Transcription Affects the Rate but Not the Extent of Repair of Cyclobutane Pyrimidine Dimers in the Human Adenosine Deaminase Gene* Preferential repair of dimers in transcriptionally active genes

To study the relationship between transcription and strand-specific repair of UV-induced cyclobutane pyrimidine dimers, dimer removal was analyzed in a cell line containing two alleles of an inactivated adenosine deaminase (ADA) gene. The cell line was derived from a patient suffering from severe combined immunode- ficiency. The disease was caused by a deletion of the complete promoter of the gene as well as the first exon of the ADA gene. This resulted in a true null allele without any detectable transcription (Berkvens, T. M., Gerritsen, E. J. A., Oldenburg, M., Breukel, C., Wijnen, J. T. H., Van Ormondt, H., Vossen, J. M., Van der Eb, A. J., and Meera Khan, P. (1987) Nucleic Acids Res. 15, 9365-9378). Despite this lack of transcrip- tion, repair of the ADA gene in this cell line was found to be very efficient with 80% of the dimers being removed within 24 h after UV irradiation. However, the initial

tion, repair of the ADA gene in this cell line was found to be very efficient with 80% of the dimers being removed within 24 h after UV irradiation. However, the initial rapid repair which is associated with the transcribed strand in normal cells is absent. Dimer removal from two inactive loci, 754 and coagulation factor IX, was much less efficient with only 40% dimers removed after 24 h. From this data, we conclude that transcription is not required for efficient repair of a gene, but forms an additional signal for accelerated repair of the transcribed strand. Furthermore, we suggest that different levels of repair exist between nontranscribed sequences in active genes and those in repressed loci. The results are discussed in terms of the current ideas about the mechanism of preferential DNA repair in human cells.
Ultraviolet light (UV) induces two major DNA adducts, the cyclobutane pyrimidine dimer and the pyrimidine (6-4) pyrimidone photoproduct. Both types of damage are removed by DNA excision repair. Especially in the case of cyclobutane pyrimidine dimers, ample evidence has been presented that repair of this lesion occurs nonrandomly in the genome. Preferential repair of dimers in transcriptionally active genes has now been well documented in a wide array of organisms ranging from Escherichia coli (Mellon and Hanawalt, 1989) and the lower eukaryote Saccharomyces cereuisiue (Terleth et al., 1989(Terleth et al., , 1990 to rodent (Bohr et al., 1985;Madhani et al., 1986) and human cells (Mellon et al., 1986;Venema et al., 1990a). An exception is found in Drosophila embryonal cell lines which exhibit efficient repair of both active and inactive loci (De Cock et al., 1991). Efficient removal of DNA damage from transcribed sequences is presumed to enhance cellular survival by enabling the cell to express essential genes before removal of the bulk of the damage. This concept is supported by the finding that Cockayne's syndrome cells are UV-sensitive due to a deficiency in preferential repair of active genes (Venema et al., 1990b).
The biological advantage of preferential repair is evident, but the mechanism by which it occurs is still largely unknown. Initially, the presumed explanation for the accelerated repair of active genes was the more open chromatin configuration of such genes which renders them more accessible to DNA repair enzymes. This model was challenged by the observation of strand-specific repair, i.e. within active genes the transcribed strand was repaired much faster and more efficiently than the nontranscribed strand (Mellon et al., 1987). The most dramatic effect was seen in Chinese hamster (Mellon et al., 1987) and xeroderma pigmentosum complementation group C cells . Both cell types exhibited efficient repair of the transcribed strand, whereas virtually no dimers were removed from the nontranscribed strand. These results appeared to favor a model in which the transcription process itself is the primary target for repair enzymes. This was supported by the finding of strand-specific repair in E. coli (Mellon and Hanawalt, 1989) which does not possess a "eukaryote-like" chromatin structure. In normal human cells, the difference in repair rates of both strands was only 2-3-fold. The nontranscribed strand was repaired at a rate and to an extent similar to that of the genome overall, suggesting that preferential repair of active genes was due to selective repair of the transcribed strand. These cells therefore also performed efficient repair of nontranscribed sequences.
To further identify the factors which are involved in preferential repair of active genes, we investigated removal of cyclobutane pyrimidine dimers in a cell line derived from a patient suffering from severe combined immunodeficiency (SCID).' The disease was caused by a deletion in the adeno-The abbreviations used are: SCID, severe combined immunodeficiency; ADA, adenosine deaminase; bp, base pair(s); kb, kilobase(s); DHFR, dihydrofolate reductase; CS, Cockayne's syndrome. sine deaminase (ADA) gene in both alleles. The deletion covered the promoter and the first exon of the gene. It was previously shown that this mutation resulted in a true null allele having no detectable transcription (Berkvens et al., 1987). This enabled us to analyze removal of dimers in two ways: first, the rate and extent of repair of the nontranscribed ADA gene were compared with those of the wild type ADA gene. Secondly, we compared dimer removal in the same cell line between the nontranscribed ADA gene and two repressed loci, 754 and coagulation factor IX.

