Strand breaks are repaired efficiently in human ribosomal genes.

We examined repair of DNA strand breaks induced by the anti-cancer drug bleomycin in both Pol I and Pol II transcribed genes in permeabilized human fibroblasts. The majority of these breaks (>80%) are single strand breaks (SSBs) thought to be repaired by base excision repair enzymes. Repair was examined in each strand of a 7. 2-kilobase fragment, completely within the Pol I transcribed region of ribosomal DNA (rDNA) and an 8.3-kilobase fragment completely within the Pol II transcribed region of the dihydrofolate reductase (DHFR) gene. Bleomycin dose-response studies revealed no bias for SSBs in either strand of the rDNA fragment. Furthermore, repair of SSBs is rapid (approximately 80% resealed in 60 min) in both the transcribed and nontranscribed strands of rDNA. Rapid repair of SSBs is also observed in both strands of the DHFR gene (approximately 60% resealed in 60 min). In contrast, little (or no) repair of UV photodimers occurs in either strand of human rDNA, regardless of whether cells are confluent or actively growing. Thus, DNA lesions in human ribosomal genes may be more accessible to base excision repair enzymes than those involved in nucleotide excision repair.

We examined repair of DNA strand breaks induced by the anti-cancer drug bleomycin in both Pol I and Pol II transcribed genes in permeabilized human fibroblasts. The majority of these breaks (>80%) are single strand breaks (SSBs) thought to be repaired by base excision repair enzymes. Repair was examined in each strand of a 7.2-kilobase fragment, completely within the Pol I transcribed region of ribosomal DNA (rDNA) and an 8.3-kilobase fragment completely within the Pol II transcribed region of the dihydrofolate reductase (DHFR) gene. Bleomycin dose-response studies revealed no bias for SSBs in either strand of the rDNA fragment. Furthermore, repair of SSBs is rapid (ϳ80% resealed in 60 min) in both the transcribed and nontranscribed strands of rDNA. Rapid repair of SSBs is also observed in both strands of the DHFR gene (ϳ60% resealed in 60 min). In contrast, little (or no) repair of UV photodimers occurs in either strand of human rDNA, regardless of whether cells are confluent or actively growing. Thus, DNA lesions in human ribosomal genes may be more accessible to base excision repair enzymes than those involved in nucleotide excision repair.
Bleomycin is a radiomimetic antitumor agent that induces (primarily) strand breaks in chromosomal DNA (reviewed by Stubbe and Kozarich (1987) and Povirk and Finley Austen (1991); also see Bennett et al. (1993)). Unlike bulky DNA lesions, repair of these breaks is resistant to aphidicolin, and the average repair patch size is less than 6 nucleotides (DiGiuseppe and Dresler, 1989;DiGiuseppe et al., 1990;Sidik and Smerdon, 1990a). Bleomycin is a glycopeptide that complexes with divalent cations (mainly Fe 2ϩ ) and O 2 , and a portion of the molecule intercalates in DNA, particularly at NGC sequences (Kuwahara and Sugiura, 1988). It produces strand breaks or apurinic/apyrimidinic sites by abstracting hydrogen from C-4Ј of deoxyribose adjoining the guanidyl-3Ј-phosphate at the site of bleomycin-DNA intercalation (Stubbe and Kozarich, 1987;Povirk and Finley Austen, 1991). At this point, the resulting free radical can partition into one of two alternate damage pathways. Upon addition of O 2 , a peroxyl radical is formed, which decomposes to yield a strand break with 5Јphosphate and 3Ј-phosphoglycolate termini, with release of a base propanal. Alternatively, hydroxylation of the C-4Ј radical can occur opening the sugar ring and releasing the base. This pathway yields an apurinic-apyrimidinic site with a chemically modified sugar. In chromatin, however, many of these modified apurinic-apyrimidinic sites undergo spontaneous cleavage, probably by reaction with histone amine groups (Bennett et al., 1993). These strand breaks cannot be ligated directly and may be repaired via removal of the modified base (and a few adjacent bases) by a 3Ј 3 5Ј exonuclease and (short patch) repair synthesis (e.g., see Mosbaugh and Linn (1984)).
