G. L. Dianov, Radiation and Genome Stability Unit, Medical Research Council, Harwell, Oxfordshire OX11 0RD, UK Fax: +44 1235 841200 Tel.: +44 1235 841134 E-mail: firstname.lastname@example.org
Ionizing radiation, oxidative stress and endogenous DNA-damage processing can result in a variety of single-strand breaks with modified 5′ and/or 3′ ends. These are thought to be one of the most persistent forms of DNA damage and may threaten cell survival. This study addresses the mechanism involved in recognition and processing of DNA strand breaks containing modified 3′ ends. Using a DNA–protein cross-linking assay, we followed the proteins involved in the repair of oligonucleotide duplexes containing strand breaks with a phosphate or phosphoglycolate group at the 3′ end. We found that, in human whole cell extracts, end-damage-specific proteins (apurinic/apyrimidinic endonuclease 1 and polynucleotide kinase in the case of 3′ ends containing phosphoglycolate and phosphate, respectively) which recognize and process 3′-end-modified DNA strand breaks are required for efficient recruitment of X-ray cross-complementing protein 1–DNA ligase IIIα heterodimer to the sites of DNA repair.
DNA single-strand breaks (SSBs) arising as a result of the disruption of phosphodiester bond linkage of nucleotides in a polymer are dangerous, because, if left unrepaired, they may block vital processes such as DNA transcription and DNA replication. SSBs arise by several distinct mechanisms including direct energy deposition by ionizing radiation, attack by reactive oxygen species, during enzymatic processing of endogenous DNA lesions, and as a result of aberrant DNA topoisomerase I activity (reviewed in [1,2]). Many of the SSBs arise as a consequence of loss of the DNA base and subsequent sugar fragmentation, which should be considered as a single nucleotide gap with modified 5′ and/or 3′ ends. For example, SSBs produced by ionizing radiation or by attack from reactive oxygen species often contain 3′-phosphoglycolate or 3′-phosphate groups . Similarly, some radiation-induced SSBs, as well as those arising during base excision repair (BER), contain a 5′-sugar phosphate residue [4–6]. Thus, SSBs produced by several genotoxic agents include a variety of termini that have to be converted into conventional 3′-OH/5′-phosphate nicks before the gap can be filled by a DNA polymerase and the DNA ends resealed by a DNA ligase.
Several BER enzymes have been shown to possess ‘end cleaning’ activities. The major mammalian endonuclease that processes abasic sites [apurinic/apyrimidinic endonuclease 1 (APE1)] was also shown to be the major 3′-phosphoglycolate activity in human cell extracts [8–11]. Human polynucleotide kinase (PNK) is the major 3′-phosphatase [12,13], and DNA polymerase β (Pol β) is the major activity in human cell extracts catalysing removal of 5′-sugar phosphate residues . Although most of the enzymes involved in end processing have been identified and this process plays a key role in maintaining genome stability, the precise mechanism governing recognition and processing of such a variety of SSBs is unclear. To address the mechanism involved in DNA SSB recognition and repair, we used a formaldehyde cross-linking assay to follow the proteins involved in processing of 3′-end-modified DNA SSBs in human cell extracts.
Outline of cross-linking assay
We have recently developed a new DNA–protein cross-linking protocol aimed at revealing the engagement of BER proteins during repair of damaged DNA . In brief, this protocol uses oligonucleotides containing a 3′-biotinylated end which are used to form a damage-containing and a control duplex oligonucleotide complete with a hairpin loop to prevent nuclease digestion of the oligonucleotide during incubation with cell extracts (Fig. 1). The oligonucleotides are subsequently bound to streptavidin magnetic beads and incubated with whole cell extracts (WCEs) in buffer containing ATP, dNTPs and NAD+ to allow repair to proceed. After incubation for the times indicated, proteins involved in repair are cross-linked to the DNA and to each other by the addition of formaldehyde. The beads are subsequently washed before reversal of the cross-links, and released proteins are analysed by gel electrophoresis and immunoblotting with the corresponding antibodies. We used this assay to follow repair protein cross-linking during incubation with human WCE for three different oligonucleotide duplexes containing a one-nucleotide gap marked with 3′-hydroxyl, 3′-phosphate or 3′-phosphoglycolate ends in comparison with a control undamaged duplex (Fig. 1).
