Trapping of DNA Nucleotide Excision Repair Factors by Nonrepairable Carcinogen Adducts¹

To cope with bulky DNA damage induced by environmental carcinogens, all living organisms ranging from bacteria to humans are endowed with a versatile DNA repair system that removes many base defects (1, 2, 3, 4). This universal repair reaction has been termed NER³ because it operates through a cut and patch mechanism that is initiated by double endonucleolytic cleavage of phosphodiester bonds. One DNA incision occurs on the 3′ side to the lesion and the other on the 5′ side, thus releasing DNA damage as the component of single-stranded oligonucleotide segments (5). The excised nucleotides are replaced by DNA synthesis and ligation, thereby generating repair patches that, in human cells, extend over ∼30 nucleotides in length (6).

All of the core NER subunits necessary for the removal of bulky DNA lesions in bacteria (7), yeast (8), and mammals (9, 10, 11) have been isolated and characterized. These repair factors lack specificity for a particular type of damage and, hence, recognize a nearly infinite range of base abnormalities. UV light, for example, generates two different types of covalent cross-links between adjacent bases, i.e., CPDs and pyrimidine-pyrimidone (6-4) photoproducts, both of which are processed by the NER system (12, 13, 14). In view of this broad repair capacity, we analyzed the distribution of active human repair complexes between mixtures of chemically and structurally unrelated DNA lesions. Surprisingly, the combination of different standard substrates in human cell extracts showed that a minor fraction of damaged sites is able to monopolize a large proportion of NER complexes, whereas other concurrent lesions remain unrepaired. This finding provided a rational basis for engineering decoy adducts that are refractory to excision but nevertheless immobilize NER factors and compete for excision activity. The formation of such decoy adducts indicates that exposure to multiple DNA-damaging agents may result in hyper-additive genotoxic effects.

A single AAF adduct at the C⁸ position of guanine was generated by reacting the 19-mer oligonucleotide 5′-ACCACCCTTCGAACCACAC-3′ with N-acetoxy-AAF (15). A single platinum adduct at N⁷ of guanine was obtained by incubating the same oligonucleotide (40 μm) with 400 μm of [Pt(dien)Cl]Cl (Ref. 16; generously provided by Dr. Giuseppe Villani, CNRS, Toulouse, France). Alternatively, this reaction was carried out with the 19-mer 5′-ACCACCCTTAGTACCACAC-3′, in which the central sequence 5′-CGA-3′ was replaced by 5′-AGT-3′. After incubations at 37°C for 18 h in 10 mm Tris-HCl (pH 8.0) and 1 mm EDTA, the platinum-damaged oligonucleotides were recovered by ethanol precipitation. Complete modification was verified by a reduced electrophoretic mobility of the adducted oligonucleotides. M13 double-stranded DNA (7287 bp) containing a site-directed lesion was constructed by ligating 19-mer oligonucleotides into a gapped circular DNA intermediate (17). Plasmids pSPORT (4110 bp; Life Technologies, Inc.) were irradiated (254 nm, 450 J/m²) on ice in an open Petri dish in aliquots of 20 μl containing 35 nm DNA, 10 mm Tris-HCl (pH 8.0), and 1 mm EDTA. Subsequently, the DNA was incubated with Escherichia coli endonuclease III to remove plasmids that contain pyrimidine hydrates, and the residual closed circular substrate was separated from nicked DNA by 5–20% sucrose gradient centrifugation (18). Internally radiolabeled DNA fragments of 139 bp with a single base adduct in the center were constructed as described (19).

