Recognition of DNA adducts by human nucleotide excision repair. Evidence for a thermodynamic probing mechanism.

The mechanism by which mammalian nucleotide excision repair (NER) detects a wide range of base lesions is poorly understood. Here, we tested the ability of human NER to recognize bulky modifications that either destabilize the DNA double helix (acetylaminofluorene (AAF) and benzo[a]pyrene diol-epoxide (BPDE) adducts, UV radiation products) or induce opposite effects by stabilizing the double helix (8-methoxypsoralen (8-MOP), anthramycin, and CC-1065 adducts). We constructed plasmid DNA carrying a defined number of each of these adducts and determined their potential to sequester NER factors contained in a human cell-free extract. For that purpose, we measured the capacity of damaged plasmids to compete with excision repair of a site-directed NER substrate. This novel approach showed differences of more than 3 orders of magnitude in the efficiency by which helix-destabilizing and helix-stabilizing adducts sequester NER factors. For example, AAF modifications were able to compete with the NER substrate ∼1740 times more effectively than 8-MOP adducts. The sequestration potency decreased with the following order of adducts, AAF > UV ≥ BPDE > 8-MOP > anthramycin, CC-1065. A strong preference for helix-destabilizing lesions was confirmed by monitoring the formation of NER patches at site-specific adducts with either AAF or CC-1065. This comparison based on factor sequestration and repair synthesis indicates that human NER is primarily targeted to sites at which the secondary structure of DNA is destabilized. Thus, an early step of DNA damage recognition involves thermodynamic probing of the duplex.

The mechanism by which mammalian nucleotide excision repair (NER) detects a wide range of base lesions is poorly understood. Here, we tested the ability of human NER to recognize bulky modifications that either destabilize the DNA double helix (acetylaminofluorene (AAF) and benzo[a]pyrene diol-epoxide (BPDE) adducts, UV radiation products) or induce opposite effects by stabilizing the double helix (8-methoxypsoralen (8-MOP), anthramycin, and CC-1065 adducts). We constructed plasmid DNA carrying a defined number of each of these adducts and determined their potential to sequester NER factors contained in a human cell-free extract. For that purpose, we measured the capacity of damaged plasmids to compete with excision repair of a site-directed NER substrate. This novel approach showed differences of more than 3 orders of magnitude in the efficiency by which helix-destabilizing and helixstabilizing adducts sequester NER factors. For example, AAF modifications were able to compete with the NER substrate ϳ1740 times more effectively than 8-MOP adducts. The sequestration potency decreased with the following order of adducts, AAF > UV BPDE > 8-MOP > anthramycin, CC-1065. A strong preference for helix-destabilizing lesions was confirmed by monitoring the formation of NER patches at site-specific adducts with either AAF or CC-1065. This comparison based on factor sequestration and repair synthesis indicates that human NER is primarily targeted to sites at which the secondary structure of DNA is destabilized. Thus, an early step of DNA damage recognition involves thermodynamic probing of the duplex.
Nucleotide excision repair (NER) 1 is an essential pathway for removing bulky base modifications from DNA. This repair mechanism involves endonucleolytic cleavage at two phosphodiester bonds, one 3Ј and the other 5Ј of the site of damage, followed by excision of DNA damage as the component of a single-stranded fragment (1)(2)(3)(4)(5). The excised oligonucleotide is replaced by DNA repair synthesis, and DNA continuity is reestablished by ligation. In mammalian cells, the major sites of incision are at the 5th phosphodiester bond 3Ј and the 24th phosphodiester bond 5Ј to the lesion (6).
Human patients deficient in NER suffer from xeroderma pigmentosum (XP), a hereditary disease characterized by photosensitivity, increased incidence of skin cancer, and frequently neurological abnormalities (7,8). At the biochemical level, XP individuals are impaired in the removal of radiation products induced by the UV component of sunlight (9). Somatic cells obtained from these patients are also defective in excision repair of bulky DNA adducts resulting from genotoxic chemicals (1,2,4). Recent in vitro studies using cell-free extracts showed that the range of base lesions processed by mammalian NER extends to nonbulky adducts, and even abasic sites are susceptible to excision repair by this pathway (10).
Biochemical reconstitution experiments demonstrated that mammalian NER is catalyzed by the coordinated action of at least 30 polypeptides (11,12). Several of these factors have been implicated in the recognition step of NER, primarily a complex made up of XPA and the three subunits of RPA (p70, p34, p11) (13); in addition, the multisubunit protein complex TFIIH and XPE protein may participate in damage recognition (1)(2)(3)(4). The molecular mechanism by which these proteins discriminate a large number of chemically unrelated DNA lesions as substrates for NER is unknown. However, the versatility of NER led to the assumption that this system recognizes conformational changes imposed on DNA at sites of damage rather than specific base modifications (2,5). In this report, we compared recognition of DNA adducts where detailed structural information is available and identified a molecular determinant triggering initiation of the mammalian NER pathway.
