The DNA Repair Endonuclease XPG Binds to Proliferating Cell Nuclear Antigen (PCNA) and Shares Sequence Elements with the PCNA-binding Regions of FEN-1 and Cyclin-dependent Kinase Inhibitor p21*

Proliferating cell nuclear antigen (PCNA) is a DNA polymerase accessory factor that is required for DNA replication during S phase of the cell cycle and for resynthesis during nucleotide excision repair of damaged DNA. PCNA binds to flap endonuclease 1 (FEN-1), a structure-specific endonuclease involved in DNA replication. Here we report the direct physical interaction of PCNA with xeroderma pigmentosum (XP) G, a structure-specific repair endonuclease that is homologous to FEN-1. We have identified a 28-amino acid region of human FEN-1 (residues 328–355) and a 29-amino acid region of human XPG (residues 981–1009) that contains the PCNA binding activity. These regions share key hydrophobic residues with the PCNA-binding domain of the cyclin-dependent kinase inhibitor p21 Waf1/Cip1 , and all three competed with one another for binding to PCNA. A conserved arginine in FEN-1 (Arg339) and XPG (Arg992) was found to be crucial for PCNA binding activity. R992A and R992E mutant forms of XPG failed to fully reconstitute nucleotide excision repair in an in vivo complementation assay. These results raise the possibility of a mechanistic linkage between excision and repair synthesis that is mediated by PCNA.

Exposure to UV light causes damage to DNA primarily in the form of cyclobutane pyrimidine dimers and (6-4) photoproducts. These types of DNA lesions, as well as bulky adducts produced by some chemical mutagens, are processed by nucleotide excision repair (NER). 1 The human genetic disorder xeroderma pigmentosum (XP) is the result of defects in this DNA damage repair pathway. Symptoms of XP include extreme sensitivity to sunlight exposure and a greatly elevated risk of skin cancer. In the past few years, much progress has been made in understanding the molecular events associated with NER (1). The DNA-binding protein XPA is involved in damage recognition. In concert with replication protein A, which binds singlestranded DNA, and helicases XPB and XPD, a ϳ27-29-base oligonucleotide segment containing the lesion is excised as the result of dual incision by structure-specific endonucleases XPF-ERCC1 and XPG. The XPF-ERCC1 complex cleaves the damaged strand at a 5Ј site about 23 nucleotides from the lesion, whereas XPG cleaves the strand approximately 5 nucleotides to the 3Ј side of the damage. The resultant gap is filled in by the action of DNA polymerase ␦ or ⑀, and then DNA ligase seals the nick to complete repair. The resynthesis step requires proliferating cell nuclear antigen (PCNA; Refs. 2 and 3), a ring-shaped homotrimeric protein that encircles DNA and acts as a "sliding clamp" that links the polymerase to the DNA template (4). PCNA performs the same essential function in replicative DNA synthesis during S phase of the cell cycle. PCNA requires replication factor C, a primer recognition protein that loads the PCNA trimer onto DNA in an ATP-dependent manner (5)(6)(7).
XPG is homologous to another structure-specific endonuclease, FEN-1. FEN-1 is involved in Okazaki fragment processing during DNA replication (8), and it is required for avoidance of duplication-type insertion mutations in yeast (9). FEN-1 binds to PCNA (10 -12), and this complex can be disrupted by p21 Waf1/Cip1 (12), a bifunctional protein that has a C-terminal PCNA-binding domain and an N-terminal domain that inhibits cyclin-dependent protein kinases (13,14). Here we report domain mapping experiments to pinpoint the PCNA-binding region of FEN-1 and show that the small region responsible for activity is conserved in XPG. This domain in XPG as well as the full-length XPG protein are shown to bind to PCNA. We also provide evidence from in vivo studies indicating that the PCNA-XPG interaction has a role in repair of UV damage. Finally, we identify a convergent evolutionary relationship between the PCNA-binding domains of the DNA damage-inducible inhibitor p21 and the repair endonuclease XPG and show that these domains compete for binding to PCNA.
‡ Supported by Los Alamos National Laboratory Director's Postdoctoral Fellowship.
