Influence of oxidized purine processing on strand directionality of mismatch repair

: Replicative DNA polymerases are high-fidelity enzymes that misincorporate nucleotides into nascent DNA with a frequency lower than 1/105, and this precision is improved to about 1/107 by their proofreading activity. Because this fidelity is insufficient to replicate most genomes without error, nature evolved postreplicative mismatch repair (MMR), which improves the fidelity of DNA replication by up to three orders of magnitude through correcting biosynthetic errors that escaped proofreading. MMR must be able to recognize non-Watson-Crick base pairs and excise the misincorporated nucleotides from the nascent DNA strand, which carries - by definition - the erroneous genetic information. In eukaryotes, MMR is believed to be directed to the nascent strand by pre-existing discontinuities such as gaps between Okazaki fragments in the lagging strand, or breaks in the leading strand generated by the mismatch-activated endonuclease of the MutL homologs PMS1 in yeast or PMS2 in vertebrates. We recently demonstrated that the eukaryotic MMR machinery can make use also of strand breaks arising during excision of uracils or ribonucleotides from DNA. We now show that intermediates of MYH-dependent excision of adenines mispaired with 8-oxoguanine (GO) also act as MMR initiation sites in extracts of human cells or Xenopus laevis eggs. Unexpectedly, GO/C pairs were not processed in these extracts and failed to affect MMR directionality, but extracts supplemented with exogenous OGG1 did so. Because OGG1-mediated excision of GO might misdirect MMR to the template strand, our findings suggest that OGG1 activity might be inhibited during MMR. Background: We studied the interplay between base excision repair (BER) of 8-oxoguanine (G O ) and mismatch repair (MMR). Results: BER and MMR interact during the processing of G O /A but not G O /C mispairs. Conclusion: BER of G O containing lesions appears to be regulated. Significance: BER intermediates were believed to be unavailable to other pathways of DNA metabolism. This hypothesis may be incorrect.

E. coli, the newly-synthesized strand remains transiently unmethylated on adenines within GATC sequences. Mismatch-activated MutS/MutL complex licences MutH to incise the unmethylated GATC, which allows the loading of the UvrD helicase, together with one of several exonucleases. This leads to the degradation of the error-containing strand from the MutH-catalyzed nick towards and some distance past the mismatch to generate a singlestranded gap that is subsequently filled-in by polymerase III. The remaining nick is sealed by DNA ligase (1)(2)(3).
In eukaryotes, DNA methylation is not used in strand discrimination during MMR. Instead, the nascent strand is distinguished from the template by transient discontinuities, such as gaps between Okazaki fragments in the lagging strand. The leading strand contains no such discontinuities, other than the 3' terminus. Because the only MMR-associated exonuclease identified to date, EXO1, has an obligate 5' to 3' polarity, it appeared unlikely that MMR would use the 3' terminus of the primer strand for initiation. However, seminal work from the Modrich laboratory demonstrated that association of the mismatch-activated MutL (a heterodimer of MLH1 and PMS2) with PCNA bound at the 3' terminus activates a cryptic endonuclease in the PMS2 subunit, which introduces nicks into the newlysynthesized strand and thus provides EXO1 with entry sites where 5' to 3' degradation of the error-containing strand can initiate (4,5).
While studying the involvement of MMR in somatic hypermutation, which is triggered by activation-induced cytidine deaminase (AID) that converts cytosines to uracils at the immunoglobulin locus of activated B cells (6), we discovered that the MMR system can hijack strand breaks generated by the base excision repair (BER) system during uracil removal for the purpose of strand discrimination (7). This was unexpected, because BER was thought to be a concerted process in which the strand break generated by AP endonuclease after removal of the aberrant base is not available for other processes of DNA metabolism (8). In later experiments, we showed that breaks generated during the RNaseH2-catalyzed removal of ribonucleotides misincorporated into DNA during replication can also be utilized by MMR as strand discrimination signals (9).
The above experiments suggested that strand breaks arising during different processes of DNA metabolism may be hijacked by MMR. Although this might lead to improved replication fidelity when breaks in the nascent DNA strand were involved, the opposite would be true if breaks in the template strand were used to direct the MMR process. In order to gain novel insight into this phenomenon, we set out to examine the interplay between MMR and oxidative DNA damage metabolism, because BER-mediated processing of oxidized DNA could potentially take place on both nascent and template strands.
Depending on cell type, human genomes have been estimated to harbor steady-state levels of the major product of DNA oxidation, 2'-deoxy-8-oxoguanosine, ranging between 1'000 and 100'000 residues (10-12). This nucleoside can be present in DNA in two distinct contexts: either paired with deoxycytidine, or mispaired with deoxyadenosine, because the base, 8oxoguanine (G O ), can adopt either an anti or a syn conformation about the glycosidic bond (13,14). Thus, while oxidation of doublestranded DNA gives rise solely to G O /C pairs, in which G O is in the anti conformation, the replicative polymerases , or could insert either a C opposite anti-G O to form a Watson-Crick-like G O /C base pair, or an A opposite syn-G O to form a Hoogsteen G O /A mispair (15)(16)(17). Oxidation affects also the nucleotide pool, where it generates dG O TP. However, because this nucleotide is hydrolyzed by the MutT homolog 1 (MTH1) protein to dG O MP (18-20), G O should not be incorporated into the nascent DNA strand. Thus, in newly-replicated DNA, all G O residues should be in the template strand and all As mispaired with G O should be in the nascent strand. This point is key to our understanding of the potential interplay between oxidative damage processing and MMR. G O /C base pairs are addressed primarily by 8oxoguanine DNA glycosylase (OGG1), a glycosylase/lyase that removes the oxidized base and cleaves the sugar-phosphate backbone to initiate a BER process that ultimately replaces the oxidized nucleotide with a dGMP (13). G O /As are not addressed by OGG1; they are recognized by the MutY homolog (MYH) glycosylase, which removes the mispaired adenines to initiate a BER process that inserts Cs opposite the oxidized guanines. The G O /Cs arising in this way can then be repaired later to G/C by OGG1-dependent BER (21,22). Thus, if MMR were to use breaks generated during MYH-initiated BER of G O /As, these discontinuities would be in the nascent strand and their hijacking by MMR might improve the efficiency of the latter process. In contrast, incisions made during OGG1-initiated BER of G O /Cs would be in the template strand, where they would not only misdirect the mismatch repair process to the wrong strand, but where they could also give rise to double-strand breaks that could cause replication fork collapse.
