Introduction

Alkylated bases in DNA are repaired by three principally different mechanisms, depending on the type of lesion. These include (1) DNA glycosylases, which initiate base excision repair at methylated bases by removing the methylated base (Laval, 1977; Chen et al, 1989); (2) suicidal methyltransferases, which transfer the methyl group from the DNA onto a cysteine residue in the transferase itself (Tano et al, 1990) and (3) the newly identified AlkB family of dioxygenases, which directly reverse DNA base damage by oxidative demethylation (Duncan et al, 2002; Falnes et al, 2002; Trewick et al, 2002; Aas et al, 2003). The importance of efficient alkylation repair is further emphasized by the existence, within the same organism, of more than one enzyme representing each of these repair strategies. An example is the well-known AlkA and Tag DNA glycosylases, which initiate base excision repair at alkylated bases in Escherichia coli (E. coli) (Clarke et al, 1984). Tag is constitutively expressed, whereas AlkA is induced as part of the adaptive response (Karran et al, 1979; Evensen and Seeberg, 1982; Karran et al, 1982). All three strategies for alkylation repair are conserved from E. coli to man (Sedgwick, 2004).

The conservation of enzymes specialized to repair methylated bases in DNA implies that naturally occurring methylating agents are widespread (Sedgwick, 2004). Such methylation can occur endogenously by the methyl donor S-adenosyl-methionine (SAM) (Barrows and Magee, 1982; Rydberg and Lindahl, 1982; Naslund et al, 1983) or by exposure to exogenous alkylating agents, originating from food, cigarette smoke, occupational chemicals or chemotherapeutics (Sedgwick, 2004). Methylating agents can react with DNA at several different sites on the DNA bases, including all the exocyclic oxygens and most ring nitrogens (Lindahl, 1993). The degree of alkylation at a given site depends not only on the mode of the chemical reaction (S_N1 or S_N2) but also on whether the DNA is in a single-stranded (ssDNA) or double-stranded (dsDNA) state (see below). AlkB has been thought to be involved in the cellular defense against methylation damage ever since it was discovered that alkB mutant E. coli cells were hypersensitive towards methylating agents (Kataoka et al, 1983; Kataoka and Sekiguchi, 1985), but it took almost 20 years to reveal the function of AlkB. A major breakthrough came when Sedgwick and co-workers demonstrated that AlkB repaired lesions that were specifically introduced in ssDNA, but not in dsDNA, and suggested that 1meA and 3meC were substrates for AlkB (Dinglay et al, 2000). Soon after, a landmark bioinformatics study placed the AlkB family of proteins in the superfamily of 2-oxoglutarate and Fe(II)-dependent oxygenases (Aravind and Koonin, 2001). By combining these clues, two independent studies showed that AlkB from E. coli was indeed a 2-oxoglutarate- and Fe(II)-dependent DNA repair enzyme capable of repairing 1meA and 3meC by oxidative demethylation (Falnes et al, 2002; Trewick et al, 2002). The substrate specificities of bacterial and human AlkB proteins have recently been extended to include 1-methylguanine (1meG) and 3-methylthymine (3meT), as well as ethylated bases (Delaney and Essigmann, 2004; Falnes, 2004; Koivisto et al, 2004). Thus, substrates for AlkB comprise all the lesions resulting from methylation at the N1 position of purines and N3 position of pyrimidines. These positions are structurally equivalent (the numbering of purine and pyrimidine bases is different) and shielded from methylation in the context of dsDNA, as they are involved in hydrogen bonding in Watson–Crick base pairing. Recently, the substrates of AlkB was expanded to include 1,N(6)-ethenoadenine and 3,N(4)-ethenocytosine (Delaney et al, 2005).

Bioinformatics indicate the existence of eight different dioxygenases of the AlkB family in mammalian cells, denoted ABH1–ABH8 (AlkB Homolog) (Kurowski et al, 2003). However, only two of the corresponding human proteins, hABH2 and hABH3, have been shown to possess a similar repair activity as AlkB from E. coli (Duncan et al, 2002; Aas et al, 2003). Besides removing lesions from DNA, AlkB and hABH3 also repair 1meA and 3meC in RNA (Aas et al, 2003). Purified AlkB and hABH3 were recently found to restore the biological function of mRNA and tRNA inactivated by chemical methylation (Ougland et al, 2004). Furthermore, chemically methylated E. coli tRNA was demethylated in an AlkB-overexpression strain, revealing a putative role for such repair enzymes in vivo (Ougland et al, 2004). In contrast, purified hABH2 displayed undetectable or very low levels of activity against methylated RNA species (Aas et al, 2003; Ougland et al, 2004). Whereas ds nucleic acids are the favored substrates for hABH2, ss substrates are preferred by hABH3 (Falnes et al, 2004; Koivisto et al, 2004), and these proteins also display distinct subnuclear localization patterns (Aas et al, 2003).