We analyzed repair of bleomycin-induced single strand breaks (SSBs) in the individual strands of Pol I-transcribed ribosomal genes (rDNA) and the Pol II-transcribed DHFR gene. Repair in the DHFR gene has been well characterized, and both preferential and strand-specific repair of UV-induced cyclobutane pyrimidine dimers (CPDs) occurs in this gene (reviewed in Friedberg et al. (1995)). In contrast, repair of CPDs in either total rDNA of human and hamster cells (Christians and Hanawalt, 1993) or the transcriptionally active fraction of rDNA in mouse cells (Fritz and Smerdon, 1995) is very inefficient in both DNA strands. We report here that, in contrast to very inefficient repair of CPDs in rDNA, repair of bleomycininduced SSBs occurs rapidly in both strands of a human 7.2-kb rDNA fragment and an 8.3-kb fragment in the DHFR gene ( Fig. 1).

MATERIALS AND METHODS
Chemicals-Bleomycin (Blenoxane) was a generous gift from Bristol Myers Laboratories (Syracuse, New York). L-␣-Lysophosphatidylcholine (LPC) was purchased from Sigma.
Cell Culture-Human diploid fibroblasts (strain AG 1518) were split 1:3 and grown in culture as described previously (Smerdon et al., 1982). Cells were prelabeled with 5 nCi/ml [ 14 C]deoxythymidine (50 mCi/ mmol; DuPont NEN) for 1 week after splitting. Prelabeling medium was replaced with fresh medium, and the cells were grown an additional 2-2.5 weeks until confluent. The medium was changed every 8 days during this period. To obtain actively growing cells used in some experiments, cells were split 1:4 and harvested 3 days after splitting. * This study was supported by National Institutes of Health Grants ES02614 and ES04106 from NIEHS. 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. ‡ Present address: Molecular Biosciences Dept., Pacific Northwest National Laboratory, P7-56, Box 999, Richland, WA 99352. § These authors contributed equally to this work. ¶ To whom correspondence should be addressed. Tel.: 509-335-6853; Fax: 509-335-9688. 1 The abbreviations used are: CPD, cis-syn cyclobutane pyrimidine dimer; bp, base pair(s); kb, kilobase pair(s); LPC, L-␣-lysophosphatidylcholine; PBS, phosphate-buffered saline; PIPES, piperazine-N,NЈ-bis(2ethanesulfonic acid; T4 endo V, T4 endonuclease V; Pol I, RNA polymerase I; Pol II, RNA polymerase II; DHFR, dihydrofolate reductase.

Preparation of Permeable Cells and Bleomycin
Exposure-Cells were permeabilized with LPC as described previously (Lorenz et al., 1988). Briefly, cells were washed twice with ice-cold PBS (16 mM phosphate buffer, pH 7.2, 5 mM KCl, 135 mM NaCl). An 80 g/ml solution of LPC (dissolved in PBS and 1 mM CaCl 2 ) was added to the cells, which were then kept on ice for 2 min. The effectiveness of cell permeabilization was determined in separate plates by the fraction of cells taking up trypan blue dye (Lorenz et al., 1988). The LPC solution was then carefully removed and replaced with repair mixture (35 mM HEPES, pH 7.4, 50 mM sucrose, 80 mM KCl, 5 mM MgCl 2 , 7.5 mM KH 2 PO 4 , 1 mM CaCl 2 , 5 mM ATP; see Smerdon (1990a, 1990b)).