In an attempt to learn whether MMR utilized strand breaks arising during MYHand/or OGG1-initiated BER processes, we generated substrates containing a single nick (a bona fide strand discrimination signal), a G O /A mispair or a G O /C pair, and studied the efficiency and directionality of MMRcatalyzed repair of a G/T mismatch situated in the vicinity. We show that, in extracts of human cells or Xenopus laevis eggs, the MMR excision machinery can use strand breaks arising upon MYH-catalyzed removal of adenines from G O /A mispairs as initiation sites for exonucleolytic degradation of errorcontaining strands on circular heteroduplex substrates. A similar phenomenon was observed neither on G O /C-containing heteroduplexes, nor on substrates containing 2hydroxyadenine (A O ), another product of purine oxidation (12), even though A O /C and A O /G mispairs were previously reported to be addressed by MYH (23,24).

EXPERIMENTAL PROCEDURES
Restriction enzymes -All restriction enzymes were purchased from New England Biolabs.
Recombinant proteins -GST-tagged MYH was expressed from the pET41a-MYH expression vector (gift of Dr. Barbara van Loon) and partially purified using Glutathione Sepharose beads (GE Healthcare) as described previously (25). In a second purification step, eluates were loaded onto a HiTrap TM Heparin HP column (GE Healthcare), washed with 10 column volumes of wash buffer (30 mM Tris-HCl pH 8, 0.1 mM EDTA, 10% glycerol, 1 mM DTT, 50 mM NaCl) at a flow rate of 0.5 ml/min. Elution was performed using a salt gradient (50-600 mM NaCl). MYH-GST was eluted with 400-500 mM NaCl. The active fractions were identified using a MYH nicking assay, pooled, aliquoted and stored at -80°C. Recombinant APE1 was purchased from New England Biolabs. Recombinant OGG1 was a kind gift of Barbara van Loon. Recombinant OGG1-GST was purchased from Trevigen (4130-100-EB). Recombinant MutL and MutS were expressed and purified in our laboratory as described previously (26). The expression constructs for human geminin (geminin-pET28) and p27 (p27-pET21) were a kind gift of Yoshi Hashimoto and Vincenzo Costanzo. Briefly, protein expression was induced with 0.5 mM IPTG in BL21 cells (Invitrogen) grown at 37°C. The cell pellets were resuspended in 20 mM Tris-HCl pH 7.5, 500 mM KCl, 10% glycerol, 1 mMmercaptoethanol, 0.1% NP40 and PMSF for lysis. After cell disruption and centrifugation of cell debris and membranes, the soluble fraction containing 10 mM imidazole was loaded onto a nickel-chelating column, which was then washed with 5-25 mM imidazole. The protein was eluted with a gradient of 50-300 mM imidazole. Fractions containing the desired polypeptides were pooled and dialyzed against EB buffer (100 mM KCl, 2.5 mM MgCl 2 , 50 mM Hepes-KOH pH 7.5, 10% glycerol).
Cell culture -HCT116 (MutL -deficient) and HCT116 + chromosome 3 (MutLproficient) cells were obtained from Richard Boland (27) and were cultured in McCoy's 5a medium (GIBCO) supplemented with 10% bovine calf serum (GIBCO). The medium for the chromosome 3-complemented cell line was supplemented with 400 g/ml G418. LoVo (MutS -deficient) cells were grown in DMEM (GIBCO) supplemented with 10% bovine calf serum (GIBCO). All media additionally contained 1% penicillin/ streptomycin. Nuclear extracts of human cells -Nuclei were isolated as previously described (28), resuspended in 1/3 of their packed volume in cold extraction buffer (25 mM HEPES-KOH pH 7.5, 292 mM sucrose, 1 mM PMSF, 0.5 mM DTT, 1 g/ml leupeptine) and transferred to a small beaker fitted with a magnetic stirrer bar. NaCl was added dropwise to a final concentration of 150 mM and extraction continued for 1 h at 4°C. The nuclei were pelleted by centrifugation at 14'500 x g for 20 min at 4°C in a tabletop centrifuge. The supernatant was transferred and dialyzed 2 x 1 h at 4°C against 1 liter of cold dialysis buffer (25 mM HEPES-KOH pH 7.5, 50 mM KCl, 0.1 mM EDTA, 231 mM sucrose, 1 mM PMSF, 2 mM DTT, 1 g/ml leupeptine). The dialyzed extract was clarified by centrifugation at 20'000 x g for 15 min at 4°C. The supernatant was aliquoted, snap frozen in liquid nitrogen and stored at -80°C. The protein concentration was determined with the Bradford assay and the salt concentration was measured using a conductivity meter.
