Nucleotide Excision Repair of the 5 S Ribosomal RNA Gene Assembled into a Nucleosome*

A-175-base pair fragment containing theXenopus borealis somatic 5 S ribosomal RNA gene was used as a model system to determine the effect of nucleosome assembly on nucleotide excision repair (NER) of the major UV photoproduct (cyclobutane pyrimidine dimer (CPD)) in DNA. Xenopus oocyte nuclear extracts were used to carry out repair in vitro on reconstituted, positioned 5 S rDNA nucleosomes. Nucleosome structure strongly inhibits NER at many CPD sites in the 5 S rDNA fragment while having little effect at a few sites. The time course of CPD removal at 35 different sites indicates that >85% of the CPDs in the naked DNA fragment have t 1 2 values <2 h, whereas <26% of the t 1 2 values in nucleosomes are <2 h, and 15% are >8 h. Moreover, removal of histone tails from these mononucleosomes has little effect on the repair rates. Finally, nucleosome inhibition of repair shows no correlation with the rotational setting of a 14-nucleotide-long pyrimidine tract located 30 base pairs from the nucleosome dyad. These results suggest that inhibition of NER by mononucleosomes is not significantly influenced by the rotational orientation of CPDs on the histone surface, and histone tails play little (or no) role in this inhibition.

A-175-base pair fragment containing the Xenopus borealis somatic 5 S ribosomal RNA gene was used as a model system to determine the effect of nucleosome assembly on nucleotide excision repair (NER) of the major UV photoproduct (cyclobutane pyrimidine dimer (CPD)) in DNA. Xenopus oocyte nuclear extracts were used to carry out repair in vitro on reconstituted, positioned 5 S rDNA nucleosomes. Nucleosome structure strongly inhibits NER at many CPD sites in the 5 S rDNA fragment while having little effect at a few sites. The time course of CPD removal at 35 different sites indicates that >85% of the CPDs in the naked DNA fragment have t1 ⁄2 values <2 h, whereas <26% of the t1 ⁄2 values in nucleosomes are <2 h, and 15% are >8 h. Moreover, removal of histone tails from these mononucleosomes has little effect on the repair rates. Finally, nucleosome inhibition of repair shows no correlation with the rotational setting of a 14-nucleotide-long pyrimidine tract located 30 base pairs from the nucleosome dyad. These results suggest that inhibition of NER by mononucleosomes is not significantly influenced by the rotational orientation of CPDs on the histone surface, and histone tails play little (or no) role in this inhibition.
In eukaryotic cells DNA is associated with histone proteins in the structural hierarchy of chromatin, required to package the enormous length of DNA into the small volume of a nucleus. The fundamental unit of this hierarchy is the nucleosome core, which consists of 147 bp 1 of DNA wrapped in 1.65 lefthanded superhelical turns around an octamer of 2 each of the 4 core histones, H2A, H2B, H3, and H4 (1). This subunit is linked to adjacent nucleosome cores by less tightly bound linker DNA, and the entire unit repeats every 170 -240 bp (2, 3).
It is believed that DNA damage and DNA processing events in eukaryotic cells such as DNA repair and transcription are modulated by the packaging of DNA into chromatin (4,5). A clear example of chromatin modulation of DNA damage is observed with formation of UV photoproducts (6). The major UV photoproduct in DNA (cis-syn-cyclobutane pyrimidine dimer (CPD)) forms with a striking 10.3-base average periodicity in mixed sequence nucleosome cores (7). This periodic pattern of CPD formation reflects the bending of DNA molecules on the histone surface (8 -10), where the periodic compression of the minor grove of DNA may modulate the [2 ϩ 2]cyclo addition of adjacent 5-6 double bonds after UV photon absorption (11,12). Indeed, CPDs cause an overall bend of 7°-9°in the long axis of a double-strand DNA molecule (13,14), with possibly significant distortion of the phosphodiester backbone on each side of the CPD site (15), and this may be facilitated by compression of the minor groove of adjacent pyrimidines.
DNA repair in chromatin has been extensively examined in human diploid fibroblasts, where two distinct phases exist in the time course of repair in the genome overall (6,16). There is an early rapid phase (lasting 3-6 h after irradiation) and a late slow phase starting between 5 and 16 h after irradiation, depending on the cell strain. It was shown that repair synthesis is randomly distributed in nucleosome core DNA during the late repair phase, whereas there is a distinct bias toward the 5Ј ends of core DNA during the early repair phase (17,18). Surprisingly, during each of these phases, CPDs are repaired at almost equal rates on the inner and outer faces of the DNA helix, relative to the histone surface (18). This suggests that DNA repair enzymes make little distinction between CPDs on different sides of the DNA helix in most of the nucleosomes in human chromatin.