Plasmid pET-FCH produces full-length FEN-1 (amino acids 1-380) with six histidine residues appended to the C terminus (15). The polyhistidine tag binds tightly to metal chelation affinity resin. The FEN-1 codon arginine 339 in this plasmid was replaced with either an alanine or glutamate codon by QuickChange Mutagenesis to create R339A and R339E single point mutant derivatives using mutagenic primer pairs 5Ј-CCAAGGCAGCACCCAGGGCGCGCTGGATGATTTCTTCAA-3Ј and 5Ј-CCTTGAAGAAATCATCCAGCGCGCCCTGGGTGCTGCCTT-3Ј, 5Ј-CCAAGGCAGCACCCAGGGCGAGCTCGATGATTTCTTCAA-3Ј and 5Ј-CCTTGAAGAAATCATCGAGCTCGCCCTGGGTGCTGCCTT-3Ј, respectively. These primer pairs created a BssHII or SacI site to facilitate screening. The second pair of mutagenic primers was also used to generate the R339E derivative of plasmid GST-FEN 328 -363 , whose product was used in PCNA bead competition experiments. Truncated FEN-1 proteins comprising amino acids 1-328 or 1-363 with C-terminal polyhistidine tags were created by QuickChange Mutagenesis of pET-FCH to replace codon 329 or 364, respectively, with six histidine codons followed immediately by a stop codon. The mutagenic primer pairs used to create the truncated FEN-1 constructs were 5Ј-GCAGTGGGGTCA-AGAGGCTGCACCATCACCACCATCACTAGTGCAGCACCCAGGGC-CGCC-3Ј and 5Ј-GGCGGCCCTGGGTGCTGCACTAGTGATGGTGGT-GATGGTGCAGCCTCTTGACCCCACTGC-3Ј, 5Ј-AGCCAGAACCCAA-GGGATCCCACCATCACCACCATCACTAGTGGGGCAGCAGGGAAG-T-3Ј and 5Ј-ACTTCCCTGCTGCCCCACTAGTGATGGTGGTGATGGT-GGGATCCCTTGGGTTCTGGCT-3Ј. These primer pairs each created a SpeI site to facilitate screening.
Human PCNA was amplified by polymerase chain reaction with forward primer 5Ј-GGAATTCATGAGTCACCACCACCACCACCACAT-GTTCGAGGCGCGCCTGG-3Ј and reverse primer 5Ј-TTGCGAAGCTT-ACTCGAGAGATCCTTCTTCATCCTCG-3Ј using plasmid kindly provided by Dr. Suk-Hee Lee (St. Jude Children's Research Hospital, Memphis, TN) as template. The AflIII/HindIII-digested product was ligated to NcoI/HindIII-linearized pET28b (Novagen, Madison, WI) to produce a subcloning intermediate. The first three nucleotides of a C-terminal XhoI site (introduced to facilitate generation of tagged variants of PCNA for other experiments) were converted to a stop codon by QuickChange mutagenesis with primers 5Ј-ATCGAGGATGAAGAA-GGATCTTAGGAGTCAGCTTGCGGCCGCACTCGA-3Ј and 5Ј-TCGAG-TGCGGCCGCAAGCTGACTCCTAAGATCCTTCTTCATCCTCGAT-3Ј. This primer pair destroyed a HindIII site to facilitate screening. The resulting plasmid contained a stop codon in the naturally occurring position and directed high level inducible expression of wild-type human PCNA without tags or extensions. All subcloning and site-directed mutagenesis procedures were confirmed by DNA sequencing.