Xenopus laevis egg extracts -S-phase extract was prepared as previously described (29). Briefly, eggs were dejellied, activated with calcium ionophore (Sigma-Aldrich), rinsed with S-buffer (50 mM HEPES-KOH pH 7.5, 50 mM KCl, 2.5 mM MgCl 2 and 250 mM sucrose), transferred to 2 ml Eppendorf tubes and crushed by centrifugation for 12 min at 13'200 rpm. The cytoplasmic layer was removed and after addition of CytB (Sigma-Aldrich) cleared by centrifugation for 25 min at 70'000 rpm (Sorvall TL55 swinging bucket rotor). The extract was supplemented with 250 mg/ml cycloheximide, 25 mM phosphocreatine and 10 mg/ml creatine phosphokinase before use.
Western blotting and antibodies -Whole cell extracts of the cell lines were prepared using 2 x Lämmli buffer (120 mM Tris-HCl pH 6.8, 4% SDS, 20% glycerol). Upon determination of the protein concentration by the Lowry assay, proteins were separated on SDS-PAGE. Western blot analyses were carried out using standard procedures. The following antibodies and dilutions were used: MYH (mouse monoclonal, Abcam, ab55551, 1:333), OGG1 (rabbit monoclonal, Abcam, ab124741, 1:10'000), MSH6 (mouse monoclonal, BD Transduction Laboratories, 610919, 1:1'000), MSH2 (mouse monoclonal, Calbiochem, NA-27, 1:500), MLH1 (mouse monoclonal, BD Transduction Laboratories, 554073, 1:500) and TFIIH (rabbit polyclonal, Santa Cruz, sc-293, 1:1'000). The anti-MTH1 antibody was a generous gift of Yusaku Nakabeppu and was used at a dilution of 1:250. Horseradish peroxidase (HRP)-coupled antimouse and anti-rabbit secondary antibodies (GE Healthcare) were used at a dilution of 1:5'000. Substrate generation -The detailed procedure was described previously (30). Briefly, hetero-and homoduplexes were constructed by primer extension using the oligonucleotides listed below as primers and the single-stranded phagemid DNA listed in brackets as template. The single-stranded DNA templates differed in the position of the Nt.BstNBI site, which was situated at nucleotide 2850 either in the viral (top) or complementary (bottom) strand. Incubation of the substrates with the nickase yielded substrates in which MMR occurred either 5 to 3 , or 3' to 5', respectively. The mismatches were located within a SalI and/or AclI restriction site, which was restored upon repair. The desired closed-circular heteroduplex substrates were purified on caesium chloride gradients. Primers: The G O -or A O -containing primers were obtained from Eurogentec (Seraing, Belgium). All other primers were obtained from Microsynth (Balgach, Switzerland). The SalI (GTCGAC) and AclI (AACGTT) restriction sites are highlighted in bold and grey, respectively. PvuI restriction sites (CGATCG) are italicised and mispaired residues are underlined. Primer sequences correspond to the outer strand sequence of the substrate. G/T (pRichi-2850topAclI or pRichi-2850bot AclI): 5' CCAGACGTCTGTCGACGTTGGG AAGCTTGAG 3' T/G (pRichi-2850topSalI): 5' CCAGACGTCT Additionally, the T/G primer was used for T/G mismatch generation.
Additionally, the T/G primer was used for T/G mismatch generation.
Prenicking of substrates -MYH+APE1: 200 ng of the substrates were incubated with 10 ng purified recombinant MYH-GST and 10 U of APE1 (New England Biolabs) in 1x MMR buffer (20 mM Tris-HCl pH 7.6, 40 mM KCl, 5 mM MgCl 2 , 1 mM glutathione, 50 g/ml BSA, 0.1 mM dNTPs) for 4 h at 37°C. The reaction was stopped by heat inactivation and 50 ng of the pre-nicked substrate were analyzed on a 1% GelRed-stained agarose gel for nicking efficiency. The remaining 100 ng of the prenicked substrate were used in a MMR assay. OGG1: 1 g G O /C-G/T substrate was incubated with 0.8 g purified recombinant OGG1 in 20 mM Tris-HCl pH 8, 1 mM DTT, 1 mM EDTA and 0.1 mg/ml BSA for 2.5 hours at 37°C and subsequently purified on a MinElute Spin column (Qiagen). 100 ng of the prenicked substrate were used in a MMR assay. Nt.BstNBI: Closed-circular DNA substrates (100 ng) were incubated with 1 U of Nt.BstNBI (New England Biolabs) according to the recommendations of the manufacturer. Subsequently, they were purified on a MinElute Spin column (Qiagen) and used in MMR assays.
In vitro MYH nicking assay -100 ng of closed-circular G O /A, A O /G, A O /C or homoduplex substrate were incubated with 10 ng purified recombinant MYH-GST, 10 U of APE1 (New England Biolabs) and 1.5 mM ATP in 1x MMR buffer in a total volume of 10 l. For time course experiments, 5 l aliquots were withdrawn at the indicated time points. The reactions were stopped by the addition of 2 l of 6x loading dye (37.5 mg/ml Ficoll 400, 23% glycerol, 0.03% bromophenol blue) and subsequent heat inactivation. The samples were separated on 1% agarose gels and visualized with GelRed. The nicking efficiency was quantified from the ratio of the amount of open-circular DNA product to the total amount of DNA (closed-circular + open-circular).
