Transcription-coupled DNA Repair Is Genomic Context-dependent*

DNA damage is preferentially repaired in the transcribed strand of many active genes. Although the concept of DNA repair coupled with transcription has been widely accepted, its mechanisms remain elusive. We recently reported that in Chinese hamster ovary cells while ultraviolet light-induced cyclobutane pyrimidine dimers (CPDs) are preferentially repaired in the transcribed strand of dihydrofolate reductase gene, CPDs are efficiently repaired in both strands of adenine phosphoribosyltransferase (APRT) locus, in either a transcribed or nontranscribed APRT gene (1). These results suggested that the transcription dependence of repair may depend on genomic context. To test this hypothesis, we constructed transfectant cell lines containing a single, actively transcribedAPRT gene, integrated at different genomic sites. Mapping of CPD repair in the integrated APRT genes in three transfectant cell lines revealed two distinct repair patterns, either preferential repair of CPDs in the transcribed strand or very poor repair in both strands. Similar kinetics of micrococcal nuclease digestion were seen for all three transfectant APRT gene domains and endogenous APRT locus. Our results suggest that both the efficiency and strand-specificity of repair of an actively transcribed gene are profoundly affected by genomic context but do not reflect changes in first order nucleosomal structure.

Both prokaryotic and eukaryotic cells have the capacity to repair DNA damage preferentially in the transcribed (T) 1 strand of transcriptionally active genes (for review, see Refs. [2][3][4][5]. Although the role of transcription in nucleotide excision repair (NER) has been elucidated in detail in Escherichia coli cells in which the mechanism of transcription-coupled repair (TCR) has been established both genetically and biochemically (6 -8), the mechanistic relationships between transcription and excision repair in mammalian cells remain unclear (1,4,9). Since gene expression is essential for cell survival it seems plausible that mechanisms for rapid preferential repair of damage in the T strand of actively transcribed genes may have evolved as an economical and expedient way to cope with transcription-blocking damage produced by endogenous or environmental mutagens. TCR also provides a logical explanation for observations of higher frequencies of mutations arising from DNA damage in the nontranscribed (NT) strand than damage occurring in the T strand of several genes (10 -16).
We recently found that several different types of bulky DNA damage, including helix-destabilizing cyclobutane pyrimidine dimers (CPDs) or benzo(a)pyrene diol epoxide (BPDE)-guanine adducts and helix-stabilizing CC-1065-adenine adducts, are repaired with equal efficiency in both the T and NT strands of a 3.9-kb DNA fragment encompassing the adenine phosphoribosyltransferase (APRT) gene in Chinese hamster ovary (CHO) AT3-2 cells, even though only the template strand is transcribed (1). We further found that in two different mutant cell lines (ATS-88 and T2S-24) in which the entire promoter region and first two exons of the APRT gene have been deleted CPDs are also repaired with equal efficiency in both strands of the APRT gene domain even though neither strand is transcribed in these cell lines (1). Our results suggested that NER of DNA damage in the APRT gene is not dependent on transcription. These findings raise the following important questions. (i) Is the endogenous APRT locus intrinsically more accessible to DNA repair than other genomic regions? (ii) If so, would translocation of the APRT gene to different genomic locations change either the overall efficiency or strand-specificity with which it is repaired? (iii) Finally, what are the factors that determine the role of the transcription process in repair?
To address these questions we transfected APRT Ϫ T2S-24 cells with genomic DNA from wild-type (APRT ϩ ) Chinese hamster cells to obtain stable APRT ϩ transfectants that contained single-copy random integrations of a functional APRT gene integrated into different genomic sites. In this study we used three such independently derived APRT ϩ transfectants (G-37, G-47, and G-69) to examine the effects of genomic context on the efficiency and strand-specificity of DNA repair.