Fusion Protein Binding Assay-BL21(DE3) Escherichia coli was used as the host strain to express PCNA, polyhistidine-tagged FEN-1 and derivatives, and all GST fusion proteins. Expression was induced with 0.8 mM isopropyl-␤-D-thiogalactopyranoside. Cells were lysed in wash buffer (50 mM Tris-HCl, 150 mM NaCl, pH 7.4, for FEN-1 and its derivatives and 50 mM Tris-HCl, 100 mM KH 2 PO 4 /K 2 HPO 4 , 150 mM NaCl, pH 7.4, for GST-XPG and its derivatives) supplemented with 2 mM EDTA and 0.2 mg/ml lysozyme at a ratio of 1 ml/25-ml culture. The cell lysates were clarified by centrifugation at 16,000 ϫ g. For GST fusion protein binding assays, 80 l of 40% glutathione-agarose beads (Sigma) was mixed with 450 l of GST fusion protein cell lysate and 450 l of PCNA cell lysate. The same procedure was used for polyhistidinetagged protein binding assays, except that 80 l of 40% NiSO 4 -charged iminodiacetic acid metal chelation resin (HisBind; Novagen) replaced glutathione-agarose beads, EDTA was omitted from lysis and wash buffers, and 60 mM imidazole was added to the wash buffer. Mixtures were incubated for 2 h at 4°C and then washed six times with 0.8 ml of wash buffer. Protein complexes were eluted by heating to 100°C with 80 l of 2 ϫ Laemmli sample buffer (18) and analyzed on 12% gels by SDS-polyacrylamide gel electrophoresis (PAGE) and Coomassie Blue staining using Mark12 (Novex, San Diego, CA) molecular weight standards. In the first competition experiment, 0.24 mg of synthetic peptide LKQLDAQQTQLRIDSFFRLAQQEKEDAKR (Research Genetics, Huntsville, AL), corresponding to residues 981-1009 of XPG, was added to one of the lysate mixtures. In this set of assays, NaCl concentration was 300 mM, 150 l of each lysate was used, and 150 l of wash buffer was added to increase the volume to improve mixing. In the salt dependence experiment, low salt wash buffer (20 mM Tris-HCl, 60 mM NaCl, pH 7.4) plus 0.2 mg/ml lysozyme was used for lysis, and low salt wash buffer supplemented as indicated was used for subsequent binding and washing steps. In the second competition experiment, XPG 981-1009 peptide or biotinylated synthetic peptide GRKRRQTSMTDFYHS-KRRLIFS corresponding to residues 139 -160 of human p21 (kindly provided by Dr. Jerard Hurwitz, Memorial Sloan-Kettering Cancer Center) was added to lysate mixtures at the concentrations indicated.
PCNA Bead Binding Assay-Purified human PCNA or protease-free bovine serum albumin (BSA; Boehringer Mannheim) at 4 mg/ml in 25 mM NaHCO 3 , 200 mM NaCl, pH 8.3, was added to cyanogen bromideactivated agarose beads (Sigma) and incubated for 16 h at 4°C to covalently attach the proteins to the beads. Coupling efficiency was 80% or higher for each protein. Unbound protein was removed, and then remaining reactive sites were blocked with 0.1 M glycine, 65 mM Tris-HCl, 150 mM NaCl, pH 8.0. Washed beads were stored on ice. Human XPG was expressed from recombinant baculovirus as described previously (17), except that Trichoplusia ni (BTI-TN-5B1-4) insect cells ("High Five" cells, Invitrogen, San Diego, CA) were used as the host. Cells were infected with XPG recombinant baculovirus and collected at 35.5 h post-infection. The cell pellet from 50 ml of culture was lysed in 4 ml of 0.5% Nonidet P-40, 100 mM potassium phosphate, 300 mM KCl, 2 mM EDTA, 1 mM benzamidine, 0.5 mM phenylmethylsulfonyl fluoride, 10 M pepstatin A, 10 g/ml chymostatin, 10 g/ml aprotinin, pH 7.4. After lysis, an equal volume of the same buffer without detergent was added, reducing the final concentration of Nonidet P-40 to 0.25%. The lysate was centrifuged at 90,000 ϫ g, and the supernatant was 0.45 m filtered. For binding assays, 80 l of 20% PCNA or BSA beads was mixed with 430 l of XPG-containing insect cell supernatant. In competition experiments, 0.4 mg of purified BSA, PCNA, GST-XPG 981-1009 , GST-XPG 981-1009 (R992E), GST-FEN 328 -363 , or GST-FEN 328 -363 (R339E) was added to the lysate mixture. The mixtures were incubated for 2.5 h at 4°C and then washed five times with 0.8 ml of 100 mM KH 2 PO 4 / K 2 HPO 4 , 300 mM KCl, pH 7.4. In some experiments, a DNase I treatment was performed between the third and fourth washes. Noncovalently bound proteins were eluted with Laemmli sample buffer, resolved on 4 -12% polyacrylamide gradient gels, and transferred to Immobilon-P (Millipore, Bedford, MA) polyvinylidene fluoride membrane for Western blot analysis. The membrane was sequentially incubated with 10% nonfat dry milk, 1:2000 rabbit polyclonal XPG1322 (IgG fraction) raised against an XPG amino acid 1147-1163 peptide (20,21), and 1:2000 peroxidase-conjugated goat anti-rabbit IgG (Life Technologies, Inc.), and then XPG bands were detected by enhanced chemiluminescence (Amersham Corp.).