XRCC1 is not essential for Pol β binding to a gap-containing DNA
Although it is obvious that DNA SSBs containing modified 3′-end lesions will require specific proteins for ‘cleaning’ the ends and preparing them for DNA repair, the time and mechanism of entry of these proteins and other BER partners into the repair process is unknown. This mechanism should include identification of these lesions as a strand break, verification of the nature of the 3′ end, and identification of a satisfactory pathway for repair. The XRCC1 component of the XRCC1–DNA ligase IIIα heterodimer is thought to be a protein providing such a mechanism by acting as a nick sensor and a docking platform for formation of a multiprotein complex capable of repairing various strand breaks . Consequently, to fulfil these functions, the XRCC1–DNA ligase IIIα heterodimer should be the first protein to bind to a variety of strand breaks, and recruitment of other repair proteins should depend on it. We tested this model in a direct experiment by analysing the interdependence of BER protein binding/cross-linking to different DNA substrates.
Two proteins are required to repair a single-nucleotide gap containing substrate with a 3′-hydroxyl end: Pol β, which will add one nucleotide to the 3′ end and fill the gap, and XRCC1–DNA ligase IIIα heterodimer, which will seal the DNA ends. In agreement with this, using HeLa WCE, we were able to cross-link these proteins to the gap-containing substrate to a greater extent than to the control substrate (Fig. 2A), indicating damage specificity of cross-linking. When 5′-labelled gap-containing oligonucleotide attached to the beads was incubated with WCE, gap filling by Pol β can be observed as early as 0.25–0.5 min, at a point where substantial cross-linking of Pol β is observed, and was nearly completed within 2 min (Fig. 2B). Ligation was mostly accomplished between 1 and 2 min, and the XRCC1–DNA ligase IIIα heterodimer can be efficiently cross-linked to the substrate during this period. Approximately 80% of the repair of the substrate is achieved within 4 min at a point where the proteins are dissociating from the DNA. Therefore, cross-linking of Pol β and the XRCC1–DNA ligase IIIα heterodimer correlates well with the kinetics of repair.
We next immunodepleted the XRCC1–DNA ligase IIIα heterodimer from WCE using XRCC1 antibodies and tested whether immunodepletion will affect Pol β binding/cross-linking to the gapped DNA. As Pol β interacts with and partially coprecipitates XRCC1, immunodepleted extracts contained slightly less Pol β (20%; Fig. 3A), and therefore slightly reduced (10%) cross-linking was observed (Fig. 3B). However, we found that, when normalized to the protein amount in the cell extract, Pol β retained full ability to recognize and bind a single-nucleotide gap in the absence of XRCC1–DNA ligase IIIα (Fig. 3C), suggesting that Pol β probably binds first, processes the gapped DNA to the ligatable stage by filling the gap, and then recruits XRCC1–DNA ligase IIIα heterodimer to accomplish the repair. To test this, we immunodepleted Pol β from WCE and monitored efficiency of XRCC1–DNA ligase IIIα binding/cross-linking to the gap-containing substrate. We found that, although the amount of XRCC1 remains the same in mock or Pol β-depleted cell extracts (Fig. 3D), cross-linking of XRCC1 in the latter was reduced by approximately twofold (Fig. 3E,F), indicating that Pol β is required for efficient XRCC1–DNA ligase IIIα binding to a one-nucleotide gap containing substrate. As Pol β processing of a single-nucleotide gap containing the 3′-hydroxyl end was required for efficient XRCC1–DNA ligase IIIα binding, we speculate that all damage-processing proteins bind to the SSB before XRCC1–DNA ligase IIIα.
Processing of the modified 3′ end is required for efficient XRCC1–DNA ligase IIIα binding
We further explored this model in which a damage-specific protein binds first, processes the lesion, and then recruits the end-joining machinery. According to this model, repair of the 3′-phosphoglycolate end would involve sequential processing by APE1 (the major 3′-phosphoglycolate activity in human cell extracts ), and Pol β before XRCC1–DNA ligase IIIα would seal the strand break. Therefore, recruitment of the XRCC1–DNA ligase IIIα heterodimer should depend on APE1. To test this model we immunodepleted APE1 from HeLa WCE. The amounts of XRCC1 and Pol β were unaffected by the immunodepletion protocol (Fig. 4A). Furthermore, removal of APE1 did not affect binding/cross-linking of XRCC1 to the gapped DNA substrate where Pol β, but not APE1 processing, is required (Fig. 4B). However binding/cross-linking of XRCC1 to the 3′-phosphoglycolate-containing substrate was approximately twofold reduced in APE1-immunodepleted but not mock-immunodepleted cell extracts (Fig. 4C,D). To demonstrate that deficient XRCC1 cross-linking was exclusively due to the immunodepletion of APE1, we complemented depleted cell extracts with purified recombinant human APE1 which stimulated XRCC1 binding/cross-linking (Fig. 4E). Taken together these experiments indicate that APE1 is required for efficient XRCC1–DNA ligase IIIα heterodimer binding to the 3′-phosphoglycolate-containing substrate.
In a similar set of experiments, we immunodepleted PNK and investigated cross-linking of XRCC1 and PNK to a 3′-phosphate-containing substrate. Immunodepletion of PNK notably reduced the 3′-phosphatase activity of cell extracts (Fig. 5A) and further demonstrates that PNK is the major 3′-phosphate-processing activity in human cell extracts as previously described . Furthermore, immunodepletion of PNK did not affect the amounts of XRCC1 and Pol β in the extract (Fig. 5B). Using the same extract, we found efficient cross-linking of XRCC1 to the 3′-OH gapped substrate (Fig. 5C) for which no processing by PNK but only processing by Pol β, which remains in the immunodepleted extract, is required. Using these extracts, we were able to efficiently cross-link both XRCC1 and PNK in mock-immunodepleted cell extracts, but a threefold reduction in XRCC1 cross-linking was observed using the PNK-depleted cell extracts (Fig. 5D,E), even though the two extracts contained equal amounts of XRCC1. Finally, addition of purified human PNK to the PNK-immunodepleted cell extracts stimulated XRCC1 cross-linking (Fig. 5F). These experiments suggest that XRCC1–DNA ligase IIIα heterodimer would not efficiently bind to the 3′-phosphate-containing substrate before removal of the phosphate group by PNK. Although immunodepletion of XRCC1–DNA ligase IIIα heterodimer reduced the level of endogenous PNK (data not shown), when PNK was brought to the mock immunodepletion level by addition of the recombinant human protein, it was efficiently cross-linked to the 3′-phosphate-containing substrate in XRCC1-depleted extracts (data not shown), indicating that, as in the case of Pol β, PNK binding is not dependent on XRCC1–DNA ligase IIIα heterodimer.
Interaction between PNK and XRCC1 is required for effective XRCC1–DNA ligase IIIα recruitment to the site of DNA damage
PNK interacts with the phosphorylated form of XRCC1–DNA ligase IIIα through its FHA domain, and a mutation (R35A) in this domain disrupts this interaction . Using site-directed mutagenesis, we generated the R35A PNK mutant and demonstrate that it is deficient in its interaction with XRCC1 compared with the wild-type protein (Fig. 6A). However, the mutant retains full 3′-phosphatase activity (Fig. 6B). Using the cross-linking assay, we demonstrate that the complementation of PNK-depleted cell extracts with wild-type PNK restores XRCC1–DNA ligase IIIα heterodimer binding/cross-linking (Fig. 6C). However, complementation with the R35A PNK mutant has very little effect on XRCC1–DNA ligase IIIα heterodimer binding/cross-linking (Fig. 6C), indicating that interaction between PNK and XRCC1 is important for recruitment of XRCC1–DNA ligase IIIα heterodimer.
Poly(ADP)-ribose polymerase-1 (PARP-1) has a very high affinity for strand breaks and binds them before any other repair proteins . Binding of PARP-1 to the SSB stimulates formation of poly(ADP)-ribose polymers and dissociation of PARP-1 from the DNA . A recent study from our group showed that PARP-1 is always involved in BER of DNA base lesions and SSBs and is important for preventing degradation of excessive SSBs by cellular nucleases as previously proposed . Blocking of poly(ADP)-ribosylation and consequent dissociation of PARP-1 from SSB results in inhibition of SSB repair and repair foci formation [18,21]. However, when there are sufficient amounts of repair enzymes present, they efficiently substitute PARP-1 from the nicked DNA . Therefore, when repair enzymes are in excess over the amount of DNA SSBs, neither the cross-linking efficiency of XRCC1–DNA ligase IIIα heterodimer and Pol β nor the rate of repair are affected in PARP-1-deficient cells [22,23], suggesting that in this situation PARP-1 is not essential for recruitment of other BER proteins. That is why XRCC1–DNA ligase IIIα heterodimer plays a central role in the current models for SSB repair. As XRCC1–DNA ligase IIIα heterodimer interacts with APE1 , Pol β and PNK  it was proposed that XRCC1 serves as a docking platform to accommodate proteins required for repair of SSBs . It was further suggested that XRCC1–DNA ligase IIIα heterodimer nucleates assembly of the ‘multitask’ repair complex that would be able to repair SSBs of any complexity , although some data indicates that efficient recruitment/stability of PNK or Pol β at sites of SSB repair in cells would involve not only interaction with XRCC1, but also the recognition of the DNA substrate by PNK/Pol β.
Using cell extracts immunodepleted of individual BER proteins and monitoring binding/cross-linking of the remaining proteins to the substrate DNA, containing SSBs with various 3′-end modifications, we investigated which proteins initiate recognition and repair of such lesions. We found that, after PARP-1 binding/dissociation, repair of the SSBs is always followed by a specific protein that is required to progress the particular lesion to the next stage of repair. Therefore, Pol β initiates repair of single-nucleotide gaps, PNK is required for the initiation of repair of 3′-phosphate-containing SSBs, and APE1 initiates repair of 3′-phosphoglycolate-containing SSBs. Removal of these proteins by immunodepletion extensively decreased binding/cross-linking of XRCC1–DNA ligase IIIα heterodimer to the corresponding substrates. However, removal of Pol β had a less pronounced effect, because to a certain extent XRCC1 is able, although less efficiently than in the presence of Pol β, to bind gapped DNA with 3′-hydroxyl and 5′-phosphate ends because it resembles nicked DNA . We have also recently demonstrated that Pol β stimulates binding/cross-linking of XRCC1–DNA ligase IIIα heterodimer during the repair of incised AP sites that are intermediates of BER . Conversely, immunodepletion of XRCC1–DNA ligase IIIα heterodimer had little or no effect on the amount of binding/cross-linking of the damage-specific proteins required for repair initiation. Taken together, our data do not support the model in which XRCC1–DNA ligase IIIα heterodimer binding follows PARP-1. Alternatively, they support the idea that damage-specific proteins bind before XRCC1–DNA ligase IIIα and direct repair via a pathway that will include only a subset of proteins required for specific SSB processing.
On the basis of our findings, we propose a model in which any of the BER proteins mentioned in this study (APE1, PNK, Pol β and XRCC1–DNA ligase IIIα) are able to independently initiate the repair process by binding to the corresponding DNA lesion (3′-phosphoglycolate, 3′-phosphate, single-nucleotide gap and SSBs containing 3′-hydroxyl and 5′-phosphate groups, respectively; Fig. 7). After initiation, repair may proceed either by the ‘passing the baton’ mechanism by handing the substrate to the next enzyme required [29,30] or by nucleation of a transient damage-specific complex including a subset of enzymes that are required for repair of a particular lesion. Our data do not provide strong evidence in favour of transient complexes (although we can see simultaneous cross-linking of several repair proteins on damaged DNA). However, the formation of nuclear foci by BER proteins supports this model [31,32]. It is quite clear that the lifetime of transient repair complexes, and therefore their lifetime within DNA damage foci, depends on multiple interactions between the proteins involved. XRCC1 protein, which is involved in multiple interactions with other BER proteins, may play an essential role in assembly and stability of such complexes. Therefore, as recently reported, disruption of the interaction of XRCC1 with PNK or Pol β leads to deficiency in accumulation of PNK and Pol β in nuclear foci induced by hydrogen peroxide [17,32] and increased sensitivity of deficient cells to DNA damage . These data indicate that the mechanism involved in SSB repair in living cells may be more sophisticated than in cell extracts.
