Prereplicative repair of oxidized bases in the human genome is mediated by NEIL1 DNA glycosylase together with replication proteins

Significance

Repair of mutagenic oxidized bases in the genome is required before replication to prevent mutations. It is unknown how such base lesions, which do not block replication, are flagged for repair in the single-stranded replicating template. We demonstrate here that the repair-initiating, S-phase–activated Nei-like (NEIL) 1 DNA glycosylase binds to but does not excise the base lesion and cleave the template DNA strand, which would lead to a lethal double-strand break. Instead, NEIL1 blocks progression of the replication fork, which then regresses to allow lesion repair. In the absence of NEIL1, the related glycosylase NEIL2 serves as a backup enzyme.

Abstract

Base oxidation by endogenous and environmentally induced reactive oxygen species preferentially occurs in replicating single-stranded templates in mammalian genomes, warranting prereplicative repair of the mutagenic base lesions. It is not clear how such lesions (which, unlike bulky adducts, do not block replication) are recognized for repair. Furthermore, strand breaks caused by base excision from ssDNA by DNA glycosylases, including Nei-like (NEIL) 1, would generate double-strand breaks during replication, which are not experimentally observed. NEIL1, whose deficiency causes a mutator phenotype and is activated during the S phase, is present in the DNA replication complex isolated from human cells, with enhanced association with DNA in S-phase cells and colocalization with replication foci containing DNA replication proteins. Furthermore, NEIL1 binds to 5-hydroxyuracil, the oxidative deamination product of C, in replication protein A-coated ssDNA template and inhibits DNA synthesis by DNA polymerase δ. We postulate that, upon encountering an oxidized base during replication, NEIL1 initiates prereplicative repair by acting as a “cowcatcher” and preventing nascent chain growth. Regression of the stalled replication fork, possibly mediated by annealing helicases, then allows lesion repair in the reannealed duplex. This model is supported by our observations that NEIL1, whose deficiency slows nascent chain growth in oxidatively stressed cells, is stimulated by replication proteins in vitro. Furthermore, deficiency of the closely related NEIL2 alone does not affect chain elongation, but combined NEIL1/2 deficiency further inhibits DNA replication. These results support a mechanism of NEIL1-mediated prereplicative repair of oxidized bases in the replicating strand, with NEIL2 providing a backup function.

Several dozen oxidatively modified, and mostly mutagenic, bases are induced in the genomes of aerobic organisms by endogenous and environmentally induced reactive oxygen species (ROS) (1, 2). For example, 5-hydroxyuracil (5-OHU), a predominant lesion generated by oxidative deamination of C, is mutagenic because of its mispairing with A (3). The bases in the single-stranded (ss) replicating DNA template are particularly prone to oxidation (4); the lack of their repair before replication could fix the mutations. The bulky base adducts if formed in the template strand would block replication and trigger DNA damage-response signaling. In contrast, oxidized bases with minor modifications, which are continuously formed in much higher abundance than the bulky adducts, would mostly allow replication. This raises the question of how these bases are marked for repair before replication to avoid mutagenic consequences. Oxidized base repair in mammalian genomes occurs primarily via the base excision repair (BER) pathway which is initiated with lesion base excision mediated by one of five major DNA glycosylases belonging to the Nth or Nei families, with distinct structural features and reaction mechanisms (1). Nei-like (NEIL) 1 and NEIL2 DNA glycosylases (5, 6) of the Nei family (which also contains the less characterized NEIL3; ref. 7) are distinct from NTH1 and OGG1 of the Nth family because the NEILs can excise damaged bases from ssDNA substrates (8). Furthermore, NEIL1 is activated during the S phase (5). Our earlier studies also showed that NEIL1 functionally interacts with many DNA replication proteins including sliding clamp proliferating cell nuclear antigen (PCNA), flap endonuclease 1 (FEN-1), and Werner RecQ helicase (WRN) via its disordered C-terminal segment (9–12). Importantly, mammalian ssDNA-binding replication protein A (RPA), essential for DNA replication and most other DNA transactions, inhibits NEIL1 or NEIL2 activity with primer-template DNA substrates mimicking the replication fork, presumably to prevent double-strand break formation (13). Although they collectively implicate NEIL1 in the repair of replicating DNA, those observations did not provide direct evidence for NEIL1’s role in prereplicative repair, nor did they address whether NEIL1 is unique for this function. In this report, we document that NEIL1 binds to the lesion base in an RPA-coated ssDNA template in vitro, without excising the lesion and cleaving the DNA strand, and blocks primer elongation by the replicative DNA polymerase δ (Polδ). This strongly suggests that the replication complex at the lesion site is stalled in vivo in the presence of NEIL1, which provides the signal for repair of lesions in the template strand before replication.

Results

NEIL1 Depletion Inhibits DNA Replication Fork Progression After Oxidative Stress.