Nucleosome structure can modulate nucleotide excision repair (NER) in the non-transcribed strand of an active gene in yeast (19). Furthermore, nucleosome assembly of specific sequences significantly inhibits the action of both UV photolyase and T4 endonuclease V (T4 endo V) at specific CPD sites in vitro (20,21). Thus, even though these enzymes have much different catalytic activities at CPD sites (11), nucleosome assembly strongly inhibits the activities in each case.
In this report, a 175-bp fragment containing the Xenopus borealis somatic 5 S rRNA gene was used to examine the effects of nucleosome assembly on DNA repair by Xenopus oocyte nuclear extracts. This combination of "substrate" and repair extract was chosen because the 175-bp Xenopus somatic 5 S rRNA gene fragment forms just a few well positioned nucleosomes in vitro (10,22), and Xenopus oocyte nuclear extracts have a robust NER activity (23). These features allowed us to examine the efficiency of removal of CPDs at specific sites within subdomains of a rotationally positioned nucleosome.

MATERIALS AND METHODS
Enzymes and Chemicals-The [␥-32 P]ATP (3000 Ci/mmol) used in this study was obtained from NEN Life Science Products. SexAI and SalI were purchased from Roche Molecular Biochemicals. T4 endo V was a generous gift from Dr. R. S. Lloyd (University of Texas Medical Branch, Galveston, TX). 5 S rDNA Preparation-After linearization of plasmid pKS-5S (24) with either SexAI (for labeling the transcribed strand) or SalI (for labeling the non-transcribed strand), DNA was dephosphorylated with calf intestinal alkaline phosphatase (Roche Molecular Biochemicals) and labeled with [␥-32 P]ATP using T4 polynucleotide kinase (U. S. Biochemical Corp.). A second restriction enzyme digestion (SalI for transcribed strand labeling and SexAI for non-transcribed strand labeling) produced a single end-labeled DNA fragment containing 108 bp of the 5 S rRNA gene (see Fig. 1). The resulting 175-bp DNA fragment was recovered from a 2% agarose gel using a QIAEXII gel extraction kit (QIAGEN Inc.) and dissolved in 10 mM Tris-HCl, 1 mM EDTA, pH 8.0.
Irradiation with UV Light-The DNA samples were diluted with 10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride (PMSF) to a final DNA concentration of 50 ng/l, irradiated under two low pressure Hg lamps (Sylvania, model G30T8) providing predominantly 254-nm light at a flux of 4.3 W/m 2 . UV flux was measured with a Spectroline DM-254N UV meter (Spectronics Corp.).
Trypsin Digestion of Core Particles-Core histones were depleted of their N-terminal tails by trypsin digestion as described by others (20,26). Briefly, core particles (7.7 mg/ml in 10 mM Tris-HCl, pH 7.2, 10 mM cacodylate, 0.2 mM EDTA) were adjusted to 50 mM NaCl, mixed with 0.74 mg/ml trypsin (type XIII, tosylphenylalanyl chloromethyl ketone (TPCK)-treated, Sigma) to make a final trypsin concentration of 6 g/ml. After a 30-min incubation at room temperature, the reaction was stopped with the addition of chicken egg white trypsin inhibitor (Roche Molecular Biochemicals) to a final concentration of 60 g/ml. The digestion time was determined by visualizing the tailless histones on SDS gels as above.
Nucleosome Reconstitution-Reconstitution was achieved by histone octamer transfer as described previously (24). Briefly, 50 ng of endlabeled 5 S rDNA was mixed with 42 g of chicken erythrocyte core particles (ϳ20 g of DNA) in 1 M NaCl, 10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.2 mM PMSF for 30 min at 4°C. The final concentrations of 5 S rDNA and chicken erythrocyte core DNA in the reconstitution mixture was 0.63 ng/l and 0.53 g/l, respectively. The samples were dialyzed against 0.6 M NaCl, 10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.2 mM PMSF for 12 h at 4°C and then 50 mM NaCl, 10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.2 mM PMSF for 4 h at 4°C to complete the reconstitution. As a mock control, naked 5 S rDNA was also mixed with core particles in 10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.2 mM PMSF, and 50 mM NaCl without stepwise dialysis.
