Stereoselectivity of Human Nucleotide Excision Repair Promoted by Defective Hybridization*

To assess helical parameters that dictate fast or slow removal of carcinogen-DNA adducts, we probed human nucleotide excision repair (NER) activity with DNA containingl-deoxyriboses. Unlike natural lesions such as pyrimidine dimers or base adducts, l-deoxyribonucleosides (the mirror images of normal d-deoxyribonucleosides) involve neither the addition nor the loss of covalent bonds or functional groups and hence exclude modulation of repair efficiency by adduct chemistry and size. Previous studies showed that singlel-deoxyribonucleosides distort DNA backbones but are accommodated in the double helix with intact hydrogen bonding between complementary strands. Here, we found that such singlel-enantiomers are rejected as excision repair substrates in a NER-proficient cell extract. However, the samel-deoxyribose moiety stimulates NER activity upon incorporation into a nonhybridizing site of one or, more effectively, two base mismatches. In contrast to single l-deoxyriboses, multiple consecutive l-deoxyriboses interfere with normal hybridization; in this case, the intrinsic derangement of base pairing was sufficient to promote the excision of a cluster of three adjacentl-deoxyribonucleosides without any requirement for mismatches. Thus, using stereoselective substrates, we demonstrate the participation of a recognition subunit that guides human NER activity to sites of defective Watson-Crick strand pairing. This conformational sensor detects labile hydrogen bonds irrespective of the type of deoxyribonucleotide modification.

To assess helical parameters that dictate fast or slow removal of carcinogen-DNA adducts, we probed human nucleotide excision repair (NER) activity with DNA containing L-deoxyriboses. Unlike natural lesions such as pyrimidine dimers or base adducts, L-deoxyribonucleosides (the mirror images of normal D-deoxyribonucleosides) involve neither the addition nor the loss of covalent bonds or functional groups and hence exclude modulation of repair efficiency by adduct chemistry and size. Previous studies showed that single L-deoxyribonucleosides distort DNA backbones but are accommodated in the double helix with intact hydrogen bonding between complementary strands. Here, we found that such single L-enantiomers are rejected as excision repair substrates in a NER-proficient cell extract. However, the same L-deoxyribose moiety stimulates NER activity upon incorporation into a nonhybridizing site of one or, more effectively, two base mismatches. In contrast to single L-deoxyriboses, multiple consecutive Ldeoxyriboses interfere with normal hybridization; in this case, the intrinsic derangement of base pairing was sufficient to promote the excision of a cluster of three adjacent L-deoxyribonucleosides without any requirement for mismatches. Thus, using stereoselective substrates, we demonstrate the participation of a recognition subunit that guides human NER activity to sites of defective Watson-Crick strand pairing. This conformational sensor detects labile hydrogen bonds irrespective of the type of deoxyribonucleotide modification.
Mammalian nucleotide excision repair (NER) 1 eliminates DNA damage with several overlapping levels of heterogeneity. Active genes are processed faster than inactive loci, and the template strand of RNA polymerase II-transcribed genes is repaired faster than the coding strand (1)(2)(3)(4)(5). However, excision efficiencies may also vary in a damage-specific manner. For example, UV-induced cyclobutane pyrimidine dimers are processed at lower rates than the less frequently occurring pyrimidine(6 -4)pyrimidone lesions (6 -8). Additionally, a sequence-dependent variability occurs between closely related genomic sites. Along the p53 or phosphoglycerate kinase 1 gene of UV-irradiated human fibroblasts, repair rates of cyclobutane pyrimidine dimers differ substantially with nucleotide position (9,10). In the nontranscribed strand of these genes, some sites are less than 5% repaired within 24 h after irradiation, whereas other neighboring positions are repaired with up to 90% efficiency during the same period. Similarly, an analysis of repair kinetics along the nontranscribed strand of the human hypoxanthine phosphoribosyltransferase (HPRT) gene shows that excision repair in response to guanine adducts of benzo-[a]pyrene diol-epoxide varies by more than 1 order of magnitude (11). The excision of 1-nitrosopyrene-induced guanine adducts in the same region of the HPRT gene is equally variable but displays a distinctly different repair pattern (12).