Plasmids pUC19 (2686 bp) containing 10.2 AAF adducts, 11.5 bulky UV lesions, 5.0 or 16.7 B[a]P adducts, and undamaged plasmids pUC19 were prepared as described (20). To obtain competitor DNA with 16.6 platinum cross-links per plasmid, pUC19 DNA (16 nm) was incubated with cis-Pt(NH₃)₂Cl₂ (5.4 μm) in 10 mm Tris-HCl (pH 8.0) and 1 mm EDTA for 24 h at 37°C. Competitor DNA containing 17.2 monofunctional platinum adducts per plasmid was obtained by reacting pUC19 DNA (16 nm) with [Pt(dien)Cl]Cl (4.3 μm). The extent of platinum modification was determined by atomic absorption spectroscopy (Elan 5000; Perkin-Elmer). CPDs were removed by photoreactivation using Anacystis nidulans photolyase and irradiation at 366 nm (21). The efficacy of photoreactivation was confirmed by incubating the plasmids with T4 endonuclease V (22).

Extracts were prepared from HeLa cells (23). Repair reactions (50 μl) were performed with 80 μg of cell extract proteins, M13 DNA substrate (0.2 nm), and competitor pUC19 DNA as reported (20). After repair at 30°C for 3 h, reactions were stopped by the addition of EDTA, and the samples were processed by incubation with RNase A and proteinase K. DNA was extracted, digested with AvaII, PstI, and SmaI, resolved on 20% native polyacrylamide gels, and [³²P]dCMP incorporation was visualized by autoradiography followed by laser scanning quantification. Alternatively, repair assays in human cell extracts were performed for 3 h with plasmids pSPORT (2 nm) and pUC19 (3 nm), as indicated in the legend to Fig. 3,C. To measure DNA repair synthesis, both plasmids were recovered and analyzed as outlined before except that DNA was linearized by BamHI and separated on 1% agarose gels. The oligonucleotide excision assay of Fig. 4 was performed by incubating internally labeled substrates (0.2 nm) in cell extract (19). Competitor fragments were added at a concentration of up to 1.8 nm. Excision products were separated on 10% denaturing polyacrylamide gels and visualized by autoradiography.

The relative affinity of the NER complex for different adducts was determined from the molar excess of damaged pUC19 DNA, at which repair of the M13 substrate is reduced to 50%, multiplied by the number of lesions per pUC19 molecule. For example, pUC19 DNA containing both CPDs and (6-4) photoproducts (a total of 11.5 bulky UV lesions per plasmid) was able to immobilize NER complexes 2.5 times less efficiently than a single AAF adduct, yielding a relative affinity that was 2.5 × 11.5 = 29 times lower for UV lesions compared with the AAF standard. To calculate the relative affinity of NER factors for CPDs (Aff_CPD), the following equation was used:

where Aff_UV = 1/29 is the relative affinity of NER factors for the mixture of all of the bulky UV lesions, including CPDs and (6-4) photoproducts, and Aff_(6-4) = 1/10 is the relative affinity for (6-4) photoproducts.

We developed a repair competition assay to analyze the distribution of human NER complexes between multiple forms of damage (Fig. 1,A). This assay requires substrate DNA, a source of human NER enzymes and competitor DNA. M13 double-stranded DNA containing a single AAF-C⁸-dG adduct was used as the NER substrate (Fig. 1,A). To allow for quantitative analysis of repair reactions, this AAF lesion was flanked by restriction sequences for SmaI (23 nucleotides away from the adduct in the 5′ direction) and PstI (13 nucleotides away from the adduct in the 3′ direction). The site-specific substrate was incubated in a standard human cell extract that contains the entire repertoire of NER factors (24, 25). After incubations of 3 h, DNA repair synthesis was monitored by measuring the incorporation of ³²P-labeled deoxyribonucleotides in the region of the substrate where the AAF adduct was located before excision, i.e., in the 37-bp SmaI-PstI fragment of M13 DNA (Fig. 1 B, Lanes 3 and 4). No DNA repair synthesis was detected in the SmaI-PstI region of undamaged M13 DNA (Lanes 1 and 2).