Characterization of Reconstituted 5 S rDNA Nucleosomes-DNA mobility shift was used to examine the fraction of 5 S rDNA nucleosomes in the reconstituted population. DNA band shifts were observed on 6% non-denaturing polyacrylamide gels (8.3 ϫ 10.2 cm) in TBE buffer (89 mM Tris, 89 mM borate, 2 mM EDTA, pH 8.3) run for 1.2 h at 120 V. The fraction of 5 S rDNA reconstituted was determined from integration of the two bands, as described below.
Both enzymatic and chemical cleavage methods were carried out to further characterize the 5 S rDNA nucleosomes. Exonuclease III digestions were carried out with 9 l (ϳ6 ng of 5 S rDNA) of naked 5 S rDNA or reconstituted nucleosomes and incubated with 100 units of exonuclease III (Roche Molecular Biochemicals) at 37°C for different times in a buffer containing 50 mM NaCl, 10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 3 mM MgCl 2 , and 1 mM 2-mercaptoethanol. The reaction was stopped with 0.1ϫ volume of 200 mM Tris-HCl, pH 8.0, 5% SDS, and 50 mM EDTA, as described (27). For DNase I digestions, 9 l of reconstituted nucleosomes (6 ng of 5 S rDNA) were treated with 7 units of DNase I (Life Technologies, Inc.) for 1 or 2 min at room temperature, whereas 0.35 units of DNase I were used to digest the same amount of naked DNA for 1.5 or 3 min at room temperature (28). The reaction was stopped by the addition of 0.1 ϫ volume of 25 mM EDTA. Hydroxyl radical footprinting was carried out on 9 l of reconstituted nucleosomes (ϳ6 ng of 5 S rDNA) after dilution to 20 l with TE (10 mM Tris-HCl, 1 mM EDTA, pH 8.0) (29). After incubation with 2 l of 20 mM sodium ascorbate, 2 l of 0.4 mM Fe-EDTA, and 2 l of 1.5% or 3% H 2 O 2 for 2 min at room temperature, the reaction was stopped with 5 l of 0.1 M thiourea and 2 l of 0.2 M EDTA. After each of the reactions described above, the DNA samples were purified by phenol extraction and ethanol precipitation and run on an 8% polyacrylamide, 8 M urea DNA sequencing gel (see below).
Repair Reactions-Repair incubation with Xenopus oocyte nuclear extracts was performed as described previously (23,30). Naked DNA or reconstituted nucleosomes containing ϳ6 ng of 5 S rDNA were mixed with 2 l of Xenopus oocyte nuclear extracts in 20 l of J-buffer (10 mM HEPES, pH 7.4, 70 mM KCl, 7 mM MgCl 2 , 0.1 mM EDTA, 2.5 mM dithiothreitol, 1% polyvinylpyrrolidone, and 10% glycerol). The repair mixture also contained 0.25 mM each of dATP, dGTP, dTTP, and CTP. All repair reactions were incubated at 25°C for the time periods indicated in the text. Reactions were stopped by freezing the samples at Ϫ80°C.
Mapping and Quantitation of CPDs-Before T4 endo V digestion, DNA was phenol extracted, ethanol-precipitated, and resuspended in T4 endo V digestion buffer (20 mM Tris-HCl, pH 7.4, 10 mM EDTA, pH 8.0, 100 mM NaCl, and 100 g/ml bovine serum albumin). T4 endo V (1 l of a 200-fold dilution of enzyme stock (400 ng/l)) was then added, and the reaction was incubated at 37°C for 30 min. After the addition of sequencing gel loading buffer (0.05% bromphenol blue, 0.05% xylene cylanol, 20 mM EDTA in deionized formamide) to stop the reaction, samples were heat-denatured and run on (29 ϫ 96 cm) 8% polyacrylamide, 8 M urea DNA sequencing gels for 2.5 h at 1600 volts (31). The gels were then vacuum-dried, exposed to (43 ϫ 35 cm) PhosphorImager screens, and visualized on a Molecular Dynamics (model 445-P90) Phos-phorImager (Sunnyvale, CA). Band intensities were quantified using ImageQuaNT (Molecular Dynamics) and PeakFit 4.0 (SPSS, Inc.) software. Broad scans of each gel lane (ϳ3/4 of the width) were obtained with ImageQuaNT software, and nested peaks were deconvoluted before integration with PeakFit software using basis spectrum line shapes of Gaussian plus Lorentzian curves (32). In most cases, loading differences were corrected by normalizing the intensity of each band to the sum of all bands in a lane. The fraction of CPDs at each site was plotted for different incubation times, and the t1 ⁄2 values (time for 50% of the initial damage to be removed) was determined from fits to these data. For Fig. 8, the absolute value rather than the relative percentage of each band was used due to the fact that most of the bands run off the gel. The relative percentage of a particular band (corresponding to a specific CPD site) within each lane was used to calculate the percent repair at different times. The average CPDs per strand was calculated from the intensity of the intact fragment resistant to T4 endo V, assuming a Poisson distribution of UV damaged fragments, as described by Bohr et al. (33).