EXPERIMENTAL PROCEDURES
Cell Lines and Culture Conditions-The CHO-AT3-2 cell line is hemizygous for the endogenous APRT gene locus; these cells contain a single, actively transcribed copy of the APRT gene that is located on chromosome Z7 (17). T2S-24 is a spontaneous APRT Ϫ mutant with a 25-kb 5Ј-extending deletion that has eliminated the entire APRT promoter region and the first two exons of this gene as well as ϳ24 kb of the upstream (5Ј-flanking) sequence (Fig. 1A). The three APRT ϩ transfectant cell lines used in this study, G-37, G-47, and G-69 were obtained by transfection of T2S-24 cells with genomic DNA from wild-type (APRT ϩ ) Chinese hamster cells. These cell lines were obtained by selection for * This work was supported by Public Health Service Grants ES03124, ES08389 (M. -s. Tang) and GM56165 (G. M. Adair) from the National Institutes of Health and service core support from National Institute of Environmental Health Sciences Grants ES077784 and ES00260. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
APRT ϩ clones in 25 M alanosine, 50 M azaserine, and 100 M adenine (ALASA) selection medium. The G-37, G-47, and G-69 transfectants each contain a random single-copy integration of a functional APRT gene (along with ϳ12-24 kb of its original 5Ј-and 3Ј-flanking sequences) stably integrated at a different ectopic site in each cell line. For the experiments described in this study, cells were grown to 50 -70% confluence in ␣-minimum Eagle's medium supplemented with 10% fetal calf serum or in ALASA selection medium to maintain selection pressure for retention and expression of the ectopically integrated APRT gene.
Ultraviolet (UV) Irradiation and Genomic DNA Isolation-For UV irradiations the culture medium was washed with Hanks basic salt solution, and cells were irradiated at a fluence rate of 1 J/m 2 /s using GE1518 germicidal lamps (predominate emission, 254 nm) as the UV source. The cells were then incubated for various periods of time in fresh medium containing 5-bromo-2Ј-deoxyuridine (10 M) and 5-fluorodeoxyuridine (1 M) and lysed for DNA isolations using the same method as previously described (1). Unreplicated DNA was separated from replicated DNA by cesium chloride density gradient centrifugation (18).
Mapping DNA Damage at the Nucleotide Level in Exons 1 and 2 of the APRT Gene-A known quantity of purified unreplicated genomic DNA was treated with T4 endonuclease V or subjected to Maxam and Gilbert sequencing reactions (19) followed by ligation-mediated polymerase chain reaction (LMPCR) as previously described (1). A specific amount of 32 P-labeled linearized pBR322 plasmid DNA (about 20,000 dpm) was added to each sample (including Maxam and Gilbert sequencing reaction samples) at the beginning of the reaction as an internal standard. Primers APRT 1-1 to 1-3 were used for detecting CPDs in the T strand of APRT exon 1, and primers APRT 1-4 to 1-6 were used for detecting CPDs in the NT strand. Primers APRT 2-1 to 2-3 were used for detecting CPDs in the T strand of APRT exon 2, and primers APRT 2-4 to 2-6 were used for detecting CPDs in the NT strand. The sequences of these primers are shown in Table I.
After LMPCR, the resultant DNAs were denatured, and equivalent counts of 32 P, representing equivalent amounts of template DNA, were loaded into each lane of the sequencing gel. After electrophoresis in an 8% denaturing polyacrylamide gel, the DNA was electrotransferred to a nylon membrane and hybridized with 32 P-labeled probes. The intensities of the well-separated T4 endonuclease V incision bands were quantitated by a PhosphorImager (Cyclone, Packard Instrument) allowing calculation of the relative levels of CPD formation at dipyrimidine sites along each DNA strand in exons 1 and 2 of the APRT gene.
RNA Isolation and Electrophoresis-Total RNA was isolated using the guanidinium method (20). In brief, cells were lysed with a guanidinium solution (4 M guanidinium isothiocyanate, 20 mM sodium acetate, pH 5.2, 0.1 mM dithiothreitol, 0.5% N-lauroylsarcosine), and genomic DNA was sheared by passage through a 20-gauge needle. Total RNA was isolated by centrifugation through a 5.7 M CsCl gradient, at 35,000 rpm for 12-20 h. The pelleted RNA was then dissolved in TES (10 mM Tris, pH 7.4, 5 mM EDTA, 1% sodium dodecyl sulfate). The RNA was then run in a glyoxal-Me 2 SO-NaPO 4 gel, transferred to an Oncor membrane, and hybridized with APRT strand-specific probes.
Strand-specific DNA Probes-pGEM-zf11 (ϩ)-APRT and pGEM-zf11(Ϫ)-APRT vectors were constructed by inserting a 3.9 kb BamHI fragment containing the CHO APRT gene into the pGEM-zf11(ϩ) or (Ϫ) vectors. These two constructs allowed us to isolate either the T or NT strand of the APRT gene. Single-stranded pGEM-zf11(ϩ)-APRT or pGEM-zf11(Ϫ)-APRT DNA phages containing either the T or NT strand of the APRT gene, respectively, were isolated by infecting the vectorcontaining cells with carrier phage M13 K07 and 32 P-labeled as described previously (1).