In summary, our data support the idea that BER enzymes participate in the repair of various SSBs through participation in multiple repair pathways/complexes involving distinct subsets of repair proteins. For example, strand breaks containing 3′-phosphate ends would be initiated by PNK, followed by Pol β and XRCC1–DNA ligase IIIα heterodimer, or alternatively, binding of PNK would initiate formation of a transient DNA repair complex containing all three enzymes required for repair of this substrate. However, the detailed molecular mechanism of repair events after strand break binding by end-damage-specific protein is unclear and requires further investigation.
Synthetic oligodeoxyribonucleotides were purchased from Eurogentec (Seraing, Belgium) and purified by electrophoresis on a 20% polyacrylamide gel. [32P]ATP[γP] (3000 Ci·mmol−1) was purchased from PerkinElmer Life Sciences (Boston, MA, USA). XRCC1 (ab144) and DNA ligase III (ab587) antibodies were purchased from Abcam Ltd (Cambridge, UK). Antibodies against rat Pol β, human APE1 and human PNK were raised in rabbit and affinity purified as described . His-tagged PNK and R35A PNK generated by site-directed PCR mutagenesis were purified on a nickel chelating resin (Novagen, Madison, WI, USA) as recommended by the manufacturer.
HeLa cell pellets were purchased from Paragon (Aspen, CO, USA). WCEs were prepared by the method of Manley et al.  and dialysed overnight against buffer containing 25 mm Hepes/KOH, pH 7.9, 100 mm KCl, 12 mm MgCl2, 0.1 mm EDTA, 17% glycerol and 2 mm dithiothreitol. Extracts were divided into aliquots and stored at −80 °C.
Immunodepletion of WCEs and western blots
Immunodepletion of WCEs was performed as described  and verified by SDS/PAGE (10% gel). Proteins were transferred to poly(vinylidene difluoride) membranes and immunoblotted using the corresponding antibodies. Blots were visualized using the ECL plus system (Amersham, Chalfont St Giles, Bucks, UK).
The cross-linking assay with hairpin oligonucleotide substrates attached to streptavidin magnetic beads was performed as described . For direct comparison, proteins cross-linked from different substrates or extracts were analysed on the same immunoblot.
Repair assays of hairpin oligonucleotide substrates attached to streptavidin magnetic beads were performed as described previously .
To analyse the interactions of PNK and R35A PNK with XRCC1, the protocol described by Caldecott et al.  was followed with modifications. Briefly, 50 µg HeLa WCE was incubated with 1 µg His-tagged PNK/R35A PNK in 40 µL buffer A (50 mm Tris/HCl, pH 8, 0.1 m NaCl, 2% glycerol, 1 mm dithiothreitol, 25 mm imidazole, pH 7.5) for 30 min at room temperature, and then 160 µL buffer A and a 25 µL bed volume of nitrilotriacetate/agarose (Novagen) was added. Reactions were incubated on ice with frequent gentle mixing, and after 20 min the nitrilotriacetate/agarose was gently pelleted by brief centrifugation. The supernatant was removed, and the nitrilotriacetate/agarose beads washed with 3 × 190 µL buffer A. After the final wash, the beads were boiled in 60 µL SDS/PAGE sample buffer [25 mm Tris/HCl, pH 6.8, 2.5% (v/v) 2-mercaptoethanol, 1% (v/v) SDS, 5% (v/v) glycerol, 1 mm EDTA, 0.15 mg·mL−1 bromophenol blue], and 20 µL loaded on to an SDS/10% polyacrylamide gel. Proteins were transferred to a poly(vinylidene difluoride) membrane and immunoblotted with the corresponding antibodies.
We thank Dr Keith Caldecott for critical reading of the manuscript and for communicating his data before publication.