Characterization of Reconstituted 5 S rDNA Nucleosomes-A
diagrammatic representation of the 175-bp 5 S rDNA fragment used in this study is shown in Fig. 1. The ovals show the two major positions of the 5 S nucleosome relative to the various elements of the 5 S rRNA gene determined by Panetta et al. (34) for a longer DNA fragment (249 bp) with the Xenopus somatic 5 S rRNA gene. Nucleosomes were reconstituted onto 5 S rDNA fragments by histone octamer exchange from chicken erythrocyte (CE) core particles isolated from H1/H5-stripped chromatin (25). Gel electrophoresis demonstrated that these CE core particles contained stoichiometric amounts of intact core histones and were depleted of histones H1 and H5 (data not shown). The CE histone octamers were reconstituted onto 5 S rDNA by exchange in high salt followed by stepwise dialysis (see "Material and Methods"). DNA band shift analysis was used to monitor nucleosome formation with both irradiated and non-irradiated 5 S rDNA fragments, and under these conditions, at least 90% of the fragments were reconstituted into nucleosomes ( Fig. 2A). Moreover, band shift experiments The thick arrow designates the location of the 5 S rRNA gene, and the ovals are the two predominate nucleosome formation sites found previously in larger fragments (34). Vertical arrows (ϩ7 and Ϫ3) denote the positions of the dyad axis of these nucleosomes, and the numbers denote positions relative to the 5 S rRNA gene transcription start site (ϩ1). The location of the 14-nt long pyrimidine tract is indicated by the hatched box, and the open box indicates the TFIIIA binding region ICR, internal control region.
showed that nucleosomes are stable for at least 4 h during repair incubation (data not shown).
To characterize these 5 S rDNA nucleosomes, three different types of footprinting were performed (Fig. 2B). Exonuclease III is a 3Ј 3 5Ј exonuclease that can be used to map the translational position of nucleosomes on DNA (27). This enzyme will pause at the edge of nucleosomes and proceed more slowly into nucleosome core DNA with ϳ10-bp pauses (27). As shown in Fig. 2B, naked DNA was quickly digested by exonuclease III into short fragments (lanes 1 and 2), whereas the 5 S rDNA nucleosomes were digested much more slowly (lanes 3 and 4). A major block to digestion occurred at position ϩ65 Ϯ 3 followed by two (less prominent) bands at 10-base intervals from this band, indicating the nucleosome core edge is at (or near) this position. This observation indicates that the most prominent nucleosome core setting spans bases Ϫ67 (5Ј end of fragment) to ϩ65 in the 175-bp fragment. This yields a 131-bp nucleosome core with a translational setting closest to the position having a nucleosome dyad at Ϫ3 (shaded oval in Fig. 1; Ref. 34).
Both DNase I and hydroxyl radical digestion provide information about the rotational setting of DNA in nucleosomes. DNase I cuts the phosphate backbone from the minor groove, which is restricted when facing the histone surface, yielding an ϳ10-base repeat pattern on denaturing gels (reviewed in Ref. 28). Naked DNA is cut more randomly than nucleosome DNA, although the enzyme has a strong sequence bias (Fig. 2B, lanes  5 and 6). With hydroxyl radical cleavage, however, there is much less sequence specificity (29), and these two methods complement each other when evaluating the rotational setting of DNA in nucleosomes. The most intense cleavage sites on the 5 S nucleosome by hydroxyl radical occurred at Ϫ46, Ϫ36, Ϫ25, Ϫ15, Ϫ5, ϩ6, ϩ17, ϩ27, ϩ37, ϩ47, and ϩ57 (Fig. 2B, lanes 11  and 12). DNase I cuts preferentially within 1 nucleotide (nt) of most of these same nucleosome 5 S rDNA sites (Fig. 2B, lanes  7 and 8). We note that an intense DNase I cut site, but not hydroxyl radical cut site, also occurs at Ϫ31, being strong in both naked 5 S rDNA and the 5 S nucleosome (star in Fig. 2B, lanes [5][6][7][8]. This reflects the strong sequence specificity of DNase I (28). These results are in good agreement with previous studies on the rotational setting of 5 S rDNA in nucleosomes (22).
