactions and gel conditions were as in B except that 2 mM MnCl 2 was used instead of 2 mM MgCl 2 in reactions displayed in lanes 2, 4, and 6.
FIG. 3. Incision activity of mutant XPG proteins in the reconstituted repair system. Reactions contained a DNA substrate containing a single cisplatin lesion, XPA, RPA, XPC-hHR23B, ERCC1-XPF, and TFIIH, and the XPG protein to test. DNA was recovered from the repair reactions and was annealed with an oligonucleotide complementary to the excised oligonucleotide and with a 5Ј-GGGG overhang (32). Excised oligonucleotides carrying the platinum adduct were labeled by a Sequenase reaction using [ 32 P]dCTP and then were separated on a denaturing 14% polyacrylamide gel and visualized by autoradiography. The results show that vaccinia-expressed wild-type XPG (Fig.  4B, lanes 5 and 6) and the mutants D77E (lane 9), E791D (lane 11), A792V (lanes 12 and 13), C794A (lane 14), and C794S (lane 15) are all capable of carrying out 3Ј incision during this 15-min period, in the context of the rest of the repair complex. Two mutants, D77A and E791A, gave no detectable 3Ј incision (Fig.  4B, lanes 8 and 10). However, bands indicating opening were even more intense, indicating accumulation of open intermediates (Fig. 4A, lanes 7 and 9). The band at T(ϩ5) was particularly intense, and further bands at T(ϩ7) and T(ϩ8) were  (except for lanes 2, 3, and 16). This specifically reveals sites of enzymatic incision. All samples were treated with piperidine before gel analysis. C, scheme showing opening around the Pt-GTG lesion. Sensitive T residues are marked in boldface. Arrows indicate the 5Ј and 3Ј incision sites. The asterisk indicates the site of 3Ј 32 P labeling of the adducted strand at the AvaII site. detected. These are the most 3Ј T residues before the sites of the major 3Ј incisions (Fig. 4C). These bands were not detectable in the absence of KMnO 4 treatment (Fig. 4B), showing that they represent oxidation-sensitive T residues with singlestranded character.
Because DNA is labeled at the 3Ј end in these experiments, it is also possible to detect "uncoupled" 5Ј incisions that occur in the absence of 3Ј incision. These are normally rare events (31). With most XPG mutant proteins, uncoupled 5Ј incisions occurred at similarly low levels. However, 5Ј incisions readily accumulated in reaction mixtures with the D77A and E791A XPG proteins (Fig. 4B, lanes 8 and 10). Indeed, the intensity of these uncoupled 5Ј incisions was similar to the intensity of the 3Ј incisions seen with wild-type XPG protein. This indicates that the NER reaction proceeds normally with the alanine substituted XPG proteins, except that the 3Ј incision does not occur. The D77A and E791A proteins therefore are well folded and assemble normally into the repair complex. Uncoupled 5Ј incisions were barely if at all detectable with the A792V protein (Fig. 4B, lane 12), even with a 4-fold excess (lane 13), suggesting that this mutant XPG protein does not remain stably bound in the repair complex.

The Acidic Amino Acid Substitutions, but Not the Alanine Substitutions, Permit Complementation of XP-G Cell Lines in
Vivo-To test the in vivo relevance of these results, we cloned the various mutated XPG cDNAs into plasmid EBO-pLPP, an Epstein-Barr virus-based episomal vector (33), and transfected them into two lymphoblastoid cell lines derived from XP-G patients XP3BR and XP125LO (23,24). Following hygromycin selection, stable transfectants were irradiated with 254-nm light, and their survival was assayed 48 h later with Alamar blue, a redox indicator that both fluoresces and changes color in response to chemical reduction of the medium resulting from cell growth (35). As revealed by immunoblotting, all mutated forms of XPG were expressed in the transfectants, and in levels comparable to that of XPG in a wild-type lymphoblastoid line (data not shown).