Control and NEIL1-depleted (siRNA-mediated) HEK 293 cells (Fig. 1D) were subjected to DNA fiber analysis to measure DNA replication chain elongation rate after induction of oxidized bases in the genome (14, 15). The cells were sequentially treated with chlorodeoxyuridine (CldU), H₂O₂, and iododeoxyuridine (IdU). DNA fibers from lysed cells were spread on microscopic slides, and the progression of replication forks was visualized by detecting CldU (red) and IdU (green) tracks by using appropriate Abs, as described in Materials and Methods (Fig. 1A). The double red/green staining represents elongation of preexisting replication forks, whereas the red-only and green-only tracks represent terminated and newly initiated forks, respectively. Quantification of red and green track lengths in double-labeled DNA showed significant inhibition of fork progression after H₂O₂ treatment, as indicated by a higher proportion of shorter IdU tracks in NEIL1-depleted cells compared with controls (Fig. 1B). The global rate of fork progression, assessed by measuring the lengths of 100 fork fibers, decreased by ∼20% to 30% in NEIL1-depleted cells under oxidative stress (Fig. 1C). In contrast, depletion of NEIL2 did not significantly affect the fork progression rate. However, the inhibition of DNA synthesis in combined (i.e., NEIL1 + NEIL2)-depleted cells was markedly higher than that caused by NEIL1 depletion alone. These results strongly suggest efficient NEIL1-mediated repair of oxidized bases in the template DNA whereas repair initiated by other DNA glycosylases including NEIL2 is delayed. It is likely that NEIL2 and possibly other DNA glycosylases serve as a backup in the absence of NEIL1 (16–18).

Fig. 1.

NEIL1 deficiency inhibits replication fork progression in H₂O₂-treated human cells. NEIL1- or NEIL2-depleted and control HEK 293 cells in the log phase were sequentially pulse-labeled with CldU (red) and IdU (green) separated by a 10-min treatment with H₂O₂. DNA fibers were spread on to microscope slides (14, 15) and immunostained with Abs recognizing CldU and IdU (Materials and Methods). Schematic of CldU/IdU labeling (A) and representative immunofluorescence images of DNA fibers (B) are shown. (C) Track lengths (in micrometers) of 100 control or NEIL1- or NEIL2-depleted cells were plotted by using ImageJ software. NEIL1- (but not NEIL2-) depleted cells had 20% to 30% inhibition of fork progression, which was increased by simultaneous NEIL2 depletion (∼40%). FEN-1 depletion served asa positive control. (D) More than 75% reductions in NEIL1 and NEIL2 level after siRNA treatment were confirmed by Western analysis of whole cell extracts. The quantitation of protein levels from three separate experiments is represented as a histogram.

NEIL1 and PCNA Colocalize with the Replication Foci, Whereas NEIL2 Colocalizes with PCNA only in NEIL1-Deficient Cells.

We analyzed immunofluorescence to examine colocalization of NEIL1 or NEIL2 with BrdU at discrete foci in replicating DNA (19, 20). Log-phase HEK 293 cells cultured on microscope coverslips were pulse-labeled with BrdU (10 µM/10 min), fixed with 4% (vol/vol) formaldehyde, and then immunostained with anti-BrdU (mouse; Abcam) and NEIL1 or NEIL2 Abs (rabbit) (11, 21). Secondary Abs were Alexa Fluor anti-mouse 488 and anti-rabbit 568. Confocal microscopy revealed strong nuclear colocalization of NEIL1 with BrdU fluorescence (Fig. 2A). The sliding clamp PCNA, an S phase marker, served as a positive control. Colocalization of NEIL2 with BrdU was insignificant in control cells, but significantly higher in NEIL1-deficient cells (Fig. 2B). NEIL1 was depleted by using siRNA as in Fig. 1D. These results indicate strong association of NEIL1 with replicating DNA in WT cells; NEIL2 was associated with replicating DNA only in NEIL1-deficient cells.

Fig. 2.

NEIL1 colocalizes with DNA replication foci. (A) NEIL1 colocalization with incorporated BrdU (replication foci). (B) NEIL2 colocalization with BrdU is observed only in NEIL1-depleted cells. Log-phase HEK 293 cells on microscope coverslips were pulse-labeled with BrdU (10 min); the cells were fixed, stained with Abs for BrdU (Abnova) and NEIL1/2 (11, 21) or PCNA (Santa Cruz) and distinct fluorescence-tagged secondary Abs. PCNA was used as a positive control. Images were captured by using a Carl Zeiss LSM 510 confocal microscope.

Association of NEIL1 with DNA Replication Proteins Bound to the S-Phase Cell Genome.

By using chromatin immunoprecipitation (ChIP)/re-ChIP analysis, we tested whether NEIL1 and DNA replication proteins colocalize in the same sequences of the replicating genome. After synchronizing by using double thymidine block treatment (22), HEK 293 cells were harvested at the G₁/S boundary or in the S phase (Fig. 3D) and fixed with 1% formaldehyde for ChIP analysis using the NEIL1 immunoprecipitation (IP) with normal rabbit IgG control. A 250-bp region in the GAPDH or β-actin gene was PCR-amplified (Fig. 3A; primer sequences in Fig. S1B). Separate aliquots of NEIL1-ChIP were subjected to a second IP with Abs for PCNA, Polδ, PCNA clamp loader replication factor-C (RF-C), FEN-1, or IgG control. PCR analysis indicated the presence of these replication proteins bound to the same genomic segment as NEIL1, but only in the S-phase cells. The absence of similar PCR products in G₁ cells shows that the association of NEIL1 with replication proteins was S phase-specific. The lack of amplification with control IgG further confirmed the specificity of binding. In a reverse re-ChIP analysis, we first carried out ChIP with RF-C Ab, and the eluate was subjected to a second ChIP with NEIL1, NEIL2 Ab, or IgG control. Again, RF-C and NEIL1 were found to be bound to the same segments of β-actin and GAPDH genes in the S-phase chromatin (Fig. 3B), whereas NEIL2 level colocalized with RF-C in G₁- or S-phase cells was presumably too low to be detected under our condition of quantitative PCR. The difference in the level of immunoprecipitated DNA was further confirmed by comparison of cycle threshold (Ct) values after real-time PCR analysis of the products (plotted as fold enhancement compared with IgG in Fig. 3C). Taken together, these results show that NEIL1 but not NEIL2 colocalized with the DNA replication complex in S-phase cells.