In vitro MMR assay in nuclear extracts of human cells -The MMR assays were described previously (7,30,31). Unless otherwise specified, the reactions were carried out with 100 ng substrate and 100 g nuclear cell extract in a total volume of 25 l in a buffer containing 20 mM Tris-HCl pH 7.6, 5 mM MgCl 2 , 1 mM glutathione, 50 g/ml BSA, 0.1 mM dNTPs, 1.5 mM ATP and 80 mM KCl (in experiments involving MYH-addressed substrates) or 110 mM KCl (in experiments involving OGG1-addressed substrates). 1.8 pmol of purified MutL or MutS were added where indicated. The reactions were incubated at 37°C for 1 hour. The reaction was stopped by adding an equal volume of STOP solution (50% Poteinase K, 1 mM EDTA, 3% SDS) and subsequent incubation for 1 h at 50°C. Substrates were purified on MinElute Spin columns (Qiagen), subjected to SalI/DraI or AclI restriction digest and the DNA fragments were separated on 1% agarose gels stained with GelRed. The percentage of mismatch repair was quantified from the ratio of the repaired bands (c+d) to the total amount of DNA [c+d+(c+d)] using ImageQuantTL. The band intensities were corrected according to their respective DNA fragment sizes. Some assays were supplemented with 2 Ci [ -32 P]dATP. To monitor incorporation of the radiolabeled nucleotide, the agarose gels were vacuum dried, exposed to a PhosphoScreen and scanned with a Typhoon Scanner (FLA 9500, GE Healthcare).
In vitro MMR assay in Xenopus laevis egg extracts -Briefly, the reaction mixture containing 150 ng substrate, 26 l S-phase extract and 2 Ci [ -32 P]dATP in a total volume of 30 l was incubated at 23°C for 45 min. To inhibit replication, 500 nM geminin and 40 g/ml p27 were added. The reaction was stopped by the addition of 70 l STOP solution (76 mM EDTA, 1.5% SDS) and 40 g RNase (Sigma-Aldrich) and incubated for 30 min at 37°C. Subsequently, 200 g Proteinase K (AppliChem) were added and incubation continued at 37°C overnight. The substrates were purified on MinElute Spin columns (Qiagen) and subjected to SalI/DraI or AclI restriction digest in the presence of RNase. The digested DNA was cleaned up again and analyzed on a 1% agarose gel stained with GelRed.
Immunodepletions -20 l Protein A/G PLUS-agarose beads (sc-2003, SantaCruz) were washed twice with 750 l binding buffer (30 mM HEPES-KOH pH 7.6, 7 mM MgCl 2 ) and spun down at 2'700 g for 2 min at 4°C. The beads were resuspended in binding buffer and 1 g anti-MYH (ab55551, Abcam) or anti-MSH6 (610919, BD Transduction Laboratories) antibody was added. The mixture was then incubated for 3 h at 4°C, the beads were washed 3x with 750 l binding buffer and subsequently used to immunodeplete 100 g nuclear cell extracts for 30 min at 4°C. MMR assays were performed immediately after depletion. Mock-depleted nuclear cell extracts were obtained by incubation with beads only.
MMR-and/or BER-dependent incorporation assays -This assay was used to test MMR or BER activities. It was performed similarly to the MMR assay, except for the following modifications: In order to track MMR-, MYH-or OGG1-dependent nucleotide incorporation, the reactions were supplemented with 2 µCi [ -32 P]dATP, [ -32 P]dCTP or [ -32 P]dGTP, respectively. Finally, the substrates were digested with NotI/BsaI, and analyzed on a GelRed-stained 1% agarose gel. Repair tracts of up to 330 bp in Nt.BstNBI nicked substrates, up to 29 bp in the G O /A substrates and up to 22 bp in the G O /C substrates are seen in the 1808 bp fragment a. Longer repair tracts appear in the 1387bp fragment b (see Figure 1a). While MMR-dependent [ 32 P]dAMP incorporation gave rise to a strong radioactive signal in both bands, MYH-dependent [ 32 P]dCMP or OGG1dependent [ 32 P]dGMP incorporation occured only in DNA fragment a. Quantification of [ 32 P]dCMP or [ 32 P]dGMP incorporation in MYH-or OGG1-induced BER was determined from the ratio between the band intensities of fragment a on the autoradiographs and on the GelRed-stained agarose gels.
BER assay -To determine A O /G to C/G or A O /C to G/C repair by MYH-dependent BER, the substrates were incubated with the extracts as described in the MMR assay. After the reaction, 50% of the purified, eluted substrate were digested with SalI/DraI (MMR assay), while the other half was used for PvuI digestion (BER assay).

G O /A and G O /C mispairs are not addressed by canonical MMR
BER and MMR are mechanistically distinct biochemical pathways. They differ principally in their substrate specificity and in repair patch size. The specificity of BER is dictated by the enzymes that initiate the process, DNA glycosylases, which recognize and excise a limited number of damaged or modified bases (32). The repair process is thus initiated at the site of the base modification and involves the replacement of one (short-patch BER) (33) or only 2-6 (long-patch BER) nucleotides (34,35). Canonical MMR addresses non-Watson-Crick base pairs that arise during replication, but, in contrast to BER, the excision process initiates at a site distal to the mispair, at a strand discontinuity that marks the nascent DNA strand. The repair tracts can thus be several hundred nucleotides long (36,37). Because we wished to study the potential interplay of BER and MMR in the processing of substrates that could conceivably be addressed by both repair systems, we first needed to learn whether both repair pathways were active in our cell extracts and to establish experimental conditions that would allow us to differentiate between the two processes (Figs. 1A, 2B, 5F). We first tested whether MMR was active in our cell extracts. To this end, we generated covalently-closed circular phagemid substrates containing a single G/T, G O /C or G O /A base pair, or a control substrate containing a G/C at the same site. By subsequent nicking of the substrates at their Nt.Bst.NBI sites, we generated an initiation site for MMR (Fig. 1A). We then incubated the nicked substrates with extracts of MutLdeficient HCT116 human cells (Fig. 1B) that were supplemented with purified recombinant MutL and [ -32 P]dATP. Following recovery of the phagemids, restriction digest with NotI and BsaI, and separation of the fragments on agarose gels, we anticipated that long-patch repair events (MMR) would give rise to heteroduplexes labeled in both a and b fragments (Fig. 1A).
