Substrate specificity and sequence preference of G:T mismatch repair: incision at G:T,O6-methylguanine:T,and G:U mispairs in DNA by human cell extracts

Extracts of two human glioma cell lines (lacking O6-methylguanine DNA-methyltransferase) (i.e., A1235 and its alkylation-resistant derivative A1235-MR4) were examined for their ability to execute strand incision at different base mismatches in model (45-bp) DNA. These heteroduplex substrates were of the same sequence except for the presence, at the same site, of one of three mispairs: G:T, O6-methylguanine:T (m6G:T), and G:U. The parental (A1235) extract, when supplemented with ATP and human thymine DNA glycosylase (TDG), acted proficiently on all three substrates, incising immediately 5' to the mismatched thymine or uracil residue. In contrast, the derivative extract, under the same conditions, recognized only the G:U substrate. The activity of the A1235 extract toward the G:T (or m6G:T) substrate was markedly reduced in the absence of ATP, whereas the G:U substrate was incised rapidly by both extracts irrespective of the addition of ATP. These combined data confirm and extend our earlier findings demonstrating that human cells possess two G:T incision activities, one efficient and ATP-dependent and the other inefficient and ATP-independent. The derivative extract lacks the former activity but retains the latter activity. In substrate competition assays, the G:U substrate inhibited the ATP-dependent G:T incision activity to a greater extent than did the G:T substrate itself. Given the well-known substrate preference of TDG for G:U as compared to G:T, this unexpected result implies that TDG may be an integral component of the ATP-dependent G:T incision machinery in human cells. Finally, the base 5' to the mismatched G in the G:T mispair conferred sequence preference on the A1235 extract in the presence of ATP and TDG, with a pyrimidine (especially cytosine) being much favored over a purine. This latter observation suggests that the ATP-dependent G:T incision activity is designed to repair deaminated 5-methycytosine lesions in CpG islands, the methylation of which is linked to control of gene expression.     

Incision at O6-methylguanine:thymine mispairs in DNA by extracts of human cells.

Human cell-free extracts were used to detect activities specifically incising O6-methylguanine (m6G) paired with C or T in DNA. A 45-bp double-stranded DNA containing one m6G across from a T (m6G:T) was the test substrate. Extracts from glioblastoma cell lines A172 and A1235 (lacking the m6G-specific repair protein m6G-DNA methyltransferase, MGMT) and colon carcinoma cell line HT29, containing MGMT, showed incision activities specific for the T strand of m6G:T [and G:T, as reported previously by Wiebauer and Jiricny (1989)] substrates, but did not cleave m6G:C (or G:C) substrates. Competition experiments showed that the activity was similar to, if not identical with, the activity in human cells that incises G:T mismatches. The incision sites were similar to those recognized by human G:T- or G:A-specific mismatch enzymes, i.e., the phosphodiester bonds both 3' and 5' to the poorly matched T, suggesting the glycolytic removal of the poorly matched T followed by backbone incisions by class I or II AP endonucleases. Three experiments in which MGMT was inactivated showed that the m6G:T incision activity was not simply due to a two-step mechanisms in which MGMT would first mediate conversion of the m6G:T substrate to a G:T substrate which would serve as a substrate for G:T incision. Extracts from HT29 contained a DNA-binding factor, possibly DNA sequence-specific, that inhibited incision of the m6G:T (but not the G:T) substrate, that was removed by the addition of synthetic DNA to the reaction.     

Multiplicity of strand incision at G:T base mismatches in DNA by human cell extracts

Cell extract from the HT29 human colon carcinoma cell line (lacking mutator phenotype) was used to study the ATP-dependent G:T mismatch repair. We found that when a 45-bp (model) DNA with a single CpG/TpG mispair was incubated with the cell extract and ATP, it was incised immediately 5' and 3' to the mismatched T, and we noted that the actual 5'- and 3'-labeled fragments were similar to the cleaved products of thymine DNA glycosylase (TDG). This TDG-like cleavage product was enhanced (5-fold) with stimulation of several novel fragments, as inferred from the effect on incision at CpG/TpG site of the addition of G:U competitor DNA and ATP to the HT29 extract. The novel fragments were compatible with a strand incision on both sides of the mismatch (the third phosphodiester bond 5' and the second phosphodiester bond 3' to the mismatched T) and an incision 3' to the mismatched T, respectively. This suggests that while the ATP-dependent (TDG-like) incision activity, contrary to expectation, shows a lack of substrate competition, its catalytic property is likely modified by an interaction with G:U mispair. These multiple ATP-dependent incision events were not detected when extracts of the mismatch repair (MMR) defective HCT15 or HCT116 cell line were augmented with ATP and G:U. We postulate that these multiple ATP-dependent incision events possibly require the same MMR factors, and together they constitute a modified single ATP-dependent G:T incision activity. This activity toward the CpG/TpG was competitively inhibited by a 45-bp DNA with an ApG/TpT mispair; incision at a single site 5' to the latter mismatch compares with one of the multiple sites incised 5' to the former mismatch. These results suggest that one of several mismatch-incision factors is required by the human ATP-dependent G:T incision activity, in addition to MMR factors and ATP.     