In some cases, sites that are slowly repaired coincide with critical mutation hotspots in human cancer cells (9,13). Thus, a consequence of excision repair heterogeneity is that mutations are generated in a biased manner, i.e. at positions of longer-persisting DNA lesions. Despite this important clinical implication, differential repair in mammalian genomes is not completely understood; in particular, the mechanistic basis of damage-specific and sequence-dependent hierarchies of excision remains obscure. Undoubtedly, DNA repair is influenced by local chromatin compaction (14,15); however, at least in the human p53 gene, no clear correlation between the pattern of DNase I or micrococcal nuclease cleavage and the repair efficiency of UV-induced lesions could be detected (16). In view of these findings, we elaborated on the hypothesis that site-specific changes in DNA conformation may contribute significantly to the observed excision repair variability (17,18). As illustrated in Fig. 1, the mirror image of natural D-deoxyribonucleosides was introduced into DNA to generate excision repair substrates without adding new chemical bonds (such as in pyrimidine dimers) or new functional groups (such as in carcinogen-DNA adducts) and without removing normal DNA components (such as in abasic sites, urea residues, formamidopyrimidines, or other base fragmentation products). This strategy was prompted by the expectation that the repairability of a model lesion that differs from normal DNA constituents only in its inverted stereochemistry might be particularly susceptible to manipulation by altering the local double helical conformation. As a source of human NER factors, we used a standard soluble extract from HeLa cells that performs excision repair of DNA lesions in the absence of concurrent RNA polymerase II activity and is therefore thought to exhibit the same substrate selectivity as global genome repair operating on nontranscribed strands (19 -21).
Nuclear magnetic resonance studies showed that single Ldeoxyribonucleosides are accommodated into the DNA double helix, where they establish regular hydrogen bonds with complementary partners. This intrahelical insertion coupled to correct base pairing is achieved by extensive rotation of the backbone torsion angles not only on either side of the L-deoxyribose residue in the modified strand, but also in the same region of the opposite strand (22)(23)(24). Despite such substantially distorted DNA backbones, we found that single replacements of L-deoxyribose for D-deoxyribose are not detectably processed by the human NER system. However, exactly the same modification was a substrate of NER enzymes when combined with one or two base mismatches, implying that the site-specific loss of complementary hydrogen bonding is absolutely necessary to recruit human NER activity to L-deoxyribose residues. In support of this conclusion, we observed that three consecutive L-deoxyriboses, which disturb proper hydrogen bonding on their own, were excised even when located in a sequence of matching bases. This study, which was conducted in the complete absence of base adducts or base degradation, indicates that repair heterogeneity is determined by the involvement of a NER subunit that senses improper base pairing at sites of DNA damage.

MATERIALS AND METHODS
Enzymes-T4 polynucleotide kinase and T4 DNA ligase were purchased from Life Technologies, Inc. Restriction enzymes were from New England Biolabs. Creatine phosphokinase, ribonuclease A, and proteinase K were from Boehringer Mannheim.
Oligonucleotide Modifications-Heterochiral oligonucleotides containing single or triple L-deoxyriboses in the 19-mer sequence 5Ј-AC-CACCCTTCGAACCACAC-3Ј were synthesized using the cyanoethylphosphoramidite methodology as described previously (23). After deprotection with NH 4 OH, each product was purified by dissolving the crude compound in 1 ml of H 2 O followed by extraction with 10 ml of butanol. A site-directed acetylaminofluorene-C 8 -guanine (AAF-C 8 -dG) adduct was produced by incubating the 19-mer oligonucleotide 5Ј-AC-CACCCTTCGAACCACAC-3Ј or its analogs containing a single L-deoxyribose with N-acetoxy-2-acetylaminofluorene (purchased from the National Cancer Institute Chemical Carcinogen Reference Standard Repository) as described previously (25). Alternatively, a site-specific AAF-C 8 -dG was generated at the single guanine in the 11-mer sequence 5Ј-CCATCGCTACC-3Ј (18). The resulting AAF-modified oligonucleotides migrated on a 20% polyacrylamide gel markedly slower than the undamaged controls.
