The general transcription-repair factor TFIIH is recruited to the excision repair complex by the XPA protein independent of the TFIIE transcription factor.

Recent studies have revealed that the general transcription factor TFIIH is also a general excision repair factor which, along with several other proteins, is required for transcription-independent excision reaction. As a general transcription factor, TFIIH is recruited to RNA polymerase II-promoter complex by another general transcription factor called TFIIE. We were interested in knowing whether TFIIE is also involved in recruiting TFIIH to the excision repair complex. We found that cell-free extract depleted of TFIIE carried out excision repair at a normal rate, leading us to conclude that TFIIE is not involved in recruiting TFIIH to the damage site and has no role in general excision repair. In contrast, the human damage recognition protein XPA specifically binds to TFIIH and apparently recruits it to the damage site. The carboxyl-terminal half of XPA is responsible for specific interaction with TFIIH. The C261S/C264S mutant of XPA bound the ERCC1-XPF complex normally, but failed to bind TFIIH and failed to complement an XP-A mutant cell-free extract indicating that the XPA-TFIIH interaction is essential to effecting the excision reaction. Interestingly, XPA also binds to the p34 subunit of TFIIE specifically and in competition with the p56 subunit of TFIIE. This latter interaction has no apparent role in general excision repair but may be relevant in the transcription-coupled repair reaction.

XPG and ERCCI of mammalian excision repair proteins (see Prakash et aZ. (1993) and Tanaka and Wood (1994)). However, recent biochemical studies have revealed that the entire TFIIH general transcription factor complex (at least 8 subunits including the XPB and XPD proteins) is a repair factor suggesting that additional proteins, not defined by XP complementation groups, may play essential roles in excision repair (Drapkin et aZ., 1994a;Schaeffer et aZ., 1993Schaeffer et aZ., , 1994 and that there may be an intimate relation between transcription initiation by RNA Pol II and transcription-independent excision repair. This raised the possibility that other general transcription factors may participate in excision repair. An obvious candidate is TFIIE because it appears that all TFIIH transactions in transcription are mediated through specific interactions with TFIIE (Flores et aZ., 1989;Goodrich and Tjian, 1994;Ohkuma and Roeder, 1994).
During transcription initiation TFIIE interacts with several proteins assembled at the promoter, including RNA Pol II itself, before recruiting TFIIH (Conaway and Conaway, 1993;Zawel and Reinberg, 1993;Maxon et aZ., 1994); thus we reasoned that it may function in an analogous manner in repair by first interacting with the damage binding proteins at the lesion site and then recruiting TFIIH to carry out its repair function. Therefore, we tested the interactions of the human damage binding protein XPA with TFIIE and TFIIH and the requirement for these interactions to carry out the excision repair reaction. We found that XPA binds to the p34 subunit ofTFIIE and to TFIIH directly. Cell-free extracts (CFEs) depleted of both subunits of TFIIE are capable of full strength excision repair indicating that the interaction ofXPA with TFIIE plays no role in general (transcription-independent) excision repair. In contrast, the XPA-TFIIH interaction is crucial for general excision repair. Two point mutations in the carboxyl-terminal half ofXPA protein do not affect the binding of the ERCC1·XPF complex to XPA but abolish the binding of TFIIH to XPA and the excision activity of the in vitro system reconstituted with the mutant XPA protein. We conclude that, most likely, the XPA protein recruits TFIIH to the lesion site and that TFIIE is not involved in the general nucleotide excision reaction.

RESULTS
Binding of XPA to TFIIH and TFIIE Subunits-We have found that under mild conditions TFIIH in HeLa CFE is retained by an XPA affinity column (see below). To find out if TFIIH directly binds XPA or if it is bound via TFIIE in a manner analogous to its binding to RNA Pol II-promoter complex, we investigated the interactions of three subunits of TFIIH and the two subunits of TFIIE with XPA protein by expressing these proteins in an in vitro transcription-translation system and measuring binding by a GST pull-down assay. The results are shown in Fig. 1. XPA did not bind to the three subunits of TFIIH tested, the XPBIERCC3, XPDIERCC2, and p62 proteins. It did not bind to the p56 subunit ofTFIIE (p56E) either; but, it specifically bound to the p34 subunit of TFIIE (p34E). This finding raised the possibility that TFIIH is recruited to the lesion site by this TFIIE subunit. We decided to investigate the interaction of TFIIE and XPA in more detail.