As shown in Fig. 1C, intense radiolabelling was seen only in the bona fide MMR substrate G/T (lane 2). That both fragments a and b were labeled with similar intensity indicates that the repair patch spanned at least the distance of 361 nucleotides between the nick and the mispair. Only background amounts of [ 32 P]dAMP were detected in the a and b fragments of the control homoduplex, G O /A or G O /C substrates (lanes 1, 3-4), which confirmed that the G O /C and G O /A pairs failed to activate nick-dependent long-patch excision and resynthesis characteristic of MMR (38,39) and exemplified in lane 2.

MYH-dependent BER in human nuclear cell extracts
Having shown that MMR was active in the extracts, but that it did not address the G O /A and G O /C substrates, we wanted to see whether they were processed by BER. G O /A repair should be initiated by MYH, which should remove the mispaired adenine. The resulting AP-site should then be incised by apurinic endonuclease 1 (APE1). We therefore asked whether incubation of the supercoiled G O /A substrate with purified, recombinant MYH-GST and APE1 gave rise to nicked circular molecules. As shown in Fig. 2A (lanes 3-6), the G O /A substrate was efficiently converted to the open-circular form upon incubation of the plasmid with the recombinant proteins.
MYH has been reported to excise, in addition to adenine from mispairs with G O , also the oxidation product of adenine, 2hydroxyadenine (A O ), from mispairs with guanine or cytosine (23,24). We therefore included the covalently-closed A O /G and A O /C substrates in this assay. As shown in Fig. 2A, only limited nicking was detected on the A O /G (lanes 7-10) and A O /C (lanes 11-14) substrates, with only 42% or 38% of the plasmids having been nicked after 8 hours, respectively. This was only slightly above the non-specific nicking levels observed on the homoduplex (26%).
We next set out to test BER activity in the extracts. MYH-initiated BER of G O /A should give rise to G O /C. Thus, upon incubation of the covalently-closed G O /A substrate with extracts supplemented with [ -32 P]dCTP, specific incorporation of [ 32 P]dCMP into the a fragment of the substrate (Fig. 2B) should be detectable. This was indeed the case (Fig. 2C, lanes 4-6). As in the case of the purified enzymes, incubation of the A O /G (Fig. 2C, lanes 7-9) or A O /C (Fig. 2D, lanes 4-6) substrates with [ -32 P]dCTP-or [ -32 P]dGTP-supplemented HCT116 extracts, respectively, the amount of radionucleotide incorporated into the A Ocontaining fragment a of the two substrates was only slightly greater than that detected in the homoduplex substrate.

MYH-dependent mismatch repair in human nuclear cell extracts
Having obtained preliminary evidence that both MYH-dependent BER and MMR were active in the extracts, we next wanted to learn whether there was a cross-talk between these processes. To address this question, we deployed a phagemid substrate containing a G/T mismatch within the unique SalI restriction site (G/T; covalently-closed or nicked with Nt.BstNBI 376 nucleotides 3' from the mispaired T), a phagemid containing a single G O /A mispair in addition to the G/T (G O /A-G/T) and a control phagemid G O /A-G/C. Upon incubation with the cell extracts, the recovered phagemids were digested with SalI and DraI to give rise to the pattern of bands shown in Fig. 3A. Successful repair of the G/T mismatch to G/C regenerates the SalI site and the plasmid is cut into 4 instead of 3 fragments (the smallest DraI fragment is ran out of the gel and is therefore not seen). This process should be dependent on the presence of a nick in the strand containing the mispaired T, as canonical MMR requires a strand discontinuity for initiation of the EXO1dependent excision step (40,41).
When the above substrates were incubated with MMR-deficient HCT116 nuclear cell extracts supplemented with [ -32 P]dATP, only background levels of repair and [ 32 P]dAMP incorporation were detected in the bona fide MMR substrate, nicked G/T (Fig. 3B, lane 2). However, the same substrate was repaired with high efficiency when the extracts were supplemented with purified recombinant MutL (lane 3), particularly when compared to the unnicked control G/T plasmid (lane 1), which was processed to a limited extent by non-canonical MMR (42). Strikingly, while only low levels of [ 32 P]dAMP incorporation were detected in this system when the closedcircular G O /A-G/C control substrate was used (lanes 6-7), around 60% of the covalentlyclosed G O /A-G/T phagemid were repaired in the extract supplemented with recombinant MutL (lane 5). This result implied that A opposite G O can serve as a cryptic strand discrimination signal for MMR.
We postulated that the MMR machinery might be hijacking intermediates of G O /A processing for EXO1 loading. We therefore set out to confirm that the above-described phenomenon was dependent on MYH by immunodepleting it from the extract (Fig. 3C) prior to incubation with the substrates. This reduced MMR efficiency on the G O /A-G/T substrate from 52% (lane 7) to ~20% (lane 8). The latter level was comparable to the efficiency of non-canonical MMR on the covalently-closed G/T substrates (~30%) (lanes 1-3). That the observed inhibition was due to MYH depletion was confirmed by complementation of the depleted extracts with purified, recombinant MYH-GST (Fig. 3D), which restored the MMR efficiency on the G O /A-G/T substrate to 43% (Fig. 3C, lane 9). As anticipated, MMR efficiencies on the G/T (lanes 1-3) and nicked G/T (lanes 4-6) substrates were unaffected by the amount of MYH-GST in the extracts.