UV Melting Experiments-The melting temperature of unmodified or modified 19-mer duplexes (2.7 mM) was determined in 0.1 M Tris-HCl, pH 7.0, and 0.1 M NaCl using a Perkin-Elmer 554 UV-visible spectrophotometer equipped with a MGW Lauda RC5 temperature controller and a MGW Lauda R40/2 digital thermometer. Shortly before each melting experiment, the sample cuvettes were heated to 80°C and then slowly cooled (for 2 h) to 6°C to allow for reannealing. An electronic device generated a temperature gradient of 0.5°C/min from 10°C to 80°C, and UV absorbance was recorded at 260 nm. Moisture condensation at low temperatures was prevented by flushing the cell holder with nitrogen. Under these conditions, unmodified 19-mer duplexes yielded a melting temperature of 70.2°C.
Excision Repair Substrates-Internally radiolabeled DNA duplexes of 147 base pairs were obtained by ligating unmodified or modified 19-mer oligonucleotide (5Ј-ACCACCCTTCGAACCACAC-3Ј) with five other partially overlapping oligonucleotides ( Fig. 2A) as described previously (25,26). The sequence of the complementary 27-mer oligonucleotide was 5Ј-GCTCGTGTGGTTCGAAGGGTGGTTCAG-3Ј. Alternatively, this 27-mer sequence was replaced by 5Ј-GCTCGTGTGGTTTG-AAGGGTGGTTCAG-3Ј or 5Ј-GCTCGTGTGGTTTTAAGGGTGGTTCA-G-3Ј to obtain substrates containing one or two mismatches. The four flanking oligonucleotides of 59 -64 residues had exactly the same sequence as that described in Matsunaga et al. (27). For the construction of linear double-stranded substrates, the central 19-mer oligonucleotide (70 pmol) was 5Ј end-labeled with [␥-32 P]ATP (7,000 Ci/mmol; ICN Pharmaceuticals) and mixed with 100 pmol of each of the other five oligonucleotides that were phosphorylated with cold ATP. The mixture was annealed and ligated in the presence of T4 DNA ligase, followed by electrophoretic purification of the full-length fragments of 147 base pairs as described previously (21,26). In the experiment described in Fig. 5, we also employed a duplex substrate of 139 base pairs that was constructed using exactly the same method but starting with the acetylaminofluorene-modified 11-mer 5Ј-CCATCGCTACC-3Ј and the complementary 19-mer sequence 5Ј-GCTCGGTAGCGATGGTCAG-3Ј (18). Full-length substrates were stored in small aliquots at Ϫ80°C. Circular M13 DNA molecules containing site-directed modifications between the SmaI and PstI restriction sites were obtained by ligating unmodified or modified 19-mer oligonucleotides into a gapped M13 intermediate (25,28). Double-stranded covalently closed M13 substrates were purified by CsCl gradient centrifugation and stored in small aliquots at Ϫ80°C.
Human Excision Repair Assays-HeLa cells (American Type Culture Collection, Rockville, MD) were grown in RPMI 1640 medium supplemented with 7% fetal bovine serum (Life Technologies, Inc.). Cell-free extracts with typical protein concentrations of 10 -15 mg/ml were prepared by ammonium sulfate fractionation according to previously published methods (29). Oligonucleotide excision reactions (21, 26) contained (in 25 l) 35 mM HEPES-KOH, pH 7.9; 60 mM KCl; 40 mM NaCl; 5.6 mM MgCl 2 ; 2 mM ATP; 80 M each of dATP, dCTP, dGTP, and TTP; 0.8 mM dithiothreitol; 0.4 mM EDTA; 3.4% (v/v) glycerol; 5 g of bovine serum albumin; 5 fmol (75,000 dpm) of radiolabeled DNA substrate; and 50 g (in protein equivalents) of HeLa cell-free extract. After the indicated incubation times at 30°C, reactions were stopped by the addition of SDS to 0.3% (w/v) and proteinase K to 200 g/ml, followed by proteinase K digestion for 15 min at 37°C. DNA was purified by phenol-chloroform extraction and resolved by electrophoresis in 10% polyacrylamide denaturing gels, after which oligonucleotide excision products were visualized by autoradiography. The relative levels of excision were quantified by densitometric analysis of oligonucleotides in the 24 -32-mer size range on appropriately exposed x-ray films. The linearity of this densitometric quantification was confirmed by counting Cerenkov radiations of representative gel slices.