EXPERIMENTAL PROCEDURES
Cell Lines and Culture Conditions-The cell line used in this study was derived from a female SCID patient. Fibroblasts were cultured in Ham's F10 medium (without hypoxanthine and thymidine) supplemented with 15% fetal calf serum and antibiotics in a humidified 2.5% COS atmosphere. Routinely, cells were grown to confluence and then split 1:3. To prelabel the cells 0.1 pCi/ml [3H]thymidine (100 mCi/mmol) was added to the medium, and the cells were cultured for 3 days. The medium was then replaced by fresh label-free medium and cell growth was continued to confluence. Prior to UV irradiation, prelabeled cells were seeded in 94-mm petri dishes and cultured for 10 days with regular medium changes to reach stationary phase.
Analysis of Cyclobutane Pyrimidine Dimer Remoual-Removal of cyclobutane pyrimidine dimers from specific DNA sequences was analyzed as described previously (Venema et al., 1990a) except for one modification, which is presented below. Cells were irradiated with 10 Jm-' UV (254 nm) and incubated for up to 24 h. After incubation, cells were lysed and high molecular weight DNA was purified by phenol and chloroform extractions followed by ethanol precipitation.
The DNA was digested with either BclI, EcoRI, or KpnI (Pharmacia LKB Biotechnology Inc.) and purified by phenol and chloroform extraction, followed by concentration with n-butanol and ethanol precipitation (since the cells used in this study were in the confluent Go state, it is not necessary to density label the DNA with bromodeoxyuridine in order to separate replicated DNA by CsCl density gradient centrifugation). Equal amounts of DNA were either treated or mock-treated with the dimer-specific enzyme T4 endonuclease V (Nakabeppu et al., 1982) and electrophoresed in parallel on an alkaline agarose gel. The DNA was transferred to Hybond-N+ membranes (Amersham) by Southern blotting and hybridized with 32P-labeled gene-specific probes. After autoradiography, films were scanned using a Bio-Rad video densitometer. T4 endonuclease V causes nicking of fragments which contain cyclobutane pyrimidine dimers. The number of dimers present in a fragment at each time point was calculated from the amount of fragments not nicked by T4 endonuclease V, using the Poisson expression.
Preparation of Strand-specific Probes-Three ADA cDNA PstI fragments (Berkvens et al., 1987) and a 690-bp HindIII-EcoRI genomic fragment from intron V of the DHFR gene (Will and Dornick, 1986) were subcloned in SSEV18/19 vectors (Biernat et al., 1989). The orientation of all DNA fragments was confirmed by sequence analysis. The SSEV vector contains a polylinker which is able to form a stem-loop structure in the single-stranded form. This stemloop structure contains an EcoRI site and thereby allows for the separation of single-stranded cloned inserts from vector sequences. Isolation and purification of probe fragments was performed as described earlier . The fragment was labeled by filling in the 3' recessed end of the EcoRI site using [cY-~*P]~ATP and Klenow DNA polymerase.

RESULTS
Transcriptional Status of the ADA Gene-The cell line used in this study was derived from a female SCID patient and was homozygous for a deletion of 3.2 kb in the 5' end of the ADA gene which removed the promoter and the first exon (Berkvens et al., 1987). A partial restriction map is shown in Fig. 1 which also shows the location and extent of the deletion. The deletion runs from position -1480 to +1770 relative to the ADA cap site and is presumed to originate from unequal crossing over between two direct Alu repeats (Berkvens et al., 1990). It completely abolished transcription which was demonstrated by the absence of any detectable ADA-specific RNA in heterogeneous nuclear RNA (hnRNA) isolated from fibroblasts of the patient (Berkvens et al., 1987). Even after long overexposure no hybridization signal could be detected, whereas the low expression in normal fibroblasts was readily observed. Correspondingly, no ADA enzymatic activity was detected in blood cells from the patient.
Dimer Removal in the ADA Gene-We wanted to establish the effect of the absence of transcription on the efficiency and strand specificity of dimer removal from the ADA gene. This was measured using T4 endonuclease V to specifically detect cyclobutane pyrimidine dimers. Digestion of DNA from UVirradiated cells with this enzyme results in a single-stranded nick at the site of a dimer. This degradation is visualized on a denaturing agarose gel as a decrease in the amount of fulllength DNA. Removal of dimers will result in less degradation of the T4 endonuclease V-treated DNA. Repair in specific DNA fragments can be analyzed by Southern blotting and hybridization with a gene-specific probe.