DNase I Treatment-PCNA affinity beads were incubated with insect cell lysate and washed three times. 10 mM MgCl 2 and 0 or 250 units of DNase I (Boehringer Mannheim) were added to the washed beads. The beads were incubated for 20 min at 10°C and were mixed periodically to keep the beads in suspension. After two more washes, proteins were eluted and analyzed. To confirm the activity of DNase I under the conditions used, HindIII-digested Lambda DNA (Life Technologies, Inc.) and DNase I were added at various ratios to washed beads and incubated as described. These confirmatory reactions were terminated by addition of 25 mM EDTA and 0.5% SDS, and the DNA was evaluated by agarose gel electrophoresis and ethidium bromide staining.
Luciferase Reporter Plasmid Repair Assay-The repair assay using UV-damaged luciferase reporter plasmid has been described previously (22). Briefly, luciferase expression plasmid pGL-2 (Promega, Madison, WI) was irradiated with 0, 400, or 800 J/m 2 of 254-nm UV light from a calibrated source. XPG function was provided by pBactin-XPG, a mammalian expression plasmid containing human XPG cDNA under the control of a ␤-actin promoter. A truncated derivative of this plasmid was created by BamHI digestion to remove amino acids 948 -1186 of the XPG coding region and a downstream polyadenylation signal and then recircularizing the vector fragment. R992A and R992E mutations were generated in pBSK-XPGA using mutagenic primers as described for bacterial expression plasmids. These mutations were introduced into pBactin-XPG by swapping EcoRV-KpnI fragments between plasmids. XPG plasmid concentrations were determined by measurement of absorbance at 260 nm and verified by agarose gel electrophoresis and ethidium bromide staining. The XPG-deficient CHO cell line UV135 (16,23) was transfected with a mixture of 150 ng of luciferase plasmid, 150 ng of ␤-galactosidase plasmid (22), and 30 ng of wild-type or mutant XPG plasmid by calcium phosphate precipitation (16). For each XPG transfection set, three to six 60-mm dishes were used. Cells were lysed 48 h after transfection for analysis of luciferase and ␤-galactosidase activity (22). Activity assays were performed in duplicate, and the values were averaged. For each transfection, the luciferase activity was divided by the corresponding ␤-galactosidase activity to give relative luciferase activity, a measure of DNA repair that is normalized for transfection efficiency. Statistical analyses of the data were performed using Microsoft Excel version 4.

RESULTS
The PCNA Binding Activity of FEN-1 Resides within a Short Region near the C Terminus-We produced a series of GST fusion proteins that contain various regions of human FEN-1 and assayed their PCNA binding activity using a "pull-down" affinity bead interaction assay (Fig. 1). The binding activity was contained entirely within amino acids 328 -355 of FEN-1. All fusion proteins that contained this region bound to PCNA, whereas none of the proteins tested that lacked the complete 28 amino acid sequence displayed binding activity. For example, the fusion protein containing only residues 328 -348 of FEN-1 did not bind PCNA.
It has been reported previously that the PCNA-binding domain of human FEN-1 is contained within residues 307-380 and that residues 364 -380 are essential for PCNA binding (12). The former conclusion is consistent with the observations reported here, but the latter conclusion is not. Because we observed no requirement for 364 -380 in our domain mapping studies, we generated truncated FEN-1 proteins to address the importance of this region in more detail (Fig. 2). Deletion of the PCNA-binding region to give a truncated form of FEN-1 comprising only amino acids 1-328 abolished PCNA binding activity. However, a truncated FEN-1 comprising amino acids 1-363 bound PCNA as effectively as full-length FEN-1 (amino acids 1-380). Thus, we conclude that the C-terminal region from residues 364 to 380 of FEN-1 is not essential for PCNA binding activity and in fact makes little if any contribution to this activity. This contrasts with the previous report that truncated FEN-1 (amino acids 1-363) is unable to bind to PCNA in gel filtration and affinity bead pull-down assays (12). The reason for the difference between those observations and our own is not apparent. We next sought to identify specific residues of FEN-1 that are most important for interaction with PCNA. Arginine 339 of FEN-1, lying within the PCNA-binding region, was found to be crucial for PCNA binding activity. Single point mutagenesis of FEN-1 to convert Arg 339 to either alanine or glutamate dramatically decreased the ability of the protein to bind PCNA, showing the importance of this region of FEN-1 and of this residue in particular. Although PCNA binding activity of the R339A and R339E mutants of FEN-1 was severely impaired, each mutant retained endonuclease activity as determined by rapid kinetic flow cytometry (15, 24) using a fluoresceinated 5Ј-flap DNA substrate (data not shown).