Repair of 5 S rDNA Nucleosomes by Xenopus Oocyte Nuclear Extracts-Contrary to the low repair efficiency found in cultured mammalian cell extracts (e.g. 35), Xenopus oocyte nuclear extracts remove almost all CPDs from irradiated exogenous DNA within a few hours (23). Therefore, these extracts were chosen for examining repair of CPDs in the 5 S rDNA nucleosome, since CPD removal at specific sites could be followed. For these experiments, the 175-bp 5 S rDNA fragment was UV irradiated, reconstituted into nucleosomes, and incubated with oocyte nuclear extracts as described in Conconi et al. (30). After different incubation times, 5 S rDNA was purified, cut with T4 endo V, and run on DNA sequencing gels as above. As shown in Fig. 3, repair of most CPDs in either strand of 5 S rDNA (or the disappearance of bands) was strongly inhibited by nucleosome assembly (compare D and N lanes). The decreased intensity of CPD bands should be accompanied by an increased intensity of the whole length (or uncut) fragment, which reflects the overall repair of the 5 S rDNA fragment. As shown in Fig. 4, the overall repair efficiency of each strand decreases upon nucleosome formation (compare solid to dashed lines), in complete agreement with the trend at individual CPD sites observed in Fig. 3.
These results can be compared by determining the percentage of CPDs remaining at each site after different times of repair incubation. Three examples are shown in Fig. 5 for different short tracts of CPD sites (denoted by stars in Fig. 3). Clearly, naked DNA has different repair rates than nucleosomal DNA at two of these sites ( ϩ21 CTTTC ϩ25 and ϩ54 TTC ϩ56 ). Furthermore, these slower repaired CPDs in nucleosomal DNA were removed at different rates from each other (compare top and middle panels in Fig. 5). Interestingly, a few CPD sites in nucleosomes were repaired almost as rapidly as naked DNA (e.g. site ϩ58 TCCC ϩ61 , Fig. 5, bottom panel). The structural features of these sites are discussed below (see "Discussion").
The t1 ⁄2 value of each CPD site was determined from its repair curve, and these values are compared graphically over the entire 5 S rDNA fragment in Fig. 6. Most CPD sites in naked DNA have t1 ⁄2 values of less than 2 h (Fig. 6, closed bars). However, careful inspection of several different gels and sub- For hydroxyl radical reactions, either 1.5% (lanes 9 and 11) or 3% (lanes 10 and 12) H 2 O 2 was used to react with Fe-EDTA to generate hydroxyl radicals (see "Materials and Methods"). Numbers on the right denote the most frequent cleavage sites on the reconstituted nucleosomes by both DNase I and hydroxyl radicals. Numbers on the left denote the three major exonuclease III blockage sites. sequent quantitation of the results indicates that the rate of repair at individual sites in naked 5 S rDNA is quite variable. For example, extremely fast repair is observed at sites CT ϩ42 , ϩ54 TTC ϩ56 , ϩ58 TCCC ϩ61 , ϩ64 CCC ϩ66 , CC ϩ71 , and TC ϩ75 of the TS, whereas sites Ϫ59 TCTCCT Ϫ54 of the TS and ϩ28 CCC ϩ30 of the NTS are repaired much more slowly. This variation in site-specific repair does not correlate with initial yields of CPDs at these sites (data not shown) and, therefore, may reflect an influence by neighboring DNA sequence and/or local DNA structure on damage recognition or removal.
Compared with naked DNA, repair of CPDs in 5 S rDNA nucleosomes is slower at almost all sites, with most t1 ⁄2 values being Ͼ2 h (Fig. 6, shaded bars). Moreover, some CPD sites just outside of the predominant nucleosome location (i.e. positions larger than ϩ80 of the TS) are also repaired slower. This may in part be due to the variation in nucleosome translational settings and/or binding of the DNA ends by exposed basic residues on the histone tails. However, some CPD sites continue to be repaired efficiently in the nucleosome complex (e.g. ϩ58 TCCC ϩ61 , Figs. 5 and 6). In addition, the time course of repair at individual CPD sites relative to the other sites observed for naked DNA are different from those observed for nucleosomes. For example, one slow repair site of naked DNA, ϩ28 CCC ϩ30 of NTS, is repaired at an intermediate rate in nucleosomes. Repair at sites Ϫ28 TTCC Ϫ25 and ϩ21 CTTTC ϩ25 of the TS is fast in naked DNA but very slow in nucleosomes.