We combined M13 DNA containing the single AAF adduct with pUC19 DNA containing, on the average, 10.2 AAF adducts per molecule. When this mixture was incubated for 3 h in cell extract, repair patch synthesis in the 37-bp SmaI-PstI region of the substrate was suppressed (Fig. 1,B, Lanes 5–8). In dose response experiments, 50% inhibition of substrate repair was detected by adding AAF-damaged pUC19 DNA at a concentration of 0.02 nm, corresponding to a 10-fold molar excess of M13 substrate over pUC19 competitor (Fig. 1,C, Lane 6). Because each M13 DNA molecule carried only one AAF adduct, whereas pUC19 DNA contained on the average 10.2 adducts, these results showed that 50% inhibition of substrate repair is observed when there is a 1:1 distribution of DNA lesions between substrate and competitor DNA. Thus, the single site-directed AAF lesion in M13 substrate immobilizes NER factors as efficiently as the multiple AAF modifications located on pUC19 competitor. No reduction of substrate repair was observed when the reactions were incubated with undamaged pUC19 DNA (Fig. 1 C, Lane 3).

A comparison between base adducts of either cisplatin or a monofunctional platinum analogue confirmed that immobilization of NER proteins is strictly dependent on the ability of the NER complex to recognize a competitor lesion. Exposure of DNA to cis-Pt(NH₃)₂Cl₂ (cisplatin) results in the formation of 1,2-intrastrand platinum cross-links between the N⁷ positions of purine bases, accompanied by minor fractions of 1,3-intrastrand cross-links and interstrand cross-links (26). In contrast, the [Pt(dien)Cl]Cl analogue is unable to cross-link the bases and generates only monofunctional adducts at N⁷ of guanine (16). Plasmids pUC19 with ∼17 covalent modifications per molecule, either cross-links or monoadducts, were tested in the repair competition assay. We found nearly 50% inhibition of substrate repair in the presence of an 8-fold molar excess of cisplatin-damaged pUC19 over M13 DNA (Fig. 2,A, Lanes 4 and 5), but a stronger inhibition was induced with a 25-fold excess of cisplatin-damaged competitor (Fig. 2,A, Lanes 8 and 9). In contrast to cisplatin cross-links, monofunctional Pt(dien) adducts were essentially unable to immobilize NER factors such that repair of the AAF substrate was not reduced even when Pt(dien)-modified pUC19 was added in a 25-fold excess over M13 DNA (Fig. 2 B, Lanes 7 and 8).

The relative immobilization of NER factors because of a particular lesion was determined from the molar excess of damaged pUC19 DNA at which repair of the M13 substrate is reduced to 50%, multiplied by the number of adducts per pUC19 molecule (Fig. 2,C). Considering that half-maximal inhibition was detected at an 8-fold molar excess of cisplatin-damaged pUC19 DNA and that each pUC19 molecule contained, on the average, 16.7 cisplatin modifications, these results indicate that cisplatin cross-links immobilize NER factors 8 × 16.7 = 134 times less efficiently than AAF adducts. For comparison, we tested pUC19 plasmids containing DNA adducts generated by diol epoxide derivatives of B[a]P, a widespread environmental contaminant (Fig. 2,C). Repair of the AAF substrate was reduced to 50% when pUC19 plasmids with 5.0 B[a]P adducts per molecule were added in an 8-fold molar excess (Fig. 2,C). This ratio translates to a 5.0 × 8 = 40-fold preference of NER factors for AAF adducts compared with B[a]P adducts. After increasing the B[a]P adducts from 5.0 to 16.6 per pUC19 molecule, repair of the AAF substrate was already inhibited to 50% when B[a]P-damaged competitor was added in a 2.5-fold excess. This ratio indicates a 16.6 × 2.5 = 42-fold preference for AAF adducts compared with B[a]P adducts. Thus, the repair competition assay yields the same relative affinity of NER factors for a particular lesion, independently of the frequency of pUC19 modification. Consistent with previous observations (27), a single Pt(dien)-dG monoadduct in the SmaI-PstI region of M13 DNA induced DNA repair synthesis only when the platinum lesion was combined with base mismatches that enhance local helical distortion (Fig. 2 D, compare Lane 2 with Lanes 3 and 4). Thus, the human NER system fails to process a monofunctional platinum adduct in duplex DNA and this finding correlates with the failure of Pt(dien)-modified pUC19 plasmids to immobilize NER factors in the competition assay. We concluded that the repair competition assay provides a quantitative tool to monitor the immobilization of human NER proteins during their involvement in bulky lesion repair.