Treatment of DNA with 8-MOP and long wavelength UV light (Ͼ320 nm) yields psoralen monoadducts and a small proportion of psoralen diadducts (27). This photoaddition reaction occurs between the 5,6 double bond of pyrimidine bases and either the 3,4 (pyrone) or the 4Ј,5Ј (furan) double bond of the psoralen. Modifications with psoralen induce helical distortion by unwinding the duplex and enhancing backbone flexibility (28) but fail to destabilize the secondary structure of DNA. On the contrary, thermostability measurements showed that both pyroneside and furanside monoadducts stabilize the helix by mediating stacking interactions between the psoralen moiety and the surrounding base pairs (19,20,28). Helix-stabilizing adducts were also obtained using anthramycin, a pyrrolo- [1,4]benzodiazepine antibiotic, and CC-1065, a composite compound consisting of three pyrroloindole subunits joined by amide linkages (Fig. 1). Anthramycin binds selectively to N 2 of guanine through aminal bonds and forms covalent adducts with essentially no distortion of the DNA helix (21,29). CC-1065 displays a cyclopropyl ring that alkylates DNA at position N 3 of adenine, generating covalent adducts that cause bending and winding of the double helix (22,30). Both anthramycin and CC-1065 adducts enhance duplex stability through noncovalent interactions derived from hydrogen bonds (anthramycin) or van der Waals and hydrophobic forces (CC-1065) within the minor groove of DNA (21,22,30).
Recognition of these bulky adducts was compared by moni-toring their capacity to sequester NER factors. To that end, we have developed a competition assay that measures the efficiency by which damaged plasmids compete for NER factors operating on a site-directed substrate (31). As a source of NER activity we exploited a standard soluble extract from human cells (32,33). Under the conditions used in our study, this cell-free extract does not support chromatin assembly or transcription, thereby eliminating nuclear activities that modulate the intrinsic capacity of NER to recognize DNA damage (34,35). This system revealed Ͼ1000-fold differences in the capacity of the tested adducts to sequester human NER factors. Those lesions that destabilize the DNA helix (AAF or BPDE adducts, UV radiation products) were effective competitors. In contrast, those adducts that stabilize the helix (8-MOP, anthramycin, and CC-1065 adducts) displayed minimal competing effects. In parallel, site-specifically placed CC-1065 adducts were unable to detectably stimulate synthesis of DNA repair patches. These results indicate that an early subset of NER recognition factors is attracted to structural defects associated with unfavorable thermodynamic changes of the DNA double helix. As a consequence, human NER is preferentially targeted to sites of helical instability.
Cell Lines and Extracts-HeLa cells were kindly provided by the Institute of Virology (Zü rich). The human lymphoid cell lines GM2250 (XP-A) and GM2634 (XP-C) were obtained from the NIGMS Human Genetic Mutant Cell Repository. HeLa and lymphoid cells were grown in RPMI 1640 supplemented with 7 and 15% fetal bovine serum, respectively. Cell extracts were prepared as described (32,33).
Site-directed Substrates-M13 double-stranded DNA containing a site-directed AAF adduct (M13-AAF) was constructed using a gapped DNA intermediate as outlined previously (36). To obtain site-specific CC-1065 adducts, a short double-stranded fragment containing a single CC-1065 modification was inserted into pUC19 by a previously described method (37), with the following modifications. The oligonucleotide 5Ј-GATCTCCGGATTAGGCCCGGGC was phosphorylated at its 5Ј end (38) and annealed with an equal amount of the oligonucleotide 5Ј-TCGAGCCCGGGCCTAATCCGGA, which still had a 5Ј-OH terminus. The resulting double-stranded fragment (0.21 mg/ml) was reacted with CC-1065 (160 M) in 10 mM Tris-HCl, pH 8.0, 1 mM EDTA, 100 mM NaCl for 24 h at room temperature. This incubation results in covalent modification of the most 3Ј-adenine of the 5Ј-GATTA sequence in the upper strand, whereas the lower strand does not contain such a high affinity site for CC-1065 binding (39). The fragment was recovered by ethanol precipitation and ligated with BamHI-digested pUC19 using a 50-fold molar excess of the fragment. The reaction products were digested by SalI, precipitated with 6% (w/v) polyethylene glycol 6000 and 450 mM NaCl, phosphorylated with ATP and T4 polynucleotide kinase, and circularized as indicated (37). The obtained pUC19 derivative (designated pUC19-CC-1065) was purified as covalently closed circular DNA by CsCl centrifugation (40). Nondam- aged pUC19 derivative (pUC19-Control) was prepared with nonmodified fragments following the identical procedure. A small amount (1 g) of these derivatives was analyzed by cleavage with SmaI, followed by treatment with calf intestinal phosphatase and 5Ј end labeling of the resulting fragments with polynucleotide kinase and [␥-32 P]ATP (38). After incubation of these fragments at 95°C for 20 min to induce strand breaks at sites of CC-1065 modification (22), the resulting products were resolved on 20% denaturing polyacrylamide gels.