TFIIE Is Not a Repair Factor-In Fig. 2 the interaction of XPA with the individual subunits ofTFIIE and with the TFIIE complex was investigated by a pull-down assay using GST-XPA resin. XPA bound to p34E but not to p56E or to either subunit when the two subunits were co-expressed in the in vitro transcription-translation system and hence were presumably in a complex. This result suggested that the XPA-and p56E-binding sites on p34E may overlap. To test this prediction we conducted a competition assay (Fig. 3). Radiolabeled p34E was incubated with the GST-p56E resin in the presence of inereasing amounts of MBP-I3-galactosidase a or MBP-XPA proteins. The presence of XPA in the reaction mixture interfered with binding of p34E to the GST-p56E resin (lane 10 versus 11 and 12 versus 13), consistent with the finding that XPA and p34E make a complex which is unable to bind p56E. Hence if TFIIE participates in excision repair it would be through its p34 subunit.
The carboxyl-terminal deletion mutants were constructed as follows. The N13 was made by digestion of pGEX18-XPA with HindIII and self-ligation to generate pGEX18-XPAN138. The N192 was constructed by digesting pGEX18-XPA with Bsa.! and NdeI, filling-in with Klenow fragment, and religating it to give pGEX18-XPAN192. The N256 was made a follows: pGEX18-XPA was digested with NdeI, filled-in with Klenow fragment, redigested with Sac!, and then ligated with Sac! and RsaI-dige ted fragment of pGEX18-XPA to generate pGEX18-XPAN256.
The point mutant XPA2CS (C261S, C264S) was constructed in the plasmid pCHA801 (Park and Sancar, 1993) by site-directed mutagenesis using the oligonucleotide 5'-TACCGTAAGACTAGTACTATGAGTG-GCCATGAACTG-3' as the mutagenic primer (Sambrook et al., 1989). The mutation was confirmed by restriction digestion of the newly introduced restriction site. An 857-base pair fragment carrying the entire mutant XPA2CS was amplified by polymerase chain reaction and subcloned into pMAL-c2 to generate pMAL-XPA2CS, with which pGEX18-XPA2CS was constructed as described for pGEX18-XPA.
Protein-Protein Interaction Assays-Interaction ofXPA with the subunits ofTFIIE or TFIIH were studied using glutathione-agarose beads containing 10 ILg of GST or 4 ILg of GST-XPA and 5 ILl of the in vitro translated products in buffer N (40 mM Hepes-NaOH, pH 7.5, 100 mM KCl, 5 mM MgCI 2 , 0.2 mM EDTA, 0.4% Nonidet P-40, 1 mM dithiothreitol). After incubation for 1 h at 4°C with mild agitation, the beads were extensively washed with the arne buffer and resuspended in SDS-PAGE sample buffer. The bound proteins were analyzed by SDS-PAGE and visualized by autoradiography.
Competition ofXPA with p56E for binding to p34E was studied using glutathione-agarose beads containing 4 ILg of GST-p56E, 5 ILl of the in vitro translated product of p34E, and the indicated amounts of MBP-f3-galactosidase a or MBP-XPA.
Interaction of XPA with TFIIH was assayed using glutathione-agarose beads containing 10 ILg of GST or 2-4 ILg of wild-type or mutant GST-XPA proteins and 50 ILg of HeLa CFE or 2 ng of purified TFIIH. The bound proteins were visualized by Western blotting using anti bodie against XPB or p62 subunits ofTFIIH or anti-ERCC1 antibodies.
Immunodepletion of TFIIE from HeLa CFE-Protein A-agarose beads (40 ILl) were wa hed with phosphate-buffered saline and incubated with 2 ILg of anti-p56E or anti-p34E antibodies for 1 h at 4°C. The bead were washed three times with storage buffer (Park and Sancar, 1993) and incubated with HeLa CFE (200 ILg) for 1 h at 4°C. After spinning down the beads, the supernatant was carefully removed and used in Western blotting and the in vitro excision assay.
Excision Assay-The human excinuclease activity was tested by the excision assay as described by Huang et al. (1994a) with some modifications. The 140-base pair substrate containing cholesterol in the cen-2 3 4 5 6 7 8 9 10 11 12 13 14 15 28-1 2 3 4 5 6 7 8 9 10 FIG. 5. XPA binds directly to TFIIH. HeLa CFE (50 IJ-g) or purified TFIIH (2 ng) were incubated with 10 IJ-g of GST or 4 IJ-g of GST-XPA linked to glutathione-agarose beads. The unbound and bound proteins were separated by SDS-PAGE and analyzed by Western blot with anti-XPB antibody. The Western blot with anti-p62 (TFIIH) antibody gave identical results (data not shown). I, input proteins; U, unbound proteins; B, bound proteins. The diffuse bands in the 25-30 kDa region of lanes 5 and 10 are GST-XPA products which react non pecifically with anti-XPB antibodies.