The results of such an analysis for the ADA BclI fragment are shown in Fig. 2. When repair was measured with probes recognizing the template and nontemplate strand, we observed efficient repair of dimers in both strands. Within 24 h after UV treatment, the band in the lane containing T4 endonuclease V treated DNA has almost returned to the density of the band in the nontreated lane. The results were quantified by densitometry scanning of the autoradiograms (Fig. 3). Both the template and the nontemplate strand are  Fig. 2 were quantified by densitometry and the dimer frequency was calculated from the relative band densities in the lanes containing treated and nontreated DNA. The percentage of repair was then calculated from the dimer frequency at each time point. 0, template strand; 0, nontemplate strand. The values are the average of three independent experiments. The standard error is indicated. Removal of cyclobutane pyrimidine dimers from the 18.6-kb ADA EcoRI fragment. Data were obtained as described in Fig. 3 . 0 , ADA template strand; 0, non-ADA template strand.
repaired to a high extent, with 80% dimers removed within 24 h after UV irradiation. However, the initial repair rate is low, since hardly any repair is detectable in the first 2 h after treatment. This is different from the situation in the wild type ADA gene, where the transcribed (ie. template) strand is rapidly repaired with 20 and 40% dimers removed within 2 and 4 h after UV, respectively .
The results of a similar analysis of the ADA EcoRI fragment are shown in Fig. 4. As in the BclI fragment, repair of the ADA template strand is slow, but by 24 h repair is substantial. The rate and extent of repair of this strand are very similar to those of the two strands in the BclI fragment. However, the complementary strand is repaired at a faster rate. This "reverse" strand specificity is probably caused by transcription of this strand (Lattier et al., 1989). In normal cells, both strands of the ADA EcoRI fragment are repaired fast and efficiently, consistent with the presence of transcription units on either strand .
Dimer Removal in the DHFR Gene-To show that the cell line used in this study was able to perform preferential repair of the transcribed strand, repair was analyzed in a 20-kb KpnI fragment of the DHFR gene (Fig. 1). Fig. 5 shows that there was a large difference in the rate of repair of the two strands in this fragment. The transcribed strand was repaired 2-to 3fold faster than the nontranscribed strand. Moreover, the nontranscribed strand of the DHFR gene showed repair ki- netics similar to the ADA gene. Also here, there is virtually no repair in the first 2 h after UV treatment. The repair values obtained for either strand of the DHFR gene in this cell line are comparable to those obtained previously with another human cell line .
Repair in Two Inactive Genes-To find out whether there was a difference in repair between the nontranscribed ADA gene and inactive tissue-specific genes, dimer removal was analyzed in two X-chromosomal loci: 754 and Factor IX. The coagulation factor IX gene is well characterized and completely sequenced (Yoshitake et aL, 1985), whereas 754 is a less defined locus which is located in the proximity of the Duchenne muscular dystrophy locus (Hofker et al., 1986). Both loci are not expressed in fibroblasts (Hofker et al., 1986;Antonarakis, 1988;Palmer et aL, 1989). Repair in these two genes was analyzed in a 16.7-kb KpnI fragment of the factor IX gene (Fig. 1) and a 14-kb EcoRI 754 fragment, respectively. The use of these fragments enabled us to measure dimer removal on the same Southern blots which were used for the active DHFR and ADA genes. This procedure excluded possible experimental variation by unequal loading of different gels. Removal of dimers in both loci was very slow and incomplete (Fig. 6). No repair was observed in the first 2 h after irradiation, and only 40% of the dimers were removed after 24 h. Repair values were comparable in both loci.

DISCUSSION
The finding of preferential repair of transcribed genes and (later on) the selective repair of the transcribed strand has posed questions about the mechanism of this repair. Data obtained with E. coli and Chinese hamster cells seemed to indicate that the transcription machinery itself was somehow involved in DNA repair or perhaps in the targeting of repair enzymes. Human repair deficient XP-C cells only repair the transcribed strand of active genes, indicating that also in human cells repair of the transcribed strand is (at least partly) independent of repair of the bulk of the genome (Venema et ul., 1991). In contrast, Cockayne's syndrome (CS) cells are deficient in preferential repair of active genes. The transcribed and nontranscribed strand of active genes are repaired to the same extent, which is characteristic of that of X-chromosomal, inactive genes.* In CS cells, the repair level of active genes appears to drop to a level which is even lower than that observed in the nontranscribed strand of normal cells.