To confirm that processing of G O /A intermediates by the BER machinery indeed generates DNA termini that can be utilized by EXO1 in the strand degradation step of MMR, we pre-incubated the closed-circular G O /A-G/T substrate with purified, recombinant MYH-GST and APE1 (Fig. 3E, upper panel, lane 3), which cleaves the sugar-phosphate backbone at the MYH-generated abasic site. We then incubated the MYH/APE1-nicked substrate with HCT116 extracts supplemented with MutL . As shown in Fig. 3E (lower panel), the MYH/APE1-generated nick in the G O /A-G/T substrate (lane 3) was efficiently recognized by MMR factors. That the latter substrate was even more efficiently repaired than the Nt.BstNBI-nicked G/T phagemid (lane 2) is most likely indicative of the shorter distance between the nick and the mispair in the G O /A-G/T substrate (54 nt) as compared to the Nt.BstNBI-nicked G/T phagemid (361 nt).
We extended the in vitro MMR assays also to substrates containing an A O /G or an A O /C pair in the vicinity of a T/G mismatch. These substrates differed from those described above, inasmuch as the A O residues were located within a PvuI site (see Experimental Procedures). Successful repair of A O /G to C/G or of A O /C to G/C by BER would restore the restriction site, such that PvuI digestion of the recovered DNA would allow quantification of BER activity at the A O sites ( Figure 3F). As anticipated from the absence of detectable BER activity on these substrates, incubation of the A O /G-T/G and A O /C-T/G substrates with MMR-proficient HCT116 extracts supplemented with recombinant MutL yielded only 15% and 12% PvuI-sensitive products respectively ( Figure 3G, lanes 4, 5 in lower panel), while the G/T phagemid was completely digested under identical conditions (lane 3). These results were reflected in MMR efficiency determined from the same assays by digestion of the recovered phagemids with SalI/DraI (Fig. 3G, upper panel). Repair of the T/G mismatch in the A O /G-T/G and A O /C-T/G substrates was very inefficient (20% and 18%, respectively), while the nicked G/T (lane 3) and the covalently-closed G O /A-G/T (lane 1) substrates were efficiently repaired (93% and 59%, respectively). Taken together, these data show that A O /G or A O /C pairs are inefficiently processed by MYH in cell extracts and therefore fail to act as entry sites for mismatchactivated excision.

G O /A mispairs act as MMR initiation sites also in X. laevis egg extracts
Because G O /A mispairs arise during replication, MYH should be able to act on newly-replicated DNA and, indeed, available experimental evidence shows that this glycosylase is more abundant during S-phase (43). In an attempt to learn whether MYH is also more efficient during S-phase, we decided to make use of Xenopus laevis (X. laevis) egg extracts, which are enriched in S-phase proteins (44).
In the first experiment, we wanted to learn whether the extracts were proficient in MYHinitiated BER. We decided to make use of the assay described above (Fig. 2B), in which the homoduplex or G O /A closed-circular substrates were incubated with X. laevis egg extracts supplemented with [ -32 P]dCTP. Digestion of the recovered phagemid DNA with NotI/BsaI revealed a three-fold greater incorporation of [ 32 P]dCMP into the G O -containing fragment of the G O /A substrate than into the corresponding fragment of the homoduplex phagemid (Fig.  4A). This demonstrated that MYH-dependent BER was active in the X. laeavis extracts.
We also had to show that the extracts supported nick-directed MMR, because an earlier report indicated that mismatch processing in X. laevis oocyte extracts was efficient, but not nick-directed (45). We therefore incubated the covalently-closed and the nicked isoforms of the G/T phagemid with the extracts supplemented with [ -32 P]dATP. As shown in Fig. 4B, we were able to detect nick-directed MMR on the G/T substrate (lane 2), whereas the covalently-closed phagemid was only inefficiently processed (lane 1). Under these conditions, the covalently-closed G O /A-G/T substrate (lane 3) was almost as efficiently repaired as the nicked G/T phagemid (lane 2). Since extracts of X. laevis eggs support plasmid replication under certain conditions, which would result in considerable [ 32 P]dAMP incorporation and conversion of 50% of the G/T mismatches to G/C in the absence of MMR, we carried out a control experiment in extracts supplemented with the DNA replication inhibitors p27 and geminin. No detectable differences between the levels of radionucleotide incorporation in the presence and absence of these inhibitors were observed, which indicated that the plasmid did not replicate in the assay (cf lanes 4 and 5).

OGG1-dependent mismatch repair is inefficient in human nuclear cell and X. laevis egg extracts
Effect of oxidative damage processing on MMR directionality Assuming that MTH1 hydrolyzes oxidized dGTP in the nucleotide precursor pool with a 100% efficiency, all G O residues in oxidized DNA should be in the template strand following replication, irrespective of whether they are paired with A or C. As discussed above, MYH-initiated BER of G O /A mispairs arising during replication would be directed to the A-strand that is also the nascent DNA strand. In contrast, BER-mediated repair of G O /C pairs arising during replication would have the opposite effect on MMR, because OGG1-dependent G O /C repair would introduce breaks in the template DNA strand and thus provide MMR with an incorrect strand bias.
We set out to test the above hypothesis by studying the effect on MMR efficiency and directionality of a G O /C pair situated in the vicinity of a G/T mismatch. We generated a G O /C-G/T substrate, in which the G O was positioned 54 nucleotides 5' from the mispaired G. OGG1-initiated BER of the G O would give rise to a break that should activate MMR to correct the G/T mismatch to A/T and thus regenerate an AclI site in the substrate. This repair bias should be identical to that introduced by a Nt.Bst.NBI-generated nick in the G/T substrate (Fig. 5A).