DNA Repair Synthesis Assay-Reactions (50 l) were slightly modified from a previous report (33) and contained 90 g human cell extract (based on protein), 200 ng of M13 or pUC19 DNA substrate, 45 mM HEPES, pH 7.8, 70 mM KCl, 7.4 mM MgCl 2 , 0.9 mM dithiothreitol, 0.4 mM EDTA, 3.4% glycerol, 2 mM ATP, 20 M each of dATP, dGTP, and dTTP, 8 M dCTP, 2.0 Ci of [␣-32 P]dCTP, 40 mM phosphocreatine, 2.5 g of creatine phosphokinase, and 18 g of bovine serum albumin. After incubation at 30°C for 3 h, reactions were stopped by the addition of EDTA to 20 mM. The samples were incubated at 37°C with ribonuclease A (80 g/ml) for 10 min, SDS to 0.5% and proteinase K to 190 g/ml were then added, and the mixtures incubated for a further 45 min at 37°C. DNA was extracted, digested with restriction enzymes to obtain the appropriate fragments, and resolved on 20% native polyacrylamide gels. After electrophoresis, the gels were dried and exposed to Typox RP x-ray films. DNA repair synthesis was quantified by scanning densitometry of the x-ray films on a Molecular Dynamics Computing Densitometer using ImageQuant software. The densitometry was calibrated with radioactive bands that were excised from the dried gels and quantified by Cerenkov counting. The obtained values were used to calculate dCMP incorporations.
To prepare anthramycin-DNA adducts, pUC19 (333 g/ml) was incubated with 3.3 M anthramycin in 1.5 mM sodium citrate, pH 7.0, and 15 mM NaCl for 90 min at 25°C. The modified DNA was purified by ethanol precipitation and 5-20% sucrose gradient centrifugation. The reaction of anthramycin with DNA is accompanied by a gradual change in the anthramycin absorption peak (43), which shifts from 333 to 343 nm and decreases in amplitude (Fig. 2D). Thus, the fractions of free and DNA-bound anthramycin were measured from the absorbance (A) at 321 nm (the maximum of the difference spectrum) and at 349 nm (the isosbestic point). With the equations A free, 321 ϭ 1.134⅐A 349 and A bound, 321 ϭ 0.709⅐A 349 (43), we calculated that 68% of anthramycin reacted with DNA, resulting in the formation of 11.9 anthramycin adducts/pUC19 molecule.
CC-1065-DNA adducts were prepared by incubating pUC19 (333 g/ml) with 3.3 M CC-1065 in 1.5 mM sodium citrate, pH 7.0, and 15 mM NaCl for 120 min at 25°C, followed by ethanol precipitation and 5-20% sucrose gradient centrifugation. The modified DNA was analyzed by UV spectroscopy (Fig. 2E), and the amount of covalently bound CC-1065, 9.6 adducts/pUC19 molecule, was measured from its optical density at 365 nm, assuming a molar extinction coefficient of 48,000 (44). Covalent modification by CC-1065 was additionally confirmed by heating the plasmids to 95°C and analysis of the resulting strand breakage products (22) on alkaline agarose gels (data not shown).
Plasmid pUC19 containing 11.5 UV radiation products was prepared by irradiation at a dose of 450 J⅐m Ϫ2 using a germicidal lamp with peak output at 254 nm (31). Irradiation was performed on ice in an open Petri dish and in aliquots of 20 l containing pUC19 DNA (100 g/ml), 10 mM Tris-HCl, pH 8.0, and 1 mM EDTA. The UV fluence was monitored with a Steritest dosimeter. All plasmids were stored in small aliquots at Ϫ80°C.
Repair Competition Assay-Standard 50-l reactions contained 50 ng of M13 DNA substrate, various amounts, between 2.5 and 450 ng, of pUC19 competitor DNA, 2 mM ATP, 20 M each of dATP, dGTP, and dTTP, 8 M dCTP, 2.0 Ci of [␣-32 P]dCTP, 40 mM phosphocreatine, 2.5 g of creatine phosphokinase, and the same buffer components as in the previously described DNA repair synthesis assay. A mixture of 450 ng of pUC19 and 50 ng of M13 DNA translates to a molar ratio of competitor to substrate of ϳ25:1. To obtain an even higher molar ratio of ϳ120:1 in the experiments of Fig. 6A, reactions were performed with 10 ng of M13 substrate, whereas the amount of pUC19 DNA was maintained at 450 ng. Incubations were at 30°C for 3 h, and the reaction products were digested with AvaII, SmaI, and PstI, followed by 20% native polyacrylamide gel electrophoresis. The dried gels were exposed to x-ray films and analyzed by computing densitometry as indicated before. DNA repair synthesis in the 37-base pair SmaI-PstI fragment was expressed as the percentage of nucleotide incorporation obtained in parallel control reactions containing identical amounts of nondamaged pUC19 DNA.

