Replication Protein A Confers Structure-specific Endonuclease Activities to the XPF-ERCC1 and XPG Subunits of Human DNA Repair Excision Nuclease*

XPF-ERCC1 and XPG proteins are nucleases that are involved in human nucleotide excision repair. In this study, we characterized the structure-specific junction- cutting activities of both nucleases using DNA substrates containing a bubble or loop structure. We found that the junction-cutting activities of XPF-ERCC1 and XPG were greatly stimulated by human replication pro- tein A (RPA), while heterologous single-stranded DNA-binding proteins could not substitute for human RPA. To test for specific interaction between RPA and XPF-ERCC1 as is known to occur between RPA and XPG, we employed a pull-down assay with immobilized “bubble” substrate. We found that the binding of XPF-ERCC1 complex to the bubble substrate was enhanced by RPA, suggesting a possible mechanism for RPA in the excision nuclease system, that is the targeting of the nuclease subunits to their specific sites of action. Furthermore, the RPA-promoted junction cutting by XPF-ERCC1 and XPG nucleases was observed with “loop” substrates as well, raising the possibility that XPF-ERCC1, XPG, and RPA may function in removing loop structures from DNA, independent of the other subunits of the human excinuclease.

XPF-ERCC1 and XPG proteins are nucleases that are involved in human nucleotide excision repair. In this study, we characterized the structure-specific junctioncutting activities of both nucleases using DNA substrates containing a bubble or loop structure. We found that the junction-cutting activities of XPF-ERCC1 and XPG were greatly stimulated by human replication protein A (RPA), while heterologous single-stranded DNAbinding proteins could not substitute for human RPA. To test for specific interaction between RPA and XPF-ERCC1 as is known to occur between RPA and XPG, we employed a pull-down assay with immobilized "bubble" substrate. We found that the binding of XPF-ERCC1 complex to the bubble substrate was enhanced by RPA, suggesting a possible mechanism for RPA in the excision nuclease system, that is the targeting of the nuclease subunits to their specific sites of action. Furthermore, the RPA-promoted junction cutting by XPF-ERCC1 and XPG nucleases was observed with "loop" substrates as well, raising the possibility that XPF-ERCC1, XPG, and RPA may function in removing loop structures from DNA, independent of the other subunits of the human excinuclease.
Nucleotide excision repair is a general repair system that plays an important role in maintaining genetic integrity (1,2). This repair system removes damaged nucleotides from DNA by dual incisions on both sides of the lesion in the damaged strand (3). Recently, the human and the highly homologous yeast nucleotide excision repair systems have been reconstituted from extensively purified proteins, which demonstrated that 14 -15 polypeptides comprising replication protein A (RPA, 1 also known as HSSB or RF-A) and the general transcription factor, TFIIH, in addition to XP and ERCC proteins were essential for the dual incision step (4 -6).
It was found that the two subunits of the excision nuclease had intrinsic endonuclease activities in the absence of the other components of the repair system. XPG was reported to possess a single-stranded DNA endonuclease activity (7,8) and an exonuclease activity with 5Ј to 3Ј directionality (9). Similarly, it was found that XPF-ERCC1 had a single strand-specific endonuclease activity and a weak activity on double-stranded UVirradiated DNA, which was stimulated by RPA (10). Furthermore, it has been reported that the RAD1-RAD10 complex (11), the yeast counterpart of the human XPF-ERCC1 nuclease, and XPG nuclease (12) have structure-specific junction-cutting activities with unique polarities. These results obtained with the model "undamaged" substrates suggested that XPF-ERCC1 and XPG were the 5Ј-and 3Ј-endonucleases of the dual incisions, respectively. A recent study using damaged DNA and specific antibodies supported this model (13).
Recently, it has been reported that RPA, which is absolutely required for dual incisions (4), specifically binds to XPA protein and XPG nuclease (14). These findings raised the possibility that RPA may play an important role in recognizing DNA lesions and then targeting the nuclease subunits of human excision nuclease system to their proper sites. Hence, we decided to investigate the effect of RPA on the nuclease subunits in the absence of other components of excision nuclease.
Proteins-The XPA cDNA tagged with 6 ϫ histidine using pRSET vector (Invitrogen) was overexpressed in Escherichia coli DR153 strain (15) and purified by a modified procedure of Jones and Wood (16). The DNA construct bearing the three-subunit human RPA was obtained from Dr. M. Wold (University of Iowa), and the recombinant RPA was expressed in E. coli and purified according to Henricksen et al. (17). Purification of XPF-ERCC1 complex was carried out as described previously (10). Protein concentrations were ϳ4 ng/l (the heparin-agarose column fraction), ϳ10 ng/l (the DEAE-column fraction), and 33 ng/l (the XPA-affinity column fraction).