Nuclei Isolation and Digestion of Chromatin by Micrococcal Nuclease-Methods for nuclei isolation and subsequent micrococcal nuclease digestion were the same as those described by Kuhnert et al. (21). DNAs were purified from micrococcal nuclease-digested nuclei and separated by electrophoresis in a 1.5% agarose gel. After staining with ethidium bromide the separated DNAs were transferred to a nylon membrane and hybridized with 32 P-labeled APRT exon 2 DNA fragments.  CHO cell lines containing a single functional APRT gene domain that has been randomly integrated into different chromosomal sites, APRT Ϫ T2S-24 cells were transfected with genomic DNA isolated from wild-type (APRT ϩ ) Chinese hamster cells. APRT ϩ transfectants were then selected in ALASA selection medium. The recipient cell line used for these experiments, T2S-24, is a spontaneous nonrevertible APRT Ϫ mutant with a 25-kb 5Ј-extending deletion that has eliminated the entire APRT promoter region, the first two exons, and ϳ24 kb of upstream 5Ј-flanking sequence (Fig. 1A). Therefore, only APRT ϩ transfectants containing a functional actively expressed APRT gene will survive and form colonies in ALASA selection medium. A total of 58 such independently derived APRT ϩ transfectant clones were screened by restriction digestion and Southern blot hybridization analysis to identify transfectants that contained only a single-copy integrated APRT gene domain. To assess the relative stability of APRT gene integration and expression in these APRT ϩ transfectants in the absence of selection, each of these cell lines was grown for 1 week in nonselective medium and then plated into 8-azaadenine selection medium to determine the frequencies of spontaneous reversion from an APRT ϩ to an APRT Ϫ phenotype. Three such stable single-copy APRT ϩ transfectants, G-37, G-47, and G-69, were chosen for use in the studies presented here. The results in Fig. 1B show that in addition to the ϳ13-kb APRT hybridizing BamHI fragment band that contains the 5Ј-truncated T2S-24 APRT gene at the endogenous gene locus in each of these APRT ϩ transfectant cell lines, all three lines contain a 3.9-kb APRT hybridizing BamHI fragment band that is identical in size and intensity to the BamHI fragment containing the endogenous APRT gene in CHO-AT3-2 cells. CHO-AT3-2 cells are hemizygous for the endogenous APRT gene locus; they contain a single actively transcribed copy of the APRT gene, which is located on chromosome Z7 (17). Thus, the G-37, G-47, and G-69 cell lines each appear to contain a singlecopy APRT gene integration.

Construction of Cell Lines Containing a Single Functional
Spontaneous frequencies of reversion from an APRT ϩ to an APRT Ϫ phenotype in G-37, G-47, and G-69 cells ranged from 1 ϫ 10 Ϫ5 to 3 ϫ 10 Ϫ5 indicating that in each of these APRT ϩ transfectant cell lines, the integrated APRT gene is stably maintained and actively expressed even in the absence of selection. Spontaneous mutation rates at the integrated APRT gene locus in these three APRT ϩ transfectant cell lines (as determined by Luria-Delbruck fluctuation analysis) were only ϳ2-5-fold higher than the spontaneous mutation rate at the endogenous APRT locus in CHO-AT3-2 cells (22).
Since the G-37, G-47, and G-69 cell lines are independently derived APRT ϩ transfectant clones, the functional APRT gene in each of these cell lines should have been randomly integrated at a different chromosomal site. We have mapped the integrated APRT gene locus in each of these cell lines by Southern blot hybridization analysis using nine restriction enzymes whose cleavage sites within the endogenous APRT gene and its surrounding 5Ј-and 3Ј-flanking sequences had been previously established. Results in Fig. 1B give the following observations. (i) The G-37, G-47, and G-69 APRT ϩ transfectant cell lines show different but partially overlapping patterns of APRT-hybridizing restriction fragments and novel junction fragments; in each case the restriction fragment patterns reflect random integration of a single intact APRT gene domain along with varying lengths of its original 5Ј-and 3Ј-flanking sequences at a different genomic site in each cell line. (ii) In addition, all three cell lines retain all APRT restriction fragments characteristic of the endogenous T2S-24 APRT locus indicating that in each case the functional APRT gene has been ectopically integrated at a site other than the endogenous T2S-24 APRT locus. The size of the integrated APRT gene domain was ϳ12 kb for G-69 cells, ϳ18.2 kb for G-47 cells, and Ͼ 23.8 kb for G-37 cells (Fig. 1C). The precise chromosomal sites of integration in these three clones have yet to be determined.