Kinetics and thermodynamics of BI-BII interconversion altered by T:G mismatches in DNA

T:G mismatches in DNA result in humans primarily from deamination of methylated CpG sites. They are repaired by redundant systems, such as thymine DNA glycosylase (TDG) and methyl-binding domain enzyme (MBD4), and maintenance of these sites has been implicated in epigenetic processes. The process by which these enzymes identify a canonical DNA base in the incorrect basepairing context remains a mystery. However, the conserved contacts of the repair enzymes with the DNA backbone suggests a role for protein-phosphate interaction in the recognition and repair processes. We have used ³¹P NMR to investigate the energetics of DNA backbone BI-BII interconversion, and for this work have focused on alterations to the activation barriers to interconversion and the effect of a mismatch compared with canonical DNA. We have found that alterations to the ΔG of interconversion for T:G basepairs are remarkably similar to U:G basepairs in the form of stepwise differences in ΔG of 1–2 kcal/mol greater than equivalent steps in unmodified DNA, suggesting a universality of this result for TDG substrates. Likewise, we see perturbations to the free energy (～1 kcal/mol) and enthalpy (2–5 kcal/mol) of activation for the BI-BII interconversion localized to the phosphates flanking the mismatch. Overall our results strongly suggest that the perturbed backbone energetics in T:G basepairs play a significant role in the recognition process of DNA repair enzymes.     

DNA-substrate sequence specificity of human G:T mismatch repair activity.

G:T mispairs in DNA originate spontaneously via deamination of 5-methylcytosine. Such mispairs are restored to normal G:C pairs by both E. coli K strains and human cells. In this study we have analyzed the repair by human cell extracts of G:T mismatches in various DNA contexts. We performed two sets of experiments. In the first, repair was sequence specific in that G:T mispairs at CpG sites at four different CpG sites were repaired, but a G:T mismatch at a GpG site was not. Cytosine hemimethylation did not block repair of a substrate containing a CpG/GpT mismatch. In the second set of experiments, substrates with a G:T mismatch at a fixed position were constructed with an A, T, G, or C 5' to the mismatched G, and alterations in the complementary strand to allow otherwise perfect Watson-Crick pairing. All were incised just 5' to the mismatched T and competed for repair incision with a G:T substrate in which a C was 5' to the mismatched G. Thus human G:T mismatch activity shows sequence specificity, incising G:T mismatched pairs at some DNA sites, but not at others. At an incisable site, however, incision is little influenced by the base 5' to the mismatched G.     

Preferential CEBP binding to T:G mismatches and increased C-to-T human somatic mutations

DNA cytosine methylation in mammals modulates gene expression and chromatin accessibility. It also impacts mutation rates, via spontaneous oxidative deamination of 5-methylcytosine (5mC) to thymine. In most cases the resulting T:G mismatches are repaired, following T excision by one of the thymine DNA glycosylases, TDG or MBD4. We found that C-to-T mutations are enriched in the binding sites of CCAAT/enhancer binding proteins (CEBP). Within a CEBP site, the presence of a T:G mismatch increased CEBPβ binding affinity by a factor of >60 relative to the normal C:G base pair. This enhanced binding to a mismatch inhibits its repair by both TDG and MBD4 in vitro. Furthermore, repair of the deamination product of unmethylated cytosine, which yields a U:G DNA mismatch that is normally repaired via uracil DNA glycosylase, is also inhibited by CEBPβ binding. Passage of a replication fork over either a T:G or U:G mismatch, before repair can occur, results in a C-to-T mutation in one of the daughter duplexes. Our study thus provides a plausible mechanism for accumulation of C-to-T human somatic mutations.     

