Nucleosome Disruption by DNA Ligase III-XRCC1 Promotes Efficient Base Excision Repair

Abstract

Each day, approximately 20,000 oxidative lesions form in the DNA of every nucleated human cell. The base excision repair (BER) enzymes that repair these lesions must function in a chromatin milieu. We have determined that the DNA glycosylase hNTH1, apurinic endonuclease (APE), and DNA polymerase β (Pol β), which catalyze the first three steps in BER, are able to process their substrates in both 601- and 5S ribosomal DNA (rDNA)-based nucleosomes. hNTH1 formed a discrete ternary complex that was displaced by the addition of APE, suggesting an orderly handoff of substrates from one enzyme to the next. In contrast, DNA ligase IIIα-XRCC1, which completes BER, was appreciably active only at concentrations that led to nucleosome disruption. Ligase IIIα-XRCC1 was also able to bind and disrupt nucleosomes containing a single base gap and, because of this property, enhanced both its own activity and that of Pol β on nucleosome substrates. Collectively, these findings provide insights into rate-limiting steps that govern BER in chromatin and reveal a unique role for ligase IIIα-XRCC1 in enhancing the efficiency of the final two steps in the BER of lesions in nucleosomes.

INTRODUCTION

Reactive oxygen species (ROS), generated as by-products of normal aerobic cellular metabolism or from exposure to exogenous agents, such as gamma irradiation, generate approximately 20,000 DNA damage events per day in each nucleated human cell. The DNA lesions produced include numerous oxidative base damages, apurinic/apyrimidinic (AP) sites, and single-strand DNA breaks (6). Base excision repair (BER) enzymes recognize and replace oxidized bases with the corresponding undamaged bases. In its simplest (“short-patch”) form, BER entails four enzymatic steps (1, 10, 21, 23, 51, 53) (Fig. 1A), beginning with the recognition and excision of a damaged base by either a mono- or bifunctional DNA glycosylase. Bifunctional glycosylases first cleave the glycosidic bond between the damaged base and the deoxyribose and then cleave the phosphodiester bond 3′ of the resulting AP site. AP endonuclease (APE) removes a residual moiety to generate a single nucleotide gap, with a 3′-OH group that can be filled by DNA polymerase β (Pol β). Finally, DNA ligase III-α (LigIIIα), in association with XRCC1, catalyzes the formation of a phosphodiester bond between the 3′-OH of the newly added nucleotide and the adjacent downstream 5′-phosphate.

Fig. 1.

Reconstitution of complete base excision repair reactions with model nucleosomes. (A) Schematic of steps in hNTH1-initiated BER. (B) Sequencing gels showing the reaction products after sequential addition of BER enzymes to Tg-out(5S) naked DNA, Tg-out(5S) nucleosomes, Tg-in(5S) nucleosomes, Tg-in (601) naked DNA, and Tg-in (601) nucleosomes. Normalized values of enzyme concentration × time for each enzyme are included below the lane numbers. Enzyme concentrations and incubation times are listed in Table SA2 in the supplemental material. The percentage of processed substrate after each enzyme addition is included as well. (C) Native gel analyses of nucleosomes after BER. Aliquots from the completed BER reactions shown in panel B were immediately loaded onto 5% native polyacrylamide gels without the addition of formamide stopping dye. Lesions containing naked DNA and nucleosomes are included for reference in lanes labeled “DNA” and “Nuc.” In lane 11, Tg-in(5S) nucleosomes were incubated sequentially with 100 nM hNTH1 for 30 min, 50 nM APE for 15 min, and 33 nM Pol β for 30 min before electrophoresis.

The nucleosomes that package most of the nuclear DNA in eukaryotes provide only minimal protection from ROS (14, 31); a small degree of protection from hydroxyl radicals is evident in DNA segments where the minor groove faces into the histone octamer (20), and histones themselves may act as a sink for ROS, thereby reducing the frequency of free-radical-inflicted DNA damage (28). Clearly, however, nucleosomal DNA is vulnerable to oxidative damage that must be made available to BER enzymes. Chromatin remodeling agents and histone chaperones facilitate most processes involving chromatin, and the other DNA repair pathways—nucleotide excision repair, mismatch repair, nonhomologous end-joining and homologous recombination-mediated repair—are all thought to require local disruption of nucleosomes (e.g., see references 18 and 38). As detailed in Discussion, we and others have reported that at least some steps in BER can occur without requiring or inducing nucleosome disruption (3, 22, 32, 35–37, 43). However, there is not yet a clear consensus on whether BER in its entirety, or a subset of specific steps in BER, requires disruption of nucleosomes; nor is it clear which if any of the known chromatin remodeling factors facilitate BER in vivo.

To better define the rate-limiting steps in BER of lesions in nucleosomes and elucidate mechanisms that govern BER in chromatin, we have conducted a series of studies using purified human BER enzymes and well-defined nucleosome substrates. We reported earlier that DNA glycosylases in both families of bifunctional DNA glycosylases can initiate BER of lesions in nucleosomes, without irreversibly disrupting or altering the translational position of the lesion containing nucleosomes (37, 43). As well, we provided evidence that spontaneous, partial unwrapping of DNA from the histone octamer enables DNA glycosylases to bind and process lesions that would ordinarily be sterically occluded (43). We have now reconstituted the entire BER reaction with nucleosomes containing discretely positioned oxidative lesions or BER intermediates. As we reported earlier for DNA glycosylases, both APE and Pol β were able to process their substrates without irreversibly altering the nucleosome or its translational position, although the efficiency of these enzymes varied depending on the helical orientation of their substrates relative to the histone octamer. In contrast, DNA ligase IIIα complexed with XRCC1 (LigIIIα-XRCC1) exhibited very little activity on intact nucleosomes at low concentrations, regardless of the helical orientation of its substrate. High concentrations of LigIIIα-XRCC1 resulted in considerable ligation activity but also led to nucleosome disruption. LigIIIα-XRCC1 proved able to bind and disrupt nucleosomes containing either a nick or a single base gap, independently of any chromatin remodeling agents. It is through this activity that LigIIIα-XRCC1 enhanced not only its own activity on nucleosomal substrates but that of Pol β as well.