A possible way to elucidate the role of transcription in preferential DNA repair would be to analyze repair in the presence of transcription inhibitors. Indeed, recently Leadon and Lawrence (1991) presented evidence that inhibition of RNA polymerase I1 transcription by a-amanitin abolished preferential repair of the transcribed strand of the human metallothionein gene. This paper describes an alternative approach. Removal of cyclobutane pyrimidine dimers was analyzed in a cell line derived from an SCID patient. The disease was caused by a deletion in the ADA gene, which completely abolished transcription (Berkvens et al., 1987). The effect of the absence of transcription on dimer removal was measured in two restriction fragments together encompassing over 38 kb of DNA. It was shown that repair of the nontranscribed ADA gene is still substantial, with about 80% of the dimers removed in 24 h. However, the initial rapid repair associated with the transcribed strand in the wild type ADA gene is absent. Instead, almost no repair is observed at early times after UV irradiation. When these results are compared to those obtained for the normal ADA gene , it can be concluded that the repair level of both strands in the absence of transcription has dropped to that seen for the nontranscribed strand in the normal ADA gene. The same slow repair kinetics are found et early times after UV irradiation, although in this study the repair levels after 8 h seem to be somewhat higher. In the EcoRI fragment, partly containing 3' flanking sequences, there appears to be strand specificity toward the complementary strand, which is likely to be caused by a convergent transcription unit (Lattier et al., 1989).
As a control, dimer removal was also analyzed in the DHFR gene. It was found that preferential repair of the transcribed strand occurred in an internal KpnI fragment. At early times after treatment, repair in this strand is clearly faster than that of the nontranscribed strand. This result is in agreement with previous data obtained with another human cell line (Venema et ul., 1991). The repair rate for the DHFR gene is slower than that of the normal ADA gene. This could be due to the fact that repair in the DHFR gene was measured in a single repair experiment which gave somewhat lower repair values. When repair in the ADA EcoRI fragment was measured in the same experiment, the repair values were comparable to those obtained for the DHFR gene. The final data for the ADA gene, however, is the average of three independent experiments. Based on these results, we conclude that the cell line used in this study is capable of performing preferential repair of the transcribed strand in an active gene. It was also important to compare the repair data found for the ADA gene with those observed in two inactive, tissue-specific genes.
The results indicate that the extent of dimer removal in the J. Venema, A. van Hoffen, A. T. Natarajan, A. A. van Zeeland, and L. H. F. Mullenders, unpublished results. X-chromosomal 754 and Factor IX loci is significantly lower than in the mutated ADA gene. The repair level of both loci is comparable and is also similar to that found in several other human cell lines (Venema et al., 1990a;1990b;Kantor et al., 1990). Hence, we find a great difference in repair efficiency between two types of nontranscribed genes, the inactive ADA housekeeping gene and the repressed 754 and factor IX loci.
Taken together, the results presented in this paper clearly show that transcription is not a prerequisite for efficient removal of cyclobutane pyrimidine dimers. In the absence of transcription of the ADA gene, we still find considerable repair which is much more efficient than that occurring in inactivated loci such as 754 and factor IX. Interestingly, similar results were recently obtained by Terleth (1991) who found no significant differences in the extent of repair of the MATa locus in the absence or presence of transcription. However, transcription apparently forms an additional signal which causes selective repair of the transcribed strand at early times after UV irradiation. This rapid repair of the transcribed strand is absent in the nontranscribed ADA gene. Therefore, other factors besides transcription must play a role in determining the repair efficiency of various genomic regions. An obvious candidate for this role is the local chromatin structure. An open or poised chromatin structure could mean a greater accessibility to repair enzymes, providing an opportunity for fast and efficient repair of (potentially) active genes. However, the term "accessibility" implies that preferential repair of active genes is a passive process, depending on diffusion of repair proteins. It is difficult to reconcile the deficiency in preferential repair of active genes in CS cells with this assumption. The results obtained with those cells strongly indicate that preferential repair of active genes needs one or more protein factors which are apparently defective in CS cells. Moreover, differences in chromatin structure between the transcribed and nontranscribed strands of active genes are not obvious. Another possible factor could be the association with the nuclear matrix, because UV-induced repair synthesis is associated with this nuclear substructure in a time-and dose-dependent manner (Mullenders et al., 1988). Attachment of genes to the matrix could ensure their proximity to repair enzymes which may be located at the matrix attachment sites. The matrix attachment sites in the ADA gene has been localized 5' to the coding sequence in a region which is still present in the ADA deletion mutant? Therefore, it is likely that despite the absence of actual transcription the promoterless ADA gene is still closely attached to the nuclear matrix.
In conclusion, this paper presents evidence for the existence of at least three repair levels in mammalian cells: 1) inactive, repressed loci (heterochromatin) which are repaired slowly and inefficiently, 2) (potentially) active genes (poised chromatin) which are repaired to a higher extent, and 3) the transcribed strand of active genes which exhibits an accelerated repair at early times after UV and which is superimposed on the already efficient repair of active genes.