We exploited the repair competition assay to examine the interaction of human NER factors with mixtures of bulky adducts. For that purpose, the M13 substrate carrying the single AAF adduct was incubated with increasing amounts of pUC19 DNA that contained 11.5 bulky UV lesions per plasmid (Fig. 3,A). Half-maximal inhibition of substrate repair was found when UV-damaged pUC19 DNA was added in a 2.5-fold molar excess (Fig. 3,B), indicating that UV photoproducts immobilize human NER factors 2.5 × 11.5 = 29 times less efficiently than AAF adducts. This evaluation is complicated by the fact that UV-C irradiation induces CPDs and (6-4) photoproducts in a ratio of 3: 1 (12). Thus, the UV-irradiated pUC19 competitor DNA contained, on the average, ∼9 CPDs and 3 (6-4) photoproducts per molecule. To determine the contribution of each photoproduct, the plasmids were photoreactivated to remove CPDs and generate pUC19 competitor DNA with only (6-4) photoproducts. However, the inhibitory effect of UV-irradiated plasmids was only marginally diminished after complete removal of CPDs (Fig. 3, A and B), indicating that (6-4) photoproducts, and not CPDs, are mainly responsible for the immobilization of NER factors. By interpolation of the data plotted in Fig. 3 B, we found that a 3.4-fold excess of photoreactivated pUC19, containing 3 (6-4) photoproducts, was required to obtain 50% inhibition of AAF substrate repair. These numbers translate to a 3.4 × 3 = 10-fold higher affinity for AAF adducts relative to (6-4) photoproducts. Additional calculations indicated that CPDs immobilize NER factors 80 times less efficiently than AAF adducts (see “Materials and Methods”).

The strong bias of human NER factors for AAF-damaged DNA implied that the presence of AAF adducts could suppress the concomitant repair of UV lesions added to the same reaction mixture. To test this prediction, UV-irradiated (450 J/m²) and endonuclease III-treated plasmid pSPORT was coincubated in cell extract with pUC19 DNA. Excision of UV photolesions was determined by monitoring repair patch synthesis in the presence of ³²P-labeled dCTP. As expected, the repair of UV-irradiated pSPORT DNA was progressively inhibited by adding AAF-damaged pUC19 DNA (Fig. 3,C, Lanes 3–6). Interestingly, the loss of repair patches in UV-irradiated pSPORT was only partially compensated by a concurrent increase of DNA repair synthesis on pUC19 (compare Fig. 3,C, Lanes 1 and 2 with Lanes 5 and 6). Pretreatment with endonuclease III (followed by purification of closed circular DNA) is required to eliminate plasmids that contain pyrimidine hydrates. When these base excision repair substrates were left in pSPORT, the repair of UV radiation products was only partially inhibited by the addition of AAF-damaged pUC19 DNA (Fig. 3 D). Thus, the reduction of repair activity is because of immobilization of NER subunits, whereas other factors, in particular those involved in base excision repair, are not immobilized by AAF adducts.