Excision of 5-hydroxymethyluracil and 5-carboxylcytosine by the thymine DNA glycosylase domain: its structural basis and implications for active DNA demethylation

The mammalian thymine DNA glycosylase (TDG) is implicated in active DNA demethylation via the base excision repair pathway. TDG excises the mismatched base from G:X mismatches, where X is uracil, thymine or 5-hydroxymethyluracil (5hmU). These are, respectively, the deamination products of cytosine, 5-methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC). In addition, TDG excises the Tet protein products 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC) but not 5hmC and 5mC, when paired with a guanine. Here we present a post-reactive complex structure of the human TDG domain with a 28-base pair DNA containing a G:5hmU mismatch. TDG flips the target nucleotide from the double-stranded DNA, cleaves the N-glycosidic bond and leaves the C1′ hydrolyzed abasic sugar in the flipped state. The cleaved 5hmU base remains in a binding pocket of the enzyme. TDG allows hydrogen-bonding interactions to both T/U-based (5hmU) and C-based (5caC) modifications, thus enabling its activity on a wider range of substrates. We further show that the TDG catalytic domain has higher activity for 5caC at a lower pH (5.5) as compared to the activities at higher pH (7.5 and 8.0) and that the structurally related Escherichia coli mismatch uracil glycosylase can excise 5caC as well. We discuss several possible mechanisms, including the amino-imino tautomerization of the substrate base that may explain how TDG discriminates against 5hmC and 5mC.     

Incorporation of dUTP does not mediate mutation of A:T base pairs in Ig genes in vivo

Activation-induced cytidine deaminase (AID) protein initiates Ig gene mutation by deaminating cytosines, converting them into uracils. Excision of AID-induced uracils by uracil-N-glycosylase is responsible for most transversion mutations at G:C base pairs. On the other hand, processing of AID-induced G:U mismatches by mismatch repair factors is responsible for most mutation at Ig A:T base pairs. Why mismatch processing should be error prone is unknown. One theory proposes that long patch excision in G1-phase leads to dUTP-incorporation opposite adenines as a result of the higher G1-phase ratio of nuclear dUTP to dTTP. Subsequent base excision at the A:U base pairs produced could then create non-instructional templates leading to permanent mutations at A:T base pairs (1). This compelling theory has remained untested. We have developed a method to rapidly modify DNA repair pathways in mutating mouse B cells in vivo by transducing Ig knock-in splenic mouse B cells with GFP-tagged retroviruses, then adoptively transferring GFP⁺ cells, along with appropriate antigen, into primed congenic hosts. We have used this method to show that dUTP-incorporation is unlikely to be the cause of AID-induced mutation of A:T base pairs, and instead propose that A:T mutations might arise as an indirect consequence of nucleotide paucity during AID-induced DNA repair.     

Human DNA glycosylase enzyme TDG repairs thymine mispaired with exocyclic etheno-DNA adducts

Lipid peroxidation directly reacts with DNA and produces various exocyclic etheno-base DNA adducts, some of which are considered to contribute to carcinogenesis. However, the system for repairing them in humans is largely unknown. We hypothesized that etheno-DNA adducts are repaired by base excision repair initiated by DNA glycosylase. To test this hypothesis, we examined the activities of the DNA glycosylase proteins OGG1, SMUG1, TDG, NEIL1, MUTYH, NTH1, MPG, and UNG2 against double-stranded oligonucleotides containing 1,N⁶-ethenoadenine (εA), 3,N⁴-ethenocytosine (εC), butanone-ethenocytosine (BεC), butanone-ethenoguanine (BεG), heptanone-ethenocytosine (HεC), or heptanone-ethenoguanine (HεG) using a DNA cleavage assay. We found that TDG is capable of removing thymine that has mispaired with εC, BεC, BεG, HεC, or HεG in vitro. We next examined the effect of TDG against etheno-DNA adducts in human cells. TDG-knockdown cells exhibited the following characteristics: (a) higher resistance to cell death caused by the induction of etheno-DNA adducts; (b) lower repair activity for εC; and (c) a modest acceleration of mutations caused by εC, compared with the rate in control cells. All these characteristics suggest that TDG exerts a repair activity against etheno-DNA adducts in human cells. These results suggest that TDG has novel repair activities toward etheno-DNA adducts.     