Fig. 3.

Enrichment of NEIL1 and DNA replication proteins in the same genomic sequences of S-phase cells. (A) ChIP/re-ChIP assay: NEIL1 and replication proteins have a stronger association with DNA in S-phase cells than in G₁ cells. ChIP assay with IgG or NEIL1 Ab (first IP) of cross-linked chromatin from HEK 293 cells at G₁ vs. S phase, PCR amplification products of a 250-bp region in the GAPDH or β-actin gene (primer sequences in Fig. S1B). Second IP of other four fractions with IgG (control) or Ab for replication proteins and PCR amplification as before. (B) Reverse ChIP was performed with the first IP by using RF-C Ab and the second IP with NEIL1, NEIL2, or PCNA Abs. (C) ChIP products were quantified by real-time PCR, and Ct values are plotted as fold enhancement after normalizing to IgG. (D) Cell-cycle distribution of G₁ vs. S phase cells showed ∼55% cells in S phase 3 h after release from double thymidine block. Other details are provided in Materials and Methods.

NEIL Immunocomplexes Contain DNA Replication Proteins.

To characterize the association of NEILs with the DNA replication proteins that may also be involved in BER, we tested for their presence in NEIL immunoprecipitates isolated from (RNase + DNase)-treated nuclear extracts of HEK 293T cell lines stably transfected with NEIL1-FLAG, NEIL2-FLAG, or empty FLAG expression plasmids. The levels of ectopic NEILs in these lines were comparable to those of the corresponding endogenous proteins. The presence of Polδ, RF-C, and LigI, in addition to PCNA and FEN-1, was observed in NEIL1-FLAG and NEIL2-FLAG IPs but not in the vector control IP (Fig. 4A). Furthermore, after normalizing to the FLAG level, the amounts of PCNA, Polδ, RF-C, FEN-1 and LigI, but not of DNA polymerase β (Polβ), were found to be three- to sixfold lower in the NEIL2-FLAG IP relative to the NEIL1-FLAG IP (Fig. 4A, Right). We also examined the complexes of endogenous NEILs with replication proteins by using NEIL1/NEIL2 Abs conjugated to IgA agarose beads (Sigma). The NEIL1 and NEIL2 IPs from (DNase + RNase)-treated HEK 293 cell extracts contained Polδ, RF-C, and FEN-1 (Fig. 4B), which were absent in the control IP with rabbit IgG.

Fig. 4.

In-cell association of NEIL1 and NEIL2 with DNA replication proteins. (A) FLAG IPs from DNase I/RNaseA-treated (100 µg/mL each) extracts of HEK 293 cells ectopically expressing empty or NEIL1- or NEIL2-FLAG (1, 21) were analyzed for replication and other repair proteins. Quantitation of immunoblot bands (histogram; Right) shows approximately three- to fivefold higher levels of replication proteins (Polδ, PCNA, RF-C, FEN-1, LigI) in the NEIL1 IP than in the NEIL2 IP, whereas the Polβ levels in these IPs are similar. (B) Endogenous IP of HEK 293 cell extracts with NEIL1 or NEIL2 Ab or IgG bound to protein A beads confirmed stable association of NEILs with Polδ, RF-C, and FEN-1. (C) In situ PLA assay (Duolink) demonstrating the association of NEIL1 with Polδ, RF-C, FEN-1, and LigI in HEK 293 cells. PLA with NEIL1 Ab (rabbit) vs. IgG (mouse) served as a control.

The in-cell physical association of endogenous NEIL1 with DNA replication proteins was further confirmed by in situ Proximity Ligation Assay (PLA; Duolink kit; cat. no. LNK-92101-KI01; Olink Biosciences). In this assay, two proteins are immunostained with distinct species-specific secondary Abs that are linked to complementary oligonucleotides. When two different Ab molecules bind in close proximity (<40 nm), the linked DNA can be amplified and visualized with a fluorescent probe as distinct foci. The assay is highly specific for physically interacting endogenous proteins in a complex (23–25). We detected a large number of nuclear foci with NEIL1–Polδ, NEIL1–RF-C, NEIL1–FEN-1, and NEIL1–LigI pairs (Fig. 4C). No significant signal was observed with control IgG as the primary Ab. Taken together, these results strongly support NEIL1’s functional association with DNA replication proteins, consistent with its preferential role in repairing the replicating genome.

NEIL2’s Enhanced Association with PCNA in NEIL1-Depleted S-Phase Cells Supports Its Backup Role in Prereplicative Repair.