Repair synthesis was measured on site-specifically modified M13 DNA by a method that was modified from that of Hansson et al. (30). Reactions of 50 l contained HeLa cell extract (80 g of proteins); 100 ng of double-stranded M13 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% (v/v) glycerol; 2 mM ATP; 20 M each of dATP, dGTP, and TTP; 8 M dCTP; 2.0 Ci of [␣-32 P]dCTP (3,000 Ci/mmol); 40 mM phosphocreatine; 2.5 g of creatine phosphokinase; and 18 g of bovine serum albumin. After 3 h at 30°C, repair reactions were stopped by the addition of EDTA to 20 mM. The samples were incubated for 10 min at 37°C with ribonuclease A (80 g/ml). SDS to 0.5% (w/v) and proteinase K to 190 g/ml were added, and the mixtures were incubated for another 45 min at 37°C. M13 DNA was extracted; digested with AvaII, SmaI, and PstI; and analyzed by 20% polyacrylamide gel electrophoresis and autoradiography. The treatment with restriction enzymes produces SmaI-PstI fragments of 37 base pairs that contain the radiolabeled repair patches.

RESULTS
Excision of Single L-Deoxyribose Substitutions-NER activity was determined using the oligonucleotide excision assay devised by Huang et al. (21,26). This cell-free assay provides high levels of sensitivity and specificity because it exploits the unique dual DNA incision pattern of human NER. Appropriate DNA substrates of 147 base pairs with a site-directed modification in the sequence 5Ј-TCGA-3Ј were constructed from six different oligomers as illustrated in Fig. 2A. Before ligation, the central 19-mer oligonucleotide was labeled with [␥-32 P]ATP at its 5Ј end such that the resulting 147-mer duplex contained an internal radiolabel in the vicinity of the modified residue (Fig.  2B). After purification, the double-stranded fragments were incubated for 40 min at 30°C in reaction mixtures that contained a standard NER-proficient HeLa cell extract (19,20), ATP, and all four deoxyribonucleoside triphosphates. Dual DNA incision by human NER generates radioactive oligonucleotides (Fig. 2B) that were separated from substrate DNA by denaturing gel electrophoresis and visualized by autoradiography. In vitro repair reactions performed with linear 147-mer substrate containing a single AAF-C 8 -dG adduct in the sequence 5Ј-TCGA-3Ј yielded characteristic excision products that migrated on polyacrylamide gels as an oligomeric ladder with lengths ranging from 24 to 30 nucleotides (Fig. 3A, lane 6;  Fig. 3B, lane 1). No excision products were released from undamaged control DNA (Fig. 3A, lanes 1 and 7), although intact 147-mer substrate as well as the radioactive bands generated by nonspecific nuclease activity can be observed at similar levels at the top of the gel.
We then introduced a single L-deoxyribose at the thymine, guanine, or adenine position of the sequence 5Ј-TCGA-3Ј, but none of the tested deoxyribose replacements were able to elicit detectable NER activity in HeLa cell extract. In fact, incubation of linear substrate containing a single L-dG paired with dC yielded no oligomeric excision products (Fig. 3A, lane 3). Similarly, L-dA paired with dT (Fig. 3B, lane 2) or L-dT paired with dA (Fig. 3B, lane 3) failed to stimulate oligonucleotide excision. To rule out the possibility that the presence of a single Ldeoxyribonucleoside may not be compatible with productive binding of NER proteins to DNA, we also tested substrate molecules containing both L-deoxyribose and a unique AAF-C 8 -dG adduct in close proximity. For that purpose, 19-mer oligonucleotides comprising either a single L-dA or a single L-dT residue in the 5Ј-TCGA-3Ј sequence were exposed to N-acetoxy-2-acetylaminofluorene to generate an AAF-C 8 -dG lesion near the L-deoxyribose site (see "Materials and Methods"). After purification, the resulting oligonucleotides were incorporated into internally radiolabeled double-stranded fragments and incubated in HeLa cell extract for 40 min. As shown in Fig. 3B, these double modifications resulted in about 2-fold higher levels of oligonucleotide excision relative to the substrate containing the AAF-C 8 -dG adduct alone (compare lane 1 with lanes 4 and 5). The changing size pattern of excision products may indicate that in some cases either the AAF adduct or the L-deoxyribose residue is repaired, leading to mixed reaction products. Thus, L-deoxyribose substitutions exert stimulatory rather than inhibitory effects on the excision of adjacent AAF-C 8 -dG lesions.
