Definition of a Short Region of XPG Necessary for TFIIH Interaction and Stable Recruitment to Sites of UV Damage

Abstract

XPG is the human endonuclease that cuts 3′ to DNA lesions during nucleotide excision repair. Missense mutations in XPG can lead to xeroderma pigmentosum (XP), whereas truncated or unstable XPG proteins cause Cockayne syndrome (CS), normally yielding life spans of <7 years. One XP-G individual who had advanced XP/CS symptoms at 28 years has been identified. The genetic, biochemical, and cellular defects in this remarkable case provide insight into the onset of XP and CS, and they reveal a previously unrecognized property of XPG. Both of this individual's XPG alleles produce a severely truncated protein, but an infrequent alternative splice generates an XPG protein lacking seven internal amino acids, which can account for his very slight cellular UV resistance. Deletion of XPG amino acids 225 to 231 does not abolish structure-specific endonuclease activity. Instead, this region is essential for interaction with TFIIH and for the stable recruitment of XPG to sites of local UV damage after the prior recruitment of TFIIH. These results define a new functional domain of XPG, and they demonstrate that recruitment of DNA repair proteins to sites of damage does not necessarily lead to productive repair reactions. This observation has potential implications that extend beyond nucleotide excision repair.

XPG is a key protein in nucleotide excision repair (NER), the highly conserved and versatile DNA repair pathway that removes UV damage and many other helix-distorting lesions from mammalian DNA (18, 27). The damage is recognized, an open DNA structure is formed around the lesion, dual incisions are made in the damaged DNA strand, an oligonucleotide of 24 to 32 nucleotides (nt) is excised, and the resulting gap is filled and sealed by a DNA polymerase and ligase, thereby restoring the DNA to an undamaged state. A subpathway of this process is known as transcription-coupled NER (TC-NER). It shares many common components with NER, but it removes damage from active genes more rapidly (6, 29), specifically from the transcribed DNA strand (30). Some 30 or so proteins are involved in these processes (2, 27).

A current model suggests that XPG acts early in NER, being one of six core factors needed for the damage recognition and dual-incision steps. The XPC-HR23B heterodimer binds to distorted DNA sites (4, 43, 49), TFIIH locally unwinds the DNA around the lesion by using its two helicases, XPB and XPD, in an ATP-dependent manner (13, 14). TFIIH, XPG, RPA, and XPA then form a stable preincision complex that apparently lacks XPC-HR23B (42, 43). Recruitment of TFIIH or XPG to the damage site does not depend on XPA, but XPA is required for the subsequent recruitment of ERCC1-XPF (54). Incisions are made 5′ and 3′ to the lesion, but it is not yet clear in which order or whether they are concurrent. The 3′ cut is made by XPG, a structure-specific endonuclease that cleaves the damaged strand at or near the junction between the 3′ end of single-stranded DNA and the 5′ end of double-stranded DNA (28, 39). The 5′ cut is made by ERCC1-XPF, another structure-specific endonuclease with opposite polarity (48).

XPG also has at least two structural roles in NER. It contributes to the stability of the preincision complex (8, 13, 14), and, even if catalytically inactive, it needs to be positioned correctly in the complex to permit ERCC1-XPF to cleave 5′ to the lesion (8, 34, 55). XPG does not rapidly dissociate after the dual incisions (43), so it may also help to recruit proteins involved in the gap-filling step (PCNA, RFC, DNA polymerase δ or ɛ, and DNA ligase I [2]), possibly through its PCNA-interacting domain (15, 56).

This multifaceted protein has been implicated in several other biological processes. XPG stimulates the in vitro activity of NTH1, the DNA glycosylase/AP lyase that removes oxidized pyrimidines in a base excision repair pathway (5, 23). The cofactor mechanism is not yet understood, but it does not require the endonuclease function of XPG (23). The Saccharomyces cerevisiae Rad2 protein appears to be required for efficient RNA polymerase II transcription in vivo (25). Since Rad2 and XPG are orthologs in most respects, XPG may have a comparable transcriptional role. XPG has also been implicated in transcription-coupled repair of oxidative DNA damage, but this issue is now controversial (7).

XPG derives its name from xeroderma pigmentosum (XP), an autosomal recessive human disease that is characterized by a very high incidence of skin cancers in sun-exposed regions and, in some cases, progressive neurological degeneration. This disorder can arise through mutations in any of seven genes (XPA to XPG) encoding NER proteins. The group G form is very rare, and only four XP-G individuals with typical but very mild XP symptoms have been reported (12, 20, 36). Defects in the XPG gene can also give rise to a second disorder, Cockayne syndrome (CS), whose features include dwarfism, mental retardation, retinal atrophy, deafness, and cranial calcification but no skin cancer predisposition (35). Most CS patients have no XP connection and instead carry mutations in the CSA and CSB genes, but XP/CS symptoms are also found in XP-B and some XP-D individuals (26). The onset of CS in XP-G/CS patients is particularly early and severe, with reported life expectancies ranging from only months to less than 7 years (12, 16, 22, 33, 53, 57).

A common pattern that can explain how XPG mutations give rise to mild XP or to severe early-onset XP/CS has emerged. Mildly affected XP-G individuals can generate from one XPG allele the full-length protein of 1186 amino acids but with single missense mutations that inhibit its endonuclease function (8, 12, 37, 38). A missense mutation has also been found in one allele of an XP-G baby with CS features (57), but this particular substitution (P72H) greatly destabilizes the protein (F. Thorel, unpublished data). All other examined XP-G/CS individuals produce truncated proteins from both XPG alleles (12, 38, 40; A. Pigni, unpublished data). Since NER is barely detectable in mild XP-G cases, the truncated and/or unstable XP-G/CS proteins are thought to be defective in another important XPG function, thereby leading to the severe XP-G/CS phenotype (38). Support for this view has come from complementary mouse studies. Xpa^−/− and Xpg^−/− mice are totally NER defective yet Xpg^−/− mice undergo premature death (17), whereas Xpa^−/− mice live for considerably longer (9). Moreover, mice expressing severely truncated XPG have retarded growth and a short life span (46), whereas mice carrying point mutations that are expected to inactivate XPG endonuclease function are viable but UV sensitive (46, 50).