Identification of a PCNA-binding Domain in XPG-The amino acid sequence of the PCNA-binding region of FEN-1 is significantly conserved in XPG (Fig. 3). We sought to determine whether the function of this region is conserved in XPG as well. We made a GST fusion protein (GST-XPG) containing the 29amino acid sequence of human XPG (residues 981-1009) that is homologous to the PCNA-binding region of FEN-1. GST-XPG bound human PCNA very efficiently when bacterial lysates containing these proteins were mixed (Fig. 4). GST alone lacking XPG sequence had no observable affinity for PCNA. The association of GST-XPG and PCNA was blocked by the addition of XPG 981-1009 peptide as competitor, confirming that the XPG moiety of the fusion protein was responsible for the binding. XPG and FEN-1 share a conserved arginine (Arg 992 in XPG) that was shown by mutagenesis to be important for PCNA binding activity in FEN-1 (Fig. 2). Replacing this arginine in GST-XPG with alanine or glutamate caused almost total loss of PCNA binding activity. The dramatic decrease in bound PCNA resulting from a single amino acid substitution in GST-XPG attests to the specificity of this assay.
The PCNA-binding domain of FEN-1 displayed robust binding activity under all conditions tested; however, the binding activity of the corresponding region of XPG exhibited profound salt dependence (Fig. 5). Binding of GST-XPG and PCNA was almost undetectable in a buffer containing 60 mM NaCl. Adding 100 mM potassium phosphate to this buffer produced maximal binding, and adding 25 mM potassium phosphate, 100 mM KCl, or 300 mM KCl increased binding to about 25, 25, and 100% of maximum, respectively. Thus, divalent anion was especially effective in aiding binding. In contrast to the behavior of GST-XPG, GST-FEN (residues 328 -355) displayed nearly maximal binding in any of these buffers. Varying pH from 6.0 to 8.0 had little effect on the association of GST-XPG and PCNA, and this complex was stable to repeated washing with either 1% Nonidet P-40 detergent or 1.0 M NaCl (data not shown).
Full-length XPG Binds to PCNA-An affinity bead assay was used to evaluate the interaction of PCNA and full-length XPG. Purified human PCNA was covalently attached to beads. Human XPG was expressed in insect cells that had been infected with a recombinant baculovirus strain that contains XPG cDNA (17). XPG present in baculovirus-infected cell lysates bound to PCNA beads but not to control beads made with BSA (Fig. 6A). The specificity of the PCNA-XPG interaction was further demonstrated in competition experiments. The binding of XPG to PCNA beads was blocked by the addition of free PCNA to the lysate but was unaffected by addition of BSA (Fig.  6B). Binding was also inhibited by the addition of GST-XPG 981-1009, the fusion protein containing the PCNA-binding fragment of XPG. However, the R992E derivative of this fusion protein, which lacks PCNA binding activity (Fig. 4), was unable to compete with XPG for binding to the beads. Similarly, GST-FEN 328 -363 , but not the R339E mutant of GST-FEN 328 -363 , competed with XPG for binding to the PCNA beads. It appears that the FEN-1 domain binds to PCNA with higher affinity than the XPG domain, because competition by GST-FEN 328 -363 was complete, whereas that by GST-XPG 981-1009 at similar concentration was partial. Because PCNA and XPG are both DNA-binding proteins, we questioned whether DNA might mediate their association indirectly. DNase I treatment of the PCNA-XPG complex failed to remove XPG (Fig. 6C), proving that the two proteins were not simply linked by DNA. Prior to treatment, the complex was washed to remove free DNA and actin, components of crude insect cell lysate that could decrease the effectiveness of DNase I. Under the reaction conditions employed, the DNase I used in the treatment was sufficient to digest 25 g of HindIII-cut Lambda DNA to completion and to convert 250 g of DNA to a low molecular weight smear (data not shown). The lack of effect of DNase I treatment, together with the competition experiments, prove conclusively that the interaction of XPG and PCNA is direct.