In both XP-G cell lines, transfectants containing the EBO-pLPP vector alone were highly sensitive to UV, whereas the presence of wild-type XPG restored UV resistance to normal levels (Ref. 26 and Fig. 5A). UV resistance was not regained by transfectants expressing either the D77A or E791A forms of XPG but was partially restored by the C794A and C794S substitutions (Fig. 5, A and B). These results were corroborated by an alternative assay for cell viability, the uptake of [ 3 H]thymidine during a short pulse 2 days after UV irradiation (36). This assay measures the proportion of cells that enter S-phase and is a more direct and sensitive assay of the effect of DNA damage on viable cell number than the redox indicator. XP125LO cells expressing the EBO-pLPP vector alone or XPG with the D77A or E791A substitutions were comparably UV-sensitive by this assay. UV resistance was again partially restored in the C794A and C794S transfectants, with the C794A protein being slightly more active than XPG carrying the C794S substitution (Fig. 5, C and D). In contrast to their inactivity on the bubble substrate (Fig. 2), the D77E and E791D XPG proteins were able to fully restore UV resistance to both XP-G cell lines in both assays (Fig. 5, A-D). Identical results were obtained (data not shown) when the mutant XPG cDNAs were transfected into lymphoblastoid line AG08802 from XP-G/CS patient XP20BE (43). These in vivo results thus corroborate the conclusion that the acidic substitutions permit XPG endonuclease activity when in the presence of the other NER proteins.
The Alanine Substitutions Render Transfected Raji Cells UVsensitive-The accumulation of uncoupled 5Ј incisions found with the D77A and E791A proteins (Fig. 4) raised the possibil-ity that these alanine substitutions may be able to sequester the NER machinery into inactive complexes to generate a dominant negative phenotype in wild-type cells. To test this possibility, cDNAs for the various XPG proteins were cloned behind the strong SR␣ promoter in the EBS-PL episomal vector (34) and then were transfected into Raji cells derived from a Burkitt lymphoma. These cells contain wild-type XP genes but carry mutations in both alleles of the p53 gene (44) that are expected to make them more resistant to UV cytotoxicity and to exhibit less UV-induced apoptosis than cells expressing wild-type p53 (45,46). Untransfected Raji cells were indeed very UV-resistant (Fig. 5E). This resistance was not impaired by transfection of the EBS-PL vector alone or by overexpression of XPG proteins having partial or complete complementation activity in the XP-G cell lines (Fig. 5, E and F, see A792V, wt, D77E, and  E791D). However, overexpression of the D77A and E791A XPG proteins made the Raji cells significantly more UV-sensitive (Fig. 5F). These results are thus consistent with the in vitro studies and suggest that these alanine substituted proteins are capable of binding both to the DNA substrate and to other components of the NER preincision complex.
The A792V Substitution Permits Weak Complementation in Vivo-The A792V protein is the only stably expressed form of XPG in XP124LO and XP125LO lymphoblasts (10). This mutant protein is of particular interest because it is unable to complement these cells by one fluorescence assay (8), yet the Alamar blue assay suggests that it is able to restore UV resistance to lymphoblasts from patient XP3BR (47). To investigate this further, we examined transfectants of both cell lines with the [ 3 H]thymidine uptake assay. This more sensitive assay revealed that the XP3BR cells are significantly more UV-sensitive than the XP125LO cells (Fig. 5G). As expected, and consistent with the earlier fluorescence assay (8), expression of the A792V protein in XP125LO cells did not alter their UV sensitivity. However, expression of this mutant protein in XP3BR cells increased their UV resistance up to the XP125LO level (Fig. 5G). We conclude that the A792V XPG protein has very low but detectable nuclease activity in cells that permits removal of some UV damage.
The Monomeric versus Multimeric State of Human XPG Protein-The 5Ј cut in human NER is made by a structure-specific nuclease formed by a heterodimer of the XPF and ERCC1 proteins (16,17). Although XPG is able to make a structurespecific 3Ј cut without additional proteins (6,14), and Fig. 2), it is conceivable that it does so as a multimer. It is also noteworthy that some related nucleases, such as FEN-1/DNase IV, load onto a single-stranded DNA end (48), whereas XPG requires an open structure within a contiguous stretch of DNA (14). For these reasons, we used two experimental approaches to investigate whether XPG could multimerize.