In Fig. 2 we present results demonstrating that the integrated APRT gene domain in each of these APRT ϩ transfectant cell lines is actively transcribed and that transcription occurs on only one strand as is the case for the endogenous APRT locus in CHO-AT3-2 cells. No APRT transcript was detected in the T2S-24 cell line, which has a deletion that encompasses the entire promoter region and first two exons of the APRT gene. However, all three APRT ϩ transfectant cell lines show approximately the same levels of APRT transcripts as the APRT hemizygous AT3-2 cell line (which contains a single actively transcribed copy of the endogenous APRT gene). In all four of these cell lines only one strand of the APRT gene domain is transcribed.

FIG. 2. Detection of APRT transcripts in CHO cell lines.
Total RNA (10 g) isolated from T2S-24, CHO-AT3-2, G-37, G-47, and G-69 cells were electrophoresed in a Glyoxal/Me 2 SO-NaPO 4 gel, transferred to an Oncor membrane, and hybridized with 32 P-labeled strand-specific APRT probes. Ethidium staining of the 28 S and 18 S RNA present in each sample is shown as a loading control.

FIG. 3. UV survival curves for CHO-AT3-2, G-37, G-47, and G-69 cells. Cells were grown to 50% confluency in modified ␣Ϫminimum
Eagle's medium with 10% fetal calf serum, reseeded at 5 ϫ 10 5 cells per 150-mm dish, and incubated for 12 h before irradiation. After UV irradiation, the cells were reseeded at 200 cells per 100-mm dish and incubated for 2 weeks at 37°C. The colonies were then stained with crystal violet and counted to determine clonogenic survival frequencies.
These results represent two independent experiments.

G-37, G-47, and G-69 Have the Same UV Sensitivity as
Wild-type CHO Cells-Since the transfected APRT gene domains in G-37, G-47, and G-69 were randomly integrated into different sites in the genome it was possible (albeit unlikely) that integration of the APRT sequence (or some other transfected DNA sequence) may have disrupted the integrity of one of the genes involved in NER, thereby affecting the overall DNA repair capacity of the cell line. To exclude this possibility we examined and compared the UV sensitivities of each of these APRT ϩ transfectant cell lines with wild-type, NER-proficient CHO cells. As shown in Fig. 3 all three transfectant cell lines display the same sensitivity to UV (as measured by clonogenic survival) as CHO-AT3-2 cells. These results demonstrate that APRT gene integrations into different genomic sites in these APRT ϩ transfectant cell lines has not affected their overall capacity to perform NER of UV-induced DNA damage. This conclusion was further supported by the Southern hybridization results that showed CPDs in both strands of the endogenous T2S-24 APRT locus in G-37, G-47, and G-69 cells were efficiently repaired the same as was found in the recipient T2S-24 cells (data not shown).
CPD Repair in Exons 1 and 2 of the APRT Gene in G-37, G-47, and G-69 Cells-In each of these APRT ϩ transfectant cell lines the T2S-24 deletion has eliminated the first two exons of the endogenous APRT gene locus. Therefore, by comparing CPD repair in exons 1 and 2 of the integrated APRT gene in each transfectant cell line with repair at the endogenous CHO-AT3-2 APRT locus, we could determine whether ectopic integration of a functional APRT gene into different chromosomal sites or genomic contexts has affected its accessibility to repair. These experiments allowed us to examine both the efficiency and strand-specificity of DNA repair.