The PCNA Binding Activity of XPG Is Needed for Maximum Repair Efficiency-PCNA and XPG both participate in excision repair, raising the expectation that the interaction of these proteins occurs in that context. Therefore, we sought to investigate the role of this interaction in NER in vivo. CHO-UV135 cells lack a functional XPG homolog. They are hypersensitive to UV-induced cell mortality and are severely defective in the repair of UV-damaged DNA. Both phenotypic defects can be corrected by transfection with an XPG expression plasmid (16,22,26). This system provides a model of NER function in vivo that is dependent upon exogenously supplied XPG. Because arginine 992 is critical for full PCNA binding activity of XPG, we transfected UV135 cells with R992A or R992E single point mutant derivatives of XPG to produce cells in which the XPG-PCNA interaction was impaired relative to wild-type XPG transfectants. XPG-dependent repair activity of wild-type and mutant transfectants was compared to assess the importance of the XPG-PCNA interaction in NER (Fig. 7). In this assay, in vivo repair of a UV-damaged luciferase reporter plasmid results in expression of luciferase commensurate with the extent of repair. Fully reconstituted repair activity is exemplified by cells transfected with wild-type XPG. Absence of repair activity is shown by cells transfected with truncated XPG lacking residues 948 -1186. The deletion of 3Ј-untranslated sequence in this mutant probably destabilizes the mRNA, and the coding region truncation removes the nuclear localization signal to encode a mutant protein unable to enter the nucleus (20,21). Lying between these benchmark indicators of maximum and minimum repair activity were those cells dependent upon R992A and R992E XPG mutants for NER function. A small but reproducible decrease in repair efficacy was observed for each of these mutants. This suggests that impairment of the PCNA-XPG interaction adversely affects NER but that the magnitude of the defect is small compared with complete absence of function.
The PCNA-binding Domains of FEN-1 and XPG Compete with That of p21-The PCNA-binding region of the cyclin-dependent kinase inhibitor p21 has been localized to amino acids 141-160 (27). A 39-mer synthetic peptide corresponding to amino acids 126 -164 of p21 competes with full-length FEN-1 for binding to PCNA (12). The high resolution domain mapping of FEN-1 and XPG reported here allowed for a useful sequence alignment between the PCNA-binding regions of these nucleases and p21 (Fig. 8). Comparison of the amino acid sequences of the PCNA-binding domains of FEN-1 and p21 reveals a structural similarity that suggests direct competition for a single binding site on PCNA. Furthermore, the mapping data predict competition between p21 and XPG for binding to PCNA. Competition between the isolated PCNA-binding domains of these proteins was confirmed using the pull-down affinity bead interaction assay (Fig. 9). Interestingly, the p21 peptide was more potent than the XPG peptide in competing with the GST-XPG fusion protein for binding to PCNA, suggesting that the p21-PCNA interaction (2.5 nM K D ; Ref. 31) is higher affinity than the XPG-PCNA interaction. As expected, the p21 peptide competed the PCNA-binding GST-FEN fusion protein (data not shown), confirming the previously reported competition between p21 peptide and full- length FEN-1 (12).

FIG. 3. Alignment of structure-specific endonucleases XPG and FEN-1 showing homologous regions. A and B
are predicted to form the nuclease domain (24,25). The PCNA-binding region (P) is expanded to show human sequence; arginine/lysine-rich region (NLS) is known to mediate nuclear localization in XPG (20,21). Arg 992 of XPG and Arg 339 of FEN-1 (targets for site-directed mutagenesis) are shown in bold type.