Thymine DNA Glycosylase Is a CRL4Cdt2 Substrate

The E3 ubiquitin ligase CRL4^Cdt2 targets proteins for destruction in S phase and after DNA damage by coupling ubiquitylation to DNA-bound proliferating cell nuclear antigen (PCNA). Coupling to PCNA involves a PCNA-interacting peptide (PIP) degron motif in the substrate that recruits CRL4^Cdt2 while binding to PCNA. In vertebrates, CRL4^Cdt2 promotes degradation of proteins whose presence in S phase is deleterious, including Cdt1, Set8, and p21. Here, we show that CRL4^Cdt2 targets thymine DNA glycosylase (TDG), a base excision repair enzyme that is involved in DNA demethylation. TDG contains a conserved and nearly perfect match to the PIP degron consensus. TDG is ubiquitylated and destroyed in a PCNA-, Cdt2-, and PIP degron-dependent manner during DNA repair in Xenopus egg extract. The protein can also be destroyed during DNA replication in this system. During Xenopus development, TDG first accumulates during gastrulation, and its expression is down-regulated by CRL4^Cdt2. Our results expand the group of vertebrate CRL4^Cdt2 substrates to include a bona fide DNA repair enzyme.     

Role of Two Strictly Conserved Residues in Nucleotide Flipping and N-Glycosylic Bond Cleavage by Human Thymine DNA Glycosylase

Thymine DNA glycosylase (TDG) promotes genomic integrity by excising thymine from mutagenic G·T mismatches arising by deamination of 5-methylcytosine, and follow-on base excision repair enzymes restore a G·C pair. TDG cleaves the N-glycosylic bond of dT and some other nucleotides, including 5-substituted 2′-deoxyuridine analogs, once they have been flipped from the helix into its active site. We examined the role of two strictly conserved residues; Asn¹⁴⁰, implicated in the chemical step, and Arg²⁷⁵, implicated in nucleotide flipping. The N140A variant binds substrate DNA with the same tight affinity as wild-type TDG, but it has no detectable base excision activity for a G·T substrate, and its excision rate is vastly diminished (by ∼10^4.4-fold) for G·U, G·FU, and G·BrU substrates. Thus, Asn¹⁴⁰ does not contribute substantially to substrate binding but is essential for the chemical step, where it stabilizes the transition state by ∼6 kcal/mol (compared with 11.6 kcal/mol stabilization provided by TDG overall). Our recent crystal structure revealed that Arg²⁷⁵ penetrates the DNA minor groove, filling the void created by nucleotide flipping. We found that the R275A and R275L substitutions weaken substrate binding and substantially decrease the base excision rate for G·T and G·BrU substrates. Our results indicate that Arg²⁷⁵ promotes and/or stabilizes nucleotide flipping, a role that is most important for target nucleotides that are relatively large (dT and bromodeoxyuridine) and/or have a stable N-glycosylic bond (dT). Arg²⁷⁵ does not contribute substantially to the binding of TDG to abasic DNA product, and it cannot account for the slow product release exhibited by TDG.     

Unique mutational profile associated with a loss of TDG expression in the rectal cancer of a patient with a constitutional PMS2 deficiency

Cells with DNA repair defects have increased genomic instability and are more likely to acquire secondary mutations that bring about cellular transformation. We describe the frequency and spectrum of somatic mutations involving several tumor suppressor genes in the rectal carcinoma of a 13-year-old girl harboring biallelic, germline mutations in the DNA mismatch repair gene PMS2. Apart from microsatellite instability, the tumor DNA contained a number of C:G→T:A or G:C→A:T transitions in CpG dinucleotides, which often result through spontaneous deamination of cytosine or 5-methylcytosine. Four DNA glycosylases, UNG2, SMUG1, MBD4 and TDG, are involved in the repair of these deamination events. We identified a heterozygous missense mutation in TDG, which was associated with TDG protein loss in the tumor. The CpGs mutated in this patient's tumor are generally methylated in normal colonic mucosa. Thus, it is highly likely that loss of TDG contributed to the supermutator phenotype and that most of the point mutations were caused by deamination of 5-methylcytosine to thymine, which remained uncorrected owing to the TDG deficiency. This case provides the first in vivo evidence of the key role of TDG in protecting the human genome against the deleterious effects of 5-methylcytosine deamination.     