In the first approach, we measured the Stokes radius and the sedimentation coefficient of purified XPG recombinant protein produced in the baculovirus system. Gel filtration chromatography on a fast protein liquid chromatography Superose 6 column revealed that in 1 M KCl, XPG protein and structurespecific endonuclease activity eluted as a single peak (Fig. 6A) with an estimated Stokes radius of 70.3Å. For S value determination, pure XPG protein was loaded onto a 10 -30% glycerol gradient in either 0.4 or 1 M KCl, and after centrifugation, XPG-containing fractions were revealed by immunoblotting. The protein sedimented at 5.1 S in both salt concentrations (Fig. 6, B and C). Overexpressed XPG in a crude vacciniainfected HeLa extract migrated in a gradient with 0.4 M KCl to exactly the same fractions as the pure protein (Fig. 6C). Combining the two measurements made in identical buffer conditions (1 M KCl), the native molecular mass of the protein was calculated to be 146 kDa by the equation of Siegel and Monty (49), assuming a partial specific volume of 0.725 cm 3 /g. This figure is close to the predicted monomeric molecular mass of 133 kDa. Thus, under these conditions, the bulk of XPG appears to exist most stably as a monomer.
In the second approach, we asked whether interactions between XPG polypeptides could be detected by immunoprecipitation. For this approach, we engineered a truncated XPG protein lacking the N-terminal 136 amino acids. This deletion eliminates the conserved N region, including the first 109 amino acids that were used previously as the immunogen to raise an anti-XPG rabbit polyclonal antibody (30). Recombinant proteins were again produced in the vaccinia expression system, but in this case, the infected-transfected HeLa cells were incubated overnight and then labeled with [ 35 S]methionine and [ 35 S]cysteine for 4 h before cell lysis. As revealed by autoradiography after gel electrophoresis, the wild-type and ⌬N136 XPG proteins were stable, easily distinguishable major labeled species, whether expressed singly or together (Fig. 7A).
When separately expressed, wild-type XPG protein, but not the ⌬N136 version without the epitope, was immunoprecipitated by the polyclonal antibody in a buffer containing 0.15 M KCl and 0.4% Nonidet P-40 (Fig. 7B, compare lane 1 with lanes  7 and 9). However, the truncated protein was found in the immunoprecipitate after the two proteins were co-expressed in the same cells (Fig. 7B, lanes 2-6 and 8). The same result was obtained when the buffer contained 50 g/ml ethidium bromide to abolish nonspecific interactions mediated via DNA (50), when MgCl 2 was omitted from the buffer, or when the salt concentration was raised to 1 M KCl (data not shown). However, ⌬N136 XPG was not found in the immunoprecipitate when the two proteins were expressed separately and then mixed (Fig.  7B, lane 10). To confirm the identities of the protein bands, unlabeled extracts were immunoprecipitated with the anti-N region polyclonal antibody and then immunoblotted with the anti-XPG 8H7 monoclonal antibody. The ⌬N136 XPG protein was again found in the immunoprecipitate with co-expressed wild-type XPG (Fig. 7C, lanes 5 and 6). We conclude that when co-expressed, the full-length and truncated XPG proteins are able to interact with one another and that this interaction does not require the conserved N region.
To estimate the stoichiometry of the interaction, infected HeLa cells were co-transfected with various ratios of the two cDNAs to try to obtain equimolar expression of the full-length and truncated proteins. When this was achieved (Fig. 7A, lane  4), the subsequent immunoprecipitate was found to contain the wild-type and ⌬N136 XPG proteins in a ϳ3:1 molar ratio (Fig.  7B, lane 4). Because this method detects wild-type/wild-type and wild-type/⌬N136 interactions, but not ⌬N136/⌬N136 interactions, this result implies an interaction between an even number of molecules, most simply that XPG can dimerize.