XPG protein was overexpressed in a baculovirus/insect cell system using pVL1392-XPG/Baculogold viral DNA (Pharmingen) and SF21 cells and extensively purified by a modified procedure of O'Donovan et al. (7). Briefly, after separation on a phosphocellulose column, XPGcontaining fractions identified by Western blotting were loaded onto an * This work was supported by National Institutes of Health Grant GM32833 and a grant from the Human Frontier Science Program. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Supported by a Human Frontier Science Program fellowship. § Supported by a grant from the Cancer Research Fund of the Damon Runyon-Walter Winchell Foundation.
Affi-Gel blue (Bio-Rad) column. The column was extensively washed with buffer A (25 mM HEPES-KOH, pH 7.9, 1 mM EDTA, 1 mM dithiothreitol, and 10% glycerol) containing 150 mM KCl and subsequently with buffer A containing 1 M KCl. Final elution was achieved with a linear gradient of NaSCN (0.5-1.5 M) in buffer A. XPG fractions were dialyzed against storage buffer (25 mM HEPES-KOH, pH 7.9, 100 mM KCl, 12 mM MgCl 2 , 0.5 mM EDTA, 2 mM dithiothreitol, and 17% glycerol) and used in most of the experiments. For further purification, the elution fractions from the Affi-Gel blue column were dialyzed against buffer A containing 1 M (NH 4 ) 2 SO 4 and loaded onto a phenyl-Superose column (HR 5/5, Pharmacia Biotech Inc.). After extensive washing, XPG was eluted with a linear gradient of (NH 4 ) 2 SO 4 (0.6 -0 M) in buffer A and dialyzed against storage buffer. The purified XPG protein was active for complementing repair deficiency of UV135 (ERCC-5 ϭ XP-G) cell-free extract in an excision assay (3, 13) (data not shown).
Pull-down Assay Using Immobilized DNA-The top strand oligomer of the substrates shown in Fig. 1a was 3Ј-labeled with biotin-16dideoxy-UTP using terminal deoxynucleotidyltransferase according to the manufacturer's protocol (Boehringer Mannheim). DS and B30 substrates were prepared as described above. Five g of Dynabeads M-280 streptavidin (Dynal) was incubated at 30°C for 1 h with the biotinlabeled DS or B30 substrates (400 ng) in nuclease reaction buffer and washed with the same buffer three times using a magnetic particle concentrator (Dynal). Immobilized substrate DNAs were incubated at 4°C for 3 h with rocking, with RPA, XPA, XPF-ERCC1 (the XPAaffinity column fraction that contained 330 ng/l protein), or a combination of these proteins as indicated, in 50 l of nuclease reaction buffer containing 3 mM CaCl 2 , in place of MgCl 2 , to inhibit the nuclease activity of XPF-ERCC1 (10) and 1% bovine serum albumin to reduce nonspecific binding of proteins to the magnetic beads. The purified beads were mixed with SDS-loading buffer and heated. The bound fractions were separated on a 10% polyacrylamide SDS gel, and Western blotting was carried out using standard procedure.

RESULTS AND DISCUSSION
To study the effect of RPA on XPF-ERCC1 and XPG, we constructed substrates with 12-or 30-nt mismatched bases (bubble), or deleted nucleotides in the bottom strand (loop) (Fig.  1a). The substrates were labeled 5Ј or 3Ј and characterized with respect to single-strandedness by S1 nuclease digestion. From the specific degradation by S1 nuclease, it is evident that the substrates used in our study contained no secondary structures in the bubble and loop regions (Fig. 1b). RPA, which has high affinity for single-stranded DNA, specifically bound to the 30-nt bubble or loop region (Fig. 1c) and protected it from DNase I (Fig. 1d); it also bound to the 12-nt bubble and loop with less affinity (data not shown), in agreement with the report that a 30-nt-long single-stranded DNA is the minimum size for binding of RPA with high affinity (18).
