Repair of Individual DNA Strands in the Hamster Dihydrofolate Reductase Gene after Treatment with Ultraviolet Light, Alkylating Agents, and Cisplatin”

We have analyzed gene-specific and strand-specific DNA damage and repair in the dihydrofolate reductase gene in hamster cells. Cells were UV-irradiated or treated with two types of chemotherapeutics, alkylat- ing agents or cisplatin. UV-induced pyrimidine dimers were detected using a previoulsy published technique in which the T4 endonuclease V enzyme is used to create nicks at the lesion sites. 6-4 photoproducts were detected in a similar assay using ABC excinuclease after prior reversal of the pyrimidine dimers with photolyase. Adducts formed by the alkylating agents nitrogen mustard and dimethyl sulfate were quantitated by generating strand breaks at abasic sites after neutral depurination. Cisplatin-induced intrastrand adducts were detected with ABC excinuclease, and cisplatin interstrand cross-links were detected using a denaturation-reannealing reaction before electropho- resis. In accord with previous reports by other investigators, we find distinct strand specificity of the re- pair of pyrimidine dimers after UV; the transcribed strand was much more efficiently repaired than the nontranscribed strand. In contrast, there was little or no strand bias in the repair of the 6-4 photoproducts.


Present address: Laboratory of Molecular Genetics, National
Institutes on Aging, NIH, 4940 Eastern Ave., Baltimore, MD 21224. the general conclusion that there is more efficient repair of pyrimidine dimers (PD) ' in transcriptionally active genes than in inactive genomic regions or in the genome overall. It was initially discovered that the active, essential DHFR gene was efficiently repaired in hamster cells (l), where there is little repair in the overall genome. This phenomenon has been termed preferential DNA repair and has been found in a number of active genes in mammalian cells, in lower eukaryotes, and in bacteria (2-4). Using the same general technique, but probing the membranes with single-stranded RNA probes (riboprobes), it was shown that the preferential DNA repair in the mammalian DHFR gene of UV-induced P D was largely due to repair in the transcribed strand, i e . there was little or no repair in the nontranscribed strand (5). Other investigators have since found strand specificity of repair of UV-induced pyrimidine dimers in other genes in yeast (6) and human cells (7). Since strand selectivity of repair of pyrimidine dimers has also been reported in Escherichia coli (8), this process might be a general feature of the fine structure of DNA repair of pyrimidine dimers.
It has been observed for several mammalian genes that there is a strand bias for mutation fixation that may be due to a strand selectivity of DNA repair. For example, in the human HPRT gene, strand bias for mutation has been shown to be due to strand selective repair resulting from the persistence of unrepaired premutagenic lesions in the nontranscribed DNA strand (9-11). In the hamster DHFR gene, mutations induced by (f)-3a,4P-dihydroxy-la,2a-epoxy-1,2,3,4-tetrahydrobenzo[c]phenanthrene are also preferentially located in the nontranscribed strand (12). There is a reasonable correlation between the strand bias of the DNA repair and the strand bias of the mutations as examined collectively in the HPRT gene after UV damage to the cells (13). One limitation in such studies is that whereas gene-specific repair can now be measured separately for the two major UV photoproducts, the type of photoproduct that caused the mutation in repairproficient cells cannot easily be distinguished. In a study by McGregor et al. (lo), a predicted strand bias for mutation fixation was found in repair-proficient cells but not in repairdeficient Xeroderma pigmentosa complementation group A cells. This supports the notion that the strand-selective repair process is responsible for the strand bias of the mutations in repair-proficient cells. It is important, however, to consider The abbreviations used are: PD, pyrimidine dimers; DHFR, dihydrofolate reductase; HPRT, hypoxanthine guanine phosphoribosyltransferase; 6-4 PP, 6-4 photoproducts; IA, cisplatin intrastrand adduct; ICL, cisplatin interstrand cross-link; T4 endo, T4 endonuclase V; CHO, Chinese hamster ovary; kb, kilobase. that since these mutations are fixed by replication, the strand bias could also arise from a bias in replication after damage.