MATERIALS AND METHODS

Construction of DNA containing BER lesions and intermediates.

The 147-bp core of the 601 nucleosome positioning sequence and 184-bp DNA containing the Lytechinus variegatus 5S ribosomal DNA (rDNA) (Lv5S) nucleosome positioning sequence, each containing a single, discretely positioned thymine glycol (Tg) residue, were prepared as previously described (43). Briefly, synthetic oligomers Tg-out, Tg-in, or Tg-in (601) (see Table SA1 in the supplemental material) were end labeled with [γ-³²P]ATP, annealed to their respective templates, and extended with (exo-) Klenow enzyme (New England BioLabs). The resulting double-stranded DNAs were gel purified and assembled into nucleosomes, as described below. Nucleosome length Lv5S DNA containing an AP site was prepared in the same manner, but with oligomers AP-out or AP-in, which contain tetrahydrofuran in place of Tg. To prepare gap- or nick-containing DNA fragments, the DNA oligomers Out (3′) and In (3′) were 5′-end labeled with [γ-³²P]ATP, annealed to equimolar amounts of Lv5S template, and extended with (exo-) Klenow enzyme (New England BioLabs) to create 154- and 149-nucleotide-long DNA segments. The extension reactions were stopped with 10 mM EDTA and then mixed with an equimolar amount of the appropriate ³²P 5′-end-labeled upstream oligomers (Gap-out, Gap-in, Nick-out, or Nick-in) and annealed to create a full-length DNA fragment containing a single, discretely positioned gap or nick.

DNA containing a 3′-phospho-α,β-unsaturated aldehyde (3′-PUA) group was prepared by incubating Tg-containing naked DNA with excess hNTH1 for 30 min at 37°C. hNTH1 was removed by phenol-chloroform extraction, and the DNA was ethanol precipitated and suspended in the appropriate volume for assembly into nucleosomes.

Nucleosome reconstitution.

Xenopus histones H2A, H2B, H3, and H4 were expressed in Escherichia coli and assembled into octamers, and the octamers were purified, all as described by Luger et al. and Dyer et al. (12, 30). To assemble nucleosomes, we added an ∼1.25-fold molar excess of purified octamer to a mixture of one part end-labeled substrate DNA and nine parts unlabeled lesion-free DNA. The resulting mixture was introduced into a button dialysis chamber constructed from a 0.2-ml PCR tube, as described by Thastrom et al. (50), and dialyzed slowly from 2 M NaCl in HED buffer (25 mM HEPES [pH 8.0], 1 mM EDTA, and 1 mM dithiothreitol [DTT]) to HED buffer lacking NaCl. Tg-out(5S) and Tg-in(5S) nucleosomes in the experiments shown in Fig. 3A to C were reconstituted by octamer transfer as described in reference 43. To quantify reconstitution efficiency, nucleosomes were fractionated through 5% native polyacrylamide gels in 1/4× Tris-borate-EDTA (TBE) buffer and visualized by either autoradiography or phosphorimaging.

Fig. 3.

Substrate handoff by BER enzymes on nucleosomes. (A) hNTH1-nucleosome ternary complexes were trapped by reduction of the Schiff base intermediate that forms transiently between hNTH1 and DNA. Reactions were allowed to proceed in the presence of 25 or 50 mM NaCNBH₃ for 30 min before the onset of electrophoresis through a 5% native polyacrylamide gel. (B) Absence of hNTH1-nucleosome ternary complexes in the presence of enzymatically active, high-turnover hNTH1 mutants. Tg-out(5S) nucleosomes were incubated with 0, 20, 60, 100, and 400 nM total concentrations of hNTH1 for 2.5 min at 22°C before the onset of electrophoresis. In lanes 1 to 5, nucleosomes were incubated with full-length hNTH1, whereas lanes 6 to 10 and 11 to 15 were incubated with Δ55 and Δ72 N-terminal truncations of hNTH1. (C) Displacement of hNTH1 from nucleosomes by APE. Nucleosomes were incubated with 35 nM APE alone (lane 2), 14 nM hNTH1 alone (lane 3), or 14 nM hNTH1 together with 14 or 35 nM APE (lanes 4 and 5) for 15 min at 22°C prior to the onset of electrophoresis.

Expression and purification of BER enzymes.

hNTH1, Pol β wild type (wt), T304I, and E309K were expressed and purified as previously described (34, 37). A manuscript describing the coexpression and purification of ligase IIIα with XRCC1 is in preparation. The expression and purification of APE are described in the supplemental material.

Enzyme assays.

All enzyme concentrations reported in the text and figures refer to total protein concentration, except in the case of hNTH1, where the active fraction was determined as described by Blaisdell and Wallace (5). Enzymes were freshly diluted into ice-cold BER reaction buffer (25 mM NaHEPES NaOH [pH 8.0], 100 mM NaCl, 5 mM MgCl₂, 0.2 mM EDTA, 1 mM DTT, and 0.1 mg/ml bovine serum albumin [BSA]) containing 20 μM dTTP, unless otherwise indicated in the figure legend, and added to substrates in reaction buffer, as indicated in the text and figure legends. The dilution and reaction buffers used with LigIIIα-XRCC1 included 1 mM ATP. Final substrate concentrations in all reactions were 4 nM, with 36 nM unlabeled lesion-free nucleosomes or naked DNA, and all reactions were conducted at 37°C, unless otherwise indicated. To monitor enzyme activity, aliquots from BER reactions were stopped by the addition of 4 volumes 0.1 N NaOH, 90% formamide, and 0.1% bromophenol blue and 0.1% xylene cyanol or, in the case of hNTH1 reactions, the same buffer minus the 0.1 N NaOH. Reaction products were resolved on 12% or 15% sequencing gels. To assess the fate and integrity of lesion-containing nucleosomes, aliquots from BER reactions were loaded immediately onto 5% native polyacrylamide gels and separated by electrophoresis in 1/2× TBE buffer; at no point were these samples exposed to formamide or other denaturants.

XRCC1 pulldown and Western blot assays.

To assess interactions between XRCC1 and either Pol β or selected Pol β mutants, we added 33 nM purified Pol β in nickel buffer (50 mM Tris [pH 7.6], 75 mM KCl, 0.1% IGEPAL CA-630 (Sigma-Aldrich), 1 mM DTT, 10 mM imidazole) to 100 μg of XRCC1-containing whole-cell extract, prepared from the 88Tag Pol β^−/− mouse embryonic fibroblast cell line (13, 48), as previously described (25). After 30 min on ice, the XRCC1-Pol β mixtures were mixed with 15 μl of Ni-nitrilotriacetic acid (NTA) agarose (Qiagen). The resin was collected by centrifugation and washed, and the resin-associated protein was recovered by suspension in SDS-PAGE loading buffer. Proteins were then fractionated through a 10% SDS-PAGE gel, transferred to nitrocellulose, and incubated with an anti-XRCC1 antibody (AbCam number ab9147). The blots were incubated with a horseradish peroxidase-conjugated secondary antibody (1:5,000 dilution), and XRCC1-antibody conjugates were visualized using a Bio-Rad Molecular Imager ChemiDoc XRS+.

RESULTS

Assembly of model nucleosome substrates.

To examine the impact of nucleosomes on each step in BER, we assembled lesion-containing nucleosomes with either of two DNA fragments that each form well-positioned nucleosomes (11, 15, 41, 43, 45, 46, 49). The first of these contained a single thymine glycol (Tg) embedded in either of two sites within the L. variegatus 5S rDNA gene (Lv5S gene). Tg-out(5S) nucleosomes contained a Tg whose minor groove faces outward from the histone octamer, while Tg-in(5S) nucleosomes contained a Tg whose minor groove faces toward the histone octamer. We showed previously that neither lesion alters the preferred helical positioning of DNA in the nucleosome (43). We showed as well that, in both the major and minor translational variants that form with 5S DNA, these lesions reside within the nucleosome (37, 43). We also assembled nucleosomes using the synthetic 601 DNA segment that Widom and colleagues selected for its capacity to form an even more stable, positioned nucleosome (45, 49). The nucleosome Tg-in (601) contained an inward-facing Tg residue 47 nucleotides from the dyad axis, a site close to that of the Tg residue in Tg-in(5S). To examine individual steps in BER, we also assembled nucleosomes with substrates for enzymes that act subsequently to DNA glycosylases (an unsaturated aldehyde for APE, gapped DNA for Pol β, and nicked DNA for LigIIIα-XRCC1).

Reconstitution of complete base excision repair reactions with nucleosomes.

Figure 1B shows the DNA products formed at each step in reactions where we added, sequentially, hNTH1, APE, Pol β, and LigIIIα-XRCC1 to Tg-out(5S) and Tg-in (601) naked DNAs and to Tg-out(5S), Tg-in(5S), and Tg-in (601) nucleosomes. As illustrated in Fig. 1A, hNTH1 first cleaves the N-glycosylic bond between the damaged base (in this case, thymidine glycol) and its associated sugar residue and then cleaves the phosphodiester bond 3′ to the AP site, generating the faster-migrating product evident in lanes 2 of Fig. 1B. Next, APE removes the 3′-unsaturated aldehyde, generating a faster-migrating primer with a 3′-OH evident in lanes 3 of Fig. 1B. As shown in lanes 4 of Fig. 1B, Pol β was able to extend this primer by 1 or more nucleotides. Finally, under the conditions used, LigIIIα-XRCC1 was able to seal most of the single nucleotide extension products in naked DNA and about half of those in the 5S nucleosomes to generate the full-length repaired DNA (lanes 5 of Fig. 1B). However, LigIIIα-XRCC1 largely failed to act on nicks in the 601-based nucleosomes (lane 5 in the right-most panel of Fig. 1B). We address the reasons for the differing ligase results in a later section.

To determine the impact of BER on nucleosome integrity, we immediately electrophoresed aliquots from the completed BER reactions on 5% native polyacrylamide gels (without the addition of formamide stopping dye). Figure 1C shows naked DNA and nucleosome templates before and immediately after the BER reactions (lanes 1, 3, 4, 6, 7, 9, 10, 12, and 13 versus lanes 2, 5, 8, 11, and 14). The presence of a major band with the mobility of an intact nucleosome in lane 5, together with the absence of a band with the mobility of naked DNA and no evidence of a high-molecular-weight aggregate, suggested that lesions in most of the Tg-out(5S) nucleosomes had been fully repaired without irreversible disruption of the nucleosome. Also evident in lane 5 of Fig. 1C is a supershifted band that likely reflects a residual ternary complex that formed between nucleosomes and Pol β. In contrast, the virtual absence of a nucleosome band together with the prominent high-molecular-weight aggregate in lane 8 of Fig. 1C suggested that Tg-in(5S) nucleosomes had been disrupted at some point during BER. Further investigation revealed that nucleosomes remain largely intact during the hNTH1, APE, and Pol β reactions (lane 11), indicating that LigIIIα-XRCC1 was responsible for the nucleosome disruption observed in lane 8. Additionally, BER reactions with Tg-in (601) nucleosomes demonstrated little disruption (lane 14), which correlates with the small proportion (∼17%) of 601 molecules that were ligated in these reactions.