Thus, repair at CPD sites in 5 S rDNA appears to be differentially modulated by nucleosome assembly.
Repair of 5 S rDNA Nucleosomes Trimmed of Histone Tails-The N-terminal tails of the core histones containing a high proportion of basic amino acids may play important roles in stabilizing nucleosome structure, particularly in polynucleosome fragments (36). Therefore, to examine their effect on NER of mononucleosomes, the histone tails were removed from CE nucleosomes by trypsin digestion, as described by Ausio et al. (26). Under these conditions, the tails are efficiently removed from the four core histones as demonstrated by their increased migration on SDS gels (Fig. 7A). As expected, when these CE nucleosomes are used for reconstitution, the 5 S rDNA nucleosomes migrate faster than the intact 5 S rDNA nucleosomes on native gels (Fig. 7B). Importantly, the fraction of nucleosomes reconstituted with the trypsin-digested CE core particles does not change appreciably (Fig. 7B). However, removal of the histone tails does not significantly enhance NER in 5 S rDNA mononucleosomes. As shown in Fig. 7C for CPDs in the 3Ј half of the fragment (Ϫ67 to ϩ65), repair of the histone tail-depleted nucleosomes is very similar to that of intact 5 S rDNA nucleosomes (compare ϩ and Ϫ trypsin lanes). Similar results were obtained for CPDs in the 5Ј half of the fragment (data not shown). Therefore, the histone tails in these mononucleosomes must play only a minor role in the modulation of NER in the 5 S rDNA nucleosome.
Finally, we examined the correlation between site-specific NER and the rotational setting of 5 S rDNA on the histone surface. For these experiments, a 14-nt long pyrimidine tract in the transcribed strand was analyzed in detail, since it spans more than one complete turn of the 5 S rDNA helix (Fig. 8A). Resolution of the 13 potential CPD-associated bands required a longer gel running time (Fig. 8B), as the pyrimidine tract is located about 130 bases away from the 32 P-labeled 5Ј end (see Fig. 1). To accurately determine repair rates from the initial slopes of repair curves, much shorter times were used for repair of naked 5 S rDNA than for 5 S rDNA nucleosomes (Fig. 8B). The area of each band was determined by peak deconvolution and used to calculate the percent of CPDs removed at each site and after each repair time. The repair rate was determined from the initial slope of each repair curve, as shown in Fig. 8C for site CC Ϫ19 (arrow in Fig. 8B). To minimize the influence of DNA sequence on the repair rates, the ratios of the slopes for nucleosomes and naked DNA (m N /m D ) were plotted versus CPD location (Fig. 8D). With the possible exception of one site (TC Ϫ14 ), no correlation exists between the rotational setting of CPD sites on the core histone surface and NER rates (Fig. 8D). The TC Ϫ14 site is at the 3Ј end of the tract facing away from the histone surface and has the highest value of m N /m D (0.58 Ϯ 0.08). However, the m N /m D ratio of the other 12 CPD sites is very similar (ϳ60% inhibition). For example, the m N /m D values for sites CC Ϫ19 (facing toward the histone surface) and TC Ϫ25 (facing away from the histone surface) are 0.31 Ϯ 0.09 and 0.40 Ϯ 0.08, respectively. The apparent small modulation of relative repair rates observed in Fig. 8D is not significant.

DISCUSSION
The X. borealis 5 S rRNA gene has been used as a model system to study the relationship between RNA polymerase III transcription and chromatin for many years. DNase I footprinting was used to show that the majority of nucleosomes are precisely positioned after in vitro reconstitution onto a 256-bp long DNA fragment containing the sea urchin 5 S rRNA gene (37). Mutation analysis indicates that a region 20 -30 bp on either side of the center of the core particle may contain the major elements responsible for nucleosome positioning (38). More recently, it was demonstrated that there is one dominant nucleosome position surrounded by minor positions, 10 bp apart, maintaining the identical rotational setting of the 5 S rDNA (39,40). In the present study, we chose a smaller fragment (175 bp) containing 108 bp of 5 S rDNA gene (41) to limit the number of molecules with different translational settings (Fig. 1). Exonuclease III digestion indicates that the predomi- nant nucleosome positions on our fragment are with dyad positions near the transcription start site (Fig. 2). Importantly, the rotational setting of the 5 S rDNA relative to the histone surface (examined by DNase I and hydroxyl radical footprinting) is in agreement with that previously reported (10,22). In a separate study, we also did extensive experiments to analyze the effect of UV irradiation on 5 S nucleosome positioning (10). After 500 J/m 2 UV (i.e. the dose used for these repair experiments), neither translational setting nor rotational positioning showed any significant changes.