NER Substrate for Repair Competition Assays-
In this report, we compared the capacity of various bulky DNA adducts to sequester NER recognition factors contained in a human cell-free extract. For that purpose, we developed a competition assay that measures the efficiency by which damaged plasmid pUC19 inhibits excision repair of a specific NER substrate. This substrate consists of M13 double-stranded DNA containing a site-directed AAF-guanine adduct (Fig. 3A) and was constructed by ligating into a gapped M13 intermediate a 19-mer oligonucleotide with a single AAF-guanine modification in the center (36). The site-directed AAF adduct is flanked by the recognition sequences for SmaI and PstI that were introduced to generate a fragment of 37 base pairs (Fig. 3A). Nondamaged control M13 DNA was prepared by ligating into the gapped intermediate a nonmodified 19-mer of the same sequence. Repair incubations were performed in a NER-proficient HeLa cell extract (33) supplemented with ATP, an ATP-regenerating system, and all four deoxynucleoside triphosphates, of which dCTP was radiolabeled. After incubations of 3 h at 30°C, DNA was restricted with SmaI, PstI, and AvaII, and NER patches were detected by monitoring the incorporation of radiolabeled dCMP in the region of the substrate where the AAF adduct is located, i.e. in the 37-base pair SmaI-PstI fragment (Fig. 3B,  lanes 1 and 2). A biochemical complementation experiment proved that damage-specific nucleotide incorporation into this 37-base pair SmaI-PstI region was catalyzed by the NER system. Incorporation of dCMP was abolished in extracts prepared from NER-deficient XP-A (GM2250) or XP-C (GM2634) cell lines (Fig. 3B, lanes 3 and 4). However, DNA repair synthesis was reconstituted by coincubating the two different XP cell extracts in the same repair reaction (Fig. 3C, lanes 5 and 6, in duplicate).
Repair Competition Assay with AAF-DNA Adducts-We first tested the capacity of pUC19 DNA containing 10.2 AAF adducts/plasmid to compete with the single AAF adduct located on M13 DNA. The formation of NER patches in the 37-base pair SmaI-PstI region of the site-directed substrate was progressively inhibited when increasing amounts of multiply AAFdamaged competitor were added to the repair reactions (Fig. 4). Partial inhibition was observed at molar ratios of AAF-damaged pUC19 to M13 DNA ranging from 0.06 to 0.54, and essentially complete inhibition was detected at molar ratios between the two DNA molecules of 1.35 or higher. In contrast, little or no inhibition was observed when nondamaged pUC19 DNA was added to the repair reactions even in a large molar excess over M13 DNA substrate. The typical gel shown in the inset of Fig. 4 illustrates that NER in the 37-base pair SmaI-PstI region of the M13 substrate was reduced to background levels when AAF-damaged pUC19 was added in a 1:1 mass ratio, corresponding to a pUC19 to M13 DNA molar ratio 2.7:1 (lane 3). We consistently noted that only the specific NER response in the 37-mer SmaI-PstI fragment was suppressed, whereas the nonspecific nucleotide incorporation unrelated to NER in the adjacent 330-base pair AvaII-SmaI region remained unaffected. These different responses of the two M13 DNA fragments confirmed that the competition between M13 substrate and damaged pUC19 was selective for NER factors.  (43). The frequency of anthramycin modification of DNA (11.9 adducts/pUC19) was calculated from the absorbance at 321 nm (the maximum of the difference spectrum) and at 349 nm (the isosbestic point). E, UV absorption spectrum of CC-1065-modified DNA (100 g/ml, upper line). The amount of DNA-bound CC-1065 (9.6 adducts/pUC19 molecule) was determined from its absorbance at 365 nm (44). The lower line shows the corresponding absorption spectrum of nondamaged pUC19. By interpolation of the data shown in Fig. 4, we estimated that approximately 50% inhibition is expected at a molar ratio of AAF-damaged pUC19 to M13 substrate of 0.09. This number translates to a 10.8-fold molar excess of M13 substrate over pUC19 competitor. Because each M13 DNA molecule contains a single site-directed adduct (Fig. 3A), whereas pUC19 contains an average of 10.2 adducts/molecule ( Fig. 2A), these calculations yield 50% inhibition of NER operating on the substrate at a 1:1 stoichiometry of AAF adducts. Thus, AAF modifications located on pUC19 are recognized, on the average, as efficiently as the single site-directed AAF adduct placed on M13 DNA, indicating that the repair competition assay is a quantitative method to assess recognition of a particular form of DNA damage by human NER.