These mouse models clearly are useful, but much of our current understanding of NER has come about through the availability of cells from compliant XP and CS individuals and their families. To cite just two examples, the hallmark of CS cells is their inability to carry out TC-NER (51). The CSA and CSB proteins are therefore needed for this process, even though the mechanistic details are not yet understood. Conversely, TC-NER works normally in XPC-deficient cells, thereby showing that XPC is dispensable for this process (52). Further insight is likely to be gained by studying more XP/CS individuals, particularly those with unusual properties. For these reasons, we have applied a battery of cellular and molecular techniques to cells and recombinant proteins from a highly unusual XP-G/CS individual, XPCS1BD, whose life expectancy is at least fourfold greater than that of other members of this group. The results suggest an explanation for his unusual longevity, they provide novel information about XPG-TFIIH interactions and recruitment to sites of UV damage, and they define a new functional domain of XPG.

MATERIALS AND METHODS

Cell strains and culture.

Fibroblasts and lymphoblastoid cells were cultured as described previously (8, 24). The primary fibroblasts used were XPCS1BD (patient), XP20BE (from a severely affected XP-G/CS individual ([33]), GM01630 (XP-A), XP11BE (XP-B), GM00671 (XP-C), XP1BR (XP-D), XP3BR (XP-G), and 250BR (wild type). Simian virus 40-transformed fibroblasts were from XP-G/CS patient XPCS1RO (10), who was previously known as 94RD27 (16, 38). Lymphoblastoid cells were repair-competent Raji cells and LBL463 from XP-G patient XP3BR (3, 8, 24).

DNA repair studies.

DNA repair studies were performed by using established procedures (16, 24, 53). Briefly, unscheduled DNA synthesis (UDS) was assayed in homo- and heterokaryons after fusion with Sendai virus of two XP cell strains that had previously been labeled with plastic beads of different sizes. Recovery of DNA synthesis was assessed 16 h after exposure to 254-nm UVC by a 2 h [³H]thymidine pulse-label. Sensitivity to UVC was measured 5 days after UV exposure of sparsely seeded fibroblast cultures. The number of proliferating cells was assessed by scintillation counting of [³H]thymidine incorporated in a 2.5-h pulse-label. UVC sensitivity of transfected lymphoblastoid XP3BR cells was assayed by Alamar Blue fluorescence (37, 45).

Mutation analyses.

Mutations were identified by previously described methods (37, 38). Briefly, poly(A)⁺ RNAs were isolated from XPCS1BD primary fibroblasts, reverse transcribed, and amplified with specific primers to obtain XPG cDNA, which was then cloned and sequenced. Nucleotide changes identified in several cDNA clones were then confirmed by sequencing the appropriate regions of XPCS1BD genomic DNA and, when feasible, by restriction analyses.

Expression and purification of recombinant His-tagged XPG proteins.

pFastBac1-XPG-His₆ and pFastBac1-E791A-His₆ plasmids were generated by fusing a six-histidine tag to the C terminus of XPG cDNA as previously described (19). The 1,310-bp HindIII-NdeI cDNA fragment encompassing the Δ225-231 deletion was excised from plasmid pBS-Δ225-231 and was used to replace the corresponding fragments (1,331 bp) in pFastBac1-XPG-His₆ or pFastBac1-E791A-His₆ to generate pFastBac1-Δ225-231-His₆ and the double mutant pFastBac1-Δ225-231::E791A-His₆.

Recombinant His-tagged XPG proteins were expressed in Sf9 insect cells, were purified by consecutive step elutions from Ni-nitrilotriacetic acid resin (Qiagen) and hydroxyapatite columns (Bio-Gel HT; Bio-Rad), and then concentrated over Centricon YM-30 membranes (Millipore). A detailed protocol is available on request.

Structure-specific endonuclease assay.

The substrate was a “bubble” formed by annealing two 90-mer oligonucleotides with a central 30-nucleotide unpaired region (39). One strand was previously 5′ labeled with polynucleotide kinase and [γ-³²P]ATP. The bubble substrate (10 ng) was incubated with increasing amounts of purified XPG proteins (6.6, 20, and 60 ng), and the products were analyzed on denaturing 12% polyacrylamide gels (8).

NER incision assays.

Covalently closed circular DNA containing a site-specific 1,3-intrastrand d(GpTpG)-cisplatin cross-link (31) was incubated in 9-μl reaction mixtures with purified proteins. Each reaction mixture contained 50 ng of RPA, 22.5 ng of XPA, 10 ng of XPC-HR23B complex, 25 ng of ERCC1-XPF complex, 1.5 μl of HeLa TFIIH (heparin fraction III), and 66 ng of purified wild-type or mutant tagged XPG in repair buffer (50 mM HEPES-KOH [pH 7.8], 70 mM KCl, 5 mM MgCl₂, 0.5 mM dithiothreitol, 0.3 mM EDTA, 2 mM ATP, 0.36 μg of bovine serum albumin [BSA] per μl, 0.02% Nonidet P-40, 3.4% glycerol). Repair proteins were preincubated for 10 min at 30°C, 50 ng of DNA containing the single cisplatin adduct was then added, and incubation was continued for a further 90 min. Following heat inactivation (5 min at 95°C), repair reaction mixtures were digested with the restriction enzyme XhoI for 1 h at 37°C. Dual incisions were detected by annealing 5 ng of oligonucleotide B5GB, a 34-mer complementary to the region around the adduct, to dual-incision products, thereby creating an overhang of four G residues (47). 3′ and 5′ incisions were detected by annealing 5 ng of B5′IN, a 38-mer (5′-GATAGCGGGGGGGCAGGAAACAGCTATGACCGAATTCC-3′) complementary to the region flanking the XhoI restriction site 44 nt 3′ to the cisplatin adduct, to excised repair products, creating in this case an overhang of six G residues. Repair products were labeled with 0.1 U of Sequenase version 2.0 polymerase (U.S. Biochemicals) and 1 μCi of [α-³²P]dCTP and were separated on denaturing 14% polyacrylamide gels.

Cell transduction with lentiviral XPG recombinants.

Recombinant lentiviruses containing wild-type or mutant XPG cDNAs under the control of the EF1α promoter were produced by transient cotransfection of three plasmids into 293T cells as described previously (44). Supernatants containing the lentiviral XPG recombinants were harvested at 35 h posttransfection. Immunoprecipitation experiments used the pWIR-CD8 vector containing mouse CD8 cDNA as selectable marker. For immunofluorescence studies, the green fluorescent protein cDNA in the pLOX/EWgfp vector was replaced by wild-type or mutant XPG cDNAs. The lentivirus vectors and protocols are described at the website of D. Trono (htpp://www.tronolab.unige.ch).

Immunoprecipitation and immunoblotting.

XPCS1RO cells were transduced with lentivectors containing an inverted XPG cDNA (as a negative control) or with cDNAs encoding wild-type XPG or the XPGΔ225-231 mutant. Cells were lysed for 1 h on ice in lysis buffer (25 mM HEPES-KOH [pH 7.9], 100 mM KCl, 17% glycerol, 12 mM MgCl₂, 1 mM EDTA, 1 mM dithiothreitol) containing 1× complete protease inhibitors (Roche). Lysates were clarified by centrifugation, and the protein concentration was determined. Goat anti-mouse immunoglobulin G (IgG) M-450 Dynabeads (Dynal) were washed with lysis buffer and were precoupled to the 8H7 anti-XPG monoclonal antibody (13) or to an anti-cdk7 monoclonal antibody (MO1-1; Novocastra Laboratories) for 1 h at 4°C (1). Identical amounts of total protein extracts were incubated with precoupled XPG beads or cdk7 beads for 2 h at 4°C. Beads were collected by using a magnetic particle concentrator (Dynal MPC) and were washed four times in washing buffer (25 mM HEPES-KOH [pH 7.9], 150 mM KCl, 10% glycerol, 0.01% Triton X-100) containing 1× complete protease inhibitors (Roche). Beads were resuspended in 2× loading buffer and were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and immunoblotting.

Samples were fractionated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to Immobilon P membranes (Millipore). The membranes were then immunoblotted with the mouse monoclonal anti-XPG 8H7 antibody (at a 1/1,000 dilution), with a rabbit polyclonal anti-p89 (XPB) antibody (at a 1/1,000 dilution), with a mouse monoclonal anti-p44 antibody (at a 1/500 dilution), or with the mouse monoclonal anti-cdk7 (at a 1/500 dilution). Detection was performed with peroxidase-conjugated anti-mouse or anti-rabbit IgGs (Promega) (at a 1/5,000 dilution). Bands were visualized by chemiluminescence (ECL; Amersham).

Local UV irradiation and immunofluorescence.

Lentiviral recombinants were added to growing cells seeded on glass coverslips. After 24 h, the supernatant was replaced by fresh medium and the cells were incubated for a further 2 days. They were then subjected to local UV irradiation as described previously (54). Briefly, cells were washed with phosphate-buffered saline (PBS) and were covered with a polycarbonate Isopore membrane filter (8-μm pore size; Millipore). Cells were irradiated through the filter by using a TUV 6-W lamp (UVGL-58; Omnilab) at a UVC dose rate of 0.5 J/m² per s, which was monitored at 254 nm with a UV radiometer (UVX-25; Omnilab). The filter was removed, and the cells were incubated in original medium and then fixed at 30 min or 24 h postirradiation.

Fixation was done with 3% formaldehyde-0.2% Triton X-100 in PBS for 15 min at 4°C. Subsequent steps were carried out at room temperature. Cells were washed in PBS and equilibrated in PBS containing 1% bovine serum albumin. Double labeling of XPG and TFIIH was carried out with the 8H7 mouse monoclonal antibody (at a 1/600 dilution) and a rabbit polyclonal antibody (at a 1/500 dilution) against XPB. Detection was performed with a goat anti-mouse IgG coupled to fluorescein isothiocyanate (Molecular Probes) and a Cy3-conjugated goat anti-rabbit IgG (Jackson) (both at a 1/600 dilution). For detection of cyclobutane pyrimidine dimers (CPDs), DNA was denaturated with 70 mM NaOH in 70% ethanol for 3 min and then labeled with mouse monoclonal antibody TDM-2 (32) at a 1/2,500 dilution. All primary and secondary antibodies were diluted in PBS-1% bovine serum albumin and were incubated, respectively, for 1 h or 30 min at room temperature. DNA was stained with DAPI (4′,6′-diamidino-2-phenylindole) (Molecular Probes) during secondary antibody incubation.

RESULTS

XPCS1BD is an XP-G/CS individual with very low UV resistance.

At 15 years of age, patient XPCS1BD, a Spanish male, was diagnosed as having CS. Cutaneous features of XP gradually developed, and at 24 years he was diagnosed with severe XP/CS. A detailed case report was written when the patient was 28 years old and is available on request. Unfortunately, no later trace exists of XPCS1BD or of the doctor in charge of his case.

The combination of XP and CS symptoms in XPCS1BD suggests that he has defects in one of three human genes, XPB, XPD, or XPG. To establish which gene is affected, complementation assays were performed in which UV-induced UDS was measured in heterokaryons obtained by cell fusion. Normal repair levels were restored after fusion of XPCS1BD fibroblasts with XP-B or XP-D cells but not with XP-G cells (Fig. 1A). This result assigns the patient to XP group G.

FIG. 1.

DNA repair characteristics of XPCS1BD primary fibroblasts. (A) XPCS1BD cells were fused with XP-B, XP-D, or XP-G fibroblasts. Repair activity was then measured as UV-induced UDS and is shown as a percentage of that in normal cells (± standard error of the mean). (B) Recovery of DNA synthesis 16 h after exposure of wild-type (WT) and XPCS1BD fibroblasts to various UV doses. Typical responses of XP-C, XP-A, and CS-B fibroblasts are indicated by the dotted lines. (C) UV-survival curves of wild-type and XPCS1BD fibroblasts assayed by [³H]thymidine incorporation. The shaded areas indicate previously reported survival characteristics of primary fibroblasts from two XP-G and four XP-G/CS individuals (24).