GST-p56E
XPA binds to p34 subunit of TFlIE but not to the p5G-p34 (TFIIE) complex. The individual subunits of TFIIE (p56E and p34E) were expressed separately or co-expressed in the in vitro transcription-translation system and incubated with 10 IJ-g ofGST or 4 IJ-g of GST-XPA linked to glutathione-agarose beads. The unbound and bound proteins were analyzed by SDS-PAGE and visualized by autoradiography. I, input proteins; U, unbound proteins; B, bound proteins.
by either anti-p56E or anti-p34E antibodies, carry out normal excision repair. In contrast, the transcription activity in the immunodepleted extract was reduced by two-thirds (data not shown). Hence, it appears that TFIIE or its individual subunits are not involved in transcription-independent (general) excision repair.
XPA Directly Recruits TFlIH to Damage Site-Having found that the XPA-p34(TFIIE) interaction is unlikely to playa role in general excision repair, we wished to know how TFIIH is recruited to the lesion site. Previous work has shown that under stringent conditions, of the tested excision repair proteins, only the ERCC1·XPF complex binds to the XPA affinity column (Park and Sancar, 1994). We used milder conditions to detect the interaction ofXPA with TFIIH either directly or via interaction with other repair or transcription proteins. Fig. 5 shows that when HeLa CFE is incubated with the XPA resin, XPB is specifically retained under these assay conditions. Since previous studies have shown that XPB is in a complex with XPD (Reardon et al., 1993) in TFIIH (Drapkin et al., 1994a;Schaffer et al., 1994) we conclude that XPA recruits TFIIH specifically. To establish whether this recruitment is direct or through another protein, we incubated TFIIH purified to nearhomogeneity with the XPA resin. As is seen in Fig. 5 the TFIIH is specifically retained by the XPA resin. Thus, we conclude that XPA interacts directly with TFIIH to recruit it to the repair complex.
The Carboxyl-terminal Domain of XPA Binds TFIIH-To further support the specificity of the XPA-TFIIH interaction, we generated a number ofXPA deletion derivatives and tested their interaction with TFIIH, and with ERCC1 as a control. The results shown in Fig. 6 reveal that ERCC1 (in the form of ERCC1·XPF complex) binds near the E cluster in the aminoterminal region ofXPA, confirming the results ofLi et al. (1994) and that TFIIH binds to the carboxyl-terminal region of XPA. In fact, it appears that the carboxyl-terminal 48 amino acids are sufficient for specific binding (Fig. 6B, lane 8). Since a previous study had shown that XPA point mutations C261S or C264S interfere with XPA function in vivo (Miyamoto et al., 1992), we wished to know whether these residues are important for interaction with TFIIH. A GST-XPA fusion carrying the double mutations (GST-XPA2CS) was constructed and tested for binding. Fig. 6, lane 12, shows that the mutant binds normally to ERCC1 but does not bind to TFIIH. Thus, the FIG. 4. Immunodepletion of TFlIE from HeLa cells does not affect its excision activity. HeLa CFE was depleted of TFIIE using anti-p56E or anti-p34E antibodies linked to protein A-agarose beads. The immunodepleted CFE (50 IJ-g) was tested in Western blotting and excision assay. A, Western blot after immunodepletion. The immunodepleted CFE was resolved by SDS-PAGE and analyzed by Western blot with anti-XPB, anti-p56E, and anti-p34E antibodies. B, excision assay with immunodepleted HeLa CFE. The immunodepleted CFE was tested for excinuclease activity. 28 indicates the major excision product.  carboxyl-terminal 48 amino acids of XPA are sufficient for binding TFIIH, and C261 and C264 within this region play crucial roles in this process.
Role of XPA-TFIIH Interaction in Excision-To provide a functional test for the XPA-TFIIH interaction detected by the pull-down assay, we conducted two types of experiments. First, both wild-type MBP-XPA and the double mutant MBP-XPA2CS (C261S/C264S) were purified (Fig. 7) and used to complement an XP-A mutant CFE. Fig. 8 shows that the wildtype but not the mutant protein is capable of complementing the XP-A CFE, consistent with the notion that the XPA-TFIIH interaction is necessary for excinuclease activity. The specificity of this interaction was further supported by the following experiment. Increasing amounts of either wild-type or mutant MBP-XPA protein were added to HeLa CFE and the excision assay was carried out. The wild-type protein had only mild inhibitory effect while the mutant strongly inhibited excision (Fig. 9). We interpret these results as follows: the complex of XPA-ERCC1'XPF-TFIIH formed with intrinsic proteins is the most efficient in excision repair; that formed with MBP-XPA is less efficient and hence causes mild inhibition. In contrast, MBP-XPA2CS binds to the ERCC1·XPF complex normally but cannot bind to TFIIH and hence it sequesters the ERCC1·XPF complex in a nonproductive XPA2CS-ERCC1·XPF complex which cannot interact with TFIIH and thus leads to severe inhibition ofrepair.