When we incubated the nicked G/T substrate, or the covalently-closed G/T, G O /C-G/T or G O /C-A/T phagemid heteroduplexes with HCT116 nuclear cell extracts supplemented with [ -32 P]dATP, no significant repair or [ 32 P]dAMP incorporation was observed (Fig. 5B, lanes 1, 3, 5, 7), but when the extract was supplemented with purified, recombinant MutL , the Nt.BstNBI-nicked G/T substrate was repaired (lane 4). In contrast to what was observed with the G O /A-G/T substrate (Fig. 3B, lane 5), only background levels of repair and [ 32 P]dAMP incorporation were detected with the G O /C-G/T substrate (Fig. 5B, lane 6) and with the control, covalently-closed G/T phagemid (lane 2). Similar results were obtained also with MutSdeficient LoVo extracts supplemented with purified recombinant MutS (Fig. 5C, lanes 1-3). Given that both extracts contained the key MMR and BER factors, as ascertained by western blotting (Figs. 1B, 5D), these results at first suggested that G O processing failed to activate MMR on the G O /C-G/T substrate.
The above result could be explained in several ways. We first considered the possibility that the MMR machinery was unable to use the BER intermediates generated during G O processing as strand discrimination signals, because OGG1 is a DNA glycosylase/lyase that generates a strand break on the 3' side of the abasic site (46,47). APE1mediated cleavage at its 5' side then generates a single nucleotide gap, rather than just a break at the 5' side of the abasic site as is the case for MYH/APE1. We therefore treated the covalently-closed G O /C-G/T substrate with purified recombinant OGG1-GST fusion protein. As shown in Fig. 5E (lane 2), the substrate was as efficiently nicked as the G/T control with Nt.BstNBI (lane 1). When the OGG1-nicked phagemid was incubated with extracts of LoVo cells supplemented with purified recombinant MutS , it was repaired with similar efficiency to the control, nicked G/T phagemid (Fig. 5C, cf lanes 2&4). This shows that MMR can use OGG1-generated strand breaks as initiation sites.
The second possibility was that OGG1 in the tested cell extracts was inactive, or that it was present in insufficient amounts. We therefore incubated the G O /C and homoduplex substrates with LoVo extracts supplemented with [ -32 P]dGTP to test for OGG1-dependent BER, which should result in [ 32 P]dGMP incorporation into fragment a (Fig. 5F). As shown in Fig. 5G, only background levels of the radiolabeled nucleotide were incorporated into fragment a of the phagemid heteroduplex, which suggested that G O /C repair was indeed inefficient, despite the fact that the extracts contained readily-detectable amounts of OGG1 (Figs. 1B, 5D).
Similar results were obtained with [ 32 P]-labeled G O /C-containing oligomers, which were inefficiently processed in LoVo extracts, while processing of U/Gcontaining oligomers was very efficient. Moreover, purified, recombinant OGG1 was able to process the G O /C oligonucleotide substrate, with ~50% cleavage of the G O strand seen after only 10 min (data not shown).
To test whether the LoVo extracts contained an inhibitor of OGG1, we supplemented them with an amount of purified recombinant OGG1-GST that was comparable to that of the endogenous protein present in the nuclear extracts (Fig. 5H). Under these conditions, the G O /C-G/T substrate (Fig. 5I, lane 3) was repaired with efficiency similar to the nicked G/T phagemid (lane 1). Importantly, the above-described phenomenon was not limited to extracts of human cells, as identical results were obtained with the X. laevis MMR system (Fig. 6). The reason underlying the low OGG1 activity in the extracts is currently unknown, but we were able to eliminate inappropriate salt concentration, lack of an activator protein in the cytoplasmic fraction, or short half-life of OGG1 in our assay as possible causes (data not shown).

DISCUSSION
The postreplicative mismatch repair system improves the fidelity of DNA replication by several orders of magnitude through removing from nascent DNA nucleotides that fail to form Watson-Crick base pairs. In order to fulfill this function, it has to satisfy two key criteria: it has to (i) recognize base-base mismatches and small insertion/deletion loops (IDLs) generated by the replicative polymerases during DNA synthesis and (ii) direct the repair process to the newly-synthesized DNA strand. How the major mismatch recognition factor MutS recognizes the different helical distortions that are caused by purine/purine, purine/pyrimidine and pyrimidine/pyrimidine mispairs, as well as by IDLs, is still poorly understood, despite the fact that several structures of protein/DNA complexes exist (48). However, because MMR deficiency leads to transition, transversion and frameshift mutations, MutS clearly has broad substrate specificity. Although this characteristic of MutS is beneficial as far as its role in the maintenance of replication fidelity is concerned, it might also be deleterious, should MutS bind to lesions that ought to be processed by other repair systems. One such example are mispairs containing O 6methylguanine; these lesions activate MMR, but because their processing does not lead to the removal of the modified nucleotide from the template strand, they trigger futile repair that eventually leads to cell death (1). Processing of G O /A mismatches arising through the incorporation of dAMP opposite G O in the template strand might have also been expected to trigger futile MMR, but the in vivo evidence for such a process is currently lacking and in vitro data showing that G O /A mispairs are very poorly addressed by the MMR system in human cell extracts [this study and (39)] would appear to argue against it. However, in vivo, the MMR system appears to remove from the nascent strand G O misincorporated opposite template A (49,50), and to increase the efficiency of MYH-dependent BER, at least in vitro (51). Thus, available experimental evidence suggests that MMR supports BER during oxidative DNA damage processing. In the present study, we asked whether the reverse might also be true, namely, whether BERdependent processing of G O affects MMR efficiency.