An elegant study by Delaney and Essigmann (2004) investigated the relative mutagenicity, genotoxicity and repair of 1meA, 3-ethylcytosine (3eC), 1meG and 3meT in E. coli. All these lesions blocked replication in repair-deficient cells, whereas their miscoding properties varied significantly. Mutations observed included G to T and G to A base substitutions, which are also observed in organisms exposed to alkylating agents (Delaney and Essigmann, 2004). It was concluded that AlkB suppresses both genotoxicity and mutagenicity at physiologically realistic levels of alkylated lesions.

In order to investigate the relevance of DNA repair by ABH2 and ABH3 in vivo, we have generated gene-targeted mice lacking functional genes coding for these two proteins. Novel assays have been developed to measure the presence of methyl lesions in mouse genomic DNA, as well as to characterize the repair proficiency of extracts and cells prepared from mouse tissues.

Results

Generation of mABH2 and mABH3 null mice

The two murine AlkB homologs, mAHB2 and mABH3, are relatively small basic proteins spanning 239 and 286 amino acids, respectively. The identity between mABH2 and hABH2 is 75.1% and the identity between mABH3 and hABH3 is 85.7%, suggesting similar activities (Lee et al, 2005). Sequence alignment shows that the three residues (His 149/191, Asp 151/193 and His 214/257), presumed to constitute the Fe(II)-binding cluster in E. coli AlkB, are conserved in the mouse proteins. The importance of some of these residues was recently verified by site-directed mutagenesis of mouse and human ABH2 and ABH3 (Lee et al, 2005). Based on comparison with other 2-oxoglutarate and Fe(II)-dependent oxygenases, a conserved arginine (Arg 226 in mABH2 and Arg 269 in mABH3) most likely binds the 5-carboxylate of the 2-oxoglutarate co-substrate (Mukherji et al, 2001; Aas et al, 2003; Valegard et al, 2004). The Fe(II)-binding cluster and the conserved arginine are encoded by exon 4 of the mABH2 gene and exons 8 and 10 of the mABH3 gene. To eliminate activity of the enzymes, exon 4 of mABH2 and exons 7 and 8 of mABH3 were targeted, as outlined in Figures 1 and 2 and Supplementary Figure 1.

Mendelian inheritance of mutated alleles in single and doubleknockouts

Genotyping of more than 40 live-born mice produced by inter-mating of F₁ heterozygotes (one litter for each genotype is presented in Supplementary Figure 1F) showed that mABH2 and mABH3 null mice are viable. Homozygous mutant mice were recovered in F₂ litters at frequencies consistent with Mendelian segregation (data not shown) and were phenotypically indistinguishable from their heterozygous and wild-type siblings. Furthermore, interbreeding of mABH2 and mABH3 null mice generated homozygous double mutants at the expected ratio in F₂. Null animals remained viable and apparently healthy into adulthood with no overt phenotype. Although some mABH2 and mABH3 null mice have died before the age of 18 months, no obvious phenotype was observed (results not shown). The aging colonies of mice are now about 18-month old, except for the double knockouts at 12 months of age. Histopathological examination of four 18-month-old mice, each of wild-type, mABH2 and mABH3, and two 12-month-old double knockouts, revealed no lesions or abnormalities that were considered to be related to genotype (Materials and methods). All lesions recorded were considered to be incidental.

Accumulation of 1meA in repair-deficient mice

To study the role of mABH2 and mABH3 for repair of erroneous endogenous methylation in DNA, genomic DNA was extracted from the livers of untreated wild-type, mABH2 and mABH3 knockout mice at different ages (Figure 3). The steady-state level of 1meA was determined at 1, 4, 8 and 12 months of age by the HPLC-MS/MS scheme, described in detail in Materials and methods. Two DNA samples, typically 100 g, were purified from each liver, and 2–4 mice of each genotype were sacrificed at each time point. The 1meA levels were below the detection limit in untreated 1-month-old mice, regardless of the genotype. Likewise, no 1meA could be detected in wild-type and mABH3 knockout mice at later time points. In contrast, 1meA accumulation was observed in 4-month-old mABH2 null mice, although only slightly above the detection limit, with a further increase at 8 months (Figure 3A and B). We estimated that at 8 months of age, wild-type and mABH3 knockout mice had accumulated less than 120 1meA per genome, whereas the corresponding number for the mABH2 knockout was within the range 150–500. Significant accumulation of 1meA was also observed in two 12-month-old mABH2 knockout mice. These results suggest that a substantial degree of aberrant methylation, producing 1meA lesions in DNA, occurs under normal physiological conditions.