UV (254 nm) irradiation of cells causes the formation of two major photoproducts in DNA. The predominant lesion (65-85% of total) is the cyclobutane PD, while the less common (10-30%) is the 6-4 photoproduct (6-4 PP). Since 6-4 P P s are cytotoxic and mutagenic in mammalian cells (14-161, we recently developed an assay for the measurement of 6-4 PP repair in specific genes. This approach involved the reversal of UV-induced PDs with photolyase followed by cutting of the remaining UV photoproducts (almost exclusively 6-4 PPs) with E. coli ABC exinuclease (17).
We have also developed techniques to detect lesions caused by alkylating agents and cisplatin in specific genes. Cisplatin is the prototypical heavy metal compound with anti-tumor activity. The most frequent cisplatin DNA lesion is the intrastrand adduct (IA) between adjacent purines (GG or AG) (18), whereas the interstrand cross-links (ICL) account for only about 1% of the total DNA lesions (19). We detect the cisplatin IA by cleavage with the ABC excinuclease as described above, and the cross-links are detected by a denaturation-reannealing method.
Alkylating agents include simple monofunctional methylating agents such as dimethyl sulfate and bifunctional agents such as nitrogen mustard (HN2), which forms both monoadducts and interstrand DNA cross-links. The predominant lesion is the alkylation of the Af position of guanine, but many other base modifications are possible, and the spectrum varies for the individual agents (20). In our assay, the I P methylations are detected by cleavage in a two-step reaction, neutral depurination followed by alkaline hydrolysis.
We have examined the strand selectivity of DNA repair for lesions other than pyrimidine dimers for two main reasons: 1) to determine whether the enzymology involved in the genespecific repair of pyrimidine dimers differs from that involved in the repair of other types of damage, which addresses the question of whether strand specificity is restricted to the repair of pyrimidine dimers or is a general feature of mammalian DNA repair; 2) if the strand specificity of DNA repair correlates with strand specificity of mutations after several types of damage, it would support the general hypothesis that the mutational bias is caused by the strand selectivity of the repair process. We have thus examined the strand specificity of the DNA repair in our model gene, DHFR, after treatment of hamster cells with UV (PD and 6-4 PP), alkylating agents (HN2 and dimethyl sulfate), and cisplatin.

MATERIALS AND METHODS
Cell Lines, Culture Conditions, and Probes-The Chinese hamster ovary (CHO) cell line CHO-B11, which contains an amplified DHFR xanthine, or thymidine (GIBCO), supplemented with 10% dialyzed gene (21), was grown in Ham's F-12 medium without glycine, hypofetal calf serum, and maintained in 500 nM methotrexate. The cells were subcultured to ensure exponential growth 1 day before irradiation. ABC excinuclease and DNA photolyase were purified as described elsewhere (22, 23). The T4 endonuclease V (T4 endo) was produced from an overproducing strain provided by Dr. DeRiel, Temple University. The DNA probes were as previously described (24). The pMB5 probe was used to detect the CHO DHFR 14-kb KpnI fragment, which is shown at the top of Fig. 1.
Construction of Plasmid pZ3d8"Plasmid pGEM-3f(-) was digested to completion with AuaI, and the resulting linear product was t.reated with bacterial alkaline phosphatase (International Biotechnologies) prior to phenol-chloroform extraction and concentration by ethanol precipitation. And AuaI fragment of plasmid pMB5 (25) containing the first two exons of the DHFR gene ( Fig. 1) was purified from low melting temperature agarose after restriction digestion and electrophoresis of pMB5 DNA and ligated to the phosphatased, AuaIcut pGEM-3Zf(-) DNA. The ligation product was diluted and used . The 14-kb KpnI fragment, which is probed for DHFR-specific repair, is shown above the scale at the very top of the figure. Plasmid pZ3d8 was constructed by cloning an AuaI (Au) fragment from plasmid pMB5 (25) into the multiple cloning site of plasmid pGEM-3Zf(-) (Promega) as described in the text. Transcription from the T7 promoter using BamHI-digested pZ3d8 as template produces a single-stranded RNA probe for detection of the transcribed DHFR strand, whereas transcription from the SP6 promoter using KpnIdigested pZ3d8 produces a probe for the nontranscribed DHFR strand. Other restriction sites indicated in the figure are SmaI ( S m ) , Sac11 ( S ) , and XmnI (X).