XPCS1BD fibroblasts had a reduced ability to recover from UV-induced inhibition of DNA synthesis; they were much more affected than XP-C fibroblasts but not as debilitated as XP-A or CS-B cells (Fig. 1B). In addition, their post-UV survival (Fig. 1C) was intermediate between those of two mildly affected XP-G siblings and four severely affected XP-G/CS individuals (24). These results indicate that XPCS1BD carries mutations in each XPG allele that lead to reduced repair of UV damage.

XPCS1BD is homozygous for a splice acceptor site mutation.

To determine the nature of this mutation(s), XPCS1BD total RNA was reverse transcribed, amplified by PCR with XPG-specific primers, and cloned in a plasmid vector. Sequencing of several clones revealed that they all lack the G at position 870 in the XPG cDNA (data not shown). This creates a frameshift that results in a truncated protein of 242 amino acids, with the last 18 being unrelated to XPG.

Interestingly, the missing G is not due to a deletion at the genomic DNA level but rather to a G-to-A transition in the last nucleotide of the preceding intron (Fig. 2A). This creates a new AG splice acceptor site 1 nt downstream, with the G being the first nucleotide of the next exon. Aberrant splicing at this new acceptor site thus accounts for the absence of G870 from the XPG mRNA and cDNA. The G-to-A transition in intron 6 is expected to create a new MseI restriction site (T/TAA). Digestion of genomic PCR fragments confirmed this prediction and further revealed that the patient is homozygous for this mutation (Fig. 2C). XPCS1BD thus produces the severely truncated E225fsX243 XPG protein (Fig. 2B).

FIG. 2.

Mutations in XPCS1BD and their functional consequences. (A) Top line: the normal splicing of the intron (intron 6) situated between positions 869 and 870 of XPG mRNA. Middle line: in patient XPCS1BD, the indicated G-to-A transition at the last nucleotide of this intron provokes aberrant splicing. This removes the first G of the downstream exon (exon 7), thereby causing a frameshift and a premature stop at position 243. Bottom line: an infrequent alternative splice removes 21 nt from exon 7 and puts the downstream XPG mRNA sequences back into frame. (B) Top line: schematic of wild-type (WT) XPG of 1,186 amino acids, showing the locations of the conserved N and I regions, nuclear localization signal (NLS), and basic C terminus (45). Remaining lines: schematic of XPG proteins made in XPCS1BD. Both XPG alleles generate a truncated protein of 242 amino acids, with the last 18 being unrelated to XPG. Both alleles also generate XPG proteins lacking seven internal amino acids with a downstream Cys-Ser substitution. (C) The G-A transition creates a new MseI site that permits the digestion of a genomic PCR product of 83 bp into two fragments of 43 and 40 bp. The patient XPCS1BD (BD) is homozygous for this mutation, which is absent from patient XPCS1RO (94RD27 [94]). RT-PCR generates the expected product of 203 bp from both patients and, in the case of XPCS1BD, a minor product of 182 bp that arises from alternative splicing. M: MseI digestions; −: undigested; L: length marker of HaeIII digest of pBluescript SK(−). (D) Lymphoblastoid LBL463 cells from patient XP3BR were transfected with the episomal EBO-pLPP vector containing the indicated mutant XPG cDNAs. Transfected or untransfected cells were irradiated with increasing doses of 254-nm UV light, and their survival 48 h later was assayed with Alamar Blue, a dye that is rendered fluorescent by living cells. Each point represents the mean from four samples. Standard deviations are shown by error bars.

XPCS1BD produces an XPG protein lacking seven internal amino acids.

Inspection of the reverse transcription-PCR (RT-PCR) products from this region of XPCS1BD XPG mRNA revealed the presence of a minor band somewhat shorter than the major product (Fig. 2C). This band was absent from the RT-PCR products of XP-G/CS patient XPCS1RO, indicating that it is specific for XPCS1BD (Fig. 2C). When the minor band was cloned and sequenced, it was found to be missing the 21 nt from position 870 to 890 (data not shown). We infer that this arises by alternative splicing to a cryptic acceptor site in the downstream exon (Fig. 2A). As a consequence, the RT-PCR product of this spliced mRNA lacks a HinfI restriction site. From the relative intensities of the HinfI-digested RT-PCR products, we estimate that this alternative splicing accounts for ∼5% of XPG mRNA in XPCS1BD cells (data not shown). Its important consequence is that the downstream mRNA sequences are now back in frame so that the resulting XPG protein is almost full length and lacks just seven amino acids from positions 225 to 231 (Fig. 2A and B).

XPCS1BD cells contain a second XPG mutation, a G-to-C transversion at position 1783, which is predicted to destroy an RsaI restriction site (G/TAC to CTAC). This was confirmed experimentally, which also revealed that patient XPCS1BD is homozygous for this transversion (data not shown). At the protein level, this changes Cys529 to serine (Fig. 2B).

XPCS1BD truncated and internally deleted XPG proteins are unable to restore UV resistance in vivo.

To determine if these various changes affect the ability of XPG to restore UV resistance to XP-G cells, we subcloned the corresponding cDNAs into an episomal expression vector and transfected them into lymphoblastoid cells from XP3BR, an XP-G individual with some CS features. His cells encode XPG proteins of 521 and 980 amino acids (24), and they are very UV sensitive (8, 24). We then examined the ability of the transfectants to resist increasing doses of UV.

As expected, the cDNA encoding the premature stop codon at position 243 did not restore UV resistance to the XP3BR cells, but neither did the cDNA corresponding to the infrequent alternative splicing event (Fig. 2D). To determine whether it is the lack of the seven internal amino acids and/or the missense mutation that is responsible for inactivating the protein, we tested them independently. Transfectants carrying the Ser529 XPG cDNA were as UV resistant as those encoding Cys529 XPG (37) whereas the internal 7-amino-acid deletion destroyed XPG complementation activity (Fig. 2D). These results suggest that C529S is a rare polymorphism with no particular functional significance, consistent with the GeneSNPs finding that this substitution is present in ≤10% of several hundred unrelated individuals (http://www.genome.utah.edu/genesnps). Nevertheless, for all subsequent experiments we retained the more common Cys529 in XPG constructs lacking the seven internal amino acids. This protein with the internal deletion is hereafter referred to as XPGΔ225-231.

XPGΔ225-231 retains structure-specific endonuclease activity.

To determine the consequences of the Δ225-231 deletion on XPG incision activity in vitro, recombinant wild-type and mutated XPG proteins were purified from baculovirus-infected Sf9 cells. To facilitate their purification, the XPG proteins were tagged at the C terminus with a His₆ tag. They were then purified by sequential Ni⁺-agarose and hydroxyapaptite chromatography.

Intrinsic structure-specific endonuclease activity was assayed with a radiolabeled substrate containing a centrally unpaired region of 30 nt flanked by two duplex regions of 30 bp. In agreement with earlier reports (8, 13, 39), purified untagged wild-type XPG cleaved this bubble substrate close to the 3′ junction between single- and double-stranded DNA, generating a labeled fragment of 61 nt (Fig. 3A, lanes 2 to 4). The C-terminal His-tagged versions of wild-type and Δ225-231 XPG also specifically cleaved on the 3′ side of the unpaired region with efficiencies comparable to that of the untagged wild-type protein (Fig. 3A, lanes 5 to 7 and 8 to 10). Hence, neither addition of a C-terminal His₆ tag nor removal of the 7 amino acids from position 225 to 231 disrupts the ability of XPG to act as a structure-specific endonuclease.

FIG. 3.

Endonuclease activities of His-tagged XPG proteins in isolation and in a reconstituted NER system. (A) Equivalent increasing amounts of the indicated purified XPG proteins were incubated with a bubble substrate that was 5′ labeled on one strand. Reaction products were analyzed on a denaturing 12% polyacrylamide gel. Structure-specific endonuclease activity releases a 61-nt labeled oligonucleotide. WT, wild type. (B) Schematic of the substrate and probes used in the reconstituted NER system. Probe B5GB is complementary to the 24- to 32-nt products of the dual incisions on either side of the cis-Pt adduct. Probe B5′IN detects 3′ incisions by XPG or, only if these do not occur, uncoupled 5′ incisions by ERCC1-XPF. These various products are labeled by fill-in reactions and resolved on denaturing gels. (C) Products of the reconstituted NER system containing XPC-HR23B, TFIIH, XPA, RPA, ERCC1-XPF, and either no XPG or the indicated His-tagged XPG proteins. Dual, 3′, and uncoupled 5′ incisions are indicated by brackets. Only wild-type XPG permits 3′ and dual incisions, and only E791A XPG permits uncoupled 5′ incisions. Bands marked with an asterisk are nonspecific artifacts that are not found in reactions lacking either substrate or primer.

XPGΔ225-231 displays no activity in a reconstituted NER system.

The in vitro endonuclease activity of XPGΔ225-231 seems surprising because this protein is unable to restore UV resistance to transfected XP-G cells in vivo (Fig. 2D). To try to resolve this apparent paradox, we examined the activity of the tagged XPG proteins in a reconstituted NER system comprising highly purified XPC-HR23B, XPA, RPA, ERCC1-XPF, TFIIH, and a DNA substrate with a uniquely located cisplatin adduct (31). Incisions 3′ to this adduct by XPG, and 5′ cuts by ERCC1-XPF, were detected by end labeling (47). Specifically, dual-incision products were annealed to the B5GP oligonucleotide (Fig. 3B) and were extended with Sequenase and [α-³²P]dCTP. Similarly, 3′ incisions and uncoupled 5′ incisions were detected with the B5′IN oligonucleotide, which is complementary to a region downstream of the major and minor XPG incision sites (Fig. 3B).

No products of either 3′ or 5′ cuts were observed when XPG was omitted from the reaction mixtures (Fig. 3C, lanes 1 and 6). In contrast, in the presence of the other purified incision components, tagged wild-type XPG generated dual and 3′ incision products (Fig. 3C, lanes 2 and 7) that are identical to those produced with the untagged wild-type protein (reference 31 and data not shown). Hence, the His₆ tag does not interfere with either the formation or the enzymatic activity of the NER incision complex in vitro. However, tagged XPGΔ225-231 was unable to promote either 3′ or 5′ cuts or, therefore, dual incisions (Fig. 3C, lanes 4 and 9). To investigate this further, we turned to E791A, a catalytic-site mutant of XPG that was previously shown to be inactive as a 3′ endonuclease in NER but permitted ERCC1-XPF to make 5′ incisions (8). Consistent with this earlier work, the His-tagged version of E791A was unable to make 3′ incisions, but it did permit uncoupled 5′ incisions in the reconstituted system (Fig. 3C, lanes 3 and 8). This uncoupled 5′ incision activity was lost, however, when the E791A mutation and the Δ225-231 deletion were present on the same XPG molecule (Fig. 3C, lane 10). This strongly suggests that, despite its intrinsic structure-specific endonuclease activity (Fig. 3A), XPGΔ225-231 is unable to carry out either its normal enzymatic or structural functions in the presence of the other NER incision components. This, in turn, strongly suggests that amino acids 225 to 231 are involved in a key interaction with another NER component.

The Δ225-231 deletion impairs the interaction between XPG and TFIIH.

One candidate for such an XPG partner is TFIIH. Coimmunoprecipitation experiments have suggested that several TFIIH subunits interact with two regions of XPG (21), the most N terminal of which includes the region missing in Δ225-231. In addition, the XPG-TFIIH interaction is sufficiently strong so that immunoprecipitation with a TFIIH-specific monoclonal antibody brings down both TFIIH and XPG, which are active in NER reactions lacking these two components (1).

To determine whether the Δ225-231 deletion compromises the ability of XPG to interact with TFIIH, we transduced simian virus 40-transformed fibroblasts from XP-G patient XPCS1RO (10) with lentiviral recombinants expressing wild-type XPG, XPG in its antisense orientation as a negative control, or XPGΔ225-231. XPG was immunoprecipitated from the cell extracts with 8H7, a mouse monoclonal antibody to XPG (13). TFIIH was detected by immunoblotting with antibodies to three of its subunits, XPB and p44 of core TFIIH and the cdk7 subunit of its cdk-activating kinase. Recombinant wild-type and Δ225-231 XPG proteins were found in comparable amounts in the anti-XPG immunoprecipitates (Fig. 4A, lanes 1 and 2). Under the same conditions, no signal was seen in immunoprecipitated extracts of cells transduced with the inverted XPG construct (Fig. 4A, lane 3). The TFIIH subunits XPB, p44, and cdk7 were coimmunoprecipitated with recombinant wild-type XPG (Fig. 4A, lane 1). By contrast, only trace amounts of these TFIIH subunits were found in the XPGΔ225-231 immunoprecipitates, and none at all were found in the inverted control (Fig. 4A, lanes 2 and 3).

FIG. 4.

Coimmunoprecipitation of TFIIH with wild-type XPG but not with XPGΔ225-231. Immortalized fibroblasts from XP-G patient XPCS1RO were transduced with lentiviral recombinants containing wild-type (WT) or Δ225-231 XPG cDNAs or, as negative control, wild-type XPG cDNA in the inverted orientation. Cell extracts were immunoprecipitated with monoclonal antibodies to (A) XPG or (B) cdk7, both at physiological salt levels (150 mM KCl). The immunoprecipitates were then analyzed by Western blotting with antibodies to XPG and the XPB, p44, and cdk7 subunits of TFIIH.

In a complementary experiment, TFIIH was immunoprecipitated from the cell extracts with a mouse monoclonal antibody to the cdk7 subunit. As expected, all three extracts yielded comparable amounts of cdk7, p44, and XPB in the immunoprecipitates (Fig. 4B, lanes 4 to 6). Wild-type XPG was also efficiently brought down with the anti-cdk7 monoclonal antibody (Fig. 4B, lane 4). However, very little of the Δ225-231 XPG protein was found in the anti-cdk7 immunoprecipitate, and, as expected, the inverted control was negative for XPG (Fig. 4B, lanes 5 and 6). Quantification of these and similar blots shows that TFIIH was present at 2 to 5% in the XPGΔ225-231 immunoprecipitates and that ∼2% of XPGΔ225-231 was present in the cdk7 immunoprecipitates, compared to the wild-type XPG controls. These results demonstrate that the interaction between XPG and TFIIH is severely destabilized by the deletion of XPG amino acids 225 to 231.

TFIIH is recruited to UV lesion sites in the absence of XPG.

To determine the in vivo consequences of this weakened interaction, we induced UV damage locally and used immunofluorescence to monitor the recruitment of XPG and TFIIH (XPB) to damage sites. Primary fibroblasts from XP-G/CS patient XP20BE (33) were subjected to local UV irradiation (54). These cells encode severely truncated XPG proteins of 10 and 137 amino acids (40; A. Pigni, unpublished data) that were not detected by Western blotting or immunofluorescence (data not shown), whereas XPG was readily detectable by immunofluorescence in wild-type 250BR fibroblasts (Fig. 5A). Strikingly, TFIIH (XPB) colocalized with CPDs after local UV irradiation of XP20BE cells (Fig. 5A). TFIIH recruitment to in vivo sites of UV damage therefore does not require the presence of XPG.

FIG. 5.

Recruitment of various forms of XPG and TFIIH (XPB) to local UV damage sites in XP-G/CS cells. (A) XP-G/CS (XP20BE) fibroblasts were exposed to 100 J of UV per m² through 8-μm-pore-size filters and 30 min after irradiation were immunolabeled with antibodies against CPDs (green), the XPB subunit of TFIIH (red), and XPG (green). Normal (250BR) fibroblasts were immunolabeled at 30 min after irradiation with the antibody against XPG (green) as a positive control. (B) XP-G/CS (XP20BE) fibroblasts were transduced with the indicated XPG cDNAs in lentiviral recombinants. The transductants were either mock irradiated and immediately fixed or exposed to 100 J of UV per m² through 8-μm-pore-size filters and fixed 30 min or 24 h after irradiation. The cells were then immunolabeled with antibodies against XPG (green) and TFIIH (XPB) (red). Merged images (yellow) indicate colocalization. WT, wild type. Bars, 10 μm.

XPGΔ225-231 is recruited to UV lesion sites.

To examine recruitment further, wild-type or mutated XPG cDNAs were cloned into a lentiviral vector and the recombinants were transduced into XP20BE cells. Just as in normal cells (54), recombinant wild-type XPG and endogenous TFIIH colocalized in nuclear foci within 30 min of local UV irradiation. Also as in normal cells, the XPG and TFIIH immunofluorescent signals then became broadly dispersed throughout the nucleus within 24 h (Fig. 5B, top row), indicating that repair of the photolesions had occurred.

Recombinant XPGΔ225-231 and endogenous TFIIH (XPB) were also recruited to nuclear foci within 30 min of local UV irradiation of transduced XP20BE cells. However, they were still found at such local damage sites at 24 h postirradiation (Fig. 5B, Δ225-231 row). Hence, the 7-amino-acid deletion in XPGΔ225-231 does not stop its recruitment to UV lesions, but it appears to prevent their repair. Recombinant E791A XPG and the double mutant XPGΔ225-231::E791A behaved in the same way in transduced XP20BE cells. Each was found together with TFIIH at local UV damage sites both 30 min and 24 h after irradiation (Fig. 5B, bottom two rows). Thus, these three mutant XPG proteins that are unable to make NER 3′ incisions in vitro (Fig. 3) also seem to be unable to promote DNA repair in vivo. However, all three migrated normally to sites of UV damage. The migration of XPGΔ225-231 is of particular interest because it demonstrates that initial recruitment does not depend on a strong XPG-TFIIH interaction.

Recruitment of XPGΔ225-231 to UV lesion sites depends on XPC but not on XPA.

To determine whether other factors that are known to act early in NER are needed for XPGΔ225-231 recruitment, we transduced primary fibroblasts from XP-C and XP-A patients with the wild-type XPG and XPGΔ225-231 lentiviral recombinants. In transduced XP-C cells, neither wild-type XPG nor the mutant with the internal deletion was recruited to local UV lesions (data not shown). Hence, overexpression of recombinant XPG does not induce artifactual recruitment of NER factors to sites of local UV damage. The result with recombinant wild-type XPG is consistent with earlier work suggesting that the damage recognition factor XPC-HR23B is needed for the subsequent recruitment of all other NER factors (54). The failure of XPGΔ225-231 to be recruited demonstrates that the internal deletion does not obviate the need for prior recruitment of XPC.

By contrast, both wild-type XPG and XPGΔ225-231 were found at UV lesion sites in transduced XP-A cells (data not shown). This also corroborates earlier work showing that XPG recruitment does not require XPA (41), and the loss of seven internal amino acids from XPG does not make its recruitment XPA dependent.

RPA is the other member of the stable NER preincision complex (42, 43). Given its crucial roles in DNA replication as well as repair, no RPA-deficient cells exist, so it is not yet known if recruitment of either wild-type XPG or XPA is dependent on RPA. However, RPA recruitment does not require the presence of either XPG or XPA (41). The available evidence thus suggests that XPA, RPA, and XPG can be recruited independently of each other and that XPC is the cornerstone for the subsequent recruitment of all other NER factors.

The recruitment of XPGΔ225-231 to UV lesion sites is unstable.

Although XPGΔ225-231, E791A XPG, and the double mutant XPGΔ225-231::E791A have impaired repair capacities in vivo (Fig. 5B), they may be weakened in different ways that depend on how well they are positioned within the incision complex. To examine the stability of their recruitment to UV lesions in the presence of endogenous XPG, we transduced the same series of lentiviral recombinants into normal (250BR) primary fibroblasts and then immunostained for XPG and TFIIH (XPB) at 30 min and 24 h after local UV irradiation.

For these experiments, the anti-XPG 8H7 antibody was diluted so that the endogenous XPG was detectable only when it was enriched at local UV damage sites. Under these conditions, yellow foci corresponding to the colocalization of XPG and TFIIH were detected at 30 min postirradiation. These foci disappeared within 24 h post-UV treatment, indicating that XPG and TFIIH are dispersed again within the nucleus following repair (Fig. 6A, top row). In unirradiated transduced cells, overexpressed XPG proteins were visible throughout the whole nucleus.

FIG. 6.

Displacement of XPGΔ225-231 from local UV damage sites in normal cells. (A) Normal (250BR) fibroblasts were either not transduced or transduced with the indicated XPG cDNAs in lentiviral recombinants. The cells were either unirradiated or exposed to 100 J of UV per m² through 8-μm-pore-size filters and at 30 min or 24 h after irradiation were immunolabeled with antibodies against XPG (green) and TFIIH (XPB) (red). Only the merged images (yellow) are shown. WT, wild type. Bars, 10 μm. (B) Normal (Raji) lymphoblasts were transfected with the indicated XPG cDNAs in EBO-pLPP recombinants. Transfectants were irradiated with increasing doses of 254-nm UV light, and their survival 48 h later was assayed by Alamar Blue fluorescence. Each point represents the mean from four samples. Standard deviations are shown by error bars.

As expected, XPG and TFIIH were found in local UV damage sites within 30 min when recombinant wild-type XPG was expressed in wild-type cells; these foci had disappeared after 24 h, indicating that DNA repair at these sites was complete (Fig. 6A, WT row). XPGΔ225-231 was also recruited to local damage sites within 30 min of UV irradiation in these transduced normal cells. However, in contrast to its behavior in XP-G/CS cells, XPGΔ225-231 then was found broadly dispersed throughout the nucleus at 24 h post-UV irradiation (Fig. 6A, Δ225-231 row). This suggests that, with time, small amounts of endogenous XPG can displace the overexpressed XPGΔ225-231 from the damage sites and permit their repair.

By contrast, foci persisted at 24 h postirradiation in wild-type cells transduced with the E791A XPG cDNA (Fig. 6A, E791A row). This result indicates that once recruited to lesion sites, the XPG E791A cannot be displaced by endogenous wild-type XPG, thereby preventing DNA repair. The double mutant protein carrying both the Δ225-231 deletion and the E791A mutation was also recruited to lesion sites soon after local UV treatment, but foci were not found at 24 h postirradiation (Fig. 6A, bottom row). This suggests that the Δ225-231 deletion destabilizes the interactions of XPG E791A at lesion sites. The double XPG mutant can then be replaced by the endogenous wild-type XPG, thereby allowing normal DNA repair.

Consistent with these results, repair-proficient Raji lymphoblastoid cells expressing XPG E791A were threefold more UV sensitive than cells transfected with vector only or with wild-type or Δ225-231 XPG cDNA recombinants (Fig. 6B). Cells expressing D77A, another mutant XPG protein with no 3′ endonuclease activity (8), also possessed UV sensitivity comparable to that of E791A transfectants (data not shown). These results indicate that the E791A and D77A catalytic-site mutants are able to compete with the endogenous XPG for stable recruitment to UV damage sites. By contrast, the incorporation of the Δ225-231 deletion into both E791A and D77A XPG proteins restored normal UV resistance to the Raji transfectants (Fig. 6B and data not shown). These results demonstrate that the Δ225-231 deletion interferes with the stable recruitment of XPG to UV damage sites, thereby allowing endogenous XPG access to the lesions and their subsequent repair.

DISCUSSION

Alternative splicing accounts for the atypical delayed onset of CS and XP symptoms in XPCS1BD.

The XP-G/CS individual XPCS1BD is remarkable because he was first diagnosed at 15 years with CS and was diagnosed only at 24 years with XP. He was last examined at 28 years, when he presented advanced XP/CS clinical symptoms. His life span is thus >4-fold greater than that of any other reported XP-G/CS individual (reviewed in the introduction and in reference 7). The question naturally arises of what caused his unusual longevity. The results presented here suggest that the explanation lies in an infrequent alternative splicing event that makes use of a cryptic acceptor site in the adjoining downstream exon. This removes 21 nt from the XPG mRNA, which, when translated, yields a protein of 1,179 amino acids that lacks the seven amino acids from XPG position 225 to 231 (Fig. 2). This is the first example of a rare splicing event that allows prolonged survival of an XP/CS individual.

A low level of normal XPG transcript splicing is sufficient for viability.