Kinetics of 1meA repair in vivo

To assess the role of 1meA repair in living cells, mouse embryo fibroblast (MEF) cell lines established from the different mouse models were treated with the DNA methylating agent methyl methane sulfonate (MMS). Before examining the repair capacities in targeted cell lines, the repair kinetics of 1meA lesions induced by three different MMS concentrations (0.1, 0.5 and 2 mM) in wild-type cells were established (Figure 4A), and a substantial degree of repair was observed upon exposure to all three concentrations. The highest concentration of MMS (2 mM) was subsequently used in all the studies owing to more robust and accurate lesion detection. Repair was allowed to continue for up to 36 h after removal of MMS. Genomic DNA was purified, and the amount of 1meA determined using HPLC-MS/MS (Figure 4). Extensive washing of the cells was essential to ensure that no residual MMS remained as this caused a substantial further increase of DNA methylation (results not shown). In wild-type cells exposed to 2 mM MMS for 1 h, the amount of 1meA present was reduced by more than 50% after 8 h; approximately 10% of the induced 1meA remained in the cells after 36 h (Figure 4A and B). Such repair kinetics are comparable to those observed for other base lesions (Klungland et al, 1999).

The repair kinetics of 1meA in mABH3-targeted cells was indistinguishable from those of wild-type cells (Figure 4B). In contrast, and in agreement with the data obtained regarding repair of 1meA and 3meC in cellular extracts presented below, only a small fraction of 1meA was repaired in mABH2-targeted cells. An approximate 10% reduction in the 1meA levels was observed after 8 h, whereas a reduction of 30% was observed after 36 h. All experiments were carried out in MEF cells grown to about 80% confluence. In order to eliminate dilution of the lesions due to DNA replication, the experiments were repeated in the mABH2-targeted cells with the addition of aphidicolin, which inhibits elongation of the replicons (Saintigny et al, 2001). The repair rate of 1meA in mABH2-targeted cells was further reduced with this treatment, and only a negligible fraction of 1meA was removed following 36 h incubation (Figure 4C). In order to exclude the possibility that aphidicolin inhibits 1meA repair, similar experiments were carried out in wild-type cells. No significant reduction in the repair rates was observed (Figure 4C).

Cellular sensitivity to methylating agent

E. coli mutants deficient in 3-methyladenine DNA glycosylase activity (alkA and alkA tag) are extremely sensitive to exposure of alkylating agents (Clarke et al, 1984), whereas the alkB mutant displays moderate sensitivity (Kataoka et al, 1983). MEFs and ES clones lacking the murine 3-methyladenine DNA glycosylase (Aag) are hypersensitive towards methylating agents (Engelward et al, 1996; Engelward et al, 1997), but much less so than in the E. coli alkA tag mutant. Based on these observations, a modest but significant sensitizing of mABH2 cells and wild-type resistance of mABH3 cells was expected (Figure 5). Wild-type growth rates are observed in all cell lines.

DNA substrates for 1meA and 3meC repair

Purified E. coli AlkB and the human homologs hABH2 and hABH3 are able to repair 1meA and 3meC in DNA by oxidative demethylation (Duncan et al, 2002; Falnes et al, 2002; Trewick et al, 2002; Aas et al, 2003). Taking advantage of the inability of the DpnII restriction endonuclease to recognize a methylated GATC sequence, we established an in vitro assay for the repair of 1meA and 3meC by cellular extracts (Figure 6A and Supplementary Table 1). Repair of these lesions was monitored using a ds oligonucleotide containing either 1meA or 3meC within the GATC recognition sequence, and DpnII cleavage would only occur if the lesion had been repaired by direct reversal. Initially, the BamHI site was used to distinguish between methylated and unmethylated GGATCC. However, some BamHI activity was observed on methylated sequences (results not shown). As illustrated in Figure 6B, 49-mer 5'-[³²P]end-labeled oligonucleotides containing 1meA or 3meC were cleaved to 22 nucleotide (nt)-labeled fragments only when the DpnII digestion was preceded by incubation with AlkB, hABH2 or hABH3. A similar approach, utilizing methylation-sensitive restriction endonucleases, has previously been used successfully for the characterization of O6-alkylguanine-DNA alkyltransferase activity (Wu et al, 1987).

Activity of hABH2 on dsDNA has been reported to be stimulated by magnesium, and maximal repair activity for hABH2 was obtained in the presence of 10 mM MgCl₂ (Falnes et al, 2004). In contrast, AlkB and hABH3 were not affected by magnesium. Consequently, we added magnesium to the reaction buffer in the case of mABH2, but not for AlkB or mABH3. Figure 6B demonstrates that hABH3 repairs 3meC more efficiently than 1meA. This is consistent with previous reports (Duncan et al, 2002; Aas et al, 2003; Lee et al, 2005). The repair of synthetic oligonucleotides by purified AlkB family enzymes was carried out to justify further use of these DNA substrates, and not to reveal the kinetics of the purified enzymes.

Enzyme activity in extracts from mABH2, mABH3 and mