Nucleosomes Suppress the Formation of Double-strand DNA Breaks during Attempted Base Excision Repair of Clustered Oxidative Damages*

Background: Ionizing radiation can produce clustered lesions in DNA; attempted base excision repair of these lesions can generate double-strand breaks (DSBs). Results: The extent to which nucleosomes suppress DSB formation is governed by their structural and dynamic properties. Conclusion: Nucleosomes suppress formation of radiation-induced DSBs. Significance: This study helps elucidate mechanisms responsible for potentially mutagenic or lethal DSBs. Exposure to ionizing radiation can produce multiple, clustered oxidative lesions in DNA. The near simultaneous excision of nearby lesions in opposing DNA strands by the base excision repair (BER) enzymes can produce double-strand DNA breaks (DSBs). This attempted BER accounts for many of the potentially lethal or mutagenic DSBs that occur in vivo. To assess the impact of nucleosomes on the frequency and pattern of BER-dependent DSB formation, we incubated nucleosomes containing oxidative damages in opposing DNA strands with selected DNA glycosylases and human apurinic/apyrimidinic endonuclease 1. Overall, nucleosomes substantially suppressed DSB formation. However, the degree of suppression varied as a function of (i) the lesion type and DNA glycosylase tested, (ii) local sequence context and the stagger between opposing strand lesions, (iii) the helical orientation of oxidative lesions relative to the underlying histone octamer, and (iv) the distance between the lesion cluster and the nucleosome edge. In some instances the binding of a BER factor to one nucleosomal lesion appeared to facilitate binding to the opposing strand lesion. DSB formation did not invariably lead to nucleosome dissolution, and in some cases, free DNA ends resulting from DSB formation remained associated with the histone octamer. These observations explain how specific structural and dynamic properties of nucleosomes contribute to the suppression of BER-generated DSBs. These studies also suggest that most BER-generated DSBs will occur in linker DNA and in genomic regions associated with elevated rates of nucleosome turnover or remodeling.

of DNA lesions in nucleosomes, albeit with varying and sometimes much reduced efficiency relative to their activities on naked DNA (for reviews, see Ref. 18; see also Refs. 19 -21). The efficiency with which a nucleosomal lesion is processed depends critically on its helical orientation relative to the underlying histone octamer. These observations suggested that the packaging of DNA in nucleosomes would substantially suppress BER-mediated DSB formation in vivo. In particular, only one of two closely spaced lesions on opposing strands could be optimally accessible to DNA glycosylases. In such cases, BER of the more readily accessible of two closely opposed lesions might go to completion before BER of the less accessible lesion commenced, thereby precluding DSB formation. Here, we have tested this prediction using nucleosomes that contain discretely positioned oxidative lesions in opposing DNA strands. We report that nucleosomes partially protect DNA from BERmediated DSB formation but that the protection is not absolute. We have also investigated the enzymatic and structural factors that influence the efficiency with which DSBs occur in nucleosomes. Our results predict that the distribution of BERmediated DSBs within the genome is non-random and very likely influenced by cellular processes that entail remodeling or transient disruption of nucleosomes.

EXPERIMENTAL PROCEDURES
Preparation of Nucleosome-length DNA Substrates Containing Oxidative Lesions-We used two methods to assemble mononucleosome-length (170 -185 bp) DNAs containing multiple thymine glycol (Tg) or 7,8-dihydro-8-oxoguanine (8-oxoG) lesions. The first began with the synthesis of a fulllength single-strand DNA template in a linear PCR reaction. The template for this reaction was prepared by linearizing the plasmid pBS-5SLv (22) at a KpnI site immediately adjacent to the 5 S ribosomal DNA-containing nucleosome positioning sequence from Lytechinus variegatus. To this DNA we added buffer, TaqDNA polymerase (Invitrogen), and a 25-fold molar excess of a lesion-containing primer (Midland Certified Reagent Co.) that had been 5Ј end-labeled with [␥-32 P]ATP by polynucleotide kinase (New England Biolabs). After 22 cycles of denaturation (95°C), annealing (53°C) and extension (72°C), we purified the resulting "top strand" DNA by denaturing PAGE, quantified it by scintillation counting, and mixed it with equimolar amounts of three "bottom strand" oligodeoxyribonucleotides. Of the bottom strand oligodeoxyribonucleotides, the 5Ј-most was ␥-32 P-end-labeled, making it possible to individually monitor excision of oxidative lesions from both top and bottom strands of DNA. In later experiments, we prepared nucleosome-length DNA substrates using the same bottom strand DNA oligodeoxyribonucleotides as before, along with three top strand oligodeoxyribonucleotides. To these, we added buffer and a thermostable DNA ligase (Ampligase, from Epicenter) and a small amount (1/20th) of undamaged template DNA (from pBS-5SLv), to "guide" the initial ligation events. Reactions were allowed to proceed for 297 cycles of denaturation and ligation. The resulting full-length products were purified by denaturing PAGE, annealed to one another, and treated with Exonuclease I (New England Biolabs) to remove any remaining single-stranded DNA.
Nucleosome Reconstitutions-In early experiments, the lesion-containing DNAs described above were assembled into nucleosomes by octamer transfer using a 100 -200-fold excess of chromatin donor prepared from chicken erythrocytes as described (22). In later experiments, lesion-containing DNAs were mixed with an 80-fold molar excess of unlabeled carrier DNA (isolated from micrococcal nuclease-digested chicken chromatin) and a slight molar excess of purified histone octamers (assembled using recombinant human or Xenopus histones from Escherichia coli; Refs. 23 and 24) in 25 mM HEPES, pH 8.0, 1 mM EDTA, 1 mM DTT, and 2 M sodium chloride and dialyzed into the same buffer lacking sodium chloride. In both methods the lesion-containing nucleosomes were adjusted to 2.5 nM and diluted to 0.5 nM as needed for activity assays. The presence of unlabeled carrier nucleosomes ensured that the total concentration remained well above concentrations at which one sees dilution-driven dissociation of nucleosomes (cf. Ref. 25). To assess reconstitution efficiencies, nucleosomes were electrophoresed through 5% native gels in 0.5ϫ Tris borate-EDTA buffer. The impact of small amounts of contaminating naked DNA on experimental results was eliminated computationally as previously described (22).
Isolation of BER Enzymes-Wild type human endonuclease III (hNTH1) and a variant lacking the first 55 amino acids (hNTH1⌬55) were expressed and purified as described (26,27). hAPE1 and E. coli formamidopyrimidine-DNA glycosylase (Fpg) were purchased from New England Biolabs.
Enzyme Activity Assays-All the glycosylase enzyme concentrations reported in this study refer to active enzyme, as determined using a Schiff base assay (29). The reported concentrations of hAPE1 refer to total enzyme, calculated from data supplied by the manufacturer. Unless otherwise noted, enzyme reactions also contained 0.5 nM lesion-containing nucleosomes or naked DNA in 25 mM HEPES, pH 8.0, 1 mM EDTA, 1 mM DTT, 100 mM sodium chloride, 5 mM MgCl 2 , 0.05% Nonidet P-40, and 0.1 mg/ml bovine serum albumin (New England Biolabs). Naked DNA control reactions contained either chicken chromatin or micrococcal nuclease-digested chicken DNA (depending upon nucleosome reconstitution method) such that carrier conditions matched those used in the nucleosome reactions. Reactions were carried out at 37°C for up to 30 min and quenched by the addition of a 1 ⁄ 5 volume of 2.5% SDS, 250 mM EDTA, and 2.5 mg/ml freshly added proteinase K. After a 30 -60-min incubation in quench buffer at 37°C, aliquots to be used for quantifying DSB formation were mixed with equal volumes of 100 mM Tris base, 25 mM EDTA, 0.2% SDS, 2 mg/ml bromphenol blue, and 12% glycerol and electrophoresed through 6% native gels. Aliquots to be used for quantifying single-strand break (SSB) formation were mixed with equal volumes of formamide containing 20 mM EDTA and 2 mg/ml each of bromphenol blue and xylene cyanol and fractionated on 6% sequencing gels. Gels were then dried and exposed to K-screens (Bio-Rad), and bands of interest were quantified using PharosFX Plus phosphoimaging with Quantity One software (Bio-Rad). After adjusting for backgrounds (using the global subtraction method), DSBs were calculated as a percentage of total DNA. SSBs were calculated for each strand by dividing the individual product band (multiplied by two to adjust for the fact that DNA was double end-labeled) by the total sum of all bands (i.e. both product bands ϩ substrate).