Lesion orientation plays a major role in the efficiency of the first three steps of hNTH1-initiated BER but not the final step.

In the experiments shown in Fig. 1, we deliberately adjusted enzyme concentrations and reaction times to achieve roughly equivalent levels of repair for all substrates. The relative “enzyme-times-time” (E × t) values given in Table SA2 in the supplemental material and at the bottom of each lane in Fig. 1B show that it was necessary to use increasing enzyme amounts and longer reaction times as we progressed from naked 5S DNA to naked 601 DNA to Tg-out nucleosomes to Tg-in nucleosomes. As these values indicate, hNTH1 was less able to excise Tg from Tg-in (601) DNA than from 5S DNA-based DNA templates, possibly because we had positioned the Tg residue such that its minor groove resided on the inner side of the intrinsically curved 601 DNA (29). E × t amounts also had to be increased 2.5- to 5-fold relative to the corresponding naked DNAs to repair Tg-out(5S) nucleosomes. Thus, nucleosomes moderately inhibit BER even for optimally oriented lesions. Conditions yielding far higher E × t values were needed for repair of Tg-in(5S) nucleosomes, indicating that helical orientation of substrates is a critical determinant of repair efficiency. It should be noted, however, that enzyme concentrations used to initiate BER of lesion-containing nucleosomes in vitro can substantially affect the extent of repair observed for lesions that, nominally, are sterically occluded. As we reported previously, increasing the concentration of hNTH1 to its estimated concentration in vivo resulted in relatively efficient excision of occluded lesions (37, 43).

As indicated in Fig. 1, the E × t values for each of the subsequent steps in BER increased in the same fashion as that described above for hNTH1, with the progression from naked DNA to nucleosomal substrates. To more precisely relate the efficiency of each enzyme on nucleosome substrates to that on naked DNA, we estimated initial reaction velocities from reactions where enzyme concentrations were adjusted to maintain single turnover conditions during the first 30 s of the reaction. These values, listed in Table 1, correlate closely to E × t values shown in Fig. 1B.

Table 1.

BER enzyme velocities with DNA and nucleosome substrates

Enzyme	Concn (nM)	Substrate	Velocity ± SD (fmoles/s)
hNTH1	5	Tg-out (5S) DNA	0.0756 ± 0.0154
hNTH1	20	Tg-out(5S) Nuc	0.0707 ± 0.0048
hNTH1	20	Tg-in(5S) Nuc	0.0329 ± 0.0034
APE	1	3′-PUA DNA	0.0751 ± 0.0034
APE	5	3′-PUA-out(5S) Nuc	0.0640 ± 0.0081
APE	5	3′-PUA-in (5S) Nuc	0.0400 ± 0.0035
Pol β	1	Gap-out(5S) DNA	0.0227 ± 0.0053
Pol β	3	Gap-out(5S) Nuc	0.0298 ± 0.0056
Pol β	3	Gap-in(5S) Nuc	0.0204 ± 0.0028
LigIII-XRCC1	1	Nick-in(5S) DNA	0.0111 ± 0.0060
LigIII-XRCC1	5	Nick-out(5S) Nuc	0.0356 ± 0.0008
LigIII-XRCC1	5	Nick-in(5S) Nuc	0.0298 ± 0.0043

To examine the effect of substrate orientation on each enzyme in more detail, we conducted additional kinetic studies, this time choosing enzyme concentrations that would facilitate comparisons between different nucleosomal substrates (Fig. 2). We prepared DNA and nucleosomes containing each of the intermediates of hNTH1-initiated BER: Tg, 3′-PUA, single nucleotide gap, and nicked DNA (see Materials and Methods). We placed each of these substrates at the positions corresponding to those in the Tg-out(5S) and Tg-in(5S) nucleosomes. As expected, hNTH1 excised lesions from Tg-out(5S) nucleosomes more efficiently than from Tg-in(5S) nucleosomes; as well, the lyase activity of hNTH1 was slower than its glycosylase activity regardless of lesion orientation (filled versus open triangles in Fig. 2A). These results are consistent with the largely one-sided binding of DNA glycosylase to DNA (Fig. 2A). Figure 2B shows that, like hNTH1, APE interacts largely with one side of the DNA helix. This led us to predict that it too would be able to bind directly to optimally oriented substrates; this proved correct, as shown in the graph in Fig. 2B. The binding of Pol β to gapped DNA induces DNA to bend away from the prospective site of nucleotide insertion, as shown in Fig. 2C. An outward-facing gap in a nucleosome might to some extent be appropriately “prebent” and thereby facilitate the initial binding of Pol β. As with hNTH1 and APE, Pol β was not as active in nucleosomes as it was on naked DNA but nonetheless was significantly more active on outward- than on inward-facing gaps (Fig. 2C).

Fig. 2.