We isolated the IP of endogenous PCNA from extracts of G₁ or S phase HEK 293 cells by using PCNA Ab (Santa Cruz). Although the NEILs were barely detectable in PCNA IP from G₁ cells, the NEIL1 level was markedly higher in the PCNA IP from S phase cells (Fig. 5A), consistent with NEIL1’s higher level in the S phase (5). Furthermore, the NEIL2 level was enhanced ∼10 fold in PCNA IP, normalized to the PCNA level, from extracts of NEIL1-depleted cells (as in Fig. 1D) compared with control siRNA-transfected cells (Fig. 5B). Together, these results show that the PCNA IP contains detectable NEIL1 in S-phase cells, but contains significant NEIL2 only in NEIL1-depleted cells. This was further confirmed by in situ PLA assay in HEK 293 cells where a strong PLA signal was detected for NEIL1–PCNA interaction in S-phase cells, and NEIL2–PCNA association was markedly higher in NEIL1-depleted cells (Fig. 5C). These results provide additional support for our conclusion that NEIL1 is preferentially associated with replication proteins in the S phase, and that NEIL2 serves as a backup for NEIL1 in repairing the replicating genome.

Fig. 5.

Enhanced association of NEIL1 with PCNA in S-phase cells and of NEIL2 with PCNA in NEIL1-depleted cells. (A and B) Endogenous PCNA was immunoprecipitated from extracts of HEK 293 cells in G₁ vs. S phase with PCNA Ab (Santa Cruz). (A)Western analysis with NEIL1 and NEIL2 Abs showed a strong association only of NEIL1 with PCNA in S phase vs. G1 cells. (B) Enhanced association of NEIL2 with PCNA in NEIL1-depleted cells. (C) (Upper) In situ PLA assays showing NEIL1’s strong interaction with PCNA in S-phase cells and NEIL2’s enhanced interaction in NEIL1-depleted HEK 293 cells. (Lower) PLA signals for NEILs and PCNA in IgG controls.

NEIL1 Coopts DNA Replication Proteins to Carry Out Repair of Oxidized Bases.

During BER, the DNA glycosylase sequentially excises the base lesion and cleaves the DNA strand at the damage site via its intrinsic lyase activity to generate a 3′-blocking phospho-α,β-unsaturated aldehyde or phosphate moiety. In mammalian cells, these groups are then removed by APE1 or PNKP, respectively, followed by the filling of the 1-nt gap by Polβ and nick sealing by LigIII during single-nucleotide incorporation or short-patch base-excision repair (BER) (1, 26). In contrast, the long-patch BER (LP-BER) subpathway, initially identified in vitro, involves repair synthesis of a longer (2–8 nt) segment by Polδ/ε after downstream strand displacement (27–29). Participation of DNA replication proteins in LP-BER includes FEN-1 to cleave the displaced flap, PCNA clamp and clamp loader RF-C, and LigI for final nick sealing, suggesting that this repair process is activated during DNA replication (30).^† We previously showed that NEIL1 (and NEIL2) complexes with PNKP, Polβ, LigIIIα, and XRCC1 are able to carry out efficient single-nucleotide BER when repair is reconstituted with these proteins in vitro (26, 31). Similarly, NEIL1’s stable complexes with replication proteins suggested that it also can carry out LP-BER by using the replication proteins. To test this, we eluted the proteins in NEIL1-FLAG IP from HEK 293 cells with a FLAG peptide (Sigma). After confirming the presence of replication proteins in the eluate by Western analysis (Fig. 3A), we monitored the repair of a circular plasmid substrate (pUC19CPD) containing a single 5-OHU lesion within a 32-nt sequence that could be isolated by digestion with Nt.BstNBI, an ssDNA-specific restriction endonuclease (Fig. 6A) (32). To distinguish between single-nucleotide and LP-BER, the repair patch size was monitored by using various ³²P-labeled dNTPs in the reaction (33). Incorporation of ³²P-labeled dNMPs by NEIL1 IP indicated a LP-BER patch size of at least 2, 3, and 4 nucleotides, respectively (Fig. 6B). The presence of a small amount of unligated intermediates suggests that the DNA ligase was limiting in the IP eluate. These results showed that the protein complex eluted from NEIL1 IP from human cells contains all the essential components to proficiently carry out repair of 5-OHU lesion via LP-BER subpathway. The dependence of NEIL1 IP-initiated LP-BER on replication proteins was further confirmed by the lack of enhanced repair by addition of recombinant Polβ; in contrast, Polδ significantly stimulated the repair (Fig. 6C).

Fig. 6.

NEIL1 coopts replication proteins to carry out LP-BER. (A) The plasmid substrate for our BER assay is shown: a 32-nt NtBstNB1 (57) restriction fragment with a 5-OHU at residue 20 is monitored for repair. Incorporation of labeled bases downstream from 5-OHU indicates LP-BER indicates LP-BER. (B) FLAG-NEIL1 IP from HEK 293 cells carries out LP-BER of 5-OHU. Total proteins eluted with FLAG peptide (1, 5, and 10 µg) were used for incorporation of [α-³²P]dAMP (lanes 2–4), dGMP (lanes 5–7), or dCMP (lanes 8–10). After digesting the plasmid with Nt.BstNB1, the 32-nt repaired fragment was analyzed by denaturing electrophoresis (32). Lane 1 shows repair using empty FLAG vector IP. The ³²P-5′-end labeled 32 and 20 nt oligos were used as size markers (lane 11). (C) Addition of recombinant Polδ but not Polβ (50 and 100 fmol) to NEIL1 IP (1 µg) enhanced LP-BER (second nucleotide incorporation). Other details are provided in Materials and Methods.

Collaboration of DNA Replication Proteins with NEIL1 vs. NEIL2 for Oxidized Base Repair.