to transfer competent DH5a cells to ampicillin resistance using Xgal (5-bromo-4-chloro-3-indolyl-~-~-galactopyranoside) screening to identify white recombinant clones (26). A lysozyme-boiling minilysis technique (27) was used to prepare DNA from small cultures and led to identification of plasmid pZ3d8. Large scale cultures were grown to produce plasmid DNA as described by Humphreys et al. (28).
Analysis of PD and 6-4 PP Repair in the DHFR Gene-The initial steps of the assay were as described by Bohr and Okumoto (29). The CHO cells, grown in monolayer, were irradiated with UV doses (254 nm) ranging from 20 to 60 J/m2. Cells a t 0-h repair were lysed immediately, whereas those for repair time points were incubated in bromodeoxyuridine and fluorodeoxyuridine to density label the DNA replicated after damage. After UV irradiation of the cells and DNA isolation, samples were treated with restriction endonuclease, and the parental DNA was separated on CsCl gradients (29). The DNA was divided into two portions. The first was treated with T4 endo to measure PD as described (29). The other portion was treated with DNA photolyase and light, which specifically and quantitatively reverts PD (30). The remaining, non-dimer photoproducts are almost exclusively 6-4 PP (31). Samples treated with photolyase were preincubated a t room temperature in the dark in a buffer containing 50 mM Tris, pH 7.5, 100 mM NaCl, 1 mM EDTA, 10 mM P-mercaptoethanol, and 0.25 fig of photolyase/pg of DNA. The samples were then placed in a small reaction tube and irradiated in a monochrometer at 405 nm for 30 min at room temperature. The DNA was extracted with phenol and phenol-chloroform followed by ether, precipitated with ethanol. and resuspended in T E (10 mM Tris, 1mM EDTA, pH 8.0) to a Concentration of approximately 0.2 pg/pI. The photoreactivated DNA was then divided into three aliquots. One aliquot was treated with T4 endo to demonstrate that virtually all of the PD were repaired by the photoreactivation treatment. Another was treated with AHC excinuclease to excise the DNA at the remaining adducts. A third was used as a control. The three samples were run in parallel on an alkaline-agarose gel, transferred to a support membrane, and probed for a specific DNA sequence. The treatment of the DNA with ' 1' 4 endo and ARC excinuclease and the preparation of the samples for alkaline gel electrophoresis, Southern transfer, and hyhridization have been described previously (29). Autoradiograms of the membranes were quantitated hy densitometry or by a Betagen hlot analyzer.
The relative intensities of parallel bands representing full-length fragments were compared to determine adduct frequency (29). We have measured the repair of 6-4 P P in a 14-kh Kpnl restriction fragment, which constitutes the 5"coding portion of the DHFR gene ( Fig. 1 ).
Gene-specific h m a g r and Repair with Alkylating Agents-Cells were preconditioned in fresh medium without methotrexate and supplemented with 1% fetal calf serum for 30 min. They were then incubated with HN2 or dimethyl sulfate at 37 "C for 30 min. After drug exposure, cells were washed twice with phosphate-buffered saline and either lysed immediately or after repair incubation in a solution o f 10 mM Tris, 1 mM EDTA, 0.5% sodium dodecyl sulfate, and 0.3 mg/ml proteinase K (16 h at 37 "C, pH 8.0). For repair, cells were incubated in the presence of 10 p~ hromodeoxyuridine and 1 p M fluorodeoxyuridine. Before running the alkaline gels, the depurination and alkaline hydrolysis were accomplished as follows. Duplicate portions of each sample, containing 2 pg of DNA, were heated a t 70 "C for 30 min in a citrate-phosphate buffer, pH 7.0, to maximize the conversion of all heat-lahile alkylated bases to apurinic sites. Then the samples were treated with 0.1 M NaOH (final concentration) a t 37 "C for 30 min to cleave the DNA at apurinic sites.