Influence of lesion orientation on the rate of each step of BER. (A) Rates of glycosylase (Glyc) and lyase activity in reactions with 20 nM hNTH1 and either Tg-out(5S) nucleosomes or Tg-in(5S) nucleosomes. (B) Rates of phosphodiesterase activity in reactions with 5 nM APE and either 3′-PUA-out(5S) nucleosomes, 3′-PUA-in(5S) nucleosomes, or 3′-PUA-in(5S) nucleosomes generated by incubating Tg-in(5S) nucleosomes for 30 min with 100 nM hNTH1 immediately prior to the addition of APE (post-hNTH1). (C) Rates of nucleotide extension in reactions with 3 nM Pol β and Gap-out(5S) nucleosomes and Gap-in(5S) nucleosomes. (D) Rates of ligation in reaction mixtures containing 5 nM LigIIIα-XRCC1 and Nick-out(5S) nucleosomes and Nick-in(5S) nucleosomes. Crystal structures of the enzymes that catalyze each step in hNTH1-initiated short-patch BER are shown to the right of their respective graphs (4, 8, 16, 33). The data points and error bars on the graphs represent the means and standard deviations from results of at least 3 independent experiments. In cases where error bars are not evident, standard deviations were similar in magnitude to the size of the symbols on the graphs.

Unlike the first three steps in BER, the efficiency with which LigIIIα (in complex with XRCC1) ligated nicks in nucleosomes was unaffected by the helical orientation of the nick (Fig. 2D). Increasing the concentration of LigIIIα-XRCC1 by 10-fold increased the rate of ligation for both inward- and outward-facing nicks (see Fig. SA1 in the supplemental material). This degree of enhancement was reminiscent of the enhanced processing of inward-facing lesions that are accessible only when transiently exposed due to partial unwrapping of DNA from the histone octamer. Structural studies, which show that LigIIIα fully encircles its substrate, as depicted in Fig. 2D, suggest that LigIIIα can bind substrates in nucleosomes only during an episode of partial DNA unwrapping. Unlike hNTH1, APE, and Pol β, however, high concentrations of LigIIIα-XRCC1 appear to drive nucleosome disruption (see “Substrate handoff to DNA ligase IIIα is accompanied by progressive nucleosome disruption” below).

Substrate handoff from hNTH1 to APE.

Several workers have proposed that enzyme substrates produced during BER are handed off from one enzyme to the next, such that a DNA glycosylase, for example, would remain bound to its product until it was displaced by APE; in this fashion, processing intermediates would remain sequestered during repair (39, 52, 54). We used two different assays to determine if enzyme handoff occurs during the processing of lesions in nucleosomes. The underlying premise behind the first kinetic assay was that binding of hNTH1 to its inward-facing product in nucleosomes would prevent nearby DNA from rewrapping around the histone octamer and thereby facilitate the subsequent binding of APE. If this occurs, however, the kinetic assay we used lacked the necessary sensitivity, as the rate of APE-mediated cleavage of an inward-facing 3′-PUA formed in situ by treating Tg-containing nucleosomes with hNTH1 was approximately the same as that seen when APE was added to nucleosomes that had been assembled with preformed 3′-PUA (Fig. 2B). An alternative possibility is that handoff does not have a kinetic signature but may still occur as a means of protecting substrates during repair. Accordingly, we decided to examine the dynamics of the ternary complexes that form with lesion-containing nucleosomes during BER.

We had previously reported that hNTH1 lesion processing entails the formation of a nucleosome-hNTH1 ternary complex and demonstrated that this complex contains DNA-processing intermediates (43). Before examining the fate of the hNTH1-nucleosome complex in a handoff assay, we tested two additional predictions that should hold if this ternary complex is a bona fide BER intermediate. The first prediction was that it should be possible to capture an hNTH1-DNA covalent intermediate. This proved to be the case, as shown in Fig. 3A, where we trapped the hNTH1-DNA Schiff base intermediate by reduction with NaCNBH₃. We had previously suggested that the ternary complex could be detected in a gel mobility shift assay only because wild-type hNTH1 has a very slow turnover rate. Hence, the second prediction was that hNTH1 mutants that exhibit elevated turnover rates would not form detectable ternary complexes. To test this, we incubated Tg-out(5S) nucleosomes with hNTH1 mutants lacking their N-terminal 55 or 72 amino acids. Removal of these tails has been shown to increase hNTH1 turnover without disrupting enzymatic activity (27). As can be seen in lane 10 in Fig. 3B, the mobility of Tg-containing nucleosomes was slightly retarded by high concentrations of the Δ55 truncation mutant, possibly due to nonspecific interactions between nucleosomes and the enzyme. However, at lower enzyme concentrations, only full-length hNTH1 formed a discrete ternary complex with Tg-containing nucleosomes.

Figure 3C shows the impact of adding APE to hNTH1-nucleosome ternary complexes. At APE concentrations in excess of hNTH1, hNTH1 was displaced from the ternary complex; this did not depend on the presence of Mg²⁺, which renders APE catalytically active. This result supports a model in which hNTH1 remains associated with its product until it is displaced by APE.

In the experiment shown in Fig. 3C, the APE-mediated displacement of hNTH1 from nucleosomes was not followed by formation of a detectable APE-containing ternary complex. It was possible that an APE-nucleosome complex formed but either could not be resolved or was not stable enough to be detected in a gel assay. As shown in Fig. SA2 in the supplemental material, changing assay conditions allowed us to detect a probable APE-containing ternary complex. However, its low stability and abundance made it impossible to conduct definitive studies of handoff from APE to Pol β.

Substrate handoff to DNA ligase IIIα is accompanied by progressive nucleosome disruption.