On the other hand, the complete lack of human NER activity in response to single L-deoxyriboses (in the absence of concurrent AAF modifications) was confirmed by measurements of DNA repair synthesis in HeLa cell extract (Fig. 4). To that end, 19-mer oligonucleotides containing AAF-C 8 -dG, L-dA, or L-dT within the same sequence context 5Ј-TCGA-3Ј were incorporated into gapped M13 DNA. The lesion site in the resulting circular duplex was flanked by restriction sites for SmaI and PstI, as illustrated in Fig. 4A. Control DNA was obtained by ligating unmodified 19-mers into the SmaI-PstI region of the M13 substrate. Covalently closed M13 DNA was then incubated with HeLa cell extract in the presence of ATP, an ATP regenerating system, and all four deoxyribonucleosides triphosphates including [ 32 P]dCTP. After 3-h reactions at 30°C, M13 molecules were recovered, and DNA repair synthesis was monitored by visualizing the amount of radioactivity incorporated into the 37-base pair SmaI-PstI region of modified or unmodified M13 substrates (Fig. 4B). Consistent with the absence of oligonucleotide excision (Fig. 3B), we found that neither L-dA (Fig. 4B, lanes 2 and 3) nor L-dT (lanes 6 and 7) stimulated DNA repair synthesis over the background value detected in the SmaI-PstI site of undamaged control substrate (lane 1). However, M13 DNA containing an AAF-C 8 -dG adduct near L-dA (Fig. 3B, lanes 4 and 5) or near L-dT (lanes 8 and 9) was subject to higher levels of NER activity than the corresponding substrate containing the AAF adduct alone (lane 10). These results support the notion that a single L-deoxyribose component does not interfere with proper interactions between human NER factors and damaged DNA.
Previous studies demonstrating that single L-deoxyribonucleosides preserve normal Watson-Crick hydrogen bonding within complementary sequences (22)(23)(24) prompted us to test the same L-residues in combination with mismatched bases. We found that HeLa cell extract catalyzed oligomeric excision when a single L-deoxyribose was incorporated into an altered sequence containing two base mismatches (Fig. 3A, lane 5). A direct comparison with appropriate markers indicated that the excised oligonucleotides range from 24 to about 30 residues in length and hence constitute characteristic products of human NER activity. Considerably lower but nevertheless detectable amounts of excised oligomers were also found when the linear substrate with a single L-deoxyribose was combined with only one mismatch (Fig. 3A, lane 4). Importantly, no measurable oligonucleotide excision occurred in control reactions performed with substrate containing two mismatches (Fig. 3A, lane 2) or one mismatch (data not shown), but in the absence of L-deoxyribose replacements. In each case, we used laser scanning densitometry of appropriately exposed x-ray films to determine the relative level of specific 24 -32-nucleotide-long excision products. The quantification of three to four independent experiments shows that human NER activity in response to a single L-deoxyribose was stimulated weakly by the concomitant presence of one mismatch, but the reaction was stimulated strongly upon combination with two mismatches. In contrast, essentially no NER activity was detected in response to one or two mismatches alone (Fig. 3C). Because we introduced Ldeoxyribonucleosides as artificial model substrates for studying NER enzymes, the question of whether they could also serve as substrates for mismatch repair or base excision repair systems has not been investigated.