Repair Competition Assay with BPDE-DNA Adducts and UV
Radiation Products-We next exploited the competition assay to test the capacity of human NER to recognize BPDE adducts. M13 DNA substrate with the site-directed AAF-guanine modification was coincubated with pUC19 containing either 5.0 or 16.7 BPDE adducts per plasmid molecule. Control reactions were performed with equivalent amounts of nondamaged pUC19 DNA. The molar ratios of pUC19 competitor to M13 substrate were 2.7 and 8.1. The results from four to six independent determinations were quantified by scanning densitometry and plotted as the percentages of DNA repair synthesis obtained in control reactions incubated with nondamaged pUC19 (Fig. 5A). At a modification frequency of 5.0 BPDE adducts/pUC19 and a competitor to substrate molar ratio of 8.1:1, NER in the 37-base pair SmaI-PstI region of the substrate was reduced by 53%. At a competitor to M13 DNA molar ratio of 2.7:1, the same plasmid carrying 5.0 BPDE adducts was unable to significantly inhibit NER operating on the substrate (Fig. 5A). We also tested pUC19 containing 16.7 BPDE adducts per plasmid. As one would expect from the 3-fold higher frequency of modification, this latter plasmid yielded approximately 50% inhibition at a molar ratio of pUC19 to M13 substrate of 2.7:1 and exerted a more pronounced inhibition of NER activity on the substrate at a molar ratio of pUC19 competitor to M13 DNA of 8.1:1 (Fig. 5A). The representative gel of Fig. 5B illustrates the partial inhibition of NER activity in the 37-base pair SmaI-PstI region of the M13 substrate when pUC19 (16.7 BPDE adducts/plasmid) was added to the reactions in a 2.7-fold (lane 3) or in an 8.1-fold molar excess over the substrate (lanes 4 and 5). In contrast, no reduction of NER in the 37-base pair region was found when the reactions were incubated with nondamaged pUC19 (lanes 6 and 7).
We then calculated the stoichiometric excess of BPDE adducts over AAF modifications required to inhibit NER of the substrate. Assuming ϳ50% inhibition at an 8.1-fold molar ex-FIG. 3. Site-directed NER substrate for competition assays. A, a single AAF adduct was site-specifically located in the polylinker of a M13mp19 DNA derivative of approximately 7300 base pairs. This unique AAF adduct was flanked by the recognition sites for SmaI and PstI, which were used to generate a fragment of 37 base pairs of which the sequence is shown (the asterisk denotes the covalently modified guanine). Additional digestion with AvaII produces an adjacent AvaII-SmaI fragment of 330 base pairs and a long PstI-AvaII fragment of 6920 base pairs. B, representative polyacrylamide gel demonstrating that dCMP incorporation in the 37-base pair SmaI-PstI region of the substrate is strictly dependent on NER. Reactions were performed for 3 h at 30°C in the presence of M13 DNA substrate (200 ng), [ 32 P]dCTP, and 90 g (in proteins) of HeLa cell extract (lanes 1 and 2) cess of pUC19 containing 5.0 BPDE adducts per molecule (Fig.  5A), these calculations yielded a stoichiometry of BPDE to AAF adducts of ϳ40:1. Nearly the same value is obtained by considering the ϳ50% inhibition observed when plasmid pUC19 containing 16.7 BPDE adducts/molecule was added to the reactions at a 2.7:1 molar ratio of competitor to substrate (Fig. 5A). In summary, an approximately 40-fold excess of BPDE adducts was required to inhibit NER of the site-specific AAF adduct by 50%. These results indicate that BPDE lesions sequester human NER factors about 40 times less effectively than AAF adducts.
UV radiation products were also tested in the competition assay. Plasmid pUC19 was exposed to an UV radiation dose of 450 J⅐m 2 , generating an average of 11.5 UV radiation products/ plasmid molecule (see "Experimental Procedures"). As illustrated in Fig. 5A, UV radiation damage inhibited NER of the substrate to levels similar to those observed with BPDE adducts. Partial inhibition of DNA repair synthesis in the 37-mer SmaI-PstI region was observed when UV-irradiated competitor was added to the reactions in a molar ratio of pUC19 competitor to M13 substrate of 2.7:1, and almost complete inhibition was observed at a competitor to substrate ratio of 8.1:1. To obtain ϳ50% inhibition, it was necessary to add UV-treated pUC19 DNA (450 J⅐m 2 ; 11.5 UV radiation products/plasmid) in a 2.7-fold molar excess over the substrate (Fig. 5A). This amount of UV-irradiated competitor translates to a stoichiometry between UV photoproducts and AAF adducts of ϳ30:1. Thus, UV photoproducts sequestered human NER factors about 30 times less effectively than AAF modifications.