As shown in Fig. 1, the addition of small amounts of LigIIIα-XRCC1 to nucleosomes resulted in moderate levels of ligation of outward-facing DNA nicks; the addition of larger amounts of LigIIIα-XRCC1 to nucleosomes containing inward-facing DNA nicks resulted in low to moderate ligation but was accompanied by loss of the canonical nucleosome band in the mobility shift assay. To examine the impact of LigIIIα-XRCC1 binding on nucleosome integrity in the absence of enzymatic activity, we took advantage of an N-terminal poly(ADp-ribose) polymerase (PARP)-like zinc finger domain (ZnF) in LigIIIα, which enables it to bind gapped as well as nicked DNA (8). We thus examined the binding of LigIIIα-XRCC1 to gap-containing nucleosomes in the presence and absence of Pol β, omitting deoxynucleoside triphosphate (dNTPs) from the BER reaction buffer to prevent extension by Pol β and its subsequent displacement by LigIIIα-XRCC1. Lane 3 in Fig. 4A shows a supershifted, ternary complex produced by the binding of the ∼39-kDa Pol β to Gap-in(5S) nucleosomes. The addition of the ∼170-kDa LigIIIα-XRCC1 heterodimer to gap-containing nucleosomes generated a dramatically slower migrating complex (LX in Fig. 4A, B, and C), and the addition of LigIIIα-XRCC1 together with Pol β generated a complex with a mobility very similar to that formed by LigIIIα-XRCC1 alone (βLX in lane 5 of Fig. 4A). The nearly identical mobilities of the LX and βLX complexes made it impossible to determine if LigIIIα-XRCC1 had displaced Pol β from the nucleosome.

Fig. 4.

Substrate handoff from Pol β to LigIIIα-XRCC1 leads to nucleosome disruption by LigIIIα-XRCC1. (A) Gap-in(5S) nucleosomes were incubated in BER reaction buffer lacking any dNTPs alone (lane 2), with 15 nM Pol β (lane 3), with 15 nM LigIIIα-XRCC1 (lane 4), or with 15 nM Pol β premixed with 15 nM LigIIIα-XRCC1 (lane 5) for 5 min prior to the onset of electrophoresis through a 5% native polyacrylamide gel. Lane 1 shows the mobility of Gap-in(5S) DNA. (B) Gap-out(5S) and Nick-out(5S) nucleosomes were incubated with 5 nM LigIIIα-XRCC1 for 5 min prior to the onset of electrophoresis. (C) Gap-in(5S) DNA and nucleosomes were incubated with 10 nM LigIIIα-XRCC1 for 2 min prior to the onset of electrophoresis. (D) Lesion-free 601 nucleosomes were incubated with 20 nM hNTH1, 5 nM APE, 15 nM Pol β, 10 nM LigIIIα-XRCC1, or 50 nM LigIIIα-XRCC1 for 5 min before electrophoresis.

Because LigIIIα-XRCC1 was active on nicks even in the absence of exogenously added ATP, it was not possible to investigate the formation of a preligation complex in nick-containing nucleosomes. Incubating nick-containing nucleosomes with LigIIIα-XRCC1 for just 2 min before resolution by electrophoresis led to the formation of a discrete supershifted particle with the same migration rate as that formed with gap-containing nucleosomes (compare lanes 2 and 3 in Fig. 4B). Once again, appreciable amounts of a slow mobility aggregate formed as well. To test whether LigIIIα-XRCC1 might displace the entire histone octamer from gap- and nick-containing nucleosomes, we incubated LigIIIα-XRCC1 with naked, Gap-in 5S DNA. This led to the formation of a complex with a mobility that was slightly faster than that associated with binding of LigIIIα-XRCC1 to nucleosomes (compare lanes 2 and 4 in Fig. 4C). This result is consistent with the residual association of at least some histones with the LX complex.

The above observations, collectively, indicate that by itself or with Pol β, LigIIIα-XRCC1 binds but then disrupts gap- or nick-containing nucleosomes. However, it is noteworthy that the sequential addition of all of the BER enzymes to nucleosomes containing outward-facing lesions resulted in moderate levels of repair with little or no evidence of nucleosome disruption (Fig. 1B and C). In this context, it is possible that the presence of the earlier enzymes ameliorates nucleosome disruption by LigIIIα-XRCC1.

Although LigIIIα-XRCC1 is not active on gap-containing nucleosomes, the addition of LigIIIα-XRCC1 to gap-containing nucleosomes produced the same low-mobility aggregate and LX complex as formed with nick-containing nucleosomes (Fig. 4B). Hence, the destabilizing effect of LigIIIα-XRCC1 on nucleosomes is independent of any ligation activity. To determine if LigIIIα-XRCC1 also destabilizes nucleosomes that do not contain gaps or nicks, we incubated lesion-free 601 DNA containing nucleosomes with each BER enzyme individually. The addition of hNTH1, APE, or Pol β to nucleosomes had no impact on their mobilities (lanes 1 to 4 of Fig. 4D), indicating that the ternary complexes shown in Fig. 1, 3, and 4 are substrate specific. Likewise, the addition of 10 nM LigIIIα-XRCC1 to nucleosomes, an amount similar to that used in earlier experiments, produced neither a nucleosome supershift nor an aggregate (lane 5 of Fig. 4D). Thus, the above-described effects of LigIIIα-XRCC1 on nucleosomes are also substrate specific. Only upon addition of a 5-fold-higher concentration of LigIIIα-XRCC1 did we observe a low mobility smear along with an aggregate that failed to migrate into the gel (lane 6 of Fig. 4D). Of note, the high LigIIIα-XRCC1 concentration used here represents a 1.25-fold molar excess of enzyme over total nucleosomes in the reaction (50 nM and 40 nM, respectively). Therefore, at high concentrations, LigIIIα-XRCC1 can bind and disrupt nucleosomes indiscriminately; however, purified XRCC1 alone was not sufficient to cause nucleosome disruption, as shown in Fig. SA3 in the supplemental material.