Bleomycin treatment was performed according to Smerdon (1990a, 1990b). For repair experiments, cells were treated in three groups. Group 1 cells (no bleomycin) were made permeable by LPC treatment followed by addition of repair mixture containing 3 M dNTPs. The cells were washed and harvested after a 30-min incubation at 37°C without bleomycin. Group 2 cells (damage, no repair) were made permeable by LPC treatment followed by addition of repair mixture with no dNTPs and various concentrations of bleomycin dissolved in 10 mM PIPES buffer. The cells were incubated for 30 min at 37°C prior to harvest. Group 3 cells (damage and repair) were made permeable with LPC, followed by repair mixture containing 3 M dNTPs. Bleomycin dissolved in PIPES buffer was also added. Cells were incubated at 37°C for 30 min. Following this incubation, the bleomycin solution was replaced with prewarmed conditioned medium, and the cells were incubated for different repair times.
DNA Isolation-DNA was prepared according to Bohr and Okumoto (1988). For repair studies in the DHFR gene, DNA was digested with PstI (5 units/g of DNA). To assay repair in rDNA, samples were digested with EcoRI (5 units/g of DNA). Following restriction enzyme digestion, the DNA was isolated on neutral CsCl density gradients (Bohr and Okumoto, 1988), and, after dialysis against TE buffer (10 mM Tris⅐HCl, 1 mM EDTA), the specific radioactivity was determined (from absorbance scans and liquid scintillation counting).
Southern Analysis-Equal amounts of samples, determined from measured values of specific radioactivity, were electrophoresed in 0.6% or 0.8% alkaline agarose gels at 35 V for 16 -17 h, according to Maniatis et al. (1982). Samples were then depurinated in the gels by soaking in 0.25 M HCl for 15 min. After acid treatment, gels were neutralized and transferred to nylon membranes (Hybond N ϩ , Amersham) using the alkaline transfer method (Ausubel et al., 1988).
Preparation of Single Strand Probes-Plasmid pZH-15 was a gift from Drs. C. Allen Smith and Phillip Hanawalt (Stanford University, Stanford, CA). It contains 788 bp of genomic DNA extending from Ϫ315 bp into intron 2 of the human DHFR gene (Fig. 1). Plasmid pSPT28S, used to probe rDNA, was a gift from Drs. Jose Sogo (ETH, Zü rich, Switzerland) and Antonio Conconi (WSU, Pullman, WA) (Fig. 1). It contains a 137-bp BamHI/SstI fragment from the 28 S region of mouse rDNA ( Fig. 1) that is very similar to the human rDNA sequence (Fritz, 1994). Plasmid pZH111, also a gift from the Hanawalt laboratory and constructed by Dr. L. Lommel, contains a 600-bp insert (into pGEM-3Z) of the human DHFR cDNA from the EcoRI site in exon II to the Sau3A site in exon V. This plasmid was used to probe DHFR mRNA. Radioactively labeled riboprobes were generated according to the manufacturer's instructions in the Riboprobe Gemini II kit (Promega; see Fritz and Smerdon (1995)).
Quantitation of Autoradiograms-Autoradiograms were quantified using a laser scanning densitometer (LKB model 2222 or Molecular Dynamics model PDSI). The integrated intensity of bands resulting from bleomycin treatment was determined using a peak deconvolution program (Peak Fit; Jandel Scientific, Corte Madera, CA). These intensities were normalized to the integrated intensities of bands resulting from no treatment to determine the fraction of fragments free of strand breaks (P 0 ). The average strand breaks/fragment was determined using the Poisson expression [Ϫln(P 0 )] (Bohr and Okumoto, 1988). Loading variations between lanes were corrected by normalization of band intensities to either an external plasmid band (e.g. see Murad et al., 1995) or to the total intensity measured in the lane.