Assays for Nucleosome Integrity during Attempted Repair of Clustered Lesions-To assess the fate of clustered lesion-containing nucleosomes during attempted BER, aliquots were removed from the above-described reactions after 30 min at 37°C, mixed with equal volumes of 25 mM HEPES, pH 8.0, and 10% glycerol, and immediately subjected to electrophoresis on a 6% native gel. A portion of each aliquot was reserved for quantification of DSB formation as described above. In some instances yet another aliquot was treated with 50 units of Age I, a restriction endonuclease whose target in 5 S rDNA nucleosomes is normally inaccessible (22). To assess Age I cleavage efficiencies, samples were processed and fractionated as in the assay of DSB formation.

RESULTS
Experimental Design-DNA glycosylases discover oxidative lesions by extending residues into the minor groove of DNA and inducing damaged bases to flip through the major groove into an extra-helical configuration (30 -32). Steric constraints imposed by the histone octamer or DNA in the adjacent gyre may impede either glycosylase binding or base flipping. As a result, DNA glycosylases can bind directly to only a small fraction of the base lesions that may form in nucleosomes (for reviews, see Ref. 18; see also Refs. 19 -21). This is illustrated in Fig. 1, B and C, which depict the region of a nucleosome where we inserted base lesions for our initial studies. In naked DNA, attempted BER of these lesions might generate a DSB, whereas in a nucleosome only the lesion located 51 nt from the dyad axis is oriented so as to permit direct binding of a DNA glycosylase. Binding of a DNA glycosylase to the other, less optimally oriented lesions (located 46 and 49 nt from the dyad axis on the opposite strand) can occur only when they are exposed to sol-vent during episodes of spontaneous partial unwrapping of DNA from the histone octamer (21,22). Because this DNA unwrapping-dependent mode of BER is much less efficient, it seemed likely that nucleosomes inhibit the near-simultaneous lesion processing events that could lead to DSB formation. This reasoning was the basis for a null hypothesis, namely that nucleosomes fully protect DNA from BER-mediated formation of DSBs. In an attempt to disprove this hypothesis, we constructed and tested an array of model nucleosomes, each containing opposing strand base lesions. To fix the helical orientation of the lesions relative to the histone octamer, nucleosomes contained either the naturally occurring L. variegatus 5 S ribosomal DNA or the synthetic 601 DNA nucleosome positioning sequence. DNA in the 601 nucleosome occupies a single discrete helical and translational position relative to the underlying histone octamer (33,34). In the 5 S rDNA nucleosomes, most DNA molecules occupy one of three translational positions, separated by 10-bp intervals (22,35,36). Although the relative abundance of these translational variants differs somewhat with reconstitution conditions (21), the fact that they are separated by an integral number of helical turns means that the DNA in each nucleosome has the same helical orientation with respect to the underlying histone octamer (22,37).
Impact of the Stagger between Opposing Lesions on Doublestrand Break Formation-In nucleosomes, a base lesion that is optimally positioned for processing will be paired with an occluded base in the opposing strand. Thus, it was likely that optimal processing of opposing strand lesions would require that they be offset or staggered relative to one another. Two factors guided our decisions on the range of offsets tested. First, the short reactive lifetimes of hydroxyl radicals produced by a single photon suggested that clustered DNA damages generally lie within ϳ10 bp of one another. This places an upper limit on the distance between opposing strand lesions. To ensure that both lesions would be at least partly accessible to DNA glycosylases in nucleosomes, we adopted an even more stringent upper limit of 5 bp between opposing strand lesions. Second, efficient DSB formation occurs in naked DNA substrates only when opposing strand lesions are offset by at least three bp (38). This observation is consistent with evidence that most DNA glycosylases interact with both the DNA lesion and two or more backbone residues on either side of the lesion (e.g. Refs. 39 and 40). Thus, an AP apurinic/apyrimidinic site or DNA gap located just one or two bp away from a base lesion in the opposing strand would likely reduce the affinity of a DNA glycosylase for its substrate. This consideration led us to construct as a negative control a nucleosome with a 2 bp stagger between opposing strand lesions. We thus replaced thymine residues in each strand of DNA with the oxidative lesion Tg, generating DNA with opposing lesions separated by either 2 or 5 bp ("Tg51-2" and "Tg51-5," respectively). When assembled into nucleosomes, the top strand Tg was located 61, 51, or 41 nt from the central dyad axis, with the majority at position 51 (22). The bottom strand Tg was either 2 or 5 nt closer to the dyad axis, as depicted in Fig. 1, B and C.