Ligase IIIα-XRCC1 enhances the activity of Pol β on nucleosome substrates.

Given that LigIIIα-XRCC1 can bind and disrupt nucleosomes containing a single nucleotide gap, we reasoned that LigIIIα-XRCC1 might enhance the rate of nucleotide extension by Pol β during BER in chromatin, particularly during the repair of inward-facing gaps (see Fig. 2C). To test this possibility, we measured the rate of nucleotide extension by Pol β in the presence and absence of LigIIIα-XRCC1. Lanes 2 to 5 of the left panel of Fig. 5A show extension of Gap-in(5S) nucleosomes by Pol β alone, and lanes 6 to 9 show both nucleotide extension and ligation by Pol β and LigIIIα-XRCC1. The predicted enhancement of Pol β extension rates by LigIIIα-XRCC1 can be seen most readily by comparing the 30-s time points in lanes 2 and 6 of Fig. 5A. Since incubating LigIIIα-XRCC1 alone with Gap-in(5S) nucleosomes did not produce any extension or ligation products (lanes 11 to 13 in the right-hand panel of Fig. 5A), the ligation products evident in lanes 6 to 9 were dependent on extension of the gap by Pol β. This made it possible to quantify Pol β extension rates, as shown in Fig. 5B. Comparison of the initial slopes indicates that LigIIIα-XRCC1 enhances the gap-filling activity of Pol β severalfold.

Fig. 5.

Nucleosome disruption by LigIIIα-XRCC1 enhances the rate of nucleotide extension by Pol β. (A) Extension and ligation products of Pol β and LigIIIα-XRCC1 with nucleosomes containing an inward-facing gap. Gap-in(5S) nucleosomes were incubated for 0, 0.5, 1.5, 5, and 20 min with 33 nM Pol β alone or 33 nM Pol β premixed with 50 nM LigIIIα-XRCC1. Lane 1 shows the migration of the Gap-in(5S) DNA alone; lanes 2 to 5 show nucleotide extension by Pol β; lanes 6 to 9 show extension by Pol β and ligation by LigIIIα, with the full-length product denoted “FL”; and lanes 10 to 13 demonstrate the absence of product formation by LigIIIα-XRCC1 alone. The DNA fragment marked with an asterisk lies immediately 3′ of the gap that was used to monitor Pol β activity, visualized by lightly labeling with ³²P. (B) Quantification of Pol β extension rates in the presence and absence of LigIIIα-XRCC1. ○, Pol β preincubated with LigIIIα-XRCC1. Data points represent the means and standard deviations of results from 5 independent experiments. (C) Western blotting against XRCC1 associated with 33 nM wt Pol β (lane 3) or 33 nM T304I and E309K Pol β mutants (lanes 5 and 7). (D) Quantification of single nucleotide extension of Gap-in(5S) nucleosomes by 33 nM Pol β T304I and E309K incubated with and without 50 nM LigIIIα-XRCC1.

Although the enhanced efficiency of Pol β in the presence of LigIIIα-XRCC1 could be rationalized by the effect of LigIIIα-XRCC1 on gap-containing nucleosomes, the fact that Pol β and XRCC1 interact with one another (9, 19) led us to ask if this enhancement depends on physical interactions between Pol β and XRCC1. To test this possibility, we expressed two epitope-tagged Pol β variants that contain mutations in loop III, which forms the interface between Pol β and XRCC1 (19). The extension efficiency of each Pol β mutant on naked DNA was similar to that of the wild-type enzyme. We next examined the effect of these mutations on the capacity of Pol β to interact with XRCC1 at the 30 nM concentration used in the Pol β enhancement assay. Figure 5C shows that both mutations reduce the ability of Pol β to bind XRCC1 present in extracts from Pol β^−/− cells. Despite this weakened interaction, LigIIIα-XRCC1 enhanced the activity of each of the Pol β mutants toward nucleosomes containing gapped DNA (Fig. 5D). Thus, the enhancement of Pol β by LigIIIα-XRCC1 is mediated primarily by the capacity of LigIIIα-XRCC1 to disrupt nucleosomes.

DISCUSSION

Complete reconstitution of BER with lesion-containing nucleosomes.

We have investigated the influence of nucleosomes on each of the four steps in short-patch BER. DNA glycosylases and APE, which catalyze the first and second steps in BER, possess similar DNA binding motifs and substrate recognition mechanisms; both enzymes bind to one face of the DNA helix and flip the substrate out of the DNA duplex and into their active sites (16, 33, 47). These shared structural features may explain why APE and representative enzymes from the two major families of bifunctional DNA glycosylases are capable of directly binding to optimally oriented substrates in nucleosomes (as inferred from the efficient processing of outward-facing substrates). Pol β, which carries out the third step in BER, also was more active on outward- than inward-facing substrates. This result suggests that Pol β also can bind directly to optimally oriented determinants in nucleosomal DNA even though, in its enzymatically active configuration, Pol β bends DNA opposite the gap by 90°. The severity of this bend suggests that DNA near the site of the gap must lift away from the histone octamer as Pol β takes on an active configuration. These same geometric considerations may account for why FEN1, which bends the DNA opposite a DNA flap by 90 to 100°, is more active on nucleosomal substrates than on naked DNA (24).