It was shown previously that NER of CPDs by human cell extracts is suppressed in plasmid DNA containing reconstituted nucleosomes (42). In that study, repair synthesis was required to detect DNA repair due to the low repair efficiency of human cell extracts, and measurement of repair at specific CPD sites was not possible. The robust DNA repair activity of Xenopus oocyte nuclear extracts allowed us to follow CPD removal at individual sites in the 5 S nucleosome. In agreement with the data from human cell extracts (42), we also detect a marked inhibition of NER at many sites in 5 S rDNA upon nucleosome formation. Curiously, repair at some sites is not inhibited by nucleosome formation (Figs. 5 and 6).
Recently, using the same in vitro system, we observed that binding of transcription factor TFIIIA to 5 S rDNA severely inhibits repair in the 50-bp TFIIIA binding region (30). This inhibition was limited to the 50-bp TFIIIA binding site (or internal control region), as repair was equal to that of naked 5 S rDNA at CPD sites just outside the ICR. In this study, however, slow repair of nucleosomal 5 S rDNA was detected at most CPD sites throughout the fragment (Fig. 6). This difference most likely reflects the much higher binding specificity and affinity of TFIIIA for the internal control region than that of the histone octamer for 5 S rDNA (10,34). Furthermore, it was shown that transcription of the 5 S rRNA gene does not occur in our repair reactions (30). Therefore, our present results reflect NER of the 5 S rRNA gene in the absence of polymerase III transcription.
In this report, we found that naked 5 S rDNA is repaired very efficiently at most CPD sites in the Xenopus oocyte nuclear extract. At these sites, CPDs are almost completely removed after a 2-h incubation (t1 ⁄2 values are Ͻ1 h). The few exceptions (e.g. Ϫ59 TCTCCT Ϫ54 in the TS where t1 ⁄2 ϭ 4.8 h) may be due to the effect of DNA sequence on repair (43). Assembly of 5 S rDNA into nucleosomes generally caused inhibition of DNA repair. Several of the very slow repair sites are in the interior 125 bp of the nucleosome core (Ϫ65 to ϩ60) (Fig.  6). Over 50% of the CPD sites in this region have t1 ⁄2 values greater than 4 h, with extremely slow repair occurring at sites Ϫ59 TCTCCT Ϫ54 , Ϫ28 TTCC Ϫ25 , and ϩ21 CTTTC ϩ25 of TS and CC ϩ15 , ϩ43 TCTC ϩ46 , and CC ϩ53 of NTS (t1 ⁄2 Ͼ 8 h). In contrast, CPDs at five sites in this region ( Ϫ10 CCT Ϫ8 , TT Ϫ7 , ϩ17 CCCT ϩ20 , and ϩ38 TCT ϩ40 of NTS and CT ϩ32 of TS) are repaired at almost the same rate as in naked DNA. Beyond ϩ60, 10 of 14 CPD sites have t1 ⁄2 values less than 4 h (i.e. approaching repair rates of naked DNA). The remaining CPD sites are more slowly repaired (t1 ⁄2 values of 4 to 7 h), and these sites may reflect multiple translational settings of the histone octamer.
The average half-life of CPDs within the interior 125-bp domain of nucleosomes was 4.8 h, whereas beyond this domain the average half-life dropped to 3.1 h. However, the striking variation of repair rates near the dyad of the nucleosome indicates that access of CPDs to repair proteins in the Xenopus nuclear extracts does not follow the dynamic nucleosome model in the simplest form proposed by Polach and Widom (44,45). In this model, CPD sites near the two edges of the nucleosome should be far more accessible (by a factor of 100 to 1000) than CPD sites in the dyad region. Presumably, the observed variations in repair rate reflect local histone-DNA interactions.