hNTH1 is one of two bifunctional human DNA glycosylases that we knew from earlier studies could excise single Tg residues from nucleosomes. As outlined above, we predicted that processing of either lesion in Tg51-2 DNA would degrade the binding site needed to process the opposing strand lesion. However, because of its exceptionally low turnover rate, it was possible that hNTH1 binding to its product might physically prevent binding of a second hNTH1 molecule to an opposing strand lesion. To eliminate this possibility, we probed Tg51-2 nucleosomes and DNA with hNTH1⌬55, an N-terminal truncation mutant that exhibits an elevated turnover rate but is otherwise fully active (26,41). As can be seen in Fig. 2, B and D, DSB formation was poor in naked DNA and virtually undetectable in nucleosomes. By contrast, hNTH1⌬55 formed DSBs in a large fraction (72 Ϯ 7%) of the naked Tg51-5 DNA molecules and in a smaller but still significant fraction (21 Ϯ 6%) of Tg51-5 nucleosomes (Fig. 2, C and D).
hNTH1-mediated, Double-strand Break Formation in Nucleosomes Is Lyase-limited-Although the reduced yield of DSBs in Tg51-5 nucleosomes as compared with naked DNA (Fig. 2) could be due entirely to steric factors, it was possible that the results were influenced by other properties of hNTH1. In particular, the lyase activity in hNTH1 is substantially lower than the glycosylase activity; this is also the case for several other bifunctional DNA glycosylases (42-47). hAPE1 will dis-place hNTH1 from abasic sites and increase production of SSBs (22) . Fig. 3, A and B, shows that the addition of hAPE1 increased both the rate and extent of DSB formation in both nucleosomes and DNA but did not significantly alter the degree to which nucleosomes suppress DSB formation. Fig. 3C shows that both full-length hNTH1 and the high turnover truncation mutant hNTH1⌬55 produced DSBs with similar efficiencies. Thus, the relatively poor lyase activity of hNTH1, but not its slow turnover rate, was limiting for DSB production in cases when opposing lesions were separated by 5 bp. In cells, hAPE1 is thought to be present in excess over individual DNA glycosylases (48,49). We, therefore, included hAPE1 in most subsequent experiments, reasoning that doing so would better approximate in vivo conditions. DSB Formation Due to Processing of Clustered Lesions in 601 Nucleosomes-The oxidative lesions in the 5 S rDNA-based nucleosomes used for the above-described experiments were located relatively close (ϳ12 to ϳ37 nt depending on translational position) to the nucleosome edge. To determine if DNA glycosylases generate DSBs when processing clustered lesions located closer to the dyad axis and to control for possible sequence context effects, we replaced two thymine residues in FIGURE 1. Double-strand break formation during attempted BER. A, schematic of attempted BER resulting in DSB formation. Ionizing radiation causes multiple, closely spaced oxidative lesions on opposing DNA strands. Near-simultaneous strand cleavages by either a glycosylase or apurinic/apyrimidinic site endonuclease result in a DSB that can no longer be repaired by the BER pathway. B, summary of 5 S rDNA-based nucleosome constructs, indicating the approximate position of lesions with respect to the histone core (H2A-H2B and H4-H3 dimers). Construct names are in parentheses and indicate the lesion, top strand lesion position with respect to the dyad axis (Ϫ50 or Ϫ51), and base pair stagger (2 or 5) between opposing strands. The predicted yield of DSBs is based upon previous work using naked DNA substrates (38,62) as well as relative accessibility of the lesions with respect to the histone core. C, crystal structure of the 1. 