XPF-ERCC1 is a single strand-specific endonuclease, which also nicks UV-irradiated supercoiled DNA in an RPA-stimulated reaction (10). Furthermore, it was found that the XPF-ERCC1 complex purified through the XPA-affinity column step contained a junction-cutting activity, which was not obvious in the purer fractions (the DEAE and the heparin-agarose fractions) (Fig. 2a). We reasoned that some other protein(s) present in the XPA-column fraction might have conferred the junctioncutting activity to XPF-ERCC1 detected with this fraction. Since RPA binds to these model substrates, we tested RPA for conferring the junction-cutting activity to the DEAE and the heparin-agarose fractions (Fig. 2b). In the absence of additional protein(s), the purer XPF-ERCC1 fractions have a weak singlestranded nuclease activity, which generates a ladder in the area of single strandedness (Fig. 2b, lanes 1, 4, and 8). The addition of RPA suppresses the nonspecific endonuclease activity while conferring a specific junction-cutting activity upon the nuclease enabling it to hydrolyze the strand which enters the bubble region in the 5Ј to 3Ј direction at the fourth phosphodiester bond before entering the junction (Fig. 2b, lanes 2, 3, 6,  and 9). RPA also stimulated the junction-cutting activity of the XPA-column fraction (see below). In contrast, XPA, which is known to interact with XPF-ERCC1 (19), does not affect the single-stranded DNA endonuclease activity (Fig. 2b, lanes 5  and 10) and has no significant effect on the RPA-conferred junction-cutting activity of XPF-ERCC1 (Fig. 2b, lanes 7 and  11). Importantly, RPA failed to confer a junction-cutting activity to the S1 nuclease (Fig. 2c), indicating that the RPA-promoted junction-cutting activity is not caused by covering the bubble with RPA and leaving only the single-stranded junction region open for a nonspecific single-stranded DNA endonuclease. The junction-cutting activity of XPF-ERCC1 was dependent on RPA concentration, and 40 ng of RPA per reaction The top strand was 5Ј-or 3Ј-labeled, and the size of the 3Ј-labeled top strand, indicated in parentheses, was 1 nucleotide longer than that of the 5Ј-labeled top strand, as described under "Experimental Procedures." b, S1 nuclease digestion of the substrate DNAs. 5Ј-Labeled substrates were incubated at 20°C for 20 min with 0.5 unit of S1 nuclease (Life Technologies, Inc.) in 30 mM sodium acetate (pH 4.6), 1 mM zinc acetate, 50 mM NaCl, and 5% glycerol, and the products were analyzed on a denaturing 10% polyacrylamide gel.

Role of RPA in Human Excision Nuclease 11048
stimulated the activity about 64-fold (Fig. 2d). We wished to know the DNA structure requirements for the RPA-stimulated junction-cutting activity. As shown in Fig. 2e, the junctioncutting activity of the XPA-column fraction is greatly stimulated by RPA, and a smaller (12-nt) bubble structure is incised in the same way as the 30-nt bubble although with lower efficiency (lanes 1-3 compared with lanes 4 -6). Interestingly, similar results were obtained with loop structures. Both the 12-and 30-nt loop substrates were incised in a manner similar to the bubble substrates (lanes [7][8][9][10][11][12]. Similar experiments were conducted with XPG nuclease, which is known to have a junction-cutting activity in a bubble substrate (12) and to interact with RPA (14). Fig. 3a shows the effects of RPA on the XPG junction-cutting activity with the 30-nt bubble and loop substrates. Under our experimental conditions, a very weak junction-cutting activity of XPG was observed with both substrates (ϳ1% incision) (Fig. 3a, lanes 1 and  7), which is specific for the junction with entering the bubble/ loop region in 3Ј to 5Ј direction. Addition of RPA stimulated the junction-cutting activity in a concentration-dependent manner (Fig. 3a, lanes 2-6 and 8 -11), although the optimal RPA concentration and the major incision sites were different between the two substrates. The junction-cutting activity stimulated by RPA can be inhibited by anti-XPG antibodies (Fig. 3b). We failed to detect a significant XPG junction-cutting activity with smaller (12-nt) bubble and loop substrates under these conditions (data not shown).