Gene-specific Damage and Repair of Cisplatin Intrastrand Adducts-The cells were treated with cisplatin for 1 h in medium with low serum (0.2%) at 37 "C at the indicated concentrations. One-h treatment was optimal to detect the IA, and 5-hr treatment was optimal in order to reach maximal formation of ICL. Repair incubation, lysis, DNA extraction, restriction, and CsCI gradients were as described above for PD. Specific DNA regions were probed for the presence of cisplatin IA by treatments with ABC excinuclease following the procedure described by Thomas et al. (32). Preparation of the samples for alkaline gel electrophoresis, Southern transfer, and hybridization have been described above.
1)etection of Cisplatin Interstrand Cross-links-Cells were treated for 5 h to get maximal formation of ICL. A denaturation-renaturation procedure modified from a reported assay (33) was used to detect interstrand cross-links in specific DNA regions. After the denaturation-renaturation treatment, cross-linked DNA readily anneals, while the noncross-linked DNA remains single-stranded. The DNA samples were mixed with 10 X loading buffer and dye for a final concentration o f 0.26% Ficoll, 0.1 mM EDTA, and 0.0025% bromcresol green and loaded on a 0.5% neutral agarose gel buffered in 40 mM Tris acetate and 2 mM EDTA. The denatured DNA samples were loaded in parallel with the corresponding, nondenatured samples. Electrophoresis was carried out a t 26 V for 16-20 h with buffer circulation. Southern transfer and hybridization were as described above. Riboprohe Analysis of Strand-specific Repair in the DHFR Gene-Plasmid pZ3d8 was digested with restriction enzymes RamHl (when using T7 RNA polymerase) and Kpnl (when using SP6 RNA polymerase) to generate templates for strand-specific rihoprobes as illustrated in Fig. 1. Linear plasmid DNA was phenol-extracted, ethanolprecipitated, and resuspended in water a t a concentration of 1 pg/pl, then frozen in small aliquots and thawed just before use. Reactions with T7 and SP6 RNA polymerases to produce transcribed (coding) and nontranscrihed (noncoding) strand probes, respectively, used 2 pg of plasmid DNA template/reaction and were carried out with ['"PI C T P using Hoehringer Mannheim SP6/T7 transcription kit reagents and protocols. After transcription reactions, radiolabeled rihoprobes were isolated using Stratagene push columns following DNase treatment (37 "C for 15 min) to digest plasmid template. Prehybridization was overnight at 48 "C in buffer consisting of 0.25 M sodium phosphate, 0.25 M NaCI, 0.0001 M disodium EDTA, 50% formamide, 7% SDS. 5% polyethylene glycol 8000. containing carrier calf thymus DNA and yeast tHNA a t 50 pg/ml each. Hyhridization was conducted overnight at 48 "C in fresh huffer after adding rihoprohe heated at 65 "C for 15 min in 1 ml of hyhridization buffer. After hyhridization. membranes were washed in 2 X SSPE for 5 min at room temperature. then twice in PSE (0.25 M sodium phosphate, 001 M EDTA, pH 7.2, containing 2% SDS) at 65 "C for 15 min, followed hy 1-2 washes in PES (0.04 M sodium phosphate, 0 . 0 1 M EDTA. pH 7.2. containing 1% SDS) at 65 "C for 20 min each wash. Heat-inactivated HNase (Boehringer Mannheim 1119915) was then added to 2 X SSI'E f 19 units/250 ml), and the membrane was washed in this solution for 2 min at room temperature to remove nonspecifically hound prohe. After autoradiography, the resulting bands were analyzed as descrihed for douhle-stranded prohing.