In our earlier studies, we could show that the MMR system could use breaks generated during the BER-mediated processing of uracil residues (7,42) or during the RNaseH2mediated excision of ribonucleotides misincorporated into DNA during replication (9) as strand discrimination signals. We postulated that MMR might also use breaks generated during the BER-dependent processing of oxidative damage. However, the outcome of such interference would be positive only in the case that the breaks were generated in the nascent DNA strand. We show here that MYH-dependent processing of a G O /A mispair in extracts of human cells or of Xenopus laevis eggs directs MMR to the A-strand (Figs. 3, 4). In vivo, the ability of the MMR system to use breaks generated during the MYH-directed processing of G O /A mispairs arising through the incorporation of dAMP opposite G O would direct mismatch correction to the nascent strand and thus improve the fidelity of replication (Fig. 7, right panel). MYHdependent BER and MMR would thus synergize during S-phase, as might have been anticipated from the physical interaction of some of the subunits of these pathways (51). It could be argued that the number of G O /A mispairs arising in the vicinity of a replication error might be too low to affect MMR efficiency. This is likely, however, this number increases substantially when strand discontinuities arising during the processing of other "non-standard" nucleotides (uracils and thymines arising through deamination of cytosine and 5-methylcytosine respectively, ribonucleotides, methylpurines, thymine glycols etc.) are considered. Indeed, as we showed previously, ribonucleotide processing alone is sufficient to affect MMR efficiency in vivo (9).
Breaks introduced into the nascent DNA strand by OGG1 during the processing of C/G O pairs arising through the incorporation of dG O MP opposite template C would also help improve MMR efficiency and replication fidelity. However, due to the sanitization of dNTP pools by MTH1, dG O TP concentrations should be extremely low, such that dG O MP incorporation into nascent DNA during replication should be minimal (Fig. 7, left  panel). As a result, most G O residues present in DNA during replication should be in the template strand. Incision of this strand by OGG1 behind the replication fork would misdirect MMR to the wrong strand (Fig. 7, right panel), and OGG1-catalyzed incision of DNA at G O /C sites in front of the replication fork could cause its collapse. It would therefore appear logical that G O /C processing should be completed prior to the onset of S-phase and that OGG1 should be inactivated during this cell cycle stage. Interestingly, we found that G O /C processing in both in vitro systems was inefficient and that the presence of a G O /C pair in the heteroduplex substrate failed to activate MMR of the G/T mismatch (Figs. 5, 6), despite the fact that OGG1 was present in considerable quantities (Fig. 1B, 5D,H). Because addition of the recombinant polypeptide to the extracts resulted in G O -dependent G/T repair (Fig. 5I), we postulate that OGG1 present in the cell extracts is inactive, either as a result of posttranslational modifications, or through complexation with an inhibitor. We are currently attempting to identify the underlying cause of this inhibition, as well as carrying out a series of in vivo experiments that should show whether the observations described above correspond to the situation in living cells. Our current findings indicate that the repair of oxidative damage in vertebrate cells is highly regulated in order to prevent genomic instability and future experiments should show how this regulation is mediated at the molecular level.    substrates were incubated with Xenopus laevis egg extracts. Upon SalI/DraI digestion of the recovered phagemids, MMR efficiencies were estimated from the GelRed-stained agarose gels, while MMR-dependent DNA synthesis was visualized by [ 32 P]dAMP incorporation seen in the autoradiograph. A control reaction (lane 5) contained the replication inhibitors p27 and geminin; since [ 32 P]dAMP incorporation was similar in the presence and absence of these inhibitors (cf lanes Effect of oxidative damage processing on MMR directionality 4 and 5), the plasmid was apparently not replicated in the extracts (52). The indicated MMR efficiencies (%) represent an average of 3 independent experiments +/-SD.

FIGURE 5. A G O /C base pair does not act as an initiation site for MMR in nuclear extracts of human cells. (A)
Schematic representation of the G O /C-G/T substrate used in the in vitro MMR assay. The circular heteroduplex substrate carries a G O /C base pair 54 nucleotides from a G/T mismatch in the recognition site of AclI endonuclease. The positions of two further AclI cleavage sites and the Nt.BstNBI site, where a nick can be introduced selectively into the outer strand, are indicated. In the absence of repair, digestion of the phagemid with AclI gives rise to fragments of 2824 (c+d) and 373 bp (e). Repair of the G/T mismatch to A/T regenerates a third AclI restriction site, such that the phagemid DNA is cleaved into 3 fragments of 1518 (c), 1306 (d), and 373 bp (e).   Should dG O MP be incorporated into the nascent strand (red) opposite C, it could be removed by OGG1-dependent BER or by MMR without deleterious consequences. In contrast, should A template /G O nascent mispairs arise during replication, MYH-mediated BER to C/G O would give rise to A to C mutations, whereas MMR would process the G O -containing strand and prevent mutagenesis. In cases where the oxidized base was in the vicinity of a replication error (e.g. a G/T mispair), BER intermediates of C template /G O nascent processing would direct MMR to the correct strand, whereas MYH-initiated BER of A template /G O nascent would misdirect MMR to the parental, template strand (black). However, the likelihood of incorporation of dG O MP into the nascent DNA strand is minimized by hydrolysis of dG O TP in the nucleotide pool by MTH1. Most G O residues in the DNA should therefore be in the template strand, as shown in the right panel. Right panel: BER of G O /C pairs would lead to the removal of the oxidized base, but, during S-phase, the OGG1-generated strand break might result in replication fork collapse, or in the misdirection of MMR to the template strand. Processing of G O template /A nascent mispairs, be it by MYH-dependent BER or by MMR would have no deleterious consequences. Moreover, in cases where the oxidized base was in the vicinity of a replication error (e.g. a G/T mispair), BER intermediates of G O template /A nascent processing would direct MMR to the correct strand.