The homozygous XPCS1BD mutation creates a suboptimal 3′ splice site that, in turn, increases the probability of alternative splicing to the cryptic downstream acceptor site. This alternative splice may not occur if intron 6 contains its usual GT-AG boundaries. Indeed, it is not found in XPCS1RO, an unrelated XP-G individual (Fig. 2C). Nevertheless, there is increasing evidence for alternative splicing of XPG transcripts contributing to the heterogeneity of XP-G clinical phenotypes. For example, XP3BR produced not only truncated XPG proteins but also a protein of 1,185 amino acids containing 44 internal non-XPG residues through alternative splicing at a rare noncanonical AT-AC intron (24). In XP20BE, a patient with severe XP-G/CS (33), the XPG alleles produced proteins of just 10 and 137 amino acids (40; A. Pigni, unpublished data). Given the severity of these truncations, it seems highly likely that alternative splicing produced some partially active XPG protein that contributed to the 6-year life span of this patient.

Alternative splicing of XPG transcripts is also well documented in normal as well as XP-G individuals (11, 37, 40). A common alternatively spliced XPG mRNA isoform lacks 55 nt and encodes a protein of only 302 amino acids (11, 37). Surprisingly, this isoform is the major XPG mRNA species in some normal individuals (T. Nouspikel, unpublished data). These results indicate that a low level of normally spliced mRNA is sufficient for correct XPG function.

Small amounts of stable XPG protein are sufficient for viability.

The above conclusion is corroborated by data on XPG protein levels in XP-G individuals. One missense mutation, P72H in patient XPCS4RO, renders XPG so unstable as to be undetectable by immunoblotting (57; F. Thorel, unpublished data). The other XPG allele produced a truncated protein of 175 amino acids. This patient died at 11 months (57). By contrast, two very mildly affected XP-G siblings who are now in their thirties, XP125LO and XP124LO, produce XPG of 959 amino acids from one allele and full-length XPG with an A792V substitution from the other (37, 38). As determined by immunoblotting, the A792V substitution reduces XPG stability to ∼20% of normal (38), and it severely impairs but does not entirely abolish XPG endonuclease activity (8). These results lead to the important general conclusion that XP-G cells and patients can survive with a functionally compromised XPG protein that is present in amounts much smaller than normal. More specifically for XPCS1BD, they suggest that the rare alternative splice permits sufficient production of the XPG protein with the internal deletion to delay the onset of XP and CS symptoms in this individual. These considerations further suggest that the amounts of normal XPG mRNA and of normal XPG protein can vary widely between individuals without markedly affecting the propensity to develop skin cancer.

NER incision components can inhibit or reveal the catalytic activity of mutated XPG proteins.

In isolation, XPGΔ225-231 cuts a bubble substrate in vitro as efficiently as wild-type XPG (Fig. 3A). Its retention of intrinsic structure-specific endonuclease activity is not entirely unexpected, because the 7-amino-acid deletion is located between the N and I regions (45). These two domains are highly conserved in the extensive FEN-1/Rad2/XPG nuclease family and are thought to juxtapose to form the endonuclease active site (8).

By contrast, when in the presence of the other NER incision components, XPGΔ225-231 makes no detectable 3′ incisions, nor does it permit ERCC1-XPF to make 5′ cuts (Fig. 3C). Moreover, although the XPG E791A active-site mutant does allow uncoupled 5′ incisions (8) (Fig. 3C), this activity is lost when XPG contains both E791A and Δ225-231 (Fig. 3C). These results suggest that the internal deletion impairs the structural functions of XPG in NER. These, in turn, inhibit expression of the intrinsic endonuclease activity of the XPGΔ225-231 protein.

Interestingly, XPG proteins with D77E or E791D substitutions have directly opposite properties. In isolation, they are unable to cut the bubble substrate, but they are highly active in the dual-incision assay and in vivo (8). The nature of the mutation thus determines whether the other NER incision components will inhibit or reveal the endonuclease activity of mutated XPG proteins.

The internal deletion in XPCS1BD severely disrupts XPG-TFIIH interaction.

A major consequence of the Δ225-231 deletion is the disruption of an interaction between XPG and the DNA repair-transcription factor TFIIH (Fig. 4). Previous work has shown XPG to interact with several TFIIH subunits (21) and to do so in a sufficiently stable way to permit complementation in a reconstituted NER system lacking XPG and TFIIH (1). Given the small size of the internal deletion, just 7 amino acids, the magnitude of the disruption of the XPG-TFIIH interaction is both impressive and surprising. It will be of interest to determine which TFIIH subunit interaction(s) is affected by this short internal deletion.

Nonfunctional proteins can be recruited to damage sites.

A strong XPG-TFIIH interaction is not needed for the recruitment of either protein to sites of local UV damage. TFIIH is recruited normally to such sites in XPG-deficient cells (Fig. 5). This corroborates in vitro evidence suggesting that TFIIH is recruited to damage sites soon after XPC-HR23B and before other incision components (43, 49). Despite its weakened interaction with TFIIH, XPGΔ225-231 is also recruited normally to UV-damaged sites (Fig. 5). But it is recruited in an unstable manner, and it can be displaced by wild-type XPG (Fig. 6). In contrast, the E791A active-site mutant of XPG cannot be so dislodged (Fig. 6). Hence, XPG does not require its endonuclease function to be recruited to damage sites, but it does need to interact normally with TFIIH to be productively used at such sites.

The recruitment of XPGΔ225-231 to UV-induced foci raises the issue of what is actually happening at these foci. A few minutes after local UV irradiation, CPDs and 6-4 photoproducts are found by immunostaining only in discrete foci within the nucleus; these foci are also the only sites of intense nonreplicative DNA synthesis (54). A few hours later, 6-4 photoproducts are no longer detectable in discrete foci (54). These results demonstrate that UV-induced foci contain the two major photolesions, and the subsequent dispersion of foci strongly suggests that these photolesions have been repaired. Until now, it has been tacitly assumed that the recruitment of NER proteins to these UV-induced foci reflects the assembly of stable and productive DNA repair complexes. The recruitment to such foci of nonfunctional XPGΔ225-231 challenges this assumption. That proteins can exist within a focus without necessarily being engaged in productive reactions is an observation with potential implications that extend beyond NER to other repair pathways and to other aspects of DNA metabolism.