Excision of Clustered L-Deoxyribose Substitutions-Our results indicate that L-deoxyribose substitutions are processed by human NER only when their presence in DNA is accompanied by the disruption of normal base pairing interactions. To confirm this conclusion, we tested the repairability of a triple L-deoxyribose modification. In contrast to single L-residues, multiple consecutive L-deoxyribonucleosides have been shown to reject hybridization with complementary D-oligonucleotides FIG. 3. Excision activity in response to single L-deoxyriboses. Internally radiolabeled linear substrates (5 fmol; 75,000 dpm/reaction) containing an AAF-C 8 -dG adduct or L-deoxyribonucleoside variants (shown in bold letters) were incubated in HeLa cell extract for 40 min at 30°C. The reaction products were analyzed by polyacrylamide gel electrophoresis and visualized by autoradiography. A, comparison between AAF-C 8 -dG and a single L-dG incorporated in the indicated sequence environments. The lengths of major excision products were estimated from a 27-mer size marker. B, excision repair assay with linear substrates containing single L-dA or L-dT residues. C, quantitative evaluation of three to four independent experiments performed with substrates containing L-dG. The levels of specific oligonucleotide excision were determined by densitometric laser scanning, and the resulting repair activity is expressed as the percentage of oligonucleotide excision obtained with the AAF-C 8 -dG adduct. (31). Here, we estimated the double helical stability of DNA molecules containing a site of three consecutive L-deoxyriboses by assessing the melting temperature of complementary 19mer duplexes carrying this modification (see "Materials and Methods"). We found that the introduction of a triple L-deoxyribose substitution decreases the melting temperature to a similar extent (ϳ15°C) as three consecutive mismatches in a central position of the same 19-mer duplex (32). Thus, a sequence of three L-deoxyriboses is not compatible with normal hydrogen bonding between complementary base pairs, raising the expectation that such a cluster of three consecutive Ldeoxyriboses may constitute a NER substrate. We therefore incubated complementary DNA duplexes containing the three L-deoxyribose residues in HeLa cell extract and found that the triple L-deoxyribose site was removed with the formation of specific oligomeric excision products ranging in length from about 22 to 30 nucleotides (Fig. 5, lanes 3 and 4). The major oligonucleotides resulting from this repair activity are compa-rable in size to the excision products elicited by an AAF-C 8 -dG adduct in the same sequence 5Ј-TCGA-3Ј (compare with Fig.  3A, lane 6). We also noted that the triple L-deoxyribose lesion generates excision products that have a broader distribution but are nevertheless centered around the same size range as the corresponding products resulting from an AAF-C 8 -dG adduct in the different sequence 5Ј-TCGC-3Ј (Fig. 5, lanes 5 and  8). Finally, time course experiments indicate that the smaller excision products (Ͻ24 residues in length) generated in response to the triple L-deoxyribose site result from progressive oligonucleotide degradation during incubations of 20 min or longer (Fig. 5, lanes 9 -11). DISCUSSION Human NER is a multisubunit DNA damage-processing system that cleaves damaged strands on either side of the targeted lesion and generates oligonucleotide excision products of 24 -32 residues in length (33)(34)(35)(36)(37). This dual DNA incision and excision reaction is followed by DNA repair synthesis (to reconstitute the correct nucleotide sequence) and DNA ligation (to reestablish double helical integrity). The present study is concerned with the question of how the conformational properties of damaged DNA molecules may cause extremely heterogeneous excision activities in response to carcinogen-DNA adducts. Recent comparisons of substrate selectivity led us to suggest the existence of a conformational sensor that attracts the human NER system to lesion sites that interfere with Watson-Crick hydrogen bonding. In fact, we observed that critical NER factors are recruited up to 3 orders of magnitude more efficiently to helix destabilizing adducts than to helix stabilizing adducts (17). The involvement of a protein sensor that detects improper FIG. 4. DNA repair synthesis in response to single L-deoxyriboses. A, circular covalently closed DNA molecules containing sitespecific modifications were constructed by ligating 19-mer oligonucleotides into appropriately gapped M13 intermediates (see "Materials and Methods"). The resulting substrates (100 ng/reaction) were incubated in HeLa cell extract supplemented with [ 32 P]dCTP. After 3 h at 30°C, DNA was recovered from the reaction mixtures, and the flanking recognition sites for SmaI and PstI were used to generate fragments of 37 base pairs expected to contain the radiolabeled NER patches (double bands are generated by partial denaturation of these short fragments). Additional digestion with AvaII produces an adjacent fragment of 330 base pairs. These restriction products were then resolved by polyacrylamide gel electrophoresis and visualized by autoradiography. B, representative gel demonstrating the formation of NER patches in response to AAF-C 8 -dG adducts (lane 10) and double AAF and L-deoxyribose modifications (lanes 4 and 5 and lanes 8 and 9), but not in response to single L-deoxyriboses (lanes 2 and 3 and lanes 6 and 7). The nucleotide positions carrying the L-deoxyribose modification are shown in bold letters.

FIG. 5. Excision of triple L-deoxyriboses.