Repair Competition Assay with 8-MOP-, Anthramycin-, and CC-1065-DNA Adducts-Psoralens, anthramycin, and compound CC-1065 form bulky base adducts that increase the thermodynamic stability of double-stranded DNA. To test recognition of these lesions by human NER, we obtained covalently modified pUC19 containing 14.5 8-MOP, 11.9 anthramycin, or 9.6 CC-1065 adducts per molecule (Fig. 2, C-E). These plasmids were unable to inhibit NER of the substrate when added to the reactions at molar ratios of pUC19 to M13 DNA of 2.7:1 (data not shown) or 8.1:1 (Fig. 6A). In light of these negative results we further increased the molar excess of pUC19 over the substrate to 25:1 and observed slight inhibition in the presence of 8-MOP adducts but not in the presence of anthramycin or CC-1065 adducts (Fig. 6A). The representative gels illustrate the lack of substantial competitive inhibition exerted by plasmid pUC19 containing 8-MOP adducts (Fig.  6B), or anthramycin and CC-1065 adducts (Fig. 6C), added to the reactions in a 25-fold molar excess. At an even higher molar ratio between pUC19 and M13 DNA of 120:1, a marked inhibition was observed in the presence of 8-MOP-damaged competitor, whereas anthramycin and CC-1065 displayed no detectable effects (Fig. 6A). In contrast, AAF-damaged plasmids (molar ratio of 2.7:1) or BPDE-damaged plasmids (molar ratio of 120:1) were able to suppress NER of the substrate almost completely (Fig. 6A).
To calculate the molar excess of 8-MOP adducts required to inhibit NER of the site-directed AAF substrate, we assumed ϳ50% inhibition at a molar ratio of pUC19 competitor to M13 substrate of 120:1 (Fig. 6A). Considering that pUC19 carried an average of 14.5 8-MOP adducts/molecule, we obtained a factor of 1740. On the other hand, competitor DNA carrying 11.9 anthramycin or 9.6 CC-1065 adducts per molecule was unable to affect NER of the substrate even at the highest molar ratio of pUC19 to M13 DNA tested (120:1). These results indicate that 8-MOP, anthramycin, and CC-1065 adducts sequester NER factors more than 3 orders of magnitude less efficiently than AAF modifications.
Excision Repair Assay with a Site-directed CC-1065 Adduct-The observation that CC-1065 modifications have an extremely low capacity to sequester NER factors led to the prediction that these adducts are poorly recognized and processed by the human NER system. To test this hypothesis, we exploited the selectivity of compound CC-1065 for the sequence 5Ј-GATTA* (22,39) and constructed circular double-stranded DNA containing a uniquely located CC-1065 adduct (the asterisk denotes the covalently adducted adenine). Briefly, a duplex oligonucleotide with the CC-1065 modification in one strand was ligated between the BamHI and SalI sites in the polylinker of plasmid pUC19 (see "Experimental Procedures" for details). The desired ligation product (designated pUC19-CC-1065) was purified as covalently closed circular DNA. A control substrate (pUC19-Control) was obtained by ligating nonmodified duplexes of the same sequence into the identical region of the pUC19 polylinker.
The single CC-1065-adenine adduct was situated between two adjacent SmaI sites (Fig. 7A). To demonstrate effective FIG. 5. Repair competition assay, comparison between BPDE-, UV-, and AAF-damaged DNA. A, standard repair competition reactions were incubated in the presence of pUC19 DNA containing BPDE adducts (5.0 or 16.7/molecule), UV radiation products (11.5/molecule), or AAF adducts (10.2/molecule). The molar ratio of pUC19 competitor to M13 substrate was 2.7:1 or 8.1:1, as indicated. After 3 h at 30°C, dCMP incorporation in the 37-base pair SmaI-PstI region of the substrate was analyzed by gel electrophoresis, quantified by scanning densitometry of the corresponding autoradiographs, and expressed as the percentage of DNA repair synthesis obtained in control reactions containing identical amounts of nondamaged pUC19 DNA (mean values of four to six determinations Ϯ S.D.). B, representative gel obtained using, as a competitor, pUC19 modified with 16.7 BPDE adducts per plasmid. DNA repair synthesis in the 37-base pair SmaI-PstI region of the M13 substrate was visualized by autoradiography. Standard competition mixtures contained M13 DNA substrate with the site-directed adduct (M13-AAF) or nonmodified M13 DNA (M13-Control), as indicated. Competitor pUC19 was either BPDE-damaged or nondamaged (N), and the molar ratio of pUC19 to M13 was 2.7:1 (lanes 3 and 6) or 8.1:1 (lanes 4,  5, and 7). modification, the 21-base pair SmaI fragment was obtained from the substrate by restriction digestion, radiolabeled at its 5Ј ends, and subjected to heating at 95°C for 20 min. This treatment induces breakage of the adducted strand at the site of CC-1065 modification (22) and is expected to generate a radiolabeled 15-mer oligonucleotide that was resolved by denaturing polyacrylamide gel electrophoresis (Fig. 7B). Quantita-tive analysis of the resulting autoradiograph showed that a considerable fraction (ϳ35%) of radioactivity shifted from the position of the 21-mer oligonucleotide to a smaller oligomer of 15 residues (Fig. 7B, lane 1). This strand breakage product was not detected when the nonmodified control substrate was heated to 95°C (Fig. 7B, lane 2). Although both 5Ј ends of the 21-mer duplex were radiolabeled, only the upper strand containing the 5Ј-GATTA sequence was modified by CC-1065 and, as a consequence, only this upper strand was subject to nicking during thermal treatment. Thus, the observed fragmentation of 21-mer oligonucleotides translates to a frequency of modification of at least 70%, indicating that a major fraction of pUC19-CC-1065 substrate contained the site-specific CC-1065-adenine adduct at the expected position. This analysis also showed that cleavage by SmaI was not inhibited by the CC-1065 modification.