Repair kinetics for removal of UV-induced CPDs from either the T or NT strands of the first two exons of the APRT gene were determined at the nucleotide level in CHO-AT3-2, G-37, G-47, and G-69 cells using the T4 endonuclease V and LMPCR techniques (1,23). Three very distinct patterns of CPD repair were observed among these cell lines (Figs. 4 and 5). In CHO-AT3-2 cells more than 80% of the CPDs formed at dipyrimidine sites along both DNA strands of exons 1 and 2 of the endogenous APRT gene were repaired after 24 h of post-irradiation incubation (compare Figs. 4A, a-d and 4B, a-d and 5A). These results were consistent with our previous findings (1) that CPDs are efficiently repaired: (i) in both strands of the endogenous CHO APRT gene locus (as determined by Southern hybridization assays), and (ii) in both strands of exon 3 of this gene (as determined by LMPCR). In contrast, in the present

FIG. 4. The time course of CPD repair in exon 1 (A) or exon 2 (B) of the APRT gene in CHO-AT3-2, G-37, and G-69 cells. Cultured cells were irradiated
with UV (20 J/m 2 ) and incubated for various time periods. Genomic DNAs were isolated, treated with T4 endonuclease V followed by photoreactivation, and then subjected to LMPCR and electrophoretic separation in an 8% denatured polyacrylamide gel. A small amount of 32 P-labeled pBR322 plasmid DNA was added to each sample as internal standards for monitoring the recovery and for adjusting the sample input into the gel. The separated DNAs were transferred to a nylon membrane and hybridized with 32 P-labeled APRT DNA fragments. The position and intensity of a T4 endonuclease V incision band correspond to the site and extent of CPD formation at a particular dipyrimidine sequence, respectively. Thus, the relative frequencies of CPD formation at various sites along either the T (upper panels) and the NT (lower panels) strand can be compared for each cell line. Lanes 1 and 2 (AG, TC) represent the Maxam-Gilbert sequencing reactions (19). Sequences of contiguous pyrimidines with potential to form CPDs are indicated on the left, and T4 endonuclease V cut sites are indicated at the right (brackets).

DISCUSSION
Enhanced or preferential repair of DNA damage on the T strand has been observed in a number of genes in mammalian cells including the dihydrofolate reductase, p53, hypoxanthine phosphoribosyltransferase, and adenine deaminase gene loci (Refs. 9, 14, 15, and 24 -28). More than 90% of G to T transversion mutations in the p53 gene in cigarette smoke-related lung cancer can be attributed to DNA damage occurring in the NT strand of this gene (29,30). Mechanisms ensuring rapid and efficient repair of transcription-blocking damage on the T strand of actively transcribed genes may play an important role in maintaining the integrity of genetic information, as well as in enhancing cell survival. Cells from individuals with Cockayne syndrome appear to be deficient in TCR (24 -27, 31).
Although the mechanism of TCR in E. coli cells was established more than 10 years ago with identification of the coupling factor and demonstration of TCR in reconstituted cell-free systems (7,8), the mechanisms and factors involved in TCR in eukaryotic cells remain elusive. Recent studies by Sweder and co-workers (32,33) found the following observations. (i) The defect in TCR displayed by the original Saccharomyces cerevisiae rad26 mutant (the yeast counterpart of the mammalian CSB Cockayne syndrome gene) is not observed when these cells are grown using galactose as a carbon source. (ii) Deletion of the RAD26 gene in two other repair-proficient yeast genetic backgrounds resulted in strains that are proficient in TCR regardless of carbon source. Furthermore, whereas it has been possible to reconstitute NER in mammalian cell-free systems, TCR has not yet been demonstrated in such systems despite the concerted efforts of many well established laboratories (5). Most attempts to demonstrate TCR in vitro in mammalian cell-free systems have used a single plasmid-borne gene that can be transcribed in vitro. However, such repair substrates lack chromatin structure and are poor models for endogenous genes whose normal genomic contexts typically involve large domains including the surrounding flanking regions. Based on our findings we would argue that TCR in mammalian cells may not be a uniform phenomenon for all gene loci but may be strongly dependent on genomic context. The APRT ϩ/0 hemizygous CHO-AT3-2 cell line (which contains a single functional copy of the APRT gene at its normal endogenous locus) together with our G-37, G-47, and G-69 APRT ϩ transfectant cell lines (in which a single functional APRT gene along with its normal 5Ј-and 3Ј-flanking sequences has been randomly integrated into different genomic sites) provide an experimental system that has allowed us to examine how TCR and repair in general are affected by placing the same actively transcribed gene into different genomic contexts. We have demonstrated the following points. (i) In each of these three APRT ϩ transfectant cell lines the integrated APRT gene is actively expressed at approximately the same level as the endogenous APRT locus in CHO-AT3-2 cells (which are hemizygous). (ii) In each case only one strand of the APRT gene is transcribed. All four cell lines have a clear-cut APRT ϩ phenotype; they grow vigorously in ALASA selection medium and rapidly die when plated into medium containing 8-azaadenine.