9 Å ␣-satellite nucleosome (PDB 1KX5), oriented to highlight the 51-2 and 51-5 lesion positions. Position Ϫ51 is located on the orange strand, with 2 bp and 5 bp opposing lesions on the purple strand, closer toward the dyad axis (lesion positions are in blue and circled). Lesion orientations in 8oxoG50-5 differ from those shown for Tg51-5 by ϳ36°, whereas those for Tg39ϩ5 differ by ϳ72°. The glycosylase must bind to the lesion via the minor groove, and then "flip" the damaged base into its active site through the major groove. the top and bottom strand lesions would lie 39 and 44 nt from the dyad axis (50). We predicted that both lesions in the resulting "Tg39ϩ5" nucleosome would be only marginally accessible to hNTH1 and thus expected to see very limited DSB production. Surprisingly, the efficiency of DSB production in the 601 nucleosomes was as high as that observed for Tg51-5 nucleosomes (Fig. 4, A and B). Fig. 4C documents the effect of increasing glycosylase concentration on the efficiency of DSB formation. For enzyme concentrations in excess over total lesion (1 nM), the efficiency of DSB formation in naked DNA quickly reached a plateau value. In contrast, the efficiency of DSB formation in nucleosomes continued to climb over the enzyme range tested. This increasing yield of DSBs at high enzyme concentrations is a hallmark associated with the capture of lesions that are accessible only during periodic unwrapping events (21,22). The fact that this phenomenon was more pronounced for the 601 Tg39ϩ5 nucleosomes than for the 5 S rDNA nucleosomes is consistent with the greater distance between the lesions in Tg39ϩ5 and the nucleosome edge and a correspondingly lower rate of unwrapping-mediated lesion exposure.
Evidence for Coupled Processing of Clustered Lesions in Nucleosomes-Given the increased distance between the lesions in 601 Tg39ϩ5 nucleosomes and the nucleosome edge, the production of DSBs was surprisingly efficient. This suggested that the two lesion processing events required for DSB formation might somehow be coupled. If instead opposing strand lesions are processed independently of one another, the frequency of DSB formation (measured using non-denaturing gels such as that shown in Fig. 5A) should equal the product of the two single-strand cleavage frequencies (measured using denaturing gels such as that shown in Fig. 5B). To distinguish between independent and coupled processing of lesions, we measured rates of single-strand break formation for both the lesion-containing 5 S rDNA nucleosomes and naked DNA controls (Fig. 5, C and D, and Table 1) and for the above-described 601 nucleosomes ( Fig. 5E and Table 1). In the case of naked, Tg51-5 DNA substrates, Fig. 5C shows that hNTH1 processed ϳ80% of the lesions in each DNA strand, leading to formation of DSBs in (0.8 ϫ 0.8) ϳ64% of molecules within the reaction timeframe. Table 2 compares observed DSB frequencies from multiple experiments to those predicted from single-strand cleavage data, assuming independent processing of single lesions. If this assumption is correct, a plot of the observed versus predicted values will produce a slope equal to 1.0. For naked DNA substrates, Table 2 reports slopes that range from 0.8 to 1.0. This is consistent with the independent processing assumption or a slightly higher efficiency of lesion processing in fully intact DNA.
Although hNTH1 processed both lesions in naked Tg51-5 DNA, it exhibited a ϳ3-fold preference for processing the top strand lesion in nucleosomes ( Fig. 5D and Table 1). Thus, the helical orientation of the top strand lesion is closer to optimal than that of the bottom strand lesion. If the processing of the top and bottom lesions occurred independently of each other, the yield of DSBs would be (0.85 ϫ 0.33) ϳ28%. This prediction falls within the margin of error for the observed DSB frequency of 33 Ϯ 11%, which is consistent with an independent processing model.

Aberrant Repair of Clustered, Oxidative Lesions in Chromatin
In contrast to the results obtained for the 5 S rDNA-based nucleosomes, we observed significant differences between the expected and observed rates of DSB formation in the 601-based Tg39ϩ5 nucleosomes. Specifically, both top and bottom strand lesions in Tg39ϩ5 nucleosomes were processed with similar efficiencies (Fig. 5E and Table 1), and the efficiency of DSB formation was more than double that expected for independent processing events ( Table 2). The simplest explanation for this observation is that one of the two lesions in Tg39ϩ5 is bound by hNTH1 during a transient DNA unwrapping event and that this binding stabilizes the unwrapped configuration long enough to facilitate the binding of a second DNA glycosylase to the opposing strand lesion.