Our results indicated that some AAF lesions exert an antagonistic activity on DNA repair without inducing any excision. This finding led us to search for DNA damage that, like the AAF adduct, is able to immobilize NER factors but, unlike the AAF adduct, is not amenable to excision. To construct such a prototype decoy lesion, we manipulated the nucleotide sequence environment at the site of damage. First, we noted that excision of AAF-dG adducts from DNA duplexes of 139 bp was reduced ∼10-fold when the dCMP residue located in the nondamaged strand opposite the modified guanine was removed (data not shown). This inhibition was even more pronounced when we performed the same experiment with B[a]P adducts. In cell extract, excision of the (+)-trans-anti-B[a]P-dG adduct (as a component of 24–32-nucleotides long products) was marginally affected when the cytosine opposite the damaged guanine was replaced by adenine (Fig. 4,A, compare Lanes 4 and 5). In contrast, excision of the same (+)-trans-anti-B[a]P-dG adduct was abolished when the dCMP residue across the lesion was deleted (Fig. 4,A, Lane 6). This observation was extended by exploiting the stereochemistry of B[a]P-dG lesions, because the adducts with cis configuration are excised more efficiently than the corresponding trans isomers (19). Interestingly, excision of the (+)-cis-anti-B[a]P-dG isomer was reduced in the dG·dA context (Fig. 4,A, compare Lanes 7 and 8), and excision of the same (+)-cis-anti-B[a]P-dG adduct was essentially abolished when the dCMP opposite the lesion was deleted (Fig. 4 A, Lane 9).

The conformation of deletion duplexes containing B[a]P-N²-dG lesions has been determined, thereby revealing that these adducts form wedge-shaped intercalation complexes (28, 29). Thus, all of the subsequent studies were performed with these highly defined B[a]P adducts instead of the AAF lesion. In particular, the complete lack of excision of adducts from the deletion duplexes allowed us to perform competition experiments in which a site-directed repair substrate was combined with these damaged deletion fragments. In the experiment of Fig. 4,B, DNA substrates (0.2 nm) containing either a (+)-trans-anti-B[a]P adduct or a (+)-cis-anti-B[a]P adduct were mixed with equimolar amounts of the respective competitor DNA fragments of identical length. Surprisingly, there was inhibition of repair of the (+)-trans-anti-B[a]P-dG substrates when the competitor DNA carried the same B[a]P adduct at a deletion site lacking the cytosine residue opposite the lesion (Fig. 4,B, compare Lanes 2 and 3). This inhibition was not observed when the substrate was coincubated with the same lesion in the dG.dA context used as a competitor (Fig. 4,B, Lane 4). Inhibition of repair was also detected when the (+)-cis-anti-B[a]P-dG substrate was mixed with equimolar amounts of competitor fragments carrying the same adduct opposite a deletion site (Fig. 4,B, compare Lanes 5 and 6). Again, this inhibition was not detected when the substrate was coincubated with the same lesion in the dG·dA context as a competitor (Fig. 4,B, Lane 7). Additional control experiments showed that deletion sites alone, in the absence of covalent B[a]P modifications, failed to inhibit excision repair activity (data not shown). Finally, internally radiolabeled substrates containing a site-specific (−)-cis-anti-B[a]P adduct were used to establish the dose response shown in Fig. 4,C. The addition of deletion duplexes containing the same (−)-cis-anti-B[a]P-dG adduct, but with a missing dCMP residue, in a 3-fold (Fig. 4,C, Lane 4) or 9-fold molar excess (Fig. 4,C, Lane 6) resulted in progressive inhibition of substrate repair. However, excision was only marginally reduced when the same amounts of unmodified duplexes were added to the assay mixtures (Fig. 4,C, Lanes 3 and 5). A quantitative evaluation of substrate repair as a function of competitor concentrations is shown in Fig. 4 D. These results demonstrate that human NER factors are susceptible to hijacking by specific decoy adducts.