Conserved Catalytic Residues in the XPG Family-
The results presented here provide insight into human XPG function. Structurally, the active site geometry of this protein is expected to be very similar to that of other nucleotide cleaving enzymes in the XPG family. Fig. 8A shows a superposition of three relevant proteins for which structures have been solved by x-ray crystallography: T. aquaticus DNA polymerase (37), T5 D15 exonuclease (3), and T4 RNase H (1). The central ␤-sheets superimpose extremely well and more closely than the surrounding ␣-helices. The metal coordinating side chains in the active sites are all clustered at the top of the ␤-sheets and are highly conserved both in sequence (Fig. 1) and in position. Fig. 8B shows a model of XPG around the conserved active site region, which consists of both the N and I regions. These come together with remarkable exactness despite the very long sequence (Ͼ600 amino acids) between these two regions. All residues mutated in this study lie within this highly conserved active site region. It is possible that XPG co-ordinates two metal ions as found for T4 RNase H (Fig. 8B). Although not necessarily expected to have high precision at the atomic level, the model provides a useful visual aid for interpreting the consequences of each mutation examined in this study, and it helps to predict additional mutations that potentially could be useful in the future to probe the XPG incision mechanism in finer detail.
Consequences of Mutating the Conserved Acidic Residues-In most nucleolytic enzymes for which detailed information is immunoblot of fractions from the 0.4 M KCl glycerol gradient (B) probed with the 8H7 monoclonal antibody. Highly purified XPG and the protein present in a crude cell extract sedimented to the same position (peak fraction 8, 1.4 ml). available, a carboxylate is required either to directly capture a proton from a water molecule (in this case the divalent ion stabilizes the reaction intermediate) or to bind a divalent ion that promotes the formation of a hydroxyl ion. In either case, the resulting OH Ϫ attacks the scissile phosphodiester bond (51)(52)(53)(54)(55). For the XPG nuclease family, there is an extensive network of charged side chain to metal interactions that link together the two metal cations (Fig. 8). The two acidic residues mutated in this study, Asp-77 and Glu-791, are both involved in this network. In this regard, the report of nuclease activity in a D77A XPG mutant (56) is very surprising because substitution of this neutral side chain would be expected to disrupt this electrostatic network. We found instead that changing either Asp-77 or Glu-791 to Ala completely inactivated the nuclease function of XPG (Figs. 2-4) and abolished the ability of XPG to restore UV resistance to XP-G cell lines (Fig. 5). Another highly conserved acidic residue, Asp-812, is also expected to be part of this electrostatic network (Fig. 8B). Conversion of Asp-812 to Ala also inactivated the nuclease function of XPG (56). These results are thus consistent with a role for all three acidic residues in catalysis.
The other mutations generated in the present study, C794A and C794S, had only mild effects on XPG function (Figs. 2-5). The Cys-794 residue is conserved within most eukaryotic members of the XPG family ( Fig. 1) but is predicted to be some distance down the ␣-helix away from the catalytic center (Fig.  8B). It thus may have a subtle role in conformation rather than being directly involved in catalysis.
Changes have been made in positions corresponding to Asp-77 and Glu-791 in other XPG family members. D73N and D125N substitutions in DNA polymerase I of Mycobacterium tuberculosis and D63A and D115A in the E. coli enzyme abolish 5Ј nuclease activity (57,58). Similarly, mutations of the equivalent residues in T4 RNase H (D71N and D132N) and human FEN-1 (D86A and E160A) cause the complete loss of nuclease activity but do not affect binding to DNA (59,60). The D77A and E791A XPG proteins are less active on the bubble substrate than endogenous XPG in the CSB control (Fig. 2), they reduce the UV resistance of transfected Raji cells (Fig. 5), and although they are inactive as a 3Ј endonuclease (Figs. 2-4), they allow the 5Ј NER cut to be made very efficiently (Fig. 4). These results strongly suggest that the D77A and E791A substitutions, like the homologous mutations in FEN-1 and T4 RNase H, still permit substrate binding, and they provide further evidence that XPG needs to be in the preincision complex to permit ERCC1-XPF to make the 5Ј cut (56).