In view of our finding that NEIL2 associates with DNA replication proteins like NEIL1, albeit at a much lower level, we investigated NEIL2’s ability to carry out LP-BER by using these proteins as before. Because lesion base excision is the rate-limiting step in overall repair, we first normalized the base excision activity of NEIL1 vs. NEIL2 by adjusting their levels in eluates from NEIL-FLAG IPs from HEK 293 cells (Fig. 7A) to achieve comparable 5-OHU excision levels (Fig. 7B). It was observed that 0.1 and 0.2 μg of NEIL1 IP had DNA glycosylase activity comparable to 0.4 and 0.8 μg of NEIL2 IP, respectively. As denoted by the dotted line in the histogram in Fig. 7B (Lower). The NEIL1 IP produced approximately four to fivefold higher level of repair than the NEIL2 IP with comparable repair-initiating glycosylase activity (Fig. 7C).

Fig. 7.

NEIL1 is more efficient in LP-BER than NEIL2 when normalized to the same specific glycosylase activity. (A) FLAG IPs of FLAG-tagged NEIL1 and NEIL2 were adjusted for equal 5-OHU excision activity, based on the amount of protein by Western blotting with anti-FLAG Ab (B) and glycosylase activity (dotted line in histogram) using a 5-OHU–containing ³²P-5′-end labeled duplex oligo substrate. (C) LP-BER assay using a 5-OHU–containing plasmid (Fig. 6) shows approximately fivefold higher repair with NEIL1 IP compared with NEIL2 IP.

Similar studies that used reconstituted systems with purified DNA replication proteins (indicated in Fig. S2) and containing comparable activities of NEIL1 and NEIL2 (Fig. S3A) confirmed approximately fourfold higher repair with NEIL1 that with NEIL2 (Fig. S3B). Taken together, these results underscored preferential association of NEIL1 with the replication proteins to carry out prereplicative LP-BER of oxidized bases, and also suggested that NEIL2 could serve as a backup glycosylase in NEIL1’s absence.

NEIL1 Blocks DNA Synthesis by Polδ in a Template Containing 5-OHU Lesion.

To gain further insight into how NEIL1 coordinates the repair of oxidized bases at the replication fork, we tested the effect of NEIL1 on DNA synthesis by Polδ with a replication fork-mimicking RPA-coated primer-template substrate containing 5-OHU in the template (Fig. 8A). Incorporation of [³²P]dCMP (together with unlabeled dNMPs) indicated that DNA synthesis past the 5-OHU lesion occurred in the absence of NEIL1 (Fig. 8A, lanes 2–3). Excess PCNA in the reaction (fivefold molar excess) significantly enhanced Polδ-mediated DNA synthesis, as expected (Fig. 8A, lane 3). Presumably, the PCNA clamp loading at the end, slides on and off the duplex region of the substrate, so that equilibrium is maintained between bound and free PCNA. The presence of NEIL1, whose strand-cleavage activity is prevented by RPA (Fig. 8B) (13), blocked DNA synthesis by Polδ past the lesion site even in the presence of PCNA (Fig. 8A, lanes 4–5). The lack of strand cleavage by NEIL1 under the optimal repair conditions was confirmed by carrying out a similar reaction containing a ³²P-5′-end labeled template strand and unlabeled dNTPs (in the presence of RPA, Polδ, and PCNA as before; Fig. 8B). These data strongly support a scenario whereby NEIL1 nonproductively binds to the 5-OHU lesion in the RPA-coated partial duplex oligo and blocks nascent chain elongation by Polδ. To further confirm this, we performed affinity coelution of NEIL1 with RPA-coated partial duplex oligo with or without 5-OHU located in the ss segment, and used streptavidin pull-down analysis (Fig. 8C). Consistent with its stalling of Polδ, NEIL1’s binding was markedly higher when the oligo contained 5-OHU. These data clearly suggest that NEIL1’s high-affinity binding to base lesions in an ssDNA template (8) stalls progression of the replication fork, thus allowing prereplicative repair.

Fig. 8.

NEIL1’s nonproductive binding to the base lesion 5-OHU in a partial duplex oligo stalls DNA synthesis by Polδ. (A) A partial duplex, as indicated, represents a primer template for DNA replication with Polδ in which the ss segment (29 nt) in the template, with 5-OHU at residue 27, is complexed with RPA. The reaction contained 5 nM substrate, 10 nM RPA, 5 nM each of Polδ and PCNA, and 10 nM of NEIL1. Addition of excess PCNA (DNA:PCNA ratio, 1:5) enhances polymerase activity as monitored by [³²P]dCMP incorporation at residue 29 (lane 3). DNA synthesis was prevented by NEIL1 (lane 5). (B) RPA inhibits NEIL1’s base excision/abasic site (AP) lyase activity. The RPA bound to the template strand (RPA:DNA molar ratio, 1:3) prevents NEIL1-catalyzed cleavage of the ³²P-5′-end-labeled template strand at the 5-OHU site under the same condition as in A. (C) Affinity coelution of NEIL1 with 5′ biotinylated partial duplex oligo (with or without 5-OHU) bound with streptavidin-agarose beads (Invitrogen). The ss segments of the oligo were coated with RPA as in A. NEIL1 binding was significantly higher when the oligo contained 5-OHU compared with the oligo without the lesion.

Similar analyses with other DNA glycosylases NEIL2 and NTH1 showed that DNA synthesis by Polδ was prevented by NEIL2 but not NTH1 (Fig. S4A). Furthermore, whereas NEIL2’s binding, like that of NEIL1, was higher for 5-OHU–containing RPA-coated partial duplex compared with the control oligo, NTH1’s binding to the two oligos were comparable (Fig. S4B). These results further support NEIL2’s backup function for NEIL1 when needed. The lack of Polδ stalling by NTH1 (which does not bind to base lesions in ssDNA) was expected, so NTH1 served as a control.