A DNA repair process in Escherichia coli corrects U:G and T:G mismatches to C:G at sites of cytosine methylation

Escherichia coli contains a base mismatch correction system called VSP repair that is known to correct T:G mismatches to C:G when they occur in certain sequence contexts. The preferred sequence context for this process is the site for methylation by the E. coli DNA cytosine methylase (Dcm). For this reason, VSP repair is thought to counteract potential mutagenic effects of deamination of 5-methylcytosine to thymine. We have developed a genetic reversion assay that quantitates the frequency of C to T mutations at Dcm sites and the removal of such mutations by DNA repair processes. Using this assay, we have studied the repair of U: G mismatches in DNA to C: G and have found that VSP repair is capable of correcting these mismatches. Although VSP repair substantially affects the reversion frequency, it may not be as efficient at correcting U: G mismatches as the uracil DNA glycosylase-mediated repair process.     

The versatile thymine DNA-glycosylase: a comparative characterization of the human,Drosophila and fission yeast orthologs

Human thymine-DNA glycosylase (TDG) is well known to excise thymine and uracil from G·T and G·U mismatches, respectively, and was therefore proposed to play a central role in the cellular defense against genetic mutation through spontaneous deamination of 5-methylcytosine and cytosine. In this study, we characterized two newly discovered orthologs of TDG, the Drosophila melanogaster Thd1p and the Schizosaccharomyces pombe Thp1p proteins, with an objective to address the function of this subfamily of uracil-DNA glycosylases from an evolutionary perspective. A systematic biochemical comparison of both enzymes with human TDG revealed a number of biologically significant facts. (i) All eukaryotic TDG orthologs have broad and species-specific substrate spectra that include a variety of damaged pyrimidine and purine bases; (ii) the common most efficiently processed substrates of all are uracil and 3,N4- ethenocytosine opposite guanine and 5-fluorouracil in any double-stranded DNA context; (iii) 5-methylcytosine and thymine derivatives are processed with an appreciable efficiency only by the human and the Drosophila enzymes; (iv) none of the proteins is able to hydrolyze a non-damaged 5′-methylcytosine opposite G; and (v) the double strand and mismatch dependency of the enzymes varies with the substrate and is not a stringent feature of this subfamily of DNA glycosylases. These findings advance our current view on the role of TDG proteins and document that they have evolved with high structural flexibility to counter a broad range of DNA base damage in accordance with the specific needs of individual species.     

Separating substrate recognition from base hydrolysis in human thymine DNA glycosylase by mutational analysis

Human thymine DNA glycosylase (TDG) was discovered as an enzyme that can initiate base excision repair at sites of 5-methylcytosine- or cytosine deamination in DNA by its ability to release thymine or uracil from G.T and G.U mismatches. Crystal structure analysis of an Escherichia coli homologue identified conserved amino acid residues that are critical for its substrate recognition/interaction and base hydrolysis functions. Guided by this revelation, we performed a mutational study of structure function relationships with the human TDG. Substitution of the postulated catalytic site asparagine with alanine (N140A) resulted in an enzyme that bound mismatched substrates but was unable to catalyze base removal. Mutation of Met-269 in a motif with a postulated role in protein-substrate interaction selectively inactivated stable binding of the enzyme to mismatched substrates but not so its glycosylase activity. These results establish that the structure function model postulated for the E. coli enzyme is largely applicable to the human TDG. We further provide evidence for G.U being the preferred substrate of TDG, not only at the mismatch recognition step of the reaction but also in base hydrolysis, and for the importance of stable complementary strand interactions by TDG to compensate for its comparably poor hydrolytic potential.     

Dependence of substrate binding and catalysis on pH, ionic strength, and temperature for thymine DNA glycosylase: Insights into recognition and processing of G·T mispairs