The Efficiency of BER-dependent Formation of DSBs in Nucleosomes Varies among Different DNA Glycosylases-To determine if DNA glycosylases other than hNTH1 generate DSBs during the repair of clustered lesions, we conducted similar experiments with the hOGG1. As with hNTH1, hOGG1 is a bifunctional enzyme possessing both a glycosylase (which removes 8-oxoG lesions; Ref. 51) and a DNA lyase activity. As before, we designed constructs with the goal of disproving the hypothesis that nucleosomes fully suppress DSB formation. To minimize the possibility of altering nucleosome positioning determinants, we replaced two naturally occurring guanosine residues in 5 S rDNA with 8-oxoG, forming the construct 8oxoG50-5. Compared with Tg51-5, the 8oxoG50-5 lesions were positioned 1 bp closer to the dyad axis and differed in helical orientation by ϳ36°relative to the underlying histone octamer. Surprisingly, hOGG1 produced very few DSBs in naked 8oxoG50-5 DNA and virtually none in nucleosomes even when used at very high concentrations (e.g. 500 nM active enzyme, a 1000-fold excess over substrate) (Fig. 6, A and B). As with hNTH1, hAPE1 had been reported to stimulate the glycosylase activity of hOGG1 (52,53). Although hAPE1 is active only in the presence of magnesium, it will displace hOGG1 from hOGG1-generated abasic sites even in the absence of magnesium. Thus, if binding of hOGG1 to one lesion were interfering with binding to the opposing strand lesion, the addition of hAPE1 with no magnesium would have enhanced DSB formation. It did not. Instead, both magnesium and hAPE1 were required for efficient hOGG1-initiated DSB formation in naked DNA (Fig. 6B, top graph).
Although the addition of magnesium and hAPE1 substantially increased the hOGG1-initiated production of DSBs in naked DNA, it only marginally increased production of DSBs in nucleosomes (Fig. 6B, bottom graph). As a result, hOGG1-and hAPE1-dependent production of DSBs in naked DNA was nearly as high as that observed in reactions with hNTH1 and hAPE1 (Table 2), whereas hOGG1-and hAPE1-dependent production of DSBs in nucleosomes remained far lower than that produced by hNTH1 (7 Ϯ 1% versus 33 Ϯ 7% DSBs, respectively). To determine if this result could be attributed to the nucleosome substrate, we treated the 8oxoG50-5 nucleosomes with the E. coli glycosylase Fpg. Fpg is a functional homolog to hOGG1 and, importantly, did not exhibit the strand preference in naked DNA that we had observed for reactions containing hOGG1 and hAPE1 (Table 1). Fig. 6C shows that 50 nM Fpg alone generated ϳ3-fold more DSBs in naked DNA than did hOGG1 together with hAPE1. As with hOGG1 and hNTH1, Fpg exhibited a clear strand preference when processing lesions in nucleosomes (Fig. 6D); this preference could be attributed to the differences in helical orientation of the opposing strand lesions.
The reasons for the very low efficiency of hOGG1-mediated DSB formation in nucleosomes became apparent after careful quantification of single lesion processing efficiencies. In nucleosomes, hOGG1 is subject to steric constraints that are similar to those evident for Fpg. However, hOGG1 sequence preferences severely limited processing of lesions in the more accessible of the opposing strand lesions in 8oxoG50-5 nucleosomes (Table 1). Meanwhile, the preferred bottom strand lesion was sterically occluded (Fig. 6D). Thus, an unfavorable helical orientation in nucleosomes trumped whatever attributes account for preferential processing of the bottom strand lesion in naked DNA.
Fate of Nucleosomes after DSB Formation-Most DSBs in cells are believed to be quickly bound by Ku proteins and channeled into DSB repair pathways. However, the detection and processing of a DSB in nucleosomal DNA might be influenced by whether the nucleosome remains intact after DSB formation. Therefore, to assess the fate of nucleosomes after DSB formation, we treated 5 S Tg51-5 and 601 Tg39ϩ5 nucleosomes with hNTH1 and hAPE1 in separate reactions for 30 min at 37°C and then immediately loaded aliquots from the reactions onto native gels (Fig. 7). Lane 4 in Fig. 7A shows the shorter of the two DNA fragments created by DSB formation in 5 S Tg51-5 nucleosomes migrating independently of the nucleosome. Thus, one end of the hNTH1-generated DSB was released, whereas the other end remained nucleosome-associated. By contrast, little if any DNA was released after DSB for-

TABLE 1
The nucleosome suppresses single-strand break formation relative to naked DNA.