Discussion

The ssDNA template at the replication fork may be more prone to oxidative base damage and strand breaks than nonreplicating DNA, thus warranting its urgent repair to prevent mutations (4). Repair of mutagenic bases incorporated during replication [i.e., postreplicative repair (34, 35)] is also essential for maintaining genomic fidelity. 8-Oxoguanine (8-oxoG), a predominant ROS-induced base lesion in DNA, could be mutagenic if not repaired before replication, because replicative DNA polymerases (i.e., Polδ/ε) often incorporate A opposite 8-oxoG in the template strand, generating an A:8-oxoG mispair (36). However, this mismatch is efficiently repaired by MYH (35), a mammalian homologue of Escherichia coli MutY that excises A from the 8-oxoG:A (and possibly FapyG:A) pairs (37). Similarly, the U:A pair generated as a result of incorporation of U during replication is repaired postreplicatively by uracil-DNA glycosylase (UNG2) (34, 38). Like NEIL1, UNG2 and MYH associate with PCNA at the replication foci (39, 40), which presumably recruit these at the replication sites, facilitating postreplicative repair of inappropriate bases (40). However, whether these glycosylases are coupled to the replication machinery or they bind to PCNA molecules at the newly replicated DNA is not clear (35). In contrast, removal of U, generated by deamination of cytosine in DNA, needs to be prereplicatively repaired. Similarly, repair of most other oxidized base lesions (e.g., 5-OHU, thymine glycol, 5-OHC, FapyA, 8-oxoA, uracil glycol) in the template DNA, which are primary NEIL substrates, could not be repaired postreplicatively in an error-free manner, and hence must be repaired before replication to prevent mutation. The 5-OHU lesion is potentially the most mutagenic among these (3). However, how prereplicative repair could be carried out in the template DNA without causing double-strand breaks has not been elucidated.

Our cumulative observations helped to identify unique features that set the NEILs apart from OGG1 and NTH1, including the NEILs’ ability to recognize ss DNA substrates (8). Furthermore, NEIL1 is active with base lesions (e.g., 8-oxoG or 5-OHU) in the primer strand recessed from the 3′-terminus by 1 to 3 nt, whereas OGG1 and NTH1 are inactive with such substrates (41). In this report, we have provided direct evidence for NEIL1’s preferential role in repairing oxidized bases in coordination with DNA replication proteins. NEIL1 is thus required for replication fork progression in oxidatively stressed cells. The observed stalling of DNA synthesis by Polδ at the lesion site in the template is presumably required for prereplicative repair. Furthermore, NEIL1’s localization in replicating DNA foci and enhanced association with replication proteins in S-phase cells, resulting in its enhanced activity, indicates its ability to coopt replication proteins to repair base lesions at the replication fork.

We propose that NEIL1 is a component of the DNA replication complex needed for surveillance of oxidized bases before replication, and thus acts as a “cowcatcher” (Fig. 9). Its preference for binding to lesions in ssDNA (8) should help target such lesions at the replication fork. RPA, which coats the ssDNA at the replication fork, inhibits NEIL1’s strand scission activity via direct interaction, presumably to prevent double-strand break formation (Fig. 9) (13), where repair is not possible because of the lack of the complementary strand. We postulate that NEIL1’s stalling of the replication complex (Fig. 8) causes regression of the replication fork to form a “chicken-foot” structure. The reannealing of the template strand restores the lesion into the duplex region, where its repair is initiated with its excision by NEIL1 followed by repair synthesis by the replication proteins (42). Replication resumes after resolution of the stalled fork. The helicase activity of WRN, which functionally interacts with NEIL1 (9) and other replication proteins, including PCNA (43), RPA (44), and FEN-1 (45, 46), has been implicated in replication fork regression and resolution (44, 47, 48). Alternatively, Smarcal1 (and other annealing helicases) could also be involved in resolving the chicken-foot structure (49, 50), although its interaction with NEIL1 has not yet been tested.

Fig. 9.

Cowcatcher model for prereplicative oxidized base repair by NEIL1. We propose that NEIL1 acts as a cowcatcher in the replication complex when it stalls progression of the replicative polymerase at the lesion site, leading to fork regression, as a result of which the damage in the reannealed duplex can be repaired by NEIL1-initiated BER using the replication proteins. WRN helicase, which interacts with NEIL1 (9), and other replication proteins (43–45) resolve the chicken-foot after completion of repair (47, 48).

We observed inhibition of chain elongation after oxidative stress in NEIL1-depleted cells (Fig. 1). Although this result appears to be counterintuitive, there are two possible explanations. First, it is likely that, even when NEIL1 bound to replicating DNA at the lesion site causes replication arrest followed by fork collapse, regression, lesion repair, and replication restart, the delay in replication is too short to be detected in the fiber growth assay. On the contrary, when NEIL2 takes over in the absence of NEIL1, and carries out the same steps but at a much slower rate, the resulting delay is large and is reflected in slowdown of the fiber growth. An alternative possibility is that thymine glycol, which blocks DNA replication, is efficiently repaired by NEIL1. In NEIL1-depleted cells, inefficient repair of this lesion by alternative means was reflected in slower nascent chain growth. We favor the first possibility because, in the absence of NEIL1, NEIL2 acts in a manner qualitatively similar to NEIL1 (Fig. S4). However, its repair efficiency is significantly lower than that of NEIL1 (Fig. 7), which could be partly caused by a lack of its activation by replication proteins (10). We previously showed that NEIL2 is preferentially involved in repair during transcription (21). Taken together, these data suggest that, in the absence of NEIL1, NEIL2 acts as a “relief pitcher” in removing cytotoxic base lesions from the replicating genome.