Repair of G·T mismatches arising from deamination of 5-methylcytosine (m(5)C) involves excision of thymine and restoration of a G·C pair via base excision repair (BER). Thymine DNA glycosylase (TDG) is one of two mammalian enzymes that can specifically remove thymine from G·T mispairs. While TDG can excise other bases, it maintains stringent specificity for a CpG context, suggesting deaminated m(5)C is an important biological substrate. Recent studies reveal TDG is essential for embryogenesis; it helps to maintain an active chromatin complex and initiates BER to counter aberrant de novo CpG methylation, which may involve excision of actively deaminated m(5)C. The relatively weak G·T activity of TDG has been implicated in the hypermutability of CpG sites, which largely involves C→T transitions arising from m(5)C deamination. Thus, it is important to understand how TDG recognizes and process substrates, particularly G·T mispairs. Here, we extend our detailed studies of TDG by examining the dependence of substrate binding and catalysis on pH, ionic strength, and temperature. Catalytic activity is relatively constant for pH 5.5-9, but falls sharply for pH>9 due to severely weakened substrate binding, and, potentially, ionization of the target base. Substrate binding and catalysis diminish sharply with increasing ionic strength, particularly for G·T substrates, due partly to effects on nucleotide flipping. TDG aggregates rapidly and irreversibly at 37°C, but can be stabilized by specific and nonspecific DNA. The temperature dependence of catalysis reveals large and unexpected differences for G·U and G·T substrates, where G·T activity exhibits much steeper temperature dependence. The results suggest that reversible nucleotide flipping is much more rapid for G·T substrates, consistent with our previous findings that steric effects limit the active-site lifetime of thymine, which may account for the relatively weak G·T activity. Our findings provide important insight into catalysis by TDG, particularly for mutagenic G·T mispairs.     

In vitro studies of DNA mismatch repair proteins

The ability to monitor and characterize DNA mismatch repair activity in various mammalian cells is important for understanding mechanisms involved in mutagenesis and tumorigenesis. Since mismatch repair proteins recognize mismatches containing both normal and chemically altered or damaged bases, in vitro assays must accommodate a variety of mismatches in different sequence contexts. Here we describe the construction of DNA mismatch substrates containing G:T or O⁶meG:T mismatches, the purification of recombinant native human MutSα (MSH2–MSH6) and MutLα (MLH1–PMS2) proteins, and in vitro mismatch repair and excision assays that can be adapted to study mismatch repair in nuclear extracts from mismatch repair proficient and deficient cells.     

Stoichiometry and affinity for thymine DNA glycosylase binding to specific and nonspecific DNA

Deamination of 5-methylcytosine to thymine creates mutagenic G · T mispairs, contributing to cancer and genetic disease. Thymine DNA glycosylase (TDG) removes thymine from these G · T lesions, and follow-on base excision repair yields a G · C pair. A previous crystal structure revealed TDG (catalytic domain) bound to abasic DNA product in a 2:1 complex, one subunit at the abasic site and the other bound to undamaged DNA. Biochemical studies showed TDG can bind abasic DNA with 1:1 or 2:1 stoichiometry, but the dissociation constants were unknown, as was the stoichiometry and affinity for binding substrates and undamaged DNA. We showed that 2:1 binding is dispensable for G · U activity, but its role in G · T repair was unknown. Using equilibrium binding anisotropy experiments, we show that a single TDG subunit binds very tightly to G · U mispairs and abasic (G · AP) sites, and somewhat less tightly G · T mispairs. Kinetics experiments show 1:1 binding provides full G · T activity. TDG binds undamaged CpG sites with remarkable affinity, modestly weaker than G · T mispairs, and exhibits substantial affinity for nonspecific DNA. While 2:1 binding is observed for large excess TDG concentrations, our findings indicate that a single TDG subunit is fully capable of locating and processing G · U or G · T lesions.     

Association of Dnmt3a and thymine DNA glycosylase links DNA methylation with base-excision repair

While methylcytosines serve as the fifth base encoding epigenetic information, they are also a dangerous endogenous mutagen due to their intrinsic instability. Methylcytosine undergoes spontaneous deamination, at a rate much higher than cytosine, to generate thymine. In mammals, two repair enzymes, thymine DNA glycosylase (TDG) and methyl-CpG binding domain 4 (MBD4), have evolved to counteract the mutagenic effect of methylcytosines. Both recognize G/T mismatches arising from methylcytosine deamination and initiate base-excision repair that corrects them to G/C pairs. However, the mechanism by which the methylation status of the repaired cytosines is restored has remained unknown. We show here that the DNA methyltransferase Dnmt3a interacts with TDG. Both the PWWP domain and the catalytic domain of Dnmt3a are able to mediate the interaction with TDG at its N-terminus. The interaction affects the enzymatic activity of both proteins: Dnmt3a positively regulates the glycosylase activity of TDG, while TDG inhibits the methylation activity of Dnmt3a in vitro. These data suggest a mechanistic link between DNA repair and remethylation at sites affected by methylcytosine deamination.