The abundance of single-strand DNA breaks generated during 30-min reactions with the indicated enzymes and templates is tabulated. Substrate concentrations in all experiments were 0.5 nM. As noted under "Experimental Procedures," reaction endpoints were either determined in triplicate or by fitting multiple time points to single exponential curves. For clarity, standard deviations are omitted. Nucleosome-dependent suppression of strand cleavages was calculated by dividing naked DNA cleavage frequency by its corresponding nucleosome strand cleavage frequency using data from reactions where single-strand cleavage extents were Ͻ90% (i.e. where cleavage frequencies were less than maximal). Y, yes; N, no.   Fig. 5A. The DNA band marked with an asterisk represents an incompletely ligated substrate; it does not contain a lesion and thus had no impact on results. C, quantification of single-and double-strand break formation in 5 S Tg51-5 naked DNA. In naked DNA, DSB formation equals the products the product of the single-strand break frequencies, which are equal for both DNA strands, whereas in 5 S Tg51-5 nucleosomes (D) the rate of DSB formation is limited by the production of SSBs in the more occluded of the two DNA strands. E, unlike Tg51-5 results, 601 Tg39ϩ5 nucleosome strand cleavage was approximately equal for each strand, suggesting that neither strand specifically limits DSB formation. Furthermore, the observed rate of DSB formation is greater than predicted based on SSB results (Table 2). This suggests that processing of one lesion may facilitate binding and processing of the second lesion. Error bars represent S.D. (n ϭ 3).

Glycosylase
mation in 601 Tg39ϩ5 nucleosomes (Fig. 7A, lane 9). In both cases, however, DSB formation altered the mobility of the residual nucleosome population. To determine if these motility changes were due to altered histone-DNA interactions, we incubated Tg51-5 nucleosomes with hNTH1 and hAPE1 as before but in the presence of a large excess of Age I. The Age I restriction site lies well within the 5 S rDNA nucleosome (Fig.  7B), and we had previously determined that this site is almost completely occluded in nucleosomes (22). As shown in Fig. 7C, hNTH1 and hAPE1 generated DSBs in ϳ30% of all nucleosomes. By contrast, Age I cleaved 9.4% of the nucleosomal sites in the absence of hNTH1 and hAPE1 and only 11.6% of the nucleosomal sites in the presence of hNTH1 and hAPE1. These results indicate that formation of DSBs at the sites tested does not invariably result in nucleosome disruption. Whether nucleosomes sequester DSB ends once they form probably depends on the distance between the cleavage site and the dyad axis.

DISCUSSION
Exposure to ionizing radiation either from x-rays, radioactive elements, or naturally occurring cosmic radiation can result in potentially lethal DSBs in DNA. DSBs can occur through direct interactions between photons and the DNA backbone or indirectly through formation of clustered, oxidative lesions followed by attempted BER. The studies in this paper indicate that the packaging of DNA into nucleosomes substantially suppresses BER-dependent formation of DSBs. The degree of suppression depends on multiple variables, including the location of opposing strand lesions relative to one another, their helical orientation relative to the underlying histone octamer, and differences in the glycosylases that initiate BER.