We should point out in this context that recently characterized NEIL3 was also shown to be activated during the S-phase (51). However, human NEIL3’s substrate preference has not been well characterized. It is possible that NEIL3 is also involved in prereplicative repair for a different set of oxidized bases. On the contrary, although not an oxidized base, U generated by oxidative deamination of C in the template DNA also needs to be repaired prereplicatively. Based on the studies of Krokan et al., it appears that nuclear UNG2 may function in an analogous fashion as NEIL1 and that SMUG1 could serve as the backup enzyme for U repair (34, 38, 40).

Together, these observations led us to propose that the broad and overlapping substrate specificity of oxidized base-specific glycosylases ensures their ability to provide a backup function when needed (1, 40, 52).^† This is consistent with the observation that mouse mutants lacking individual glycosylases, and cells derived therefrom, are viable, without strong phenotypes (53–55). In any case, the present study documents a mechanism of prereplicative repair of oxidized DNA bases in mammalian genomes in which the repair enzyme acts as a component of the replication complex.

Materials and Methods

Preparation of Single-Lesion, 5-OHU–Containing Oligonucleotide and Plasmid Substrates.

A 51-mer oligo containing 5-OHU at position 26 from the 5′ end, and undamaged complementary oligos, containing G opposite the lesion, or sequences for producing oligos for ligase and polymerase assay (Fig. S1), were purchased from Midlands. The ³²P-5′-end labeling of the oligos, annealing, and purification of labeled oligos were described previously (11). Circular plasmid substrate, pUC19CPD containing a single oxidized base lesion, 5-OHU, was generated as described previously (32, 56, 57). Covalently closed form I plasmid was purified by CsCl centrifugation, and the presence of the 5-OHU lesion was verified by agarose gel electrophoresis after treatment with NEIL1, which converted the plasmid to a nicked circle (form II).

Expression and Purification of Recombinant Proteins.

Recombinant WT NEIL1, NEIL2, PNKP, PCNA, Polδ, RF-C, DNA LigI, FEN-1, Polβ, and LigIIIα were purified as described previously (5, 6, 10, 11, 58–60).

Cell Culture and Coimmunoprecipitation.

The human embryonic kidney cell line HEK 293, grown at 37 °C and 5% CO₂ in DMEM containing 10% FBS and 100 U/mL each of penicillin and streptomycin, was transfected with empty or NEIL1- or NEIL2-FLAG–expressing plasmids. We also generated FLAG-NEIL1 or NEIL2-expressing stable HEK 293 and human colorectal tumor line, HCT116, using Zeocin as the selection marker as described previously (31). At 48 to 56 h after transient transfection, or at ∼80% confluence in the case of stably expressing cultures, the cells were harvested, lysed, and treated with 500 U/mL each of DNase I and RNase A (Ambion) at 37 °C for 30 min, cleared by centrifugation, and immunoprecipitated by rocking for 3 h at 4 °C with FLAG M2 Ab crosslinked to agarose beads (Sigma) as described before (11, 13). Similarly, endogenous NEILs were immunoprecipitated using anti-NEIL1 or NEIL2 Abs conjugated to IgA agarose beads. The beads were collected by centrifugation, washed three times with cold TBS plus 0.1% Triton X-100, and FLAG-NEILs were eluted from the beads. After elution from the beads by adding SDS loading buffer, the immunocomplex was separated in 12% SDS/PAGE and immunoblotted by using Abs for Polδ, RF-C, LigI, Polβ, LigIIIα, FEN-1 (Bethyl Laboratories), and the appropriate secondary Abs.

Cell Synchronization and Cell Cycle Analysis.

HEK 293 cells were synchronized by double thymidine block as described earlier (61). Briefly, at ∼40% confluence, cells were treated with 10 mM thymidine for 18 h, released for 4 h by adding fresh media after washing with PBS solution, and subjected to a second thymidine treatment (10 mM) for 17 h. Cells were then stimulated to proliferate with fresh media and harvested at 0, 2, 3, 4, 5, 6, and 8 h and processed for cell cycle analysis (62). Cells were suspended in a low-salt buffer containing 3% polyethylene glycol, 5 µg/mL propidium iodide, 0.1% Triton X-100, 4 mM Na citrate, and 100 µg/mL RNase, and incubated at 37 °C for 30 min. An equal volume of high-salt buffer (3% polyethylene glycol, 5 µg/mL propidium iodide, 0.1% Triton X-100, and 400 mM NaCl) was added, and the cells were kept at 4 °C overnight. The cellular DNA content was evaluated by flow cytometry by using a FACScan flow cytometer (Becton Dickinson). Histograms were analyzed by using ModFit LT cell cycle analysis software (Verity Software House) to determine the percent of cells in various stages of the cell cycle. A total of 10,000 events were collected for all samples.

DNA Fiber Analysis.