RESULTS
Strand Repair of Pyrimidine Dimers- Fig. 2 shows the Southern blot analysis of the formation and repair of PD in the transcribed and nontranscribed strand of the CHO DHFR gene using riboprobes from pZ3d8. Each blot was first probed for the transcribed strand, deprobed, and then probed for the nontranscribed strand. Lanes are run in parallel with or without treatment with the T4 endo. The control, 0-h, and no-UV doses show that there is no nonspecific cutting by T4 endo. There is increased cutting by T4 endo in the treated lunes of the 20-60 J/m2 samples. The important lanes to note are the 40 J/m2 24-h time points (+) treated with T4 endo. Whereas a band is visible indicating repair in the transcribed strand, there is no repair in the nontranscribed strand. The quantitation of this analysis is shown in Table I   were published previously (17). The new results presented here clearly show that there is strand selectivity of the repair o f P D in the DHFR gene. The repair is confined to the coding strand with almost no repair in the noncoding strand. When t h e filter was reprobed with a double stranded DNA probe generated by multiprimer labeling, the repair measurements comprised an average between the two strands.
Strand Repair of 6-4 Photoproducts- Fig. ' 3 shows the Southern blot analysis of 6-4 PP repair in the two strands of the DHFR gene, and the quantitation is shown in Table 11. For each time point, we have included the photolyased sample treated with T4 endo as a control for the efficiency of the photolyase treatment. It is evident from the autoradiogram that the photolyase reaction was complete, and this is confirmed by the quantitation that shows it went to >90% completion. The density of the photolyased, ARC excinuclease-treated band where the repair is monitored is seen to increase with time, and this occurs both in the transcribed and nontranscribed strands. As see in Table 11, the initial level of 6-4 PP is very similar in the two strands, and the repair efficiency is similar in the transcribed and nontranscribed strands. When the filter was probed with the regular double-stranded DNA probe, the repair efficiency was calculated as an average between the two strands. The table shows the results of two separate biological experiments.
In of P D in the DHFR gene, and whereas there was distinct strand selectivity for PD repair (Fig. 4H). strand selective repair of 6-4 PPs was not evident (Fig. 4A ). I t is difficult to determine whether the small difference observed between the two strands in Fig. 4A. particularly at 4 h, represents strand bias or whether it is due to variability in the assav. Fig. 5 shows blots of strand repair of HN2 adducts in the hamster DHFR gene after treatment of the cells with 200 p~ HN2 for 30 min. farallel lanm 1 and 2 are identical samples. thus providing a control for loading. T h e control 1anc.s ( C ' ) have not been depurinated, whereas all other samples have undergone neutral depurination and alkaline hydrolysis. The reappearance of hand (repair) is readily visible with time after the 0-h time point. Helow the blot. we have plotted the repair efficiency. I t is seen that at all time points there is more efficient repair in the transcribed than in the nontranscrihed strand, although the difference is minimal. When probing with the double-stranded probe, the results fall between the data for the individual strands. The results (see Table  111) with HN2 are based on two separate biological experiments and several individual gels. For dimethyl sulfate (Fig. 6), the composition of the blot is similar to that for for HN2, and the repair is also similar in the two strands, although there is again more repair in the transcribed than in the nontranscribed strand. Cells were treated with dimethvl sulfate at 1.50 p~ for 30 min. The data (Table  111) are based on several independent determinations from one biological experiment.

