Nucleotide Excision Repair Eliminates Unique DNA-Protein Cross-links from Mammalian Cells*

DNA-protein cross-links (DPCs) present a formidable obstacle to cellular processes because they are “superbulky” compared with the majority of chemical adducts. Elimination of DPCs is critical for cell survival because their persistence can lead to cell death or halt cell cycle progression by impeding DNA and RNA synthesis. To study DPC repair, we have used DNA methyltransferases to generate unique DPC adducts in oligodeoxyribonucleotides or plasmids to monitor both in vitro excision and in vivo repair. We show that HhaI DNA methyltransferase covalently bound to an oligodeoxyribonucleotide is not efficiently excised by using mammalian cell-free extracts, but protease digestion of the full-length HhaI DNA methyltransferase-DPC yields a substrate that is efficiently removed by a process similar to nucleotide excision repair (NER). To examine the repair of that unique DPC, we have developed two plasmid-based in vivo assays for DPC repair. One assay shows that in nontranscribed regions, DPC repair is greater than 60% in 6 h. The other assay based on host cell reactivation using a green fluorescent protein demonstrates that DPCs in transcribed genes are also repaired. Using Xpg-deficient cells (NER-defective) with the in vivo host cell reactivation assay and a unique DPC indicates that NER has a role in the repair of this adduct. We also demonstrate a role for the 26 S proteasome in DPC repair. These data are consistent with a model for repair in which the polypeptide chain of a DPC is first reduced by proteolysis prior to NER.

of the lesion is generally large, even compared with many bulky chemical adducts (e.g. benzo[a]pyrene diol epoxide or aflatoxin B 1 ) (1-7). Some investigations suggest that DPCs kill cells based on the arrest of DNA and RNA synthesis at these "superbulky" adducts (8 -12). Therefore, the establishment of DPCs in an uncontrolled manner can prove lethal or perhaps mutagenic (13), whereas production of DPCs targeted to tumor cells can yield therapeutic benefit (14 -19).
Much research on DPCs has targeted their formation using proteomic techniques following exposure to chemical-or radiation-induced damage (20 -25). Although the treatment of cells with chemicals or radiation induces damage that is not limited to DNA, the formation of specific DPCs is generally less than 1-3% of the total damage (25,26). Thus, the dilution of DPC effects in the milieu of other cellular damage creates a difficult system in which to elucidate answers concerning repair. An additional aspect of the difficulty in the study of DPCs formed by damaging agents is that there is not a unique adduct for this class of damage. DPCs can occur on a single DNA strand, on both strands, and/or with single or multiple attachment sites. Variations in attachment of the protein to DNA could mean that multiple repair mechanisms are employed to eradicate the damage. This further complication, although not impossible to overcome, adds to the predicament of examining DNA damage and repair when only a small amount of damage exists.
One method to examine DPC repair without the presence of other DNA damage is by making unique DPCs. DNA glycosylases with associated lyase activity are trapped on DNA at an abasic site following chemical reduction of a Schiff base intermediate using sodium borohydride (27)(28)(29)(30)(31)(32)(33)(34)(35). Such substrates have afforded a first examination of in vitro DPC repair in bacterial systems. One example of this is the trapping of the T4 UV endonuclease DNA glycosylase (T4pdg) at an abasic site, covalently linking that 16-kDa protein to DNA. Using an incision assay, the Escherichia coli UvrABC excision nuclease cleaves the DPC at a 2-fold lower rate than that for benz[a]pyrene adducts (36). Another thermally stable bacterial nucleotide excision repair (NER) system also incises the T4pdg-DPC and some related polypeptides (37). This leads to the speculation that one pathway to repair DPCs in mammalian cells could be NER, and some initial in vitro work using that DPC supports that claim (38).
As mentioned above, DPCs are a significant obstacle to cellular defense mechanisms. The elimination of these superbulky adducts is a task that must be completed by cells to ensure the performance of normal biological functions, most notably replication and transcription. Although there is in vitro evidence for DPC removal by NER in bacterial systems, the role of mammalian NER in DPC repair remains a mystery. Resolution of formaldehyde-induced DPCs does not depend on xeroderma pigmentosum groups A and F, two NER components, in contrast to the other evidence (10,20,39,40). Therefore, mammalian cellular repair pathways must be carefully evaluated for the capacity to eliminate DPCs. The use of specific damage substrates is a method that can clarify some of the discrepancies in this field.
To simplify the interpretation of data, we focused on the repair of a single well defined DNA methyltransferase-DPC that is formed under mild conditions and has relevance for understanding in vivo repair of such adducts. Many human DNMTs are inhibited by azacytosine, zebularine, and fluorodeoxycytosine (FdC) (19). Although azacytosine may directly inhibit human DNMTs (41), the enzymatic inhibition by FdC requires incorporation of the modified base into the DNA followed by formation of a covalent DNA-protein cross-link (42)(43)(44)(45)(46). The inhibition of DNMT activity is believed to sequester DNMTs in cells by the formation of a DPC. If left unrepaired, the DPC can result in cell death by obstructing DNA or RNA synthesis; therefore, it must be eliminated. To mimic these DPCs, we have used the bacterial HDnmt-DPC whose structure is known (47)(48)(49)(50). Although the GCGC recognition sequence is different from that of mammalian DNMTs, HDnmt methylates the 5-position of cytosine in the central CpG sequence, and the DPC link is formed between the 6-position of cytosine linked to Cys-81, identical to the mechanism of mammalian DNMTs (Fig. 1a). The mild reaction conditions allow the study of DPC repair without introducing collateral damage to other moieties on either DNA or protein components. The construction of such a well defined substrate should facilitate the study of DPC repair pathways. We have exploited this defined substrate to explore the elimination of DPCs using both in vitro and in vivo experiments.
XPG Purification-XPG cDNA was amplified using forward (5Ј-TGCTCTAGAATGGGGGTCCAGGGGCTC-3Ј) and reverse (5Ј-CGGGGTACCTTAGGTTTTCCTTTTTCTTCC-3Ј) primers using PCR and cloned into pFastBac HT B (Invitrogen). Positive clones were confirmed by DNA sequencing. Recombinant virus was transformed into E. coli DH10Bac. Recombinant clones of 6030 bp were obtained and verified by PCR. Recombinant virus was obtained by transfecting Sf9 cells and generating subsequent concentrated viral stocks. 500 ml of Sf9 culture was infected using the baculovirus harboring the XPG cDNA, and incubation was continued for 5 days postinfection. The protein was partially purified using nickel-nitrilotriacetic acid chromatography.