To measure replication rates in HEK 293 cells after NEIL1 and/or NEIL2 depletion, cells transfected with specific siRNA for NEIL1 (sequence: sense, 5′CCGUGAUGAUGUUUGUUUAUU3′; antisense, 5′UAAACAAACAUCAUCACGGUU3′; Sigma) or NEIL2 (25) for 48 h were sequentially treated with CldU (C6891; 20 µM/20 min; Sigma), H₂O₂ (1 mM, 10 min), and IdU (I7125; 50 µM, 20 min; Sigma) (14, 15). Cells were rinsed with PBS solution three times after each treatment. Cells were then harvested by scraping in cold PBS solution and diluted to a concentration of 5 × 10⁵ cells per milliliter. A drop of cell suspension together with spreading buffer (200 mM Tris⋅HCl, pH 7.4, 50 mM EDTA, and 0.5% SDS) was put on a microscope slide and incubated for a few minutes to allow lysis. The slide was then tilted so that DNA fibers could spread over it. Slides were air-dried and the fibers fixed on the slides with methanol/acidic acid (3:1). Fibers were stained with monoclonal rat anti-BrdU [clone BU1/75 (ICR1); Oxford Biotechnologies] for CldU and monoclonal mouse anti-BrdU (clone B44; no. 347580; Becton Dickinson) for IdU. Secondary Abs were goat anti-rat Alexa Fluor 555 and goat anti-mouse Alexa Fluor 488. DNA fibers were visualized by fluorescence microscopy. The lengths of CldU and IdU tracks were measured by using ImageJ software, and micrometer values were converted into kilobases by using Alkylation-Induced Replication Block 79 conversion factor (1 μm = 2.59 kb). At least 100 forks were analyzed for every condition.

ChIP/Re-ChIP Assay.

ChIP analysis was performed with HEK 293 cells by using a ChIP assay kit (Cell Signaling Solution; Millipore) per the manufacturer’s protocol, and the re-ChIP assays were performed as described previously (21, 63).

In Situ PLA.

HEK 293 cells grown overnight in 16-well chamber slides were fixed with 4% paraformaldehyde, then permeabilized with 0.2% Tween 20, followed by incubation with a primary Ab for NEIL1 (rabbit) (5) or DNA replication proteins (mouse monoclonal, as indicated). The PLA assay was performed using the Duolink PLA kit (cat. no. LNK-92101-KI01; OLink Bioscience) per the manufacturer’s instructions. The nuclei were counterstained with DAPI, and the PLA signals visualized in a fluorescence microscope (Olympus) at 200× magnification (56, 64).

DNA Glycosylase Assay.

DNA strand scission catalyzed by NEIL1 or NEIL2 was analyzed with ³²P-5′-end labeled 51-mer oligo substrates containing 5-OHU at position 26 from the 5′ end (11) (Fig. S1). A total of 2 pmol of substrate was incubated with NEIL1 or NEIL2 (indicated amounts) or HEK 293 cell nuclear extracts at 37 °C for 15 min in a 10-μL reaction mixture containing 40 mM Hepes, pH 7.5, 50 mM KCl, 100 μg/mL BSA, and 5% glycerol. After the reaction was stopped with formamide dye (80% formamide, 20 mM NaOH, 20 mM EDTA, 0.05% bromophenol blue, and 0.05% xylene cyanol), the products were separated on a 20% polyacrylamide gel containing 8 M urea in 1× Tris/borate-EDTA buffer, pH 8.4 (11, 58). The radioactivity was quantified by using a PhosphorImager (Amersham Biosciences) and Image Quant software.

Affinity Coelution of Biotinylated DNA Oligo.

A partially duplex oligo (with or without 5-OHU within the ss segment) with a biotin residue at the 5′ end of the longer strand was mixed with RPA (DNA:RPA at 1:2 molar ratio). The oligo was then bound to magnetic streptavidin beads (Invitrogen). After extensive washing with TBS buffer, the beads were mixed with NEIL1, NEIL2, or NTH1 (1:2 molar ratio to DNA) for 1 h with constant rotating. The bound proteins were eluted with SDS buffer and identified by Western analysis.

Reconstitution of NEIL-Initiated LP-BER.

For repair assays that used immunocomplexes isolated from empty FLAG-, NEIL1-FLAG- or NEIL2-FLAG–expressing cells (transiently or stably), the immunocomplexes were eluted from FLAG-agarose beads with FLAG peptide (1×; Sigma) at 4 °C after constant rotational mixing, in a buffer containing PBS solution, 1 mM DTT, and 10% glycerol. An aliquot of the eluate was tested for the presence of DNA replication proteins by immunoblotting. The proteins eluted from NEIL immunocomplexes were used in repair reactions with the 5-OHU–containing plasmid DNA (pUC19CPD) or duplex oligo with blocked (biotinylated) ends (Fig. S1) in the presence of dNTPs plus one radiolabeled [α-³²P]dNTP as indicated. The 20-µL reaction mixture contained 200 fmol of damage-containing plasmid substrate, 1 mmol ATP, 25 μmol unlabeled dNTPs, and 10 µmol [α-³²P]dNTPs (the concentration of the corresponding cold dNTP was lowered to 5 µM, unless otherwise specified) in BER buffer (25 mM Hepes⋅KOH, pH 7.9; 50 mM KCl; 2 mM MgCl₂; 0.5 mM DTT). After incubation for 30 min at 37 °C, the plasmid DNA was phenol/chloroform-extracted, ethanol-precipitated, recovered, and digested with N.BstNBI (New England Biolabs), then resolved on a denaturing polyacrylamide gel and analyzed with a PhosphorImager.