XPG proteins carrying the D77E and E791D substitutions were unable to cut the bubble substrate but were highly active in the dual incision assay and in vivo (Figs. 2,4,and 5). This is in accord with a strict requirement for carboxylates at positions 77 and 791. These results further suggest that other NER components are needed to reveal the endonuclease activity of these mutated proteins. Minor modifications in the active surface of the protein are likely to affect the strict relative orientations of the single-stranded DNA, the metal ion, and the attacking water molecule needed for efficient catalysis. Apparently, the repair partners of the D77E and E791D XPG proteins can modulate the impact of these mutations on nucleolysis. One way this might occur is to subtly adjust the conformation of the mutated XPG proteins to better fit the substrate. Another possibility is that even if the kinetic parameters (K m and V max ) for the acidic substitution mutants are low for the isolated enzymes, their effects are negligible compared with the thermodynamic gain of having specific contacts with other repair proteins and with an optimally positioned DNA substrate within the open complex.
Implications for the XPG Incision Mechanism-In the crystal structure of T4 RNase H, the short peptide between the N and I regions, located above the cleft with the two Mg 2ϩ ions, shows nine disordered residues that have been proposed to be involved in substrate binding (1). The corresponding sequence in the Taq DNA polymerase also shows disordered residues (37).  7 and 9), or were co-transfected with the two cDNAs at different ratios (lanes 2-6 and 8). After overnight incubation followed by 4-h label with [ 35 S]methionine and [ 35 S]cysteine, crude cell lysates were electrophoresed on 5% acrylamide gels, and XPG proteins were revealed by autoradiography. Lysates containing the two separately expressed proteins were also mixed before electrophoresis (lane 10). B, the lysates were challenged with a polyclonal antibody raised against the conserved N region of XPG and immunoprecipitated proteins were analyzed by gel electrophoresis and autoradiography as in A. C, HeLa cells infected with the vTF7-3 vaccinia virus recombinant were transfected with cDNAs for wild-type XPG (lanes 1 and 4), ⌬N136XPG (lanes 2 and 5), and the two cDNAs together (lanes 3 and 6). After overnight incubation, crude cell lysates were prepared and challenged with the anti-N region of XPG polyclonal antibody. Proteins in the crude lysates (lanes 1-3) and the immunoprecipitates (IP) (lanes 4 -6) were electrophoresed on a 5% acrylamide gel and then immunoblotted with the anti-XPG 8H7 monoclonal antibody.
In the T5 D15 exonuclease, the corresponding region is part of a helical arch that could perfectly accommodate DNA (3). It has been proposed that this arch may allow T5 D15 nuclease to thread-through single-stranded DNA until it reaches the branch point, where it cleaves. Evidence for a thread-through mechanism has been found for other members of the nuclease family. The flap endonuclease activity of FEN-1 is inhibited by protein binding or by annealing of a complementary strand to the single-stranded arm (61). Proliferating cell nuclear antigen physically associates with FEN-1 and stimulates its nuclease activity at branch substrates (48). It has been proposed that FEN-1 diffuses down the single-stranded flap from the 5Ј terminus to the single-stranded double-stranded junction, where it is stabilized by proliferating cell nuclear antigen (48). T4 gene 32 single-stranded DNA-binding protein inhibits the flap endonuclease of T4 RNase H (62). Consistent with a threadthrough model, the 5Ј nucleases of Taq and E. coli polymerase I (58, 63), FEN-1 (5), and T5 D15 exonuclease (3) absolutely require a free single-stranded terminus to display nucleolytic activity.
In contrast, XPG must load onto DNA sites where there is no free 5Ј end and therefore may use a different mechanism to accomplish the 3Ј cut. During NER, a heterodimer composed of XPF and ERCC1 proteins is responsible for the 5Ј incision (16,17). Given that XPG is able to make a structure-specific 3Ј cut without additional proteins (Refs. 6 and 15 and Fig. 2), the ability of XPG to multimerize (Fig. 7) may be relevant to its incision mechanism. Also relevant may be a helix-loop-helix motif found within the I region of XPG (64). Although no evidence was found for XPG multimers in the latter study, a recombinant peptide containing this helix-loop-helix motif was able to dimerize (64). In the present work, co-expressed fulllength XPG protein and XPG lacking the conserved N region were found together by immunoprecipitation. Such multimers, most likely dimers, were resistant to high concentrations of detergent and salt and to ethidium bromide, thereby indicating that the interactions are specific and are not mediated by DNA. However, these results are not in accord with hydrodynamic measurements indicating that XPG exists as a monomer in high ionic strength (Fig. 6). This may simply mean that the observed self-interaction is reversible. Indeed, condition-dependent variability of native structure has been observed for many proteins, including the E. coli RuvC endonuclease (65) and members of the Bcl-2 family (4,66). Further studies with XPG protein and its substrates will be required to test the potential relevance of multimerization to the XPG incision mechanism.