RESULTS
Bleomycin Dose-Response of rDNA-For many of these experiments, confluent human fibroblasts were reversibly permeabilized with LPC to improve transport of bleomycin into cells (Sidik and Smerdon, 1990b). Using this treatment, bleomycin dose-response measurements were performed to determine the level of strand breaks produced in the 7.2-kb EcoRI rDNA fragment. Total strand breaks were assayed by alkaline agarose gel electrophoresis ( Fig. 2A, inset). As shown in Fig. 2, at bleomycin doses Ͼ0.1 g/ml, there is approximately a loglinear relationship between strand breaks in the rDNA fragment and bleomycin concentration. There is also a significant number of double strand breaks, or breaks on opposite strands Confluent human cells were permeabilized with LPC (see "Materials and Methods") and treated with 0, 0.1, 1, 10, and 25 g/ml (inset lanes 1-5, respectively) of bleomycin for 30 min. Cells were harvested immediately, and DNA was prepared and separated on a 0.8% alkaline agarose gel (inset A) or a 0.8% native agarose gel (inset B). DNA was transferred to a nylon membrane and hybridized to a strand-specific pSPT28S riboprobe. Graph, comparison of total strand breaks in each strand of rDNA. DNA from cells exposed to bleomycin was electrophoresed on alkaline agarose gels, blotted, and probed (as above). Following autoradiography, membranes were stripped and reprobed with the opposite strand. (It is noted that bands in lanes 1 and 2 of the inset are overexposed, and lower exposures were used for quantitative analysis.) Graph shows a fit to the data for the transcribed strand (open symbols) and nontranscribed strand (q). Different symbols are for different experiments.
in close proximity, at bleomycin doses Ͼ1 g/ml, as measured by neutral gel electrophoresis (Fig. 2B, inset). These breaks account for Ͻ20% of the total breaks induced by bleomycin in the rDNA fragment (analysis not shown). Furthermore, there is no bias for bleomycin-induced strand breaks in either strand of this fragment (Fig. 2). A dose of 3 g/ml was used for most repair experiments where 90 -95% of the DNA fragments analyzed contain at least one single strand break.
Repair of Strand Breaks in the Ribosomal and DHFR Genes-For most repair experiments, permeabilized, confluent human fibroblasts were exposed to 3 g/ml bleomycin. At this dose, the 7.2-kb rDNA fragment (Fig. 1A) contained 1.8 -2.3 strand breaks in each strand, while the 8.3-kb DHFR fragment (Fig. 1B) contained 2.4 -2.6 strand breaks in each strand. This dose was chosen to illustrate the dramatic change in strand breaks after even short repair times (see below). Also, as observed for the rDNA fragment, no bias was observed for strand breaks in either strand of the DHFR fragment (not shown).
Repair of bleomycin-induced strand breaks in both rDNA and DHFR displayed initial rapid kinetics (Fig. 3). This is illustrated by the rapid return of intact rDNA fragments after only 40 min of repair incubation (Fig. 3A, inset). At least 80% of the strand breaks are repaired in each strand of these genes after only 90 min (Fig. 3). This value approaches 100% after 24 h of repair incubation (data not shown). The "plateau" in repair data at about 80% for rDNA ( Fig. 3A) most likely reflects a much slower repair of double strand breaks (Coquerelle et al., 1987), present at Ͻ20% the level of SSBs in rDNA (above). However, regardless of the reason for the plateau in repair kinetics, it is clear that most strand breaks are very rapidly repaired during the first hour in ribosomal genes.
It is important to note that the results in Fig. 3 represent repair in total ribosomal genes, which are present in several hundred copies in human cells (Long and Dawid, 1980). Using psoralen cross-linking to separate the transcriptionally active from the inactive fraction, we find that Ͻ30% of rRNA genes are in the active chromatin fraction in these cells (data not shown; see Fritz and Smerdon (1995)). Therefore, the results in Fig. 3 indicate that efficient repair occurs in each strand of both the active and inactive rDNA fractions.