To explain how DNA glycosylases, APE, and Pol β are all able to process nominally occluded substrates in nucleosomes (albeit at enzyme concentration × time values 6- to 66-fold higher than were needed to process sterically accessible substrates), we have systematically tested (in this study and in references 37 and 43) known mechanisms that could make substrates in nucleosomes accessible to BER enzymes, using gel mobility shift assays to rule out enzyme-induced disruption of nucleosomes, restriction endonuclease accessibility assays to rule out shifts in nucleosome position, and high-resolution DNase I footprinting to rule out lesion- or enzyme-induced changes in helical positioning. Because nucleosomes assembled with 5S DNA exhibit limited but potentially significant levels of positional heterogeneity that might have escaped detection in our positioning assays, we also conducted parallel studies of BER of lesions in more robustly positioned 601 nucleosomes. Results in all cases were in good agreement with one another, which strengthens the conclusion that processing of occluded substrates is not due to positional heterogeneity.

The single remaining known mechanism for exposure of occluded lesions in nucleosomes, which our results favor, is the binding of BER enzymes to their substrates when they are transiently exposed through spontaneous partial unwrapping of DNA from the histone octamer. The low concentration and short (10- to 50-ms) duration of the unwrapped state (26) can account for why elevated enzyme concentrations are needed to process occluded lesions. It is important here to note that we observed appreciable processing of occluded lesions by hNTH1 at concentrations no higher than its in vivo concentration (37). Inconsistent reporting of reaction conditions make it difficult to determine if different active enzyme concentrations contributed to the wide variation in results among certain earlier studies of BER on chromatin substrates. A second variable that may account for much of the variation reported in earlier studies of BER in model nucleosomes is the distance between the lesion and the dyad axis of the nucleosome. In general, DNA glycosylases (and in some instances downstream enzymes as well) were substantially more active on substrates ∼40 nt or more from the dyad axis than on substrates located at or near the dyad axis (compare, e.g., this study and references 36, 37, and 43 to references 3, 32, and 35). For inward-facing lesions, this result conforms to a prediction of the DNA unwrapping hypothesis, in that the probability of lesion exposure due to DNA unwrapping diminishes with increasing proximity of the lesion to the center of the nucleosome (2, 22, 43). Differences in the probability of substrate exposure may not fully account for why Pol β exhibited good activity in our study but virtually none at all in an earlier study (3); it may also be harder for Pol β to bend DNA substrates near the center of the nucleosome into an active conformation.

Substrate handoff during BER.

It has been hypothesized that the individual steps in BER involve recognition of a product-enzyme complex by the next enzyme in the pathway in order to avoid the cytotoxic or mutagenic effects of unrepaired BER intermediates (44, 54). Although we did not detect the kinetic signature that we hypothesized might be associated with substrate handoff, we did observe a succession of discrete ternary complexes that form between enzymes and nucleosomes during BER. The hNTH1 and Pol β complexes were easier to detect than the APE-containing complex, suggesting that the complexes differ in stability or half-life. Nevertheless, the fact that we were able to detect even a trace of these complexes in a gel mobility shift assay, without use of prior cross-linking, suggests that the lifetimes of these complexes exceed the time it takes to process oxidative lesions in vivo. Thus, our results support the idea that BER enzymes displace one another in an orderly fashion, thereby ensuring that substrate intermediates remain sequestered until repair is complete.

LigIIIα-XRCC1-driven disruption of nucleosomes during BER.

DNA ligase III was recently reported to be essential for repair of DNA in mitochondria but not in nuclei (17). This probably reflects the participation of DNA ligase I in long-patch BER, which can take the place of short-patch BER in the event of defects in LigIII (42). Structural studies, showing that both LigI and LigIII fully encircle their DNA substrates (8, 40), led us to predict that LigIIIα-XRCC1 would be unable to directly bind nucleosomal DNA. Consistent with this prediction, the low to moderate levels of activity that LigIIIα-XRCC1 exhibited toward nicks in nucleosomes was independent of the helical orientation of the nick. Thus, it is likely that partial DNA unwrapping is an obligatory first step in the binding of LigIIIα-XRCC1 to nucleosomal substrates. This probably also accounts for the capacity of LigI to act on nucleosomal substrates, albeit with an ∼10-fold-lower efficiency than that seen on naked DNA (7). Increasing concentrations of LigIIIα-XRCC1 resulted in nucleosome disruption. LigIIIα-XRCC1 appeared to more readily disrupt 5S DNA-containing nucleosomes than 601 nucleosomes, which probably accounts for the more efficient ligation of nicked DNA substrates in the 5S nucleosomes. LigIIIα-XRCC1 also significantly enhanced extension of occluded gaps by Pol β, most likely via LigIIIα-XRCC1-driven disruption of gap-containing nucleosomes. Interestingly, LigI is unable to bind DNA gaps because it lacks the ZnF of LigIII and thus may not enhance the activity of Pol β in the same manner.

The severely restricted activity of Pol β on gaps near the dyad axis of nucleosomes suggests that nucleosome movement or disruption prior to the gap-filling step is generally required to complete BER in vivo. It is possible as well that the fate of nucleosomes during earlier steps in BER differs with lesion position: nucleosomes in this study remained intact during the first three steps in BER, but rare DNA unwrapping events that expose and allow binding of DNA glycosylases or APE to substrates near the dyad axis may be sufficiently destabilizing as to result in nucleosome disruption. In any event, the poor activity of Pol β on gaps at or near the dyad axis of the nucleosomes underscores the importance of our finding that LigIIIα-XRCC1 can disrupt gap-containing nucleosomes. LigIIIα-XRCC1 by itself may prove sufficient to ensure completion of BER in vivo. Whether known histone chaperones or chromatin remodeling agents participate as well remains to be determined.