Preparation of Substrates for in Vitro NER and DPC Repair-
The ODNs (50 pmol), 12 nts with unique damage sites, were 5Ј-phosphorylated with [␥-32 P]ATP (75 pmol, 6000 Ci/pmol) and T4 polynucleotide kinase (20 U) in a total volume of 25 l of buffer as recommended by the manufacturer for 2 h at 37°C, followed by incubation at 77°C for 5 min to inactivate the enzyme. In a single reaction, the ODNs needed for ligation (100 pmol each, excluding the two 5Ј-end ODNs; see Fig. 1b) were 5Ј-phosphorylated with 10 mM ATP, T4 polynucleotide kinase (25 U) with reaction conditions identical to those of the damage-containing ODNs (PP (6-4) and 8-oxoG). Phosphorylated ODNs were combined with the 32 P-labeled ODN, and the two 5Ј-phosphorylated ODNs, ethanol-precipitated using 25 g of glycogen as a carrier, were washed with 70% ethanol, dried, and resuspended in 7.5 l of 100 mM NaCl, 50 mM Tris-HCl, pH 7.9, 10 mM MgCl 2 , 1 mM dithiothreitol (Buffer 1). ODNs were annealed by heating at 77°C for 5 min and were allowed to gradually cool to 25°C over 3 h after which they were held on ice for at least 10 min. Ligation of the ODNs was performed using T4 DNA ligase (2.5 Weiss units), 3.5 l of 5ϫ ligation buffer in a 25-l final volume. The ligation reactions were incubated at room temperature for 45 min followed by 16°C overnight. The ligated ODNs were loaded onto a 6% denaturing sequencing PAGE. Full-length ODNs were isolated and electroeluted using a commercial apparatus (Bio-Rad Electroeluter model 422). The full-length ODNs were ethanol-precipitated using glycogen carrier, dried, and resuspended in Buffer 1. The ODNs were reannealed using the same protocol prior to ligation. The duplex ODNs were then purified using 5% native PAGE, electroeluted, ethanol-precipitated, and resuspended in 21 l of Buffer 1.
Formation of HDnmt-DPC Substrates-A covalent DPC substrate was formed between the duplex FdC ODN and the HDnmt in the presence of AdoMet. The reaction was performed in a buffer (50 mM Tris-HCl, pH 7.5, 10 mM EDTA, 5 mM 2-mercaptoethanol, 2 g/ml poly(dA-dT)⅐poly(dA-dT), 200 g/ml bovine serum albumin, 80 M AdoMet, 200 fmol of duplex ODN substrate, 50 U of HhaI methyltransferase (ϳ100 U/g) in 25 l) at 37°C for 2 h. All substrate concentrations for whole cell extract assays were determined by calculation from the disintegrations/min and the radioactive half-life.
Protease Treatment of DPCs-The HDnmt covalently attached to DNA in the DPC substrate described above was reduced in size using either chymotrypsin or proteinase K. The HDnmt-DPC substrate (7 l, 63 fmol) was incubated with 3.5 l of chymotrypsin (1.25 g/l, freshly suspended in 1 mM HCl) and buffer (100 mM Tris-HCl, pH 7.7, 10 mM CaCl 2 ) and was allowed to proceed at 25°C overnight. A 1-l aliquot was analyzed using 4 -15% SDS-PAGE. The remainder was ethanol-precipitated with glycogen, dried, and resuspended in Buffer 1. For the proteinase K treatment, the DPC substrate (7 l, 63 fmol) was incubated with proteinase K (1 l, 20 g/l) at 37°C for 20 min. Following the reaction, the volume was adjusted to 15 l using 10 mM Tris-HCl, pH 8.0, 1 mM EDTA, and a 1-l aliquot analyzed as for the chymotrypsin-treated substrate.
T7 RNA Polymerase Assay for DPCs-A duplex ODN substrate (2.0 pmol) with a unique FdC was cross-linked to HDnmt using the method described above. An aliquot (0.5pmol) was subjected to proteinase K treatment at 37°C for 30 min followed by incubation at 65°C for 25 min to inactivate the protease. The sample was ethanol-precipitated, dried, and resuspended in 20 l of 10 mM Tris-HCl, pH 8.0, 1 mM EDTA. The transcription reactions were conducted as recommended by the manufacturer in buffer with 10 mM dithiothreitol, 50 U of RNasin, 0.5 mM NTPs, 20 U of T7 RNA polymerase, 0.5 mol of duplex HDnmt-DPC or HDnmt-DPC proteinase K-treated substrate in a total volume of 50 l at 37°C for 4 h. At the end of the reaction, 25-l aliquots were mixed with an equal volume of formamide loading dye and analyzed on an 8% denaturing gel. The products in the gel were stained using 20 l of SybrGreen II in 200 ml of electrophoresis running buffer for 20 -30 min in the dark, and the fluorescence was imaged using UV transillumination and a digital video camera.
Excision Assay Using CHO-K1 Extracts-The excision reactions, 25-l total volume, were conducted in 18 mM HEPES-KOH, pH 7.9, 24 mM KCl, 1.92 mM MgCl 2 , 4 mM ATP, 5 g of bovine serum albumin, 35 ng of pBR322, 20 -80-fmol duplex ODN substrate, 50 g of CFE. Reactions were preincubated at 30°C for 10 min and incubated for an additional 45 min after the addition of the duplex ODN substrate. At the end of the reaction, the mixtures were treated with 20 g of proteinase K for 15 min at 37°C and extracted once with phenol and then phenol/chloroform to reduce the size of any polypeptide still bound to DNA fragments. The reaction products were ethanolprecipitated with 20 g of glycogen and analyzed using 10% DNA sequencing PAGE. Gels were enclosed in plastic wrap, exposed to a phosphor storage screen, and scanned using a Typhoon 9410 PhosphorImager (Amersham Biosciences).
pUC19DB10 Plasmid Construct-A modified pUC19 plasmid, pUC19EN3 (2332 bp), was obtained from Weiguo Cao (Clemson University) containing a single N.BstNBI site. A second N.BstNBI site and an HhaI site (5Ј-GCGC-3Ј) were added upstream of the original N.BstNBI site. Complementary oligomers (i), 5Ј-AATTCGAGTCATCGATTAGCGCTG-3Ј and (ii) 5Ј-GATCCAGCGCTAATCGATGACTCG-3Ј were ligated into the EcoRI and BamHI restriction sites in the polylinker region, selected for ampicillin resistance, and sequenced following growth of small scale cultures. This plasmid was designated pUC19DB10.
pTGFPHha10 Plasmid Construct-The pTurboGFPN plasmid (EvroGen, San Diego) was used to generate a plasmid with a green fluorescent protein-coding sequence that has a sequence with two N.BstNB I sites located on the transcribed strand downstream from the cytomegalovirus promoter. An ODN with the sequence 5Ј-AGATCTCGAGCTCAAGCTTT-GCGCAAGCGACTCCGAATTC-3Ј was inserted into the BglII and EcoRI site of the pTurboGFPN plasmid to generate a second N.BstNB I site within 30 nt on the transcribed strand of the pTurboGFPN plasmid (the N.BstNB I site is on the transcribed strand as indicated by the underlined sequence). To prepare the pTGFPN plasmid for the introduction of a unique damage site, 6 of the 7 N.BstNB I sites were eliminated using the Stratagene QuikChange kit for site-directed mutagenesis. In coding regions, the sites were mutated without changing the amino acid sequence in the proteins.
Preparation and Purification of pUC19DB10 or pTGF-PHha10 Plasmid DNA with a Unique FdC Site-The preparation of a unique damage site was essentially as described by Wang and Hays (63), but certain modifications were required to ensure that the cross-linked base was tritium-labeled. pUC19DB10 (467 g) was nicked to completion with N.BstNBI nicking endonuclease (375 U) to produce a 30-nt singlestranded ODN using 1ϫ NEBuffer 3 in a final volume of 1.5 ml and incubating at 55°C for 2 h. An aliquot analyzed on a 1% agarose gel confirmed that all the plasmid was converted from supercoiled to nicked form, indicating that all molecules had been nicked at least once. An additional 375 U of N.BstNBI were added, along with buffer, and the reaction was continued for an additional 4 h, after which the plasmid DNA was ethanol-precipitated.
Nicked pUC19DB10 DNA was methylated with HhaI methyltransferase by incubation at 37°C for 2.5 h (425 g of DNA, 85 M AdoMet, with 7.5 U of HDnmt/g pUC19DB10, in 4 ml of 50 mM Tris-Cl, pH 7.5, 10 mM EDTA, 5 mM 2-mercaptoethanol). Methylated plasmid DNA was challenged with HhaI restriction endonuclease and analyzed on a 1% agarose gel. Methylation was judged to be complete by lack of restriction.
The nicked, methylated plasmid was heated to 84°C for 5 min followed by gradual cooling to room temperature to release the 30-nt fragment and anneal it to the complementary 30-nt oligomer, 5Ј-ATCGGACTCACATGTGGATCCAGCGCT-AAT-3Ј, present in a 50-fold molar excess. Separation of the gapped plasmid and duplex 30-nt ODN was accomplished with a Centricon 100 spin filter (Millipore) using four buffer exchanges with the final retentate being in 10 mM Tris-Cl, pH  (42, 47, 48, 89 -91). b, ODNs to examine excision of DPCs. The substrates for excision assays were constructed by annealing, ligating, and purifying the appropriately kinased ODNs and the 32 P-labeled central ODN as described under "Experimental Procedures." The vertical arrows represent the ODNs that were used for the ligation. The ODN for the DPC has an FdC residue in the GFGC sequence, with F representing FdC. The opposite strand had a 5-methylcytosine at the site of HDnmt methylation as indicated by a filled circle. The control substrates used had unique 8-oxoG or PP (6-4) ODNs. The DPC and 8-oxoG substrates had a T7 RNA polymerase promoter site, which was used for mapping either the DPC or a break caused by removal of the 8-oxoG. Only the central duplex ODNs for the 8-oxoG containing ODN is indicated. The position of the 8-oxoG is indicated by the open circle. The PP (6-4) represents the substrates with the highest activity in vitro for the mammalian excision nuclease. The position of the PP (6-4) is indicated by the link between the two central thymine residues. c, purification of HDnmt for construction of the HDnmt-DPC. Proteins were separated using 4 -20% SDS-PAGE and imaged using Coomassie Brilliant Blue staining. d, EMSA of covalently linked HDnmt-DPC. The 139-bp ODN was separated using 4 -20% SDS-PAGE following the establishment of a covalent DPC between the FdC ODN and the HDnmt with AdoMet. As a control, PP (6-4) ODN (6,4 PP) was incubated with or without the HhaI methyltransferase. The FdC ODN with or without HDnmt is shown in the two right lanes. Prior to loading the gel, the samples were denatured for 5 min at 95°C. The position of the DPC and the unbound ODNs are indicated. The size in kDa of the 139-bp and the EMSA-shifted DPC are indicated on the left of the figure. Concentrations of ODNs were calculated from radioactive decay, and although the approximation is close, this method has associated error.
8.0, 1 mM EDTA. Approximately 400 g of the gapped, methylated pUC19DB10 was recovered at this point.
The gapped, methylated pUC19DB10 plasmid DNA was split into two fractions and annealed with a 4-fold molar excess of a 30-nt ODN, 5Ј-pATTAGXGCTGGATCCACATGTGAGTC-CGAT-3Ј, where X was either dC (control) or FdC. Annealing conditions were identical to those in elimination of the original 30-nt ODN. Following annealing, ligation reactions were performed using the following conditions: 190 g of plasmid DNA, 20 mM Tris-HCl, pH 7.6, 5 mM MgCl 2 , 10 mM 2-mercaptoethanol, 2 mM ATP, and T4 DNA ligase at 2 l/6 g of DNA in a total 1.2-ml final volume and aliquoted into 12 tubes, 100 l per tube. The reactions were carried out using a PTC-100 thermal cycler (MJ Research, Inc.) for 16 h alternating between 10°C for 30 s and 30°C for 30 s. Ligation products were analyzed on 1% agarose gels. Ligation of annealed pUC19DB10 produced 50 -60% supercoiled plasmid, whereas ligation of pUC19DB10 gapped plasmid produced no supercoiled DNA indicating that 100% of the N.BstNBI-cleaved molecules without the annealed ODN were gapped. The products, supercoiled and nicked plasmids, were separated by ultracentrifugation in cesium chloride (5.1-ml Quickseal tube containing 1.55 g/ml CsCl) and ethidium bromide (740 g/ml). Samples were centrifuged in a Beckman NVT 100 rotor at 68,000 rpm overnight at 20°C in a Beckman XL-100 ultracentrifuge. Supercoiled plasmid was collected, ethidium bromide was removed by butanol extraction, DNA was ethanol-precipitated, and an aliquot was analyzed on a 1% agarose gel. The plasmids were designated pUC19DB10dC and pUC19DB10FdC. Recovery at this point was ϳ50 g of each supercoiled plasmid.
The pTGFPHha10 plasmid with a unique FdC site was prepared in a similar manner to that of the pUC19DBFdC plasmid, except that no radiolabeled AdoMet was used in the preparation. This plasmid is pTGFPHha10FdC.
Formation and Characterization of a Unique, Tritiated, Covalent DPC in pUC19DB10FdC or the Control Plasmid pUC19DB10dC without a DPC-15 g of each plasmid (10 pmol) was methylated with HDnmt (74U/pmol pUC19DB10) in 150 l containing 10 M [ 3 H]AdoMet (TRK-236, Amersham Biosciences), 5 mM Tris-HCl, pH 7.5, 10 mM EDTA, and 5 mM 2-mercaptoethanol at 37°C for 2-3 h. Aliquots of each methyltransferase reaction (0.1 g of plasmid) were digested with restriction endonuclease BsaJI (1.25 U of BsaJI, and 1ϫ NEBuffer 2) at 37°C for 1 h followed by ethanol precipitation, and analyzed using 1% agarose gel electrophoresis to determine the extent of restriction endonuclease digestion. Digestion of  32 P labeling ODNs with T4 polynucleotide kinase. The bands at 30 nt corresponding to reported NER excision products are marked by a bracket. Note that following the excision assay proteinase K is added to reduce the size of the bound protein as described under "Experimental Procedures." All products were separated using 7 M urea, 10% PAGE prior to PhosphorImager analysis. Concentrations of ODNs were calculated from radioactive decay, and although the approximation is close, this method has associated error. a, excision assays using control substrates with either PP (6-4) (6,4 PP) or 8-oxoG adducts. Substrates were incubated with 50 g of Chinese hamster CFE with different femtomole amounts of substrate as indicated. b, excision assay using different femtomole amounts of the full-length HDnmt-DPC adduct. The substrate incubated with 50 g of Chinese hamster CFE. Note that only weak bands at 30 nt are observed for excision of the full-length HDnmt-DPC in b even after overexposure.  Fig. 1d. c, inhibition of T7 RNA polymerase transcription by either the HDnmt-DPC or PROK HDnmt-DPC. The right-hand side of the figure was enhanced to observe the band because of the halt of RNA synthesis from the PROK digestion. Note the slight shift in the band indicating transcription arrest when HDnmt is degraded by the PROK, but the synthesis shows that even with the small peptide remaining on the DNA, there is no extension. The marker positions for ODNs in nt are indicated on the left of the figure. d, excision assays using the full-length HDnmt-DPC, the CHT-HDnmt-DPC (CHT-DPC), and the PROK-HDnmt-DPC. M represents the markers in nt prepared as described in Fig. 2. The excision products around 30 nt in length are indicated. Note that following the excision assay proteinase K is added to reduce the size of the bound protein as described under "Experimental Procedures." The molecular masses of the products in d depend on whether the proteinase K treatment post-excision is performed on the full-length or truncated HDnmt-DPC. In the case of the HDnmt-DPC substrate, the proteinase K digestion of the product is a 37-kDa protein tethered to a single-stranded ODN, whereas for the CHT-and PROK-HDnmt-DPCs, the substrates are short oligopeptides (Ն10 amino acids) that are poorly degraded by the post-excision proteinase K treatment. Thus, the product of the HDnmt-DPC excision reaction is slightly smaller than the other two peptide products. The amount of the excision product for the HDnmt-DPC is also low ( Fig. 2b and Fig. 3d). Concentrations of ODNs were calculated from radioactive decay, and although the approximation is close, this method has associated error.
pUC19DB10 with BsaJI produces two fragments of 2091 and 259 bp that were end-labeled using the T4 polynucleotide kinase (10 U) exchange reaction in the presence of [␥-32 P]ATP (25 pmol, 6000 Ci/mmol) and 1ϫ NEB kinase buffer in a vol-ume of 10 l. Unincorporated ATP was removed with micro BioSpin 6 columns (Bio-Rad). Samples were heat-denatured for 5 min at 95°C and then analyzed on a 5% denaturing SDS-PAGE at 120 V for 1.75 h using prestained protein standards (Bio-Rad). Reaction products were imaged using the methods described above for the duplex ODN substrate.

In Vivo HDnmt-DPC Repair Assay Using [ 3 H]HDnmt-pUC19DB10FdC and the [ 3 H]pUC19DB10dC
Control-HTERTG telomerase-immortalized human skin fibroblasts were grown until they were 90% confluent on 100-mm diameter Petri dishes. Following the removal of growth medium, the HTERTG cells were transfected by the direct addition of a mixture of 60 l of Lipofectamine 2000 (Invitrogen) and 5 g of 3 H-labeled plasmid DNA in 3 ml of Opti-MEM1 reduced serum medium (Invitrogen). After 6 h of incubation, the cells treated with DNase I (50 l of 1 mg/ml stock) in the media for 20 min to remove any plasmid that bound to the exterior of the cells. The media were removed and washed with phosphate-buffered saline (64). Cells were trypsinized and frozen at Ϫ80°C until all samples were collected. The DNA was isolated using an alkaline lysis method and the recovered plasmid assayed by digestion with BsaJI followed by proteinase K digestion (20 g of total enzyme). The 2091and 259-bp bands were separated using 1% agarose gel electrophoresis, and the two bands were excised and quantified using scintillation spectroscopy.
In Vivo HDnmt-DPC Assay Using pTGFPHha10FdC-Either pTGFPHha10FdC or pTGFPHha10dC was transfected into 24-well plates with CHO-K1 control or uv135 cells. Briefly, 0.5 g of plasmid DNA and 50 l of Opti-MEM1 were mixed. In a separate tube, 2 l of Lipofectamine 2000 was mixed with 50 l of Opti-MEM1. Both solutions were incubated at room temperature for 5 min. The DNA and Lipofectamine 2000 mixtures were added together and mixed gently, allowed to stand for 20 min at room temperature, and added to cells that had 500 l of Dulbecco's modified Eagle's medium and 10% fetal calf serum media. At several time points, cells were photographed using an inverted phase fluorescence microscope (Nikon). For quantification, the media were removed from the cells, and the cells were washed with 200 l of phosphate-buffered saline solution (PBS). To the washed cells, 100 l of In all cases, M represents marker DNAs. A control pUC19DB10 with only dC was also prepared. a, introduction of nicks in the pUC19DB10 plasmid. After nicking with NBstNBI to create two nicks separated by 30 nt, the plasmid was electrophoresed in a 1% agarose gel. The bottom panel shows the generation of nicked DNA (ϩNBstNBI) in an agarose gel along with the markers and the starting plasmid DNA (ϪNBstNBI). The hatched region in the plasmid represents the DNA nicked by NBstNBI. b, exhaustive methylation of the nicked pUC19DB10 using HDnmt with unlabeled AdoMet. The bottom panel monitors the reaction using agarose gel electrophoresis. HDnmt methylation protects against cleavage by the HhaI restriction endonuclease, whereas the unmethylated NBstNBI is cleaved. The black filled circles represent HDnmt methylation sites. There are actually 24 total HDnmt sites in the pUC19DB10 plasmid. c, preparation of the pUC19DB10 plasmid with a unique FdC or dC in the DNA. The pUC19DB10 from b is heated to 84°C in the presence of an ODN with a sequence complementary to that of the N.BstNB I-generated ODN. The short duplex ODN is removed by Centricon 100 spin dialysis, and an ODN with a unique FdC or dC is annealed into the gapped DNA and ligated using T4 DNA ligase. The black region represents the ODN with the specific FdC or dC. The small gray box represents the FdC site in DNA. Note that the Gap lane is the plasmid whose 30-nt ODN was removed by denaturation, annealing with a complementary 30-nt ODN, and spin dialysis. The gapped pUC19DB10 was also treated with DNA ligase, and the failure to form a closed circular product indicates that there is a gap in the DNA. The ligated DNA is purified using CsCl equilibrium centrifugation to provide closed circular pUC19DB10FdC or pUC19DB10dC. d, labeling pUC19DB10FdC or pUC19DB10dC plasmids using tritiated AdoMet and HDnmt to generate either tritium-labeled DPC with the structure in Fig. 1a or dC (no gel is shown for this step). e, confirmation of the formation of a unique HDnmt-DPC. The pUC19DB10FdC and pUC19DB10dC plasmids were cleaved with BsaJI to produce 259-bp (85 kDa) and 2091-bp fragments. Both strands of the 259-bp fragment were 32 P-labeled using the exchange reaction of T4 polynucleotide kinase. As shown in the lower right panel, the 259-bp fragment of the FdC containing plasmid shows a gel shift to 122 kDa, which indicates that the pUC19DB10FdC plasmid DNA of that strand was quantitatively cross-linked to HDnmt. The 37-kDa HDnmt difference in the band subject to the electrophoretic mobility shift is indicated. For simplicity, only the 259-bp BsaJI fragments are shown and not the large plasmid fragment. 1ϫ trypsin (Invitrogen) and 100 l of PBS was added, and the cells were incubated at 37°C for 5 min. The original media and all the washing solutions were combined, and the cells were centrifuged at 1000 ϫ g for 5 min at room temperature. The cells were resuspended in 150 l of PBS, fixed using 150 l of 20% paraformaldehyde, and stored at 4°C until fluorescenceactivated cell sorting (FACS). FACS analysis was conducted on a Dako CyAN (Fort Collins, CO), using 488 nm excitation from an argon laser, and signals were collected using a 530 Ϯ 15 nm filter. At least 50,000 events were collected and analyzed using Summit 4.2 (Dako). All samples were analyzed in triplicate, and separate transfections to monitor transfection efficiencies were performed using the pTurboGFPN plasmid.
MG132 Treatment of CHO-K1 Cells-CHO-K1 cells were grown to 70% confluence in a 24-well plate. A stock solution of 0.2 mM MG132 in Me 2 SO was added to half of the wells on the plate to a final concentration of 0.5 M MG132 in each well. Only Me 2 SO was added to the remaining wells. The cells were incubated for 2 h and then transfected with 0.5 l of pTGFPH-haFdC with a HDnmt-DPC. The addition of the transfection mixture reduces the MG132 concentration to 0.412 M. Time points following transfection were collected at 12, 24, and 48 h and processed as indicated above.

Construction of DPC Substrate-Covalent attachment of
HDnmt to the FdC ODN substrate was performed in the presence of AdoMet. The substrates for NER excision assays are generally of low molecular mass (59,65,66), but the HDnmt-DPC substrate is 37 kDa. The DPC substrate backbone is assembled as for the smaller substrates (59,65,66) by inserting the 5Ј-32 P-FdC-containing ODN, but the final step is the coupling reaction of the HDnmt at Cys-81 to the FdC in the presence of AdoMet (Fig. 1, a-c). The stability of this linkage is manifested by the ability to heat the HDnmt substrate to 95°C and to subject the substrate to denaturing PAGE, yet the linkage remains intact (Fig. 1d). Therefore, the HDnmt-DPC is a stable, covalent connection that is formed under mild reaction conditions. As a control, a substrate with a PP (6-4) was also incubated with the HDnmt, but no cross-linking was observed by denaturing PAGE analysis.
Mammalian CFEs Excision Assay Using a Full-length HDnmt-DPC Substrate-Previous data using the bacterial excision nucleases have shown that T4 UV endonuclease linked via a reduced Schiff base DPC at an abasic site can be repaired by UvrABC excision nuclease (36,37). Therefore, we tested the capacity of mammalian CFEs to excise the HDnmt-DPC. As controls, we used substrates containing either PP (6-4) or 8-oxoG damage bases that are substrates for the mammalian NER repair system present in CFEs (60). The control substrates both show excised ODN bands that are just under 30 nt in length, consistent with NER products (Fig. 2a) (56, 67), and the intensities of the excision bands are dependent on substrate concentration. The HDnmt-DPC was then used to determine the capacity of CFEs to release excision products. Fig. 2b shows that the amount of the HDnmt-DPC released is very small compared with the amount of excision product observed for the two other substrates. The difference is striking between the PP (6-4) and 8-oxoG compared with the HDnmt-DPC. Comparable amounts of radioactivity were loaded in all cases, but (Fig. 2b) in order to observe excision products, prolonged exposure of the gel was necessary. Therefore, the 37-kDa HDnmt covalently linked to DNA through the 6-position of cytosine is a poor substrate for the mammalian excision nuclease. Fig. 2b, we asked whether proteolysis would render the HDnmt-DPC more accessible to the NER system. To test that hypothesis we generated two other substrates from the HDnmt-DPC substrate using chymotrypsin and proteinase K. Fig. 3a outlines the excision experiment using the full-length HDnmt-DPC, the CHT-HDnmt-DPC, and the PROK-HDnmt-DPC. Partial protection of the HDnmt in the CHT-HDnmt-DPC (ϳ65 kDa) is observed, but the digestion is essentially complete for the PROK-HDnmt-DPC substrate (Fig.  3b). To show that there is still a cross-link present even in the case of the PROK-HDnmt-DPC substrate, we used an assay based on T7 RNA polymerase transcription. The FdC ODN shown in Fig. 1b has a T7 RNA polymerase promoter. T7 RNA FIGURE 5. In vivo repair of DPCs in mammalian cells. a, outline of assay used for the in vivo determination of HDnmt-DPC repair. The FdC-or dC-containing pUC19DB10 plasmid is transfected into telomerase-transformed human skin fibroblast HTERTG cells. Note that the plasmid has no mammalian origin of replication, and therefore will not undergo DNA synthesis. The closed circular DNA is harvested using a standard alkaline denaturing protocol. The circular DNA isolated is subjected to BsaJI digestion to release 259-and 2091-bp fragments and treated with proteinase K. The fragments are separated on 1% agarose gels, the bands excised, melted, scintillation fluid added, and the ratio of the 259 to 2091 bands determined. This ratio is proportional to the repair of the HDnmt-DPC. b, agarose gel analysis of tritium-labeled plasmids from the repair assay that transformed either HDnmt-pUC19DB10FdC or pUC19DB10dC into telomerase-transformed human skin fibroblast HTERTG cells. All plasmids isolated were subjected to restriction enzyme digestion using BsaJI. The Control lane corresponds to the digested dC plasmid without a DPC that was transfected and harvested after 6 h, the DPC lane is the digested HDnmt-DPC plasmid with a DPC that was transfected and harvested after 6 h, and M is a control plasmid that was digested to serve as a marker. c, table data from two independent experiments transfected into telomerasetransformed normal human skin fibroblasts. The subscripts 1 and 2 for the times (h) refer to the different experiments. The % DNA repair is calculated by assuming that 100% repair would produce a 259-bp band with 0 dpm. The percent repair is obtained by first subtracting the ratio (dpm 259bp /dpm 2091bp ) for 6h 1 from and the (dpm 259bp /dpm 2091bp ) for 0h 1 . The difference is then divided by the (dpm 259bp /dpm 2091bp ) for 0h 1 and multiplied by 100 to yield the percent repair.

Excision of Smaller DPC Substrates by Mammalian Cell Extracts-Because little excision was observed with the fulllength HDnmt-DPC in
polymerase should arrest near the site of the HDnmt-DPC or a polypeptide covalently bound to the DNA. Fig. 3c shows that the ODN with FdC but without the HDnmt-DPC does not obstruct RNA synthesis. However, if the HDnmt-DPC is used as a template for transcription, RNA synthesis is halted. The PROK-HDnmt-DPC (Fig. 3c) was also used as a template for transcription. Although there was only a limited amount of RNA synthesized from the PROK-HDnmt-DPC template, a band that is a few nt larger than the band for HDnmt-DPC was synthesized, and no full-length product was observed. This would be expected if transcription were halted closer to the FdC attachment site as a consequence of degradation of the HDnmt to 4 -5 amino acids by proteinase K. It is difficult to determine the exact size of the PROK-HDnmt-DPC, but the observation of arrested transcription and the absence of a full-length transcript indicate that amino acid residues remain attached to the DNA. For the CHT-HDnmt-DPC there is a small mobility shift. The small shift suggests that the size of the DPC is ϳ1 kDa. Examination of the most probable cleavage pattern for CHT digestion of the HDnmt with the slightly altered mobility shift indicates that a polypeptide of 9 amino acids remains bound to the ODN (68). Thus, both substrates have attached polypeptides.
Excision assays were performed using CFEs on the HDnmt-DPC, CHT-HDnmt-DPC, and PROK-HDnmt-DPC substrates to determine whether shortened polypeptides are excised efficiently compared with the full-length HDnmt-DPC. Fig. 3d and Table 1 show that the PROK-HDnmt-DPC and the CHT-HDnmt-DPC are both better substrates for excision than the original HDnmt-DPC. The differences in the product sizes for the HDnmt-DPC compared with the truncated DPCs are explained in Fig. 3d. These results suggest that the excision efficiency does not vary significantly for peptide cross-links consisting of 10 or fewer amino acids.
In Vivo Substrate for DPC Repair-In vitro excision indicates the capacity of the HDnmt-DPC to serve as a substrate but does not address questions about repair in vivo. To test the capacity for in vivo repair, we constructed a plasmid without a mammalian origin of replication and transfected it into telomeraseimmortalized human skin fibroblasts. In the formation of the HDnmt-DPC, tritium-labeled AdoMet imparts a radiolabeled methyl group at the 5-position of cytosine that will only be removed if the cytosine is replaced or repaired in the crosslink (Fig. 1a). Fig. 4 illustrates the construction of the substrate, and the details for the construction of this substrate is found under "Experimental Procedures." Briefly, the plasmid was methylated with HDnmt, cleaved with the nicking enzyme NBstNB I, and the short ODN removed, and that ODN was replaced by an ODN with either FdC or dC at the HDnmt recognition sequence, followed by ligation. The ODN was then methylated with [ 3 H]AdoMet, and the modifications were verified by using 32 P labeling.
In Vivo Assay of DPC Repair-Once the HDnmt-FdC and dC plasmids were characterized, we transfected the constructs into telomerase-immortalized human skin fibroblasts. After 6 h of incubation, the total plasmids present were isolated from the cells. Fig. 5a shows the method that was used to determine the in vivo repair of the HDnmt-DPC. The BsaJI-cleaved plasmids were separated on 1% agarose gels following proteinase K treatment, and the bands were excised from the gels and were counted using scintillation spectroscopy (Fig. 5, b and c). For the control plasmid, little change was observed during the incubation period. However, analysis of the bands from the recovered HDnmt-DPC containing plasmid showed that ϳ60% of the 3 H was removed from the DPC-containing band during the incubation. This suggests that the repair during that time is significant, because the only way that removal of the label could occur is by the removal of the [ 3 H]FdC, and there is no possibility to reduce the tritium content by replication (Fig. 1a). Thus, it is possible to monitor the in vivo repair of a unique DPC.
Unique DPCs in RNA Polymerase II-transcribed Sequences Are Repaired-In the previous section, we showed that a unique DPC can be eliminated in a transfected plasmid. A previous report (38) demonstrated that in vitro extracts of Xpg deficient CHO cells can be complemented using the recombinant XPG protein to restore NER activity. Despite that important link to DPC repair, there are conflicting data concerning the role of NER in DPC repair (10, 22, 39, 69). As a first step in the exploration of the role of different DNA repair systems in the elimination of DPC damage, we constructed a plasmid, pTGFPHha10, that permits the insertion of a unique ODN to generate DPCs (Fig. 6). This plasmid differs from the pUC19DB10, because the pTGFPHha10 tests the repair on the transcribed DNA strand of the TGFP coding sequence. There is a possibility that methylation at CpG sites could reduce expression of the gene. Therefore, we tested this plasmid for TGFP production following HDnmt methylation by transfecting either the HDnmt-methylated plasmid or the unmethylated plasmid, but the expression of the TGFP in both cases was strong even following the HDnmt reaction (data not shown). This permits the examination of DPCs using this system.
A unique DPC with the attached HDnmt was formed in the same manner as for the tritiated substrate (Fig. 5), except that AdoMet was used. To investigate the role of different DNA repair systems in the eradication of DPCs, we transfected CHO-K1 cells with either HDnmt-pTGFPHha10FdC or pTGFPHha10dC (Fig. 7). There is a significant difference in the number of TGFP-positive cells when random microscope fields are examined using fluorescence microscopy/inverted phase microscopy (Fig. 7b). Cells transfected with only the HDnmt-methylated plasmid but no DPC produce higher levels of TGFP-positive cells than cells transfected using the plasmid with an HDnmt-DPC on the transcribed strand. The production of TGFP was quantitatively monitored using FACS analysis (Fig. 7c). These experiments show that the DPC on the transcribed strand is repaired to at least 80% in 48 h (Fig. 7d). Repair is based on the production of TGFP; thus it is possible that the actual repair occurs in significantly less time, but in any case, the repair is nearly complete. This validates the use of this assay to monitor repair of DPCs on the transcribed strand of RNA polymerase II-transcribed genes.
NER Has a Role in the Elimination of DPCs from Mammalian Cells-In vitro, the HDnmt-DPC is removed by a mechanism that forms products similar to those of NER excision, but the in vivo role of NER remains unanswered. To address this, we used the CHO cell line, uv135, that was derived from the parental CHO-K1 cell line that has a mutation in the Xpg gene and is therefore NER-deficient. To first test the role of NER involvement in DPC repair in vitro, we prepared whole cell extracts of Xpg-deficient uv135 cells and tested their capacity to remove DPCs using substrates described in Figs. 2 and 3. Little excision is noted for the fulllength HDnmt-DPC substrate, even when complemented (Fig.  8). However, if the proteinase K-digested substrate is treated with Xpg-deficient extract complemented with recombinant XPG, we obtain excision, consistent with a previous study (70).
To answer the question of the role of Xpg in vivo, we transfected the Xpg-deficient cells with either the control dC-containing plasmid or the HDnmt-DPC plasmid and monitored repair by following TGFP production using FACS analysis (Fig. 9, a and  b). Production of TGFP from the Xpg cells transfected with the HDnmt-DPC plasmid remains under 30% of the production of the Xpg cells transfected with the dC plasmid. The observation that the repair of the HDnmt-DPC is low demonstrates an in vivo role for NER in the repair of these DPCs. However, the fact that repair is not completely eliminated by the absence of Xpg implies that at least one other repair pathway could function to abolish DPCs. To show that Xpg has a role in repair in vivo, we co-transfected a mammalian expression vector hosting the XPG cDNA with the HDnmt-DPC plasmid into the Xpg-deficient cells (Fig. 9, c and d). Production of TGFP is significantly increased in the cells co-transfected with the XPG cDNA expression plasmid compared with the HDnmt-DPC plasmid transfected without the XPG cDNA expression vector. These data are consistent with an in vivo role of NER in DPC repair.
The 26 S Proteasome in DPC Repair-To test the role of the 26 S proteasome in DPC repair, the CHO-K1 cell line was treated with MG132, a 26 S proteasome inhibitor, for 2 h prior to transfection of the pTGFPHha10-DPC construct. The production of TGFP was monitored using FACS analysis (Fig. 10). At both 24 and 48 h, TGFP production in cells exposed to MG132 is less than 50% of the production in the same cells transfected that were not treated with MG132. These data indicate that DPC repair involves the 26 S proteasome.

DISCUSSION
Most studies on DPC repair in mammalian cells have focused on the induction of DPCs using agents such as formaldehyde or ionizing radiation (24,39,71). In this study we have demonstrated that it is possible to repair a specific HDnmt-DPC that is formed at FdC in DNA. The HDnmt-FdC linkage is a model for the formation of such structures that could occur with human DNMTs, including DNMT1 and DNMT3. This report shows that we can prepare and study the repair of these types of crosslinks in vitro and in vivo. It is significant that the formation of the HDnmt-FdC DPC does not require any harsh chemical reactions, such as sodium borohydride, which could introduce other DNA lesions. Moreover, we demonstrate for the first time, to the best of our knowledge, in our assay that a specific DPC is dependent on the NER system for in vivo repair.
Data from bacterial UvrABC excision nucleases have shown that this repair system can recognize a 16-kDa protein attached to DNA (36,37). This DPC is extremely large compared with most chemical adducts. Therefore, we initially expected that the HDnmt-DPC would be at least partially excised by the mammalian excision nuclease. The low amount of excision we observed in our initial studies with CFEs surprised us, because we saw amounts of excision that differed only by a factor of 2 from that of undamaged DNA, suggesting that the recognition of the full-length HDnmt-DPC reaction is limited compared with the bacterial system. The HDnmt-DPC substrate is considerably different from that of the covalently bound T4pdg used with the thermal stable UvrABC (37). That is because at 55°C used for the excision reaction, the T4pdg is probably denatured. But even the native T4pdg is not efficiently excised by the human excision nuclease system (38). Nonetheless, the failure of the mammalian system to excise the full-length HDnmt-DPC indicates that there is a significant difference in the bacterial and mammalian excision nuclease systems. It is possible that the formation of the bubble following recognition in mammalian NER is impeded by the presence of a DPC (72)(73)(74)(75)(76). Thus, the reduction of the polypeptide size or denaturation of the protein in the DPC could make the DPC more susceptible to NER factors.
To address the inability of mammalian NER to excise the HDnmt-DPC using CFEs, we treated the full-length DPC with either chymotrypsin or proteinase K to reduce the size of the covalently linked HDnmt. Both the protease-treated DPCs were recognized by the mammalian excision nuclease 3-4-fold more efficiently than the full-length HDnmt-DPC. However, the PROK-HDnmt-DPC substrate was excised more rapidly FIGURE 8. In vitro excision of DPCs using NER-deficient (Xpg) and complemented cell extracts. For each assay 25 fmol of substrate was used. The reactions are described under "Experimental Procedures." WCE, whole cell extracts; nt, 32 P-labeled ODN markers. XPG stock amounts were ϳ0.05 mg/ml. Concentrations of ODNs were calculated from radioactive decay, and although the approximation is close, this method has associated error. a, FACS analysis of in vivo repair of HDnmt-pTGFPHha10FdC or the dC substrate in NER-deficient uv135 cells. Either HDnmt methylated pTGFPHha10dC or HDnmt-pTGFPHha10FdC was transfected into uv135 cells (Xpg-deficient) and assayed for production of TGFP using FACS at different times post-transfection. The designation Autofluor refers to autofluorescence and Eff to transfection efficiency of the undamaged plasmid. The transfection efficiency in uv135 cells for the pTurboGFPN plasmid that was not methylated by HDnmt at 24 h was 29.6 Ϯ 1.4%. Note that the region to the upper left of the diagonal denotes the sector that is not autofluorescence that yields the observed signal. b, in vivo repair of HDnmt-pTGFPHha10FdC in NER-deficient uv135 cells. The triplicate data in a are plotted. The filled squares and the solid lines track production of TGFP for the pTGFPHha10dC plasmid, and the open circles with the dotted lines track production of TGFP for the HDnmt-pTGFPHha10FdC DPC. c, FACS analysis of in vivo repair of HDnmt-pTGFPHha10FdC in NERdeficient or complemented uv135 cells. HDnmt-pTGFPHha10FdC was transfected alone or co-transfected with pCMV.Sport6-XPG into uv135 cells. The designation Autofluor refers to autofluorescence and Eff" to transfection efficiency of the undamaged plasmid. The transfection efficiency in uv135 cells for the pTurboGFPN plasmid that was not methylated by HDnmt at 24 h was 18.7 Ϯ 1.4%. Note that the region to the upper left of the diagonal denotes the sector that is not autofluorescence that yields the observed signal. d, in vivo repair of HDnmt-pTGFPHha10FdC in NER-deficient and complemented uv135 cells. The triplicate data in c are plotted. The mean Ϯ S.E. is indicated in the figure. Note that the transfection efficiency of these plasmids is substantially lower with the uv135 cells than with the CHO-K1 cells. The probability of a cell containing both plasmids is ϳ3.5% (0.19 ϫ 0.19 ϫ 100) close to the ϳ4% value observed for the HDnmt-pTGFPHha10FdC and pCMV.Sport6-XPG co-transfection.
than the CHT-HDnmt-DPC substrate. Some of the CHT-HDnmt-DPC substrate was protected by the cross-linking, whereas all the proteinase K-digested substrate is reduced to the molecular mass of the free ODN. Therefore, in the future, it will be important to determine whether there is a size limit for the function of the excision nuclease for the removal of such DPCs. This has been addressed in part by a recent publication (38), but it is possible that the excision of the DPC will also be dependent on the specific cross-linked protein.
Our in vitro assays demonstrate that the excision nuclease is active on the smaller fragments of the HDnmt-DPC, but the ultimate test is the capacity of living cells to eliminate the DPC. The insertion of the HDnmt-DPC in a nonreplicating plasmid results in the elimination of the DPC, independent of transcription or replication. The fact that repair in vivo occurs suggests that there must be some reduction in size of the DPC, because NER will not function on the full-length HDnmt. This reduction in size could be linked to intracellular proteases or the proteasome. The evidence from random cross-linking experiments has indicated that there is a role for proteasomal degradation based on the use of proteasomal inhibitors (39). The formation of DPCs in cells using agents such as ionizing or UV radiation or chemicals could result in induction of repair or other responses not linked directly to repair of the DPCs. Therefore, a simple assay with a defined substrate permits the evaluation of only the repair and eliminates other effects because of treatment.
The method that we have developed for in vivo repair is specific for the HDnmt-DPC and uses tritium labeling. It requires the use of a significant amount of input material (5 g) to recover enough plasmid for the assay (ϳ10% or 0.5 g). The recovery of DNA is not directly linked to the HDnmt-DPC plasmid, because we find approximately the same recovery for the tritium-dC control plasmid. We have also developed another assay to monitor DPC repair. Despite the limitation of the analysis to transcribed sequences, this assay permits the rapid evaluation of DNA repair genes in DPC repair. In the technique using host cell repair and expression, we have shown that it is possible to transfect much smaller amounts of the HDnmt-DPC plasmid DNA (ϳ0.5 g) to obtain information on the role of the DNA repair systems. The difficulty in preparing such plasmids makes this a much preferred method for the study of DPCs. Using these methods along with cell lines that have specific repair functions ablated will lead to a better understanding of DPC repair.
Because NER is implicated in DPC repair, it will also be important to determine whether there are mechanisms that differentially repair either the transcribed or nontranscribed strands. This was first noted for NER (77)(78)(79) and will need to be investigated for DPC repair also. More manipulation of the transcribed sequence to move the DPC site more than 150 nt away from the transcription start site is required to observe transcription-coupled repair, at least for NER (80). The use of techniques similar to those described in this study should make it possible to study such effects.
There is a large range of chemical-and/or radiation-induced DPCs, and eventually it will be necessary to compare at least some of these in the same experimental systems (18,39,71,(81)(82)(83)(84). The formation of histone-DPCs may be more prevalent than most other observed DPCs, because of the cellular quantity and proximity of these proteins to DNA, but other adducted proteins could also have a major effect on cellular function (e.g. transcription factors or replication factors), and the methods of their elimination will need to be compared. Although an initial analysis could suggest that DPCs are repaired in the same manner, that hypothesis must be established experimentally. The differences in cross-linking observed for proteins could have more of a role in determining which repair system(s) would function. For example, the removal of DPCs on both DNA strands would require some mechanism to preserve the genetic information on both strands and may not involve NER. The variation of DPCs generated by formaldehyde and the failure to observe NER dependence (39) could be explained by the variety of DPCs formed upon chemical or radiation treatment. The repair of other DPCs will be the subject of future studies. FIGURE 10. In vivo role of the 26 S proteasome in DPC repair. a, FACS analysis of in vivo repair of HDnmt-pTGFPHha10FdC in cells exposed to the 26 S proteasome inhibitor MG132. HDnmt-pTGFPHha10FdC was transfected into CHO-K1 cells in triplicate that were either treated or not for 2 h with MG132 (0.2 mM) prior to transfection using Lipofectamine. The designation Autofluor refers to autofluorescence and Eff to transfection efficiency of the undamaged plasmid. The transfection efficiency in uv135 cells for the pTurboGFPN plasmid that was not methylated by HDnmt at 24 h was 34.6 Ϯ 1.2%. b, in vivo repair of HDnmt-pTGFPHha10FdC in CHO-K1 cells exposed to the 26 S proteasome inhibitor MG132. The triplicate data in a are plotted. The filled squares track production of TGFP from the transfected HDnmt-pTGFPHha10FdC plasmid when the cells were not exposed to MG132, and the open circles track production of TGFP HDnmt-pTGFPHha10FdC plasmid when cells were exposed to MG132. Standard deviation for all the samples was under 1%.
Previous work has implicated the proteasome (85,86) in the elimination of DPCs, because proteasome inhibitors reduce cell survival for exposure to formaldehyde (39). However, we have direct evidence that this is the case for the specific in vivo repair of the HDnmt-DPC plasmid, based on the use of the 26 S proteasome inhibitor MG132. From these data, the fact that NER excision is observed on the DPC that is reduced in size and that repair in Xpg-deficient cells is reduced, the model is consistent with the one presented in Ref. 70. The initial protease attack on the covalently linked protein is followed by removal of a smaller polypeptide in an ODN by the mammalian NER system. Further experiments using in vivo substrates are necessary to probe the pathway(s) for eradication of DPCs.
One other use of such results is to generate DPCs in extracts either by transfection of FdC-containing plasmids or ODNs (87,88). The DPCs can be recovered and assayed for the presence of different DNMTs. This can provide an indication of the efficiency of the DPC in sequestering the DNMTs. The isolation of such DNMT-DPCs could provide in vivo results to demonstrate the necessity for incorporation of FdC into DNA to inactivate the DNMTs.
The formation of DPCs by chemicals or radiation is often accompanied by a variety of damage responses. The possibility to examine repair in vitro and in vivo with specific DNA substrates is a step toward understanding specifically the repair of these superbulky adducts. The numerous types of DPCs formed that involve multiple attachments to DNA also means that other DNA repair mechanisms may play a role in the restoration of normal cellular function following cross-linking. In the future, it will be important to understand the role of different NER components and possibly other DNA repair systems (e.g. recombination repair) in the elimination of the DNMT-DPCs as well as for other DPCs. Therefore, the complexity of DPC repair will require an effort using both in vitro and in vivo assays to comprehend how this damage is formed and eliminated.