Strand Repair of Nitrogcn Mustard and Dimethyl Sulfatc-
Strand Repair of Cisplatin Intrastrnnd Adducts and Intrrstrand Cross-links- Fig. 7 shows the damage and repair in the individual DNA strands of cisplatin IA. Nontreated DNA is not shown, but the nonspecific cutting with ARC: excinuclease did not exceed 10%. The repair is seen as the reappearance of hands in the ARC excinuclease-treated lancs from 0 to 24 h. The data are shown in Table IV. There is less repair in the nontranscrihed strand, although since the nontranscribed blot was exposed longer than the transcrihed blot, this is not readily visihle. T h e plot shows ahout 2-fold more repair in the t ranscrihed strand than in the nontranscrihed strand. The difference is not as dramatic as for pyrimidine dimers, hut we consistently found this strand hias in several gels from two different hiological experiments. Fig. 8 shows the results from the cisplat in ICL assay. Samples were run hefnre and after denaturation. In the denatured samples, the repair can he seen as the removal o f the douhle-stranded ( 1 ) ) I)NA and a corresponding increase in single-stranded ( s ) native DNA. A s seen on the plot. Fig. 8. and from the data in Tahle IV, there is no difference hetween the two DNA strands with regard t o the repair of cisplatin ICL.

DISCIISSION
In this study, we have analyzed the strand hias of D N A repair after different types of DNA damage. \Ye find that for some lesions, pyrimidine dimers, and cisplat in intrastrand adducts there is a distinct hias of the repair with a preferential removal of lesions in the transcrihed strand. For a numher of other lesions, the difference hetween the strands is minimal or not. present. Hased on this data we conclude that there are differences in the DNA repair pathways of these different t+ypes of damage.
In this study, we demonstrate that  PI' repair in the CHO DHFR gene shows little or no strand selectivity. in contrast to the pronounced strand-selective repair o f 1'1)s. Our results may have important implications for hot h I)SA repair and mutagenesis. The lack o f distinct strand specificity of the gene-specific repair of 6-4 1'1' as contrasted with the I' D supports other information to suggest that there are different pathways of repair for these t w~ I'\."indrlced photoproducts. The mode of repair may in some manner he dependent upon the precise distortion that these agents inflict upon the chromosome structure.
In addition, we have studied the strand-specific repair (~f After purification, restriction, and CsCl gradient centrifugation, the samples were treated with ABC excinuclease (ARC E N ) , which cleaves a t sites of the intrastrand adducts. The damage and repair was quantitated in both strands, and the strand bias of repair is shown in the plot. Coding (0). transcribed; noncoding (O), nontranscribed. -+ * + -+ + + ----+ + + + some alkylating agents and of cisplatin-induced lesions. For HN2 there is more efficient repair in the transcribed strand, although the bias is small. For dimethyl sulfate the bias is even less: this is in accord with a previous study (34). However, although the bias is small, there is consistently more efficient repair in the transcribed than in the nontranscrihed strand for both alkylating agents. Cisplatin ICL are repaired with equal efficiency in the two strands as would be expected. For cisplatin IA, we find a strand bias with more efficient repair of the transcribed strand. This resembles the situation for repair of PDs, although the strand bias is less distinct. In evaluating the strand bias of the repair of different lesions, it is important to consider that the hamster cells are deficient in the overall genome repair of pyrimidine dimers but not in the overall genome repair of other types of damage such as cisplatin I A (35). The difference between the strands in a repair-proficient background would thus be expected to be smaller than we find in the repair-deficient background using UV. Given this difference in overall genome repair, the relative strand bias observed here for the repair of cisplatin I A may be of the same magnitude as that of pyrimidine dimers.
For most of the lesions examined, the frequency of initial adducts formed in the transcribed and in the nontranscribed strand is very similar, but for HN2 there are slightly more lesions formed in the nontranscribed strand. This is in accord with our previous observations of differences in the genespecific frequency of induced nitrogen mustard lesions (2) (but not of lesions induced by several other agents) and with data obtained by Barlett et al. (36) that also report a small heterogeneity in the formation of these lesions within the genome.
These experiments were done at doses currently used for studies of DNA damage and repair in general as well as for studies of this type. It would be preferable, but rather unfeasible, to perform a dose-response repair analysis using a wide concentration spectrum for all the compounds we have used. Our techniques are, unfortunately, limited to these relatively high doses for these experiments, since we need to be able to detect in the range of 0.5-4 lesions in each fragment that we are analyzing. Although it is a limitation that we cannot do our studies at significantly lower levels of compounds than used, we consider it an advantage of this work that we are comparing the repair in a situation where the initial level of lesions introduced by the different compounds used is fairly similar.