Implications for XP-G Clinical Phenotypes-A striking feature of individuals belonging to XP group G is that they rarely develop skin cancers, even though their cells are among the most UV-sensitive of the eight XP complementation groups. Several XP-G individuals suffer from a very severe early onset form of Cockayne syndrome, which is correlated with an inability to produce full-length XP-G protein (10,67) and an inability to carry out transcription-coupled repair of oxidative base damage (9). Due to their hospitalization and early demise, these XP-G/CS patients had quite limited exposure to UV. In contrast, the XP124LO and XP125LO siblings have been exposed to UV for over two decades, yet despite only occasional use of sun protection, they have no skin cancers.
The results presented here suggest an explanation for this paradox. At a cellular level, transcription-coupled repair of UV damage cannot be detected in XP125LO fibroblasts (10), and unscheduled DNA synthesis is at background levels after UV treatment of XP125LO lymphoblasts (26). The thymidine uptake assay showed, however, that the XP125LO lymphoblasts, although very UV-sensitive, are significantly more UV-resistant than lymphoblasts from XP-G patient XP3BR (Fig. 5G). Moreover, when the A792V protein, the only stably expressed form of XPG detectable in the XP124LO and XP125LO siblings (10), was expressed in XP3BR lymphoblasts, it increased their UV resistance up to the XP125LO level (Fig. 5G). These results strongly suggest that the XP124LO and XP125LO siblings exhibit a very mild XP phenotype because their A792V XPG protein retains some residual repair capacity.
The latter suggestion implies that the A792V protein has some 3Ј endonuclease activity, and this is borne out by the in vitro studies. Although A792V XPG was unable to cleave the bubble substrate in the presence of Mg 2ϩ , it was able to do so to FIG. 8. Model for the active site of XPG. A, superposition of three nucleotide cleaving enzymes with significant sequence similarity to XPG: T5 D15exonuclease (magenta), Taq DNA polymerase (green), and T4 RNase H (cyan). The two Mg 2ϩ cations present in the T4 structure are represented by two yellow spheres. B, model for the XPG active site. The model includes all residues that could be equivalenced to residues superimposed to under 1.0 Å used in A. Only an area from domain I is shown. Residues from the N region are shown in green and those from the I region in cyan. The numbered yellow spheres represent the two Mg 2ϩ ions that are expected to be located in positions equivalent to those found in T4 RNase H. Conserved side chains that coordinate these cations are shown in red. Residues mutated in this and another study (56) are labeled. some extent with Mn 2ϩ as metal ion cofactor (Fig. 2). Similar Mn 2ϩ -driven partial activity has been reported for active site mutants of BamHI (68), EcoRV (69), T4 RNase H (59), and MunI (70) nucleases. Residue Ala-792 is at the beginning of an ␣-helix and just after the presumed metal coordinating side chain Glu-791 (Fig. 8B). Valine side chains are ␤-branched and are less favored within ␣-helices than non-␤-branched side chains (71). Hence, this particular ␣-helix may unwind slightly in the A792V protein, and this in turn may interfere with the ability of Glu-791 to accurately coordinate a metal cation. The 3Ј endonuclease activity in the presence of Mn 2ϩ may reflect partial correction of such a distortion in the metal binding site. This missense mutation appears to have additional effects, however, because the A792V protein exhibited the lowest level of uncoupled 5Ј incisions (Fig. 4). This indicates that it does not stay firmly bound in the repair complex, because an altered conformation either reduces its ability to bind to DNA or disrupts its interaction with another protein. These interesting properties of this naturally occurring mutation indicate that it would be very valuable to identify the XPG defects in other mildly affected XP-G patients.