FIG. 7.
Site-directed CC-1065 adduct. A, a single covalent modification with compound CC-1065 was situated on a pUC19 derivative of approximately 2700 base pairs and was flanked by two SmaI sites that yield a fragment of 21 base pairs. The sequence of this 21-mer SmaI fragment is shown with the asterisk denoting the site of adduct formation. B, effective modification of the pUC19 derivative was demonstrated by digestion with SmaI, 5Ј end labeling of the resulting 21-mer fragment, and heat treatment to induce strand breakage at the site of CC-1065 modification. The appearance of a 15-mer oligonucleotide in lane 1 indicates that at least 70% of the pUC19 derivatives contained the CC-1065 adduct at the expected position. Lane 2 shows the sample from nonmodified pUC19 after heat treatment. The radioactivity migrating near the gel origin represents the large SmaI fragment that is also radiolabeled during treatment with polynucleotide kinase. mented with radiolabeled dCTP. Parallel reactions contained the corresponding nondamaged control substrates. After repair incubations of 3 h at 30°C, pUC19 or M13 DNA was recovered and digested with the appropriate restriction enzymes to obtain short fragments containing the sites of modification. As indicated in Fig. 7A, cleavage of pUC19-CC-1065 with SmaI generates a fragment of 21 base pairs. M13 DNA was cleaved with SmaI and PstI, yielding a fragment of 37 base pairs (Fig.  3A). The representative gel of Fig. 8A demonstrates that dCMP incorporation in the 21-mer SmaI region of pUC19-CC-1065 (lanes 3 and 4) was indistinguishable from that observed in the corresponding region of control pUC19 (lanes 1 and 2). For comparison, lane 5 shows the typical level of dCMP incorporation found in the 37-base pair SmaI-PstI fragment of the M13-AAF substrate. The radiolabel incorporated into these fragments was quantified and plotted in Fig. 8B. We observed comparable levels of nonspecific dCMP incorporation into the 21-mer region of pUC19 as in the corresponding 37-mer region of the M13 substrate (compare pUC19-Control with M13-Control). The site-directed CC-1065 adduct was unable to induce DNA repair synthesis (Fig. 8B, pUC19-CC-1065). In contrast, NER was substantially stimulated by the site-directed AAF adduct (Fig. 8B, M13-AAF). This comparison confirmed the preference of human NER for DNA adducts that destabilize the double helix. DISCUSSION Mammalian NER processes a wide range of chemically and structurally distinct base adducts, but some types of damage are repaired at higher rates than others (1,2,10,45,46). The biochemical mechanisms underlying this broad and extremely heterogeneous response are poorly understood. In the present study, we have established a structure-activity relationship with respect to DNA damage recognition using a novel in vitro approach based on NER factor sequestration. Our assay requires highly purified plasmid DNA (pUC19) containing a defined number of a particular base adduct and measures the efficiency by which these damaged DNA molecules sequester human NER factors and, hence, inhibit repair of a site-directed NER substrate (Fig. 9). Using this competition assay, we have tested recognition of several bulky base adducts that are known to display potent mutagenic and/or carcinogenic effects and of which detailed structural information is available. The susceptibility of some of these bulky adducts (BPDE, anthramycin, CC-1065) to mammalian NER has not been examined in a defined in vitro system before.
Several observations led us to exploit the repair competition assay to compare damage recognition by the human NER system. First, excision repair of the site-directed AAF adduct used as a substrate in these assays is strictly dependent on the NER pathway. In fact, essentially no synthesis of DNA repair patches was observed when the site-directed substrate was incubated with extracts from NER-deficient XP-A or XP-C cell lines, but DNA repair synthesis was reconstituted by combining the two different extracts (Fig. 3B). Second, competitor pUC19 containing AAF adducts (Fig. 4), UV radiation products (31), BPDE adducts (Fig. 5), or 8-MOP adducts (Fig. 6) inhibited NER of the site-specific substrate in a dose-dependent manner. When AAF-damaged competitor was tested, we found that NER of the site-specific AAF adduct on the substrate was inhibited to 50% at a 1:1 stoichiometry of these lesions (Fig. 4), indicating that competition for NER factors occurs in a quantitative manner. Third, nondamaged pUC19 DNA caused only marginal inhibition of NER operating on the substrate. Conversely, efficient competitors such as AAF-damaged pUC19 suppressed NER activity in the covalently modified region of the substrate, without significantly reducing nonspecific nucleotide incorporation in an adjacent DNA segment (Fig. 4). Thus, the competition effect is selective for NER activity. Fourth, the principal finding obtained in the competition assay (differential recognition of thermodynamically diverse lesions) was confirmed by comparing DNA repair synthesis induced by sitespecific adducts with either a helix-destabilizing (AAF) or helix-stabilizing (CC-1065) compound (Fig. 8). Finally, previous studies have shown that damage recognition/DNA incision constitute the rate-limiting step of NER in the cell-free extract (47), indicating that competition for repair factors is likely to occur at an early, preincisional level of the pathway.