Repair of UV-induced Damage in Human Ribosomal
Genes-It has been established that rDNA transcription is greatly reduced in stationary cells (Sollner-Webb, 1986;Conconi et al., 1989). Therefore, we irradiated both confluent and actively growing human fibroblasts with 20 J/m 2 UV light (254 nm), and the newly replicated DNA in growing cells was separated from parental DNA using isopycnic centrifugation (see "Materials and Methods"). The yield of CPDs in each strand of the 7.2-kb rDNA fragment was measured using the method of Bohr et al. (1985). This assay measures the fraction of restriction fragments resistant to cutting by T4 endonuclease V (T4 endo V), which cleaves DNA strands specifically at CPD sites. As shown in Fig. 4, A and B, little change occurs in the fraction of T4 endo V-resistant rDNA fragment after 8 h of repair incubation, and there is only a small increase in this band after 24 h of repair. Indeed, quantitation of these results indicates that only 20 -30% of the CPDs are removed from each strand after 24 h (Fig. 4, C and D), and repair was inefficient during each growth state. Thus, as previously found for rodent and human cells (Christians and Hanawalt, 1993;Fritz and Smerdon, 1995), repair of CPDs is inefficient in each strand of the ribosomal genes of human fibroblasts. DISCUSSION Our results indicate that bleomycin-induced SSBs in rDNA are repaired efficiently, but not in a strand-specific manner (Fig. 3). On the other hand, repair of UV-induced CPDs is very inefficient in each strand of rDNA in both growing and confluent human fibroblasts (Fig. 4). This latter observation agrees qualitatively with past studies on repair of total rDNA in human and hamster cells Hanawalt, 1993, 1994). Recently, we have shown that inefficient repair of CPDs takes place even in the transcribed strand of the transcriptionally active fraction of ribosomal genes (Fritz and Smerdon, 1995). Thus, the rapid and extensive repair of bleomycin-induced SSBs is dramatically different from the repair of UV photodimers in rDNA.
The differences we observe between the kinetics of repair of SSBs induced by bleomycin and CPDs induced by UV light are striking. Stevnsner et al. (1993) also observed differences in repair efficiency in rDNA damaged with various agents in CHO cells. These authors reported inefficient repair of CPDs, and both monoadducts and diadducts induced by nitrogen mustard. FIG. 3. Repair of bleomycin-induced strand breaks in rDNA and the DHFR gene of human cells. Confluent AG1518 cells were permeabilized and treated with 3 g/ml bleomycin for 30 min. After various repair times, DNA was prepared, digested with PstI, and purified on CsCl density gradients. Samples were then either redigested with EcoRI (A) or not redigested (B) and separated on 0.8% alkaline agarose gels. Samples were transferred to nylon membranes and hybridized to strand-specific riboprobes. Following autoradiography, membranes were stripped and reprobed with the opposite strand. Insets show representative autoradiograms from one experiment for no bleomycin treatment (lane 1) and repair times of 0, 40, 60, and 90 min (lanes 2 through 5) in the transcribed strand (TS) and nontranscribed strand (NTS). A, data for membranes probed with the pSPT28S riboprobe to measure repair in rDNA. B, data for membranes probed with the pZH-15 riboprobe to measure repair in the DHFR gene. Data points in each panel represent the average of two independent experiments for repair of strand breaks in the transcribed strand (E) and nontranscribed strand (q).
However, efficient repair of methyl methanesulfonate adducts and cisplatin interstrand cross-links was observed in rDNA (Stevnsner et al., 1993), although repair of individual rDNA strands was not measured in that study. This report, along with the present one, suggests that the "accessibility" of rDNA may vary between different repair pathways (see below).