Comparison of repair rates at different CPD sites in the 5 S rDNA nucleosome with the crystal structure indicates that three of the slowly repaired CPD sites ( ϩ21 CTTTC ϩ25 and CT ϩ42 of the TS and site CC ϩ53 of the NTS) are in direct contact with the histone folds (1). However, other slowly repaired CPD sites are not in direct contact with the histones, such as Ϫ40 TCTTCC Ϫ35 and Ϫ28 TTCC Ϫ25 of the TS and sites CC ϩ15 , TC ϩ36 , and ϩ43 TCTC ϩ46 of the NTS. More surprisingly, almost all of the fast repair sites make direct contact with histone folds in the crystal (1). These are sites CT ϩ32 and ϩ58 TCCC ϩ61 of the TS and sites Ϫ10 CCT Ϫ8 , ϩ17 CCCT ϩ20 , and ϩ38 TCT ϩ40 of the NTS. These may represent sites that locally destabilize the nucleosome core or reduce the dominant constraint(s) on DNA by core histones (46), rendering the interior of these nucleosomes more accessible to repair enzymes.
Removal of histone N-terminal tails does not significantly change the inhibition of NER by nucleosome assembly (Fig. 7). FIG. 8. Relationship between DNA repair and rotational setting of a 14-nt-long pyrimidine tract (see Fig. 1). A, sequence of the 14-nt pyrimidine tract in 5 S rDNA fragment. Stars represent sites farthest away from the histone surface, and the triangle indicates the site closest to histone surface in the undamaged 5 S rDNA fragment. Numbers above the sequence denote nt from the transcription start site of the 5 S rRNA gene. B, similar experiments as described in Fig. 3 were performed with both naked DNA and nucleosomes, where different incubation times (denoted above each lane) were used to optimize analysis. C, time course of repair of site CC Ϫ19 (denoted by the arrow in panel B) determined from the band intensity following different repair times. Two slopes (m D and m N ) were obtained after linear regression analysis of the data for naked DNA (E) and nucleosomes (q). D, the ratio of the two slopes (m N /m D ) at each CPD site in the 14-nt-long pyrimidine tract (plotted as location from the transcription start site of the 5 S rRNA gene). Similar findings have been obtained with the activities of E. coli UV photolyase and T4 endo V on the 134-bp HISAT nucleosome (20). In that study, it was found that photoreversal and cleavage of CPDs is inefficient in nucleosomes compared with naked DNA, and removal of the histone tails did not substantially enhance these activities. The role of histone N-terminal tails in chromatin has been elusive. It was recently proposed that in chromatin the N-terminal tails are engaged primarily in protein-protein interactions rather than protein-DNA interactions (36). In addition, the core histone tails were found to have a repressive effect on transcription within oligonucleosomes (47). Thus, it was not surprising that NER was unaffected by removal of the histone tails in our study on mononucleosomes but may effect repair within oligonucleosomes.
We also analyzed the correlation of NER with rotational setting in a 14-nt pyrimidine tract on the transcribed strand (Fig. 8). Repair at all but one CPD site at the 3Ј end of this tract was very similar in 5 S nucleosomes and about 60% less than in naked 5 S rDNA after 2 h. This result indicates there is no correlation between rotational setting and repair kinetics in the 5 S nucleosome, and CPDs facing toward the histone surface are repaired with equal efficiency in the Xenopus extract as CPDs facing outward. Alternatively, CPDs at sites near the histone surface may force the DNA helix to rotate outward at these sites. These results are in agreement with previous work in this lab showing that CPDs are removed at nearly equal rates from the inner and outer faces of the undamaged DNA helix in nucleosome cores during the early, rapid phase of repair in human cells (18). One possible explanation for this lack of correlation between DNA repair and original rotational setting of the DNA helix is that nucleosomes are actively disrupted before NER (e.g. see Ref. 48). Clearly, this possibility is supported by the extensive evidence showing nucleosome rearrangement during NER in intact cells (reviewed in Refs. 4 and 6).
In conclusion, we found that (a) nucleosome assembly inhibits repair by Xenopus oocyte nuclear extracts of the 5 S rDNA sequence at many CPD sites, (b) CPD removal at specific sites varies markedly where some rapidly repaired sites are located in the interior of the nucleosome, (c) removal of histone tails does not enhance DNA repair efficiency in these mononucleosomes, and (d) CPD removal rates in a 14-nt pyrimidine tract do not correlate with the rotational setting of the undamaged rDNA fragment assembled into the nucleosome.