We have observed three different patterns of CPD repair in the APRT gene in these four cell lines. First, in CHO-AT3-2 cells CPDs are efficiently repaired in both the T and NT strands in exons 1 and 2 of the endogenous APRT gene. These results further support and reinforce our original findings that NER at the endogenous APRT gene locus is not dependent upon transcription (1). In contrast, CPDs are preferentially repaired in the T strand in exons 1 and 2 of the integrated APRT gene in G-69 cells. Finally, CPDs are poorly repaired in both the T and NT strands in exons 1 and 2 of the integrated APRT gene in G-37 and G-47 cells even though the APRT genes in both of these APRT ϩ transfectant cell lines are actively transcribed. Together these results suggest that both the efficiency and strand-specificity of CPD repair in transcriptionally active genes in mammalian cells are highly dependent upon genomic context. How can different genomic contexts that do not appear to have any noticeable effects on APRT gene transcription so profoundly affect the efficiency and strand-specificity of repair of this gene?
Using mononucleosome or dinucleosome systems, several laboratories have demonstrated that nucleosomal structure can inhibit repair of CPD (34) and Ͻ6 -4Ͼ photoproducts (35,36) in in vitro nucleotide excision assays. We therefore sought to determine whether the differences in CPD repair efficiencies we have observed for transfectant APRT gene domains and the endogenous APRT locus might be caused by differences in nucleosomal structure or nucleosome density in the APRT gene in these different genomic contexts. To investigate this possibility we examined the kinetics of micrococcal nuclease digestion at the integrated APRT gene domains in the G-37, G-47, and G-69 transfectant cell lines, compared with digestion at the endogenous APRT gene locus. As shown in Fig. 6 no differences were apparent. Essentially identical results were obtained for all four cell lines (only data from AT3-2, G-37, and G-69 are shown). These results suggest that genomic context-dependent regulation of repair does not involve first order nucleosome structure but might involve higher order chromatin structure.
Our current understanding of chromatin structure and the effects of transcription on chromatin structure have not yet provided a clear picture of how these two factors regulate DNA repair. Our results together with other data accumulated to date suggest that the efficiency of repair of a region of genomic DNA is determined primarily by chromatin structure. Active genes typically have very different chromatin structures than inactive genes or noncoding regions and may be much more accessible to or have a greater affinity for DNA repair proteins. Transcription itself may increase the regional accessibility of a localized region or enhance its affinity for damage recognition/ repair proteins, but transcription per se does not appear to be a prerequisite for remodeling; remodeling may precede transcription. If the inherent chromatin structure of a gene does not allow ready access by DNA repair proteins it is possible that chromatin remodeling preceding transcription or transcriptionmediated chromatin changes may greatly enhance the accessibility of the gene to repair or at least promote TCR in its T strand. We believe that CPD repair in the integrated APRT gene in G-69 cells may reflect such processes. However, if the local chromatin structure surrounding a gene has already rendered the region maximally accessible to repair proteins, as may be the case for the endogenous CHO APRT gene locus, then further chromatin remodeling or active transcription would not be necessary to facilitate repair, and DNA damage on both strands might be repaired with equal efficiency.
Our findings that UV-induced CPDs on both strands are poorly repaired in the actively transcribed integrated APRT gene domains in G-37 and G-47 cells are difficult to reconcile with the results obtained for the G-69 cell line. Whereas it is easy to envision how chromatin remodeling preceding transcription or chromatin changes produced by transcription might increase accessibility or enhance affinity for damagerecognition/repair proteins, it is harder to explain how a genomic environment that allows normal levels of transcription could be so refractory to repair. However, integration of actively transcribed APRT genes (which are somewhat protected by buffering 5Ј-and 3Ј-flanking sequences, including a 5Јunmethylated CpG island) into regions of heterochromatin in the G-37 and G-47 transfectants could result in progressive de novo methylation that might gradually spread from down-stream sequences into the APRT gene. Hypermethylation of CpG sites in the coding sequence of the integrated APRT genes in the G-37 and G-47 cell lines might interfere with the interactions with histones or other nucleoproteins in ways that might impede or slow down the normal repair processes. The CpG sites in exons 1 and 2 of the APRT gene are normally completely unmethylated, but their methylation would be unlikely to affect transcription unless de novo methylation has spread all the way to the APRT promoter region. However, results in Fig. 4A and B show that all the CpG sites at exon 1 and exon 2 in AT3-2, G-37, and G69 cell lines are equally sensitive to hydrazine modification indicating that these sites are not methylated. Alternatively, it is possible that the APRT gene domains in the G-37 and G-47 transfectant cell lines could have been integrated into late replicating regions that may still allow transcription but are refractory to repair during most of the cell cycle.