UV-induced CPDs and bleomycin-induced strand breaks are (presumably) repaired by different repair pathways (e.g. see Friedberg et al. (1995)). DNA polymerase inhibitor studies strongly support the involvement of polymerase ␤ (independent of polymerases ␣ and ␦/⑀) in the repair of bleomycin-induced damage (DiGiuseppe and Dresler, 1989;DiGiuseppe et al., 1990;Sidik and Smerdon, 1990a). The nucleolus, where rDNA sequences are sequestered, could act as a barrier to the bulky enzyme complex associated with nucleotide excision repair (Friedberg et al., 1995), but allow access to the 39-kDa polymerase ␤ protein Wilson et al. 1988). (Obviously, polymerases ␣ and ␦/⑀ have access to these regions during S-phase; however, these cells represent a small fraction of the total population even in exponentially growing cultures.) Furthermore, because rDNA sequences are reiterated and quite highly conserved (Gonzales et al., 1990), one can speculate that recombinational repair (rather than nucleotide excision repair) of UV-induced CPDs (and perhaps other bulky DNA lesions) would offer a selective advantage in maintaining sequence integrity among the rDNA gene copies (see Christians and Hanawalt (1994)), perhaps involving a different mechanism than used in nonreiterated sequences (Ozenberger and Roeder, 1991).
As a control for our studies, we examined repair in the DHFR gene, which is transcribed by RNA Pol II. As stated earlier, transcription repair coupling appears to require an elongating Pol II complex (Leadon and Lawrence, 1991;Christians and Hanawalt, 1992), and we were concerned that DHFR might not be actively transcribed in confluent cells. However, Northern analyses with these human fibroblasts showed that equivalent amounts of DHFR mRNA are present in actively growing cells and cells in deep confluence (Fritz, 1994). Although these studies were not performed on permeabilized cells, the fact that these cells "recuperate" rapidly from LPC permeabilization (Lorenz et al., 1988) and go on to divide normally indicates that the DHFR gene is indeed transcribed following permeabilization. Furthermore, Leys and Kellems (1981) concluded that the relative rates of DHFR transcription in growing and resting cells is the same. (We note that these authors observed a small increase of DHFR mRNA levels in growing cells, after long times (30 h) of growth stimulation, and concluded this increase is due to increased stability of the DHFR message in growing cells (Leys and Kellems, 1981).) In addition, Venema et al. (1990) reported transcription-coupled repair of CPDs in the DHFR gene in confluent human fibroblasts, which (indirectly) suggests that transcription of DHFR occurs in confluent cells.
Finally, a previous study reported efficient repair of bleomycin-induced strand breaks in an amplified c-myc gene in human cells (Bianchi et al., 1990). However, as double strand probes were used in that study, it could not be determined if this gene from newly replicated DNA (see "Materials and Methods"). Parental DNA was treated with (or without) T4 endo V and separated on 0.8% alkaline agarose gels. The samples were transferred to nylon membranes and hybridized to a strand-specific pSPT28S riboprobe. Following autoradiography, membranes were stripped and reprobed with the opposite strand. Panels show autoradiograms for the transcribed strand (TS) and nontranscribed strand (NTS) of actively growing cells (A) and confluent cells (B). Lower panels show the results of quantitative analysis of the autoradiograms in A and B. UV-induced CPDs/fragment of the transcribed (E) and nontranscribed (q) strands are shown for actively growing (C) and confluent (D) cells.
FIG. 4. Repair of UV-induced CPDs in rDNA of actively growing and confluent human cells. Human AG1518 cells, prelabeled with [ 14 C]deoxythymidine, were irradiated with 20 J/m 2 UV light either was repaired in a strand specific manner. Furthermore, Leadon and Cooper (1993) found more efficient repair of ionizing radiation-induced lesions in the transcribed strand of the active metallothionein IIA gene in both normal and xeroderma pigmentosum (complementation group A) human fibroblasts, while Cockayne's syndrome (group B) cells yielded no strand specific repair of this gene. Since bleomycin is considered a radiomimetic agent (Povirk and Finley Austin (1991); although see Affentranger and Burkart (1995)), this report suggests that more efficient repair of bleomycin-induced strand breaks may also occur in the transcribed strand of DHFR. Although we observed only a slight difference between the repair efficiency of the transcribed and nontranscribed strands of the DHFR gene (Fig. 3), this distinction could easily be missed due to the rapid repair kinetics in each strand.