Protein-DNA Complexes Formed with XPA, ERCCl·XPF, and TFIIH Proteins-The results presented thus far are consistent with the following model. XPA binds to damaged DNA and recruits ERCC1·XPF and TFIIH multiprotein complexes to help assemble the excinuclease. We conducted gel-mobility shift assay to test this model directly. First, in agreement with published reports (Robins et al., 1991;Guzder et al., 1993; The indicated amounts ofMBP-XPA or MBP-XPA2CS protein were added to the reaction mixture containing 25 IJ-g of HeLa CFE and the excision reaction was conducted under tandard conditions. A, autoradiographic analysis of excision products. Lane 1, no addition; lanes 2-4,0.8, 1.2, and 1.6 IJ-g ofMBP-XPA, respectively; lanes 5-7,0.8,1.2, and 1.6 IJ-g of MBP-XPA2CS, respectively. B, quantitative analysis of inhibition. The excision products were quantitated by PhosphorImager (Molecular Dynamics). Jones and Wood, 1993) XPA bound with higher affinity to damaged DNA, although we observed less discrimination between damaged and undamaged DNA compared to previous studies. More significantly, when the ERCC1·XPF complex purified to near homogeneity (Mu et al., 1995) was added to the XPA-DNA complex, a "supershift" was observed in the XPA-DNA band indicating formation of an XPA-ERCC1·XPF-DNA complex (data not shown). TFIIH alone caused some band shift; however, its addition to the XPA-DNA or to the XPA-ERCC1·XPF-DNA complexes did not cause a supershift (data not shown). Thus, these gel-mobility shift assays were not able to demonstrate the presence of an XPA-ERCC1·XPF-TFIIH-DNA complex either because such a complex does not form on the pathway to excinuclease assembly or, more likely, because such a complex is unable to survive the electrophoretic conditions. It is quite possible that the XPA-TFIIH-DNA interaction is quite transient and may not be detectable by standard gel retardation experiments.

DlSCUSSlO
Recent research in several laboratories has revealed that TFIIH contains XPB and XPD as two of its 8-9 subunits and that the entire factor, but not a subassembly of it, is an essential component of nucleotide excision repair (Drapkin et al., 1994a;Schaeffer et al., 1994;Wang et al., 1994). In fact, since TFIIH is required only for some promoters (Parvin et al., 1992;Parvin and Sharp, 1993;Tyree et al., 1993) yet is absolutely required for any excision repair, it might be more appropriate to consider it as a repair factor first and as a transcription factor second. In transcription, TFIIH is recruited to the promoter by TFIIE (Maxon et al., 1994;Zawel and Reinberg, 1995) and its helicase, ATPase, and RNA Pol II carboxyl-terminal domain kinase activities are modulated by TFIIE (Lu et al., 1992;Drapkin et al., 1994a;Ohkuma and Roeder, 1994). Therefore, it was hypothesized that TFIIH may be recruited to the excision complex by TFIIE. Surprisingly, we found that not to be the case. General excision repair occurs normally in the absence of TFIIE and XPA plays a crucial role in recruiting TFIIH to the damage site. However, it must be pointed out that the two assays used in this study, Western blotting for TFIIE and the excision assay for excinuclease activity, may be differentially sensitive to the presence ofTFIIE. In other words, it is possible that sufficient TFIIE for excision repair remains following immunodepletion even though the remaining amount is not seen in the Western blot. To eliminate this possibility we tested the immunodepleted extract for RNA Pol II transcription and found that its transcription activity was severely reduced. In addition, we tested our recently reconstituted excinuclease system for TFIIE activity in a defined transcription system. We found that the defined excision repair system contained less than 5% of the TFIIE activity in a TFIIEtrFIIHdependent transcription system and that the TFIIH from the defined excision repair system gave full TFIIH transcription activity when the transcription system was supplemented with TFIIE (data not shown).