These results, taken together, support a model of excision nuclease reaction where RPA may stabilize the unwound DNA region (presumably caused by TFIIH helicase activity) and recruit the two nucleases, XPF-ERCC1 and XPG, to their proper sites through physical interactions as is known to occur  7 and 11) prior to the incubation with 1 l of the heparin-agarose column fraction (lanes 1-3) or the DEAEcolumn fraction (lanes 4 -11) of XPF-ERCC1. In addition to the products caused by the XPF-ERCC1 junction-cutting activity, smaller fragments were reproducibly obtained with 5Ј-labeled substrates (lanes 2, 3, 6, and  7). These fragments might be by-products of XPF-ERCC1 junction cutting followed by 3Ј 3 5Ј exonucleolytic degradation. However, it is not clear whether this exonuclease activity is intrinsic to XPF-ERCC1. Minor effects of XPA on the RPA-promoted junction-cutting activity (lanes 7 and 11) are due to experimental variability, because it has not been reproduced in repeated experiments. c, effects of RPA on the S1 nuclease activity. 5Ј-Labeled B30 substrate was preincubated on ice for 30 min with no protein (lane 1), the indicated amounts of RPA (lanes 2  and 3), or E. coli SSB (lanes 4 and 5), followed by the addition of S1 nuclease. d, titration of the RPA-conferred junction-cutting activity.  2-4). XPG protein (ϳ10 ng) preincubated with no antibody (lanes 1 and 2), buffer only (lane 3), or affinity-purified anti-XPG antibodies (13) (lane 4) was added to the reaction mixture and incubated at 37°C for 90 min. M, size markers (X174/HinfI). Note that the apparent higher level of XPG-directed cutting with L30 -3Ј than with L30 -5Ј is due to the higher specific radioactivity of the L30 -3Ј substrate.
between RPA and XPG (14). In fact, consistent with this model, neither E. coli single-stranded DNA-binding protein (SSB) (Fig. 2d, lane 6 and Fig. 3a, lane 12) nor yeast RPA (data not shown) could stimulate the junction-cutting activities of the two human nucleases, and human RPA did not stimulate nonspecific endonuclease activities of XPG and XPF-ERCC1 on single-stranded DNA substrate (data not shown). However, direct interaction between RPA and XPF-ERCC1 has not been reported so far, despite the known interactions of XPA with RPA (14,20,21) and ERCC1 (19,22,23).
In light of our results, which demonstrate stimulation of XPF-ERCC1 by RPA in the absence of XPA, we conducted a pull-down assay to test whether XPF-ERCC1 binding to the bubble substrate can be enhanced in the presence of RPA without the mediation of XPA (Fig. 4). The immobilized 30-nt bubble substrate was incubated with RPA, XPA, XPF-ERCC1, or a combination of these proteins. RPA bound to this substrate specifically and efficiently (lanes 5-8 compared with lanes 10 and 11), while XPF-ERCC1 bound to the bubble substrate weakly in the absence of RPA (lane 4). When XPF-ERCC1 was incubated with the substrate in the presence of RPA, more XPF-ERCC1 bound to the bubble substrate (lanes 5 and 6). In contrast, XPA did not affect the XPF-ERCC1 binding to this substrate (lane 7). Yeast RPA failed to facilitate XPF-ERCC1 binding to the bubble substrate (data not shown), consistent with the failure of yeast RPA to confer a junction-cutting activity to XPF-ERCC1. These results suggest that human RPA recruits XPF-ERCC1 onto the bubble substrate DNA, possibly either through direct interaction or by creating a unique DNA structure.
One reasonable question is why does the XPA-column fraction have a junction-cutting activity but the purer fractions do not? The contamination of RPA in the XPA-column fraction seems to be unlikely, because no trace of RPA (or XPA) could be detected by either silver staining or Western blotting with high sensitivity (data not shown, also see Fig. 4). A simple explanation is that XPF-ERCC1 has an intrinsic junction-cutting activity without RPA, which becomes nonspecific at low concentrations of the enzyme. Indeed, the diluted XPA-column fraction showed a nonspecific single-stranded DNA endonucle-ase activity (data not shown) rather than a junction-cutting activity (which could be restored by adding RPA).
Our results show that RPA plays an essential role in positioning the two nuclease subunits of the human excision nuclease system. Furthermore, these results show that, with loop structures of 12-30 nt in length, RPA, XPG, and XPF-ERCC1 are capable of removing the loop by incising the strand with extra nucleotides on both sides of the loops. The human MutLS system is capable of eliminating mismatches and loops up to 4 nucleotides (24). However, it was reported that loops of 5, 8, and 16 nucleotides can be removed in a hMLH1-independent manner (25). Moreover, it has been found that the yeast Schizosaccharomyces pombe exonuclease I (26) and the Saccharomyces cerevisiae RTH1 nuclease (27), which belong to the extended family of XPG/RAD2/Rad13 and FEN-1/Rad2 proteins, play roles in mismatch corrections. Thus, it is possible that XPG and XPF-ERCC1 proteins, in addition to their roles in damage excision repair, may participate in removing the looped-out DNA with a relatively large size and hence in genomic stability.