DISCUSSION
PCNA plays a major role in DNA replication and DNA damage repair. In addition to its core function of tethering DNA polymerases ␦ and ⑀ to the DNA template, PCNA also interacts with the cell cycle regulator p21, the replication endonuclease FEN-1, and, as shown here, the repair endonuclease XPG. Similarity in the primary structures of the PCNA-binding domains of p21, FEN-1, and XPG (Fig. 8) strongly suggests that these proteins all use an analogous arrangement of key hydrophobic residues to bind to a hydrophobic pocket formed primarily by the interdomain connector loop of PCNA, as is known to be the case for p21 (28). This similarity predicts that competitive binding between these proteins would be the result of direct competition for a single binding site on PCNA, rather than steric hindrance arising from adjacent but distinct sites. Experiments using the isolated PCNA-binding domains from these proteins are consistent with this prediction. In contrast, the previously observed competition between FEN-1 and p21 was interpreted as being due to ligand-induced conformational change in PCNA, rather than occupation of the same site on PCNA (12). Because p21 is otherwise unrelated to the nucleases, the common structural elements in the PCNA-binding regions of p21 and FEN-1/XPG appear to have arisen by convergent evolution.
XPG and PCNA are involved in distinct and experimentally separable steps of NER, namely excision and repair resynthesis, respectively. The identification of PCNA binding activity in XPG suggests that the excision and resynthesis machinery might communicate with one another in some manner. The structural and functional conservation of PCNA binding in FEN-1 and XPG reflects a selective pressure to preserve this activity. The submaximal NER activity observed for the Arg 992 XPG mutants assayed in vivo suggests that the PCNA-XPG interaction may contribute to repair efficiency.
The specific function of the PCNA-XPG interaction is not obvious. It seems unlikely that PCNA recruits XPG to the NER complex, because PCNA is not necessary for the assembly of a functional exinuclease containing XPG (2, 3, 32-36). Conversely, XPG seems poorly suited to target PCNA to the NER complex. XPG cleaves the damaged DNA strand at the 3Ј end of the repair patch. Because the PCNA binding and catalytic domains of XPG are quite close to one another (Fig. 3), the PCNA-XPG interaction may take place near the 3Ј incision site. However, the PCNA-polymerase complex must initiate repair synthesis at the 5Ј end of the repair patch and move along the template in a 5Ј to 3Ј direction. Therefore, PCNA might be most likely to encounter XPG at the 3Ј end of the repair patch upon completion of resynthesis.
In excision assays, the DNA undergoing repair remains tightly associated with exinuclease proteins even after comple- FIG. 6. Specific interaction of full-length XPG and PCNA. A-C, Western blots showing XPG bound to affinity beads. Purified BSA or human PCNA was covalently coupled to CNBr-activated Sepharose to make affinity beads. Beads were mixed with lysate from insect cells infected with a recombinant baculovirus that directs expression of human XPG. After washing, noncovalently bound proteins were eluted, resolved by SDS-PAGE, and analyzed by anti-XPG Western blot. A, XPG bound to BSA-CNBr or PCNA-CNBr affinity beads. B, XPG bound to PCNA-CNBr beads after competition by exogenous purified proteins. 0.4 mg of BSA, PCNA, GST-XPG 981-1009 , the R992E mutant of GST-XPG 981-1009 , GST-FEN 328 -363 , or the R339E mutant of GST-FEN 328 -363 was added to XPG lysate prior to incubation with beads. C, PCNA-CNBr beads were incubated with XPG lysate and then washed to remove free DNA and actin (a DNase I inhibitor). 10 mM MgCl 2 was added to each of the washed bead suspensions, and one received 250 units of DNase I to degrade any tightly associated DNA that might be present. After 20 min of incubation, the Ϯ DNase I-treated beads were washed again, and bound proteins were eluted for Western analysis.