Thus, taken together, these data indicate that either alone or possibly in combination with other XP proteins, XPA is a functional substitute of TFIIE in excision repair. However, in contrast with the TFIIE-TFIIH interactions which appear to be mediated by binding ofXPB to both subunits ofTFIIE (Drapkin et al., 1994a;Maxon et al., 1994), the subunit ofTFIIH binding to XPA is not known. Only three subunits ofTFIIH (XPB, XPD, and p62) were available to us for testing for potential interaction and these subunits did not bind to XPA. It is possible that one of the other subunits mediates XPA-TFIIH binding or that the binding interface is created by several subunits. Further experiments with the other subunits ofTFIIH will resolve this issue.
The second specific interaction detected in our experiments, the XPA-p34 (TFIIE) binding, remains to be explained at present. It is conceivable that this interaction may playa role in coupling transcription to repair. Based on the interaction of TFIIE with XPBIERCC3, Maxon et al. (1994) proposed a model which involves TFIIE-TFIIH interaction in transcription-repair coupling. Even though a defined system for RNA Pol II FIG. 10. Model for excision repair in humans. The damage recognition protein XPA binds to the lesion (i) and interacts with HSSB to make a stable complex. This complex recruits the ERCC·(XPF) complex in an ATP-independent reaction (ii) and the TFIIH factor in a reaction requiring ATP hydrolysis (iii). The XPC and XPG proteins are recruited with TFIIH and at one point XPE presumably enters the complex. The DNA is partially unwound and bent by the limited helicase activity of TFIIH. This distortion enables XPG to make the 3' incision and the ERCC1'XFP complex makes the 5' incision (iv). Following the dual incisions the replication proteins displace the excinuclease subunit (v), fill-in and seal the gap (vi). transcription has been available for some time, a defined system for excision repair has only recently been developed (Mu et ai., 1995) and hence this model and several others that have been proposed by other groups (see Hanawalt and Mellon (1993) and Drapkin et al. (1994b» can now be tested under specific conditions. However, the following observations are pertinent to this question. First, mutations in XPB and XPD which are components of TFIIH (Schaeffer et al., 1993(Schaeffer et al., , 1994Drapkin et al., 1994a) and in XPG which is loosely associated with TFIIH (Mu et al., 1995), often but not always, are associated with Cockayne's syndrome (CS) which is biochemically characterized by lack of excision repair of transcribed sequences. Second, eSA and CSB mutations which have only minor effects on excision repair are totally defective in gene Acknowledgments-We thank David Hsu for constructing the plasmid pGEX18. We are grateful to D. Reinberg for his help in design execution and interpretation of experiments relating to the role of TFIIE and for critical comments on the manuscript and to C, P. Selby for testing our purified excision repair system for TFIIE activity in transcription assays.  (Venema et al., 1990). The CSBlERCC6 gene has been cloned and it corrects the defect in gene specific repair and is thought (perhaps in conjunction with CSA protein) to act as a transcription-repair coupling factor (Troelstra et al., 1992). Third, as reported in this paper, XPA bound to a lesion is necessary for directing TFIIH to the site in transcription-independent excision repair. Currently, it is difficult to propose a transcription-repair coupling mechanism that incorporates all these, at times contradictory, observations. In the past, various groups including ours, have proposed that XPBIERCC3, CSBI ERCC6, or TFIIE play a role analogous to that of the Mfd protein in E. coli which displaces stalled RNA Pol from the lesion site and targets the damage recognition subunit to the site of damage (Selby and Sancar, 1993). Clearly all of these models cannot be correct. We believe that an explicit model can be made only after reconstitution of a defined system capable both of excision repair and transcription. However, the data presented in this paper along with the data on the functional yeast homologs of XPG protein (Rad 2) and the ERCC1·XPF complex (Rad lOll) enables us to propose the following working model for human excinuclease system (Fig. 10). XPA binds to the damage site and forms a stable complex with HSSB (Coverley et al., 1992;Mu et al., 1995); then it recruits the ERCC1·XPF complex and the TFIIH factor with the associated proteins XPC and XPG. The XPE protein may enter the assembly at some point. In the assembled complex XPG makes the 3' incision (Habraken et ai., 1993;Harrington and Lieber, 1994;O'Donovan et al., 1994)3 and ERCCIIXPF makes the 5' incision (Tomkinson et al., 1993;Bardwell et al., 1994;Habraken et al., 1994). The replication proteins PCNA (proliferating cell nuclear antigen) (Nichols and Sancar, 1992;Shivji et al., 1992), RFC (replication factor C), and Pol e8 dissociate the postincision complex, release the excised oligomer, and fill-in the gap which is then ligated.