To What Extent Do Nucleosomes Protect Cells from BER-dependent DSB Formation?-As noted in the Introduction, lesion recognition and early steps in short patch BER do not invariably require or lead to the movement or disruption of nucleosomes. Hence, the nucleosome-dependent suppression of DSB formation documented in this study may occur in vivo as well. In addition to the many factors that influence rates of DSB formation in model nucleosomes, there are even more that probably influence DSB formation in cells. Hence, any numerical estimate of the magnitude of protection from DSB formation that nucleosomes offer must be somewhat speculative. With that caveat, enumerating nucleosome-related protective factors may be of heuristic value. First, nucleosomes offer limited protection from lesion formation. Hydroxyl radical footprinting of nucleosomal DNA typically reveals an ϳ10-bp pattern of susceptibility to hydroxyl radical-mediated damage (54). This pattern reflects both the periodic expansion and compression of major and minor grooves in DNA as it is wound about the histone octamer and discrete interactions every ϳ10 bp between DNA and the histone octamer (55). However, the difference between highly and poorly reactive residues in nucleosomes is modest, on the order of 2-3-fold. Therefore, on average, nucleosomes provide no more than a 2-fold protection from hydroxyl radical-mediated damage compared with naked DNA. The probability that two oxidative lesions will form close to one another in opposing strands in nucleosomes would thus be about 0.5 2 relative to that in naked DNA. Second, this study has reinforced earlier observations that BER-dependent formation of DSB requires lesions to be offset from one another by at least 3 bp. In naked DNA, the maximum offset will likely be governed by the size of oxidative lesion clusters, which in turn will reflect the reactive lifetime and diffusion rates of hydroxyl radicals. In nucleosomes the maximum offset is likely governed by steric factors that influence binding of DNA glycosylases. For example, the helical orientation chosen for one of the lesions in Tg51-5 was close to optimal but much less so for the opposing strand lesion, with the result that DSB formation was The predicted yield of DSBs, assuming that processing of lesions on opposing strands occurred independently of one another, was calculated by multiplying the SSB frequency of bottom and top strands using data from Table I. The predicted and observed yield of DSBs were then plotted against one another and fit to a linear regression with both x and y intercepts set to 0. Slope and R 2 values are shown. The slopes for all of the naked DNA substrates shown and many of the nucleosome substrates were approximately equal to 1.0, indicating a close correspondence between the predicted and observed DSB frequencies. A slope much less than 1.0 (i.e. observed DSB frequencies are lower than predicted) would have indicated that cleavage at one strand hinders cleavage of the opposite strand. A slope much greater than 1.0 (i.e. observed DSB frequencies are higher than predicted) would suggest that cleavage at one strand enhances cleavage of the opposite strand. This was the case for hNTH1 acting on lesions in 601 nucleosomes. Data in italics were omitted from linear regression calculations. Y, yes; N, no.  JULY 18, 2014 • VOLUME 289 • NUMBER 29 suppressed 3-4-fold compared with naked DNA. Thus, tighter constraints on lesion stagger will reduce the number of lesion configurations that could lead to DSB formation in nucleosomes as compared with naked DNA. The above-described factors combined suggest nucleosomes suppress BER-mediated DSB formation by at least 10-fold. The contributions of other factors are harder to quantify. For example, lesions at occluded sites in nucleosomes are made transiently accessible by periodic unwrapping of DNA from the edge of the histone octamer. The unwrapping-mediated exposure of clustered lesions appears to occur at a biologically meaningful rate for lesions located within 30 -40 bp of either edge of the nucleosome but not for more centrally located lesions (21). This phenomenon along with the coupled lesion processing that we observed in the Tg39ϩ5 nucleosomes is likely to reduce the overall protection that nucleosomes offer against DSB formation. In cells such unwrapping events, in some cases assisted or amplified by chromatin remodeling agents, may be confined to regions devoid of linker histones (56). A second factor that may also increase the vulnerability of DNA near the nucleosome edges to DSB formation relates to the fact that many nucleosomes occupy multiple overlapping translational positions. As noted earlier, this is the case for the 5 S rDNA nucleosomes used in the present study and also appears to be the case for nucleosomes in budding yeast (57). However, in more complex eukaryotes histone H1 may reduce positional mobility (58). In summary, chromatin associated with linker histones, heterochromatic regions associated with multiple repressive factors, and highly condensed chromatin in cells undergoing mitosis or meiosis probably limit BER-dependent DBS formation to an even greater extent than documented in this study. Indeed, chromatin condensation has been reported to protect DNA from damaging radiation, and decondensation (not necessarily transcription-related) is sufficient to sensitize DNA to radiation (59,60). A proposed mechanism for this phenomenon is that condensation of chromatin helps to exclude water, and thus fewer reactive oxygen species are generated near DNA as a result of irradiation. An alternative or additional mechanism is that BER machinery is unable to detect or act on lesions in highly condensed chromatin, thereby reducing the abundance of DSBs formed. On the other hand, chromatin decondensation and nucleosome remodeling during replication and transcription may increase not only the frequency with which clustered lesions form but also the efficiency of BER and the chance that concurrent BER reactions will generate DSBs. This prediction could explain why continuously transcribed regions of genomic DNA are unusually susceptible to damage induced by ionizing radiation (for reviews, see Ref. 61).