FIG. 7. Mutant XPG defective in PCNA binding is unable to fully complement nucleotide excision repair activity in a repairdeficient cell line. NER-defective UV135 cells lacking functional XPG were transiently co-transfected with XPG, luciferase, and ␤-galactosidase expression plasmids. Wild-type XPG (q), point mutant R992A (f), point mutant R992E (OE), or truncated XPG (ϫ) were assayed for ability to complement the UV135 repair defect. The luciferase reporter plasmid was unirradiated (0 J/m 2 ) or damaged by UV light (400 or 800 J/m 2 ) prior to transfection. ␤-Galactosidase was used as an internal standard to normalize co-transfection efficiency. XPG-dependent repair is plotted as the mean relative luciferase activity (luciferase activity/␤-galactosidase activity, arbitrary units) Ϯ S.E. Analysis of variance was performed for each dose of UV damage, testing the null hypothesis that the mean relative luciferase activity is equal between the transfected groups. Considering all four groups (wild type, R992A, R992E, and truncated), p ϭ 0.21 at 0 J/m 2 , p ϭ 0.000014 at 400 J/m 2 , and p ϭ 0.00091 at 800 J/m 2 . Considering only wild type, R992A, and R992E, p ϭ 0.34 at 0 J/m 2 , p ϭ 0.016 at 400 J/m 2 , and p ϭ 0.086 at 800 J/m 2 . The analyses indicate no significant differences between groups when undamaged reporter plasmid was used but significant differences between XPG transfection groups in the repair of UV-damaged plasmid. One-tailed t tests showed a significant decrease in XPG-dependent repair for each point mutant compared with wild-type XPG at the same UV damage dose: at 400 J/m 2 , p ϭ 0.011 for R992A and p ϭ 0.032 for R992E; at 800 J/m 2 , p ϭ 0.052 for R992A and p ϭ 0.051 for R992E. There was no significant difference between R992A and R992E groups at any dose. The results shown represent one out of a total of six independent experiments that were conducted; in all six experiments, R992A and R992E each exhibited a repair deficiency similar in magnitude to that shown. tion of dual incision (34,35). This indicates that factors in addition to those traditionally considered as constituents of the excision reaction may be needed to effect disassembly of the complex and facilitate reutilization of the constituents at a new damage site. PCNA has been found to stimulate the excision activity of fractionated HeLa cell extract, an observation that led to the proposal that PCNA may promote turnover of the exinuclease assembly (2). Maximum excision results from the specific inclusion of both PCNA and dNTPs in the partially fractionated system, providing strong support for a linkage between excision and resynthesis. The interaction of PCNA and XPG might provide a temporally appropriate signal to initiate the dissociation of remaining proteins. Clearly, excision and repair synthesis can proceed in isolation in vitro, but integration of excision and resynthesis may be maximally efficient. Interference with such integration might produce only a mild repair defect, perhaps like that displayed by the Arg 992 XPG mutants reported here. The interaction of PCNA and XPG provides a possible mechanism by which excision and resynthesis could be interconnected.
It is interesting that the DNA repair protein XPG and the DNA damage-inducible protein p21 appear to bind to the same site on PCNA. Following UV damage, p53-dependent transcription causes elevation of p21 levels (37)(38)(39). Acting as a cyclin-dependent kinase inhibitor, p21 promotes cell cycle arrest that is thought to ameliorate DNA damage by allowing increased time for repair to take place. The inhibition of PCNA activity by p21 is a more paradoxical aspect of the p21 response, however. Although the association of p21 and PCNA might serve to disrupt replicative synthesis and thereby contribute to cell cycle arrest, PCNA in an active state is nonetheless essential for excision repair. Thus, p21 might be expected to inhibit DNA repair by sequestering PCNA. It has been found by some investigators (40), although not all (41,42), that p21 does indeed inhibit NER, and this inhibition can be alleviated by the addition of PCNA. On the other hand, p21-deficient cells exhibit a modest UV repair defect, a phenotype that can be reversed by providing wild-type p21 but not truncated p21 lacking PCNA binding activity (43). This suggests that the interaction of p21 and PCNA somehow makes a beneficial contribution to the damage response. The situation in vivo appears complicated, perhaps involving subtle regulation of the repair process or differences in the temporal expression or subcellular localization of PCNA, p21, and XPG that allow these proteins to work together appropriately.   8. Alignment of PCNA-binding regions of FEN-1, XPG, and p21 reveals key hydrophobic residues in common and suggests a similar mode of interaction with PCNA. Hydrophobic residues of p21 shown in bold type occupy a hydrophobic pocket on the surface of PCNA when the two are complexed, as determined by structural analysis of the co-crystal (28). These residues of p21 have also been shown to be particularly important for interaction with PCNA by mutational analysis (27,29,30). Several residues in the PCNA-binding region of human p21 are identical or conservatively substituted in human FEN-1 and XPG, including the key hydrophobic residues Met 147 , Phe 150 , and Tyr 151 .