Our findings, taken together with those previously reported, suggest that strand specificity of DNA repair is not a universal aspect of the DNA repair mechanisms but rather it is limited to the repair of certain adducts. However, it is noteworthy that for the lesions where we find no significant strand bias, we consistently find more efficient repair in the transcribed than in the nontranscribed strand. For most lesions, there is good correlation between the present results and our previous findings of preferential repair in the hamster DHFR gene as compared with a nontranscribed, downstream region. For example, for cisplatin IA, we previously reported that there is preferential DNA repair (35), and we now find that there is also strand-selective repair. Preferential repair of cisplatin IA was not as distinct as for pyrimidine dimers, and the strand selectivity observed is also not as distinct. Dimethyl sulfate was not preferentially repaired (37) and is also not strand selectively repaired. Cisplatin ICLs were not preferentially repaired (35) and are also not strand selectively repaired. Pyrimidine dimers were preferentially repaired (l), and they are also strand selectively repaired. In contrast, we reported that there was preferential repair of 6-4 PP (17) and of HN2 (37) in this hamster DHFR locus. For both of these lesions, we now find that there is little or no strand selectivity of the repair. This may indicate that there are several repair mechanisms and that the nontranscribed strand is repaired in a different mode than the noncoding DNA.
Several experients have shown that the efficiency of repair in a gene can correlate with its transcriptional activity. Genespecific DNA repair efficiency of pyrimidine dimers can be modulated by transcriptional induction of structural genes such as in the case of the hamster of human metallothionein genes (38, 39), and inhibitors of transcription can block the gene-specific repair (40). As with UV-induced pyrimidine dimers, gene-specific repair of alkylation damage can also correlate with transcriptional activity; in rat cells, there is substantially more repair of N-methylpurines in the active insuline gene than when it is not transcribed (41).
The strand specificity of pyrimidine dimer repair discussed above further documents the association between DNA repair processes and the transcription complex. This association appears to apply better to the strand selectivity of the repair than to the preferential gene-specific repair. An E. coli cell-free system has recently been described (42) in which geneand strand-specific repairs can be studied. It was found that the strand-specific repair was coupled much stronger with transcription than was the gene-specific repair. Whereas the strand specificity of the repair of pyrimidine dimers appears to be closely associated with transcription, there are data suggesting that the gene-specific repair efficiency may not always correlate with transcriptional activity (43). In addition, in differentiating mouse cells, it has been demonstrated that the repair of some genes correlates with the stage of differentiation rather than with gene expression (44).
It is possible that the strand bias of repair is related to the degree of transcription blockade caused by a lesion. UVinduced pyrimidine dimers are known to block transcription, whereas there is much less information about the blockade caused by the other lesions we have studied. Another possibility is that strand bias correlates with the degree of chromatin distortion invoked by the lesion.
Recently, findings on the mutational spectrum found in specific genes after DNA damage have been related to the strand specificity of DNA repair. Our results showing that in the CHO B-11 hamster DHFR gene there is no strand bias for repair of 6-4 PPs, as opposed to a distinct transcribed strand selectivity for repair of PDs (Fig. 4), suggest that repair of 6-4 PPs should not affect the strand distribution of UV-induced base substitution mutations in mammalian cells. However, it is important to note that the studies on UV-induced mutation spectra were done at a 10-fold lower dose than our studies and that it is likely that this difference in dose makes it problematic to directly compare the results from these different approaches. At this time, it is not possible to perform direct gene-specific repair studies of 6-4 PP at lower doses than those used here because of the need to induce a sufficient number of lesions into the restriction fragment assayed. It would obviously be desirable to develop methodology that would allow such repair experiments to be conducted at lower doses. Results of such studies should aid in further understanding of the roles of PDs and 6-4 PPs in UV toxicity and mutagenesis in mammalian cells.