Previous studies showed that the NER complex removes DNA adducts with vastly different efficiencies. In UV-irradiated human cells, for example, (6-4) photoproducts are repaired rapidly with 50% excision after 2 h or less, whereas CPDs are removed with a half-life of ∼8 h (12, 13, 14). Repair half-lives in the range of 2 h were also detected by monitoring the removal of AAF-dG adducts from human genomic DNA (30). A comparable relationship resulted from reactions performed in a human cell extract that excludes possible interferences because of chromatin assembly or transcriptional activity. These in vitro reactions confirmed that (6-4) photoproducts and AAF adducts are processed by human NER enzymes with similar efficiency, whereas excision of CPDs was promoted at slower rates (31). Thus, we used the same cell extract to compare the affinity of human NER factors for a series of bulky DNA adducts. Coincubation of different standard NER substrates in the same reaction mixture demonstrated that a single site-directed AAF adduct competes for a limited pool of human NER factors ∼10 times more efficiently than (6-4) photoproducts and nearly 100 times more efficiently than CPDs. This discrepancy between the recruitment of NER factors, where the AAF adduct is 10-fold more effective than (6-4) photoproducts, and the corresponding excision rates, with no significant difference between the two lesions, indicated that some AAF adducts cause immobilization of repair proteins instead of excision of the offending lesion. The notion that a fraction of carcinogen adducts may exert an antagonistic activity on DNA repair was supported by two subsequent experiments. First, the addition of AAF-damaged DNA suppressed the concomitant repair of UV-irradiated DNA, confirming that a subset of AAF adducts negatively regulates human NER activity. Second, we were able to construct a prototype decoy adduct that retains the ability to compete for NER activity but is completely refractory to excision.

The rational basis for engineering such a decoy lesion was prompted by the conformational variability of many carcinogen-damaged sites, including the AAF-C⁸-dG lesion and several B[a]P-N²-dG adducts (32, 33, 34). This polymorphism suggested that the various local structure may determine whether damage recognition is followed by excision or, alternatively, by immobilization of NER proteins without excision. To test this hypothesis, the double helical conformation at the site of covalent binding was manipulated by altering the local base composition. Surprisingly, we found that deletion of the dCMP residue opposite the tested B[a]P-dG adducts not only abolished excision but also conferred the ability to compete with repair of normal substrates added to the same reactions. Apparently, this dCMP deletion stabilized an adduct conformation that is endowed with the expected properties of a decoy lesion.

The prototype decoy adduct consisting of a bulky lesion opposite dCMP deletions has been identified as a mutagenic intermediate after the replicative bypass of AAF or B[a]P adducts. In fact, translesion synthesis across these adducts is facilitated by a mechanism in which one strand can slip with respect to the other strand, generating misalignment frameshifts in the daughter strand. Depending on the sequence context, single nucleotide deletions across the adduct constitute the predominant source of carcinogen-induced mutations (35, 36, 37). Such single nucleotide deletions have also been found in the chromosomal hypoxanthine phosphoribosyltransferase gene after exposing human cells to B[a]P diol epoxides (38). Thus, the type of decoy lesion identified in the present study is normally generated in many cases of environmental genotoxic stress. Furthermore, the discovery of such a prototype decoy adduct shows that human NER subunits are susceptible to hijacking and, hence, raises the possibility that particular fractions of genotoxic mixtures may induce bulky DNA lesions that exert negative effects on DNA repair. The resulting antagonism between substrate and decoy adducts could be responsible for hyper-additive interactions between multiple genotoxic agents.

Trapping of nucleotide excision repair factors by nonrepairable decoy DNA adducts is a new genotoxic reaction that has not been described before. Previously, Treiber et al. (39) and McLeod et al. (40) reported that certain transcription factors may bind to cisplatin cross-links or B[a]P adducts, thus protecting these lesions from DNA repair processes. Another study showed that poly(ADP-ribose) polymerase may obstruct DNA repair by interacting with DNA ends (41). Thus, the concept of protein hijacking was introduced to indicate that specific DNA lesions elicit toxic responses through sequestration of essential transcription factors (39, 40). The present study rules out any participation of transcription factors or other chromatin components and reveals a novel mechanism of repair inhibition because of direct immobilization of NER factors. As a consequence, excision complexes are titrated away from their sites of action on substrates that are normally susceptible to DNA repair.