Internally radiolabeled linear substrates (5 fmol; 75,000 dpm/reaction) carrying three consecutive L-deoxyribonucleosides (shown in bold letters) were incubated in HeLa cell extract at 30°C. Lane 5 shows the excision of an AAF-C 8 -dG adduct located in a slightly different sequence context. Lanes 1, 2, 6, and 7 contain undamaged control substrate. The incubations of lanes 1-8 were stopped after 40 min. Lanes 9 -11 show a time course of oligonucleotide excision. Reaction products were analyzed by polyacrylamide gel electrophoresis and visualized by autoradiography. The lengths of major excision products were estimated from 19-mer and 27-mer size markers. base pairing is consistent with another study in which we established that fast excision of (ϩ)-or (Ϫ)-cis-benzo[a]pyrene diol-epoxide-N 2 -dG lesions is associated with complete disruption of local Watson-Crick hydrogen bonding interactions, whereas slow excision of (ϩ)-or (Ϫ)-trans-benzo[a]pyrene diolepoxide-N 2 -dG adducts correlates with the presence of partially normal hydrogen bonds between the modified guanine and its cytosine partner (18). One may argue, however, that the substrate selectivity observed in these previous studies may result from the different chemical structure of the tested adducts or, alternatively, from the distinct geometry of carcinogen-DNA bonding, rather than from their diverging effects on local base pairing interactions. To further dissect this problem, we have introduced the mirror image of natural D-deoxyribose constituents of DNA and obtained artificial model lesions that involve neither the addition nor the loss of covalent bonds or functional groups (Fig. 1). Earlier studies on L-deoxyribonucleosides have focused on the observation that L-dT and several L-nucleoside analogs are phosphorylated by the herpes simplex virus type 1 thymidine kinase and inhibit viral proliferation in infected cells (38). In this study, we have incorporated L-deoxyribonucleoside derivatives into DNA to analyze the mechanism of substrate discrimination by excision repair enzymes, thereby confirming that human NER activity is strictly dependent on the loss of hydrogen bonds at lesion sites.
Although there is considerable distortion of both backbones in the double helix, single L-deoxyribonucleosides maintain intact Watson-Crick pairing with complementary residues, and we observed that this stereochemical variant was processed in HeLa cell extract only upon combination with at least one base mismatch. In contrast to a single L-deoxyribose, multiple consecutive L-nucleosides are not compatible with proper Watson-Crick hydrogen bonding (31); in fact, we observed that a triple L-deoxyribose modification reduces the melting temperature of duplex DNA to a similar extent as three consecutive base mismatches within the same sequence. Molecular modeling using the Hyperchem program predicts that this local instability results from abnormal pairing geometries between the three L-nucleosides and their complementary D-residues such that the three atoms participating in each hydrogen bond (the donor, the hydrogen atom, and the acceptor) lose their straight alignment. 2 As expected from the experiments with single Ldeoxyriboses, a cluster of three L-deoxyribose residues was excised even in the absence of concomitant changes in base complementarity. Thus, the inherent instability of Watson-Crick interactions at the triple L-deoxyribose site obviates the need for mismatched sequences. Enhanced excision at sites of reduced hydrogen bonding capability was also found in recent studies in which cyclobutane pyrimidine dimers or cisplatin intrastrand cross-links were combined with base mismatches (39,40). Conversely, we observed that conformational distortions induced solely by one, two, or three adjacent mismatches were unable to elicit detectable oligonucleotide excision on their own (this study and Refs. 18 and 32). The lack of NER activity in response to mismatched bases is consistent with the recently proposed bipartite mechanism of damage recognition in which disruptions of base pairing are absolutely necessary but are not sufficient by themselves to promote NER activity. In addition to base pair disruption, DNA incision requires changes in the normal deoxyribonucleotide structure (32).
In summary, our analysis of substrate discrimination between normal D-deoxyribose and abnormal L-deoxyribose variants demonstrates that human NER activity is exquisitely sensitive to local base pairing interactions. In particular, effi-cient excision depends on reduced Watson-Crick hydrogen bonding at the lesion site and, as a consequence, varies with different adducts and different nucleotide sequences (8 -11, 41). This study provides additional evidence for the existence of a conformational sensor that monitors the thermodynamic stability of complementary hydrogen bonds and recruits the human NER system to sites of improper base pairing. Current research is aimed at the identification of this unique conformational sensor through systematic characterization of xeroderma pigmentosum group A protein, replication protein A, and other human NER factors.