The repair competition data demonstrated that DNA conformation at the site of damage is of critical importance for recognition. The capacity of the tested adducts to sequester human NER factors decreased with the following order: AAF Ͼ UV radiation products Ն BPDE Ͼ 8-MOP Ͼ anthramycin, CC-1065. The competition exerted by AAF, UV, and BPDE adducts was 2-3 orders of magnitude stronger than the competition effected by 8-MOP, anthramycin, or CC-1065 adducts. For example, AAF modifications were able to sequester NER factors 1740 times more efficiently than 8-MOP adducts. This striking hierarchy of sequestration efficiency, combined with the known thermodynamic characteristics of the tested lesions, indicates that mammalian NER is primarily targeted to structural defects that destabilize the double-helical conformation of DNA. In fact, AAF modifications have been shown to completely abolish base pairing between the adducted guanine and its complementary cytosine (14,15). Analysis with chloroacetaldehyde or osmium tetroxide revealed that this helix destabilizing effect is not limited to the site of AAF modification but rather extends to the neighboring base pairs in the duplex (23). In comparison, BPDE-N 2 -guanine adducts exert more moderate effects on double-helical stability (16,17,24). Psoralen monoadducts, on the other hand, slightly increase helical stability (19,20,28), whereas anthramycin and CC-1065 adducts produce substantial increments in helical stability (21,22). Thus, an early subset of NER factors appears to be endowed with the capacity to sense thermodynamic parameters of double-stranded DNA and preferentially interact with those sites that exhibit unfavorable changes in free energy. Among the known NER factors, the XPA-RPA complex is a possible protein candidate for executing this recognition function involving thermodynamic probing of DNA. This hypothesis is prompted by previous reports demonstrating that XPA stimulates RPA's single-stranded DNA binding activity (48) and that the two factors display strong cooperativity in binding to AAF-or UVdamaged DNA (13,49).
Several previous reports related to damage recognition in mammalian NER support the concept of thermodynamic recognition. For example, cyclobutane pyrimidine dimers are processed in vivo and in vitro at a considerably lower rate than pyrimidine(6-4) photoproducts (45,46,50). The thermodynamic properties of DNA fragments containing a cyclobutane dimer indicate that this lesion minimally affects the ability to form a duplex (51), whereas pyrimidine(6-4) photoproducts strongly favor disruption of the double helix (52). The different stereoisomers of BPDE-N 2 -guanine have distinct effects on the stability of a short DNA duplex, the (ϩ)-trans adduct being more helix-destabilizing than the (ϩ)-cis isomer (16). Consistent with the concept of thermodynamic recognition, a recent analysis of epidermal tissue obtained from benzo[a]pyrenetreated mice showed that (ϩ)-trans-anti-BPDE-N 2 -guanine ad-ducts are repaired about three times faster than (ϩ)-cis-anti-BPDE-N 2 -guanine adducts in vivo (53). Thermodynamic recognition is also consistent with the finding that certain nonbulky adducts, primarily abasic sites, constitute efficient substrates for human NER (10). In fact, the loss of a base from DNA inherently destabilizes the duplex (23,54).
Interestingly, the thermodynamic characteristics of damaged DNA are less important for recognition by (A)BC excinuclease, the prokaryotic NER system. For example, it was found that (A)BC excinuclease incises at (ϩ)-cis-anti-BPDE-N 2 -guanine adducts more efficiently than at (ϩ)-trans-anti-BPDE-N 2 -guanine adducts (16), which is exactly the opposite of what was observed in mammalian tissues (53). Similarly, the helix-stabilizing psoralen monoadduct is an efficient substrate for the (A)BC excinuclease (10) but constitutes a modest substrate for the human NER system (Ref. 10 and this study). Also, the poor sequestration capacity of anthramycin-DNA adducts observed in our competition assay using the mammalian system contrasts with a previous report indicating that prokaryotic (A)BC excinuclease processes anthramycin-containing DNA with a four to five times higher rate than UV-irradiated DNA (55). A diverging substrate preference between prokaryotic and mammalian NER has already been noted before (10) and appears to reflect the lack of homology between the two systems.
Damage recognition in NER has been proposed to occur in a sequence of partially overlapping biochemical reactions, each of which confers increased selectivity for damaged sites (56). The present report shows that human NER recognition factors are preferentially targeted to those types of DNA damage that destabilize the duplex. This observation indicates that thermodynamic probing of DNA stability constitutes an early step in damage recognition by mammalian NER. A thermodynamically biased recognition mechanism may serve, in the context of the large mammalian genome, to locate with highest priority those lesions with more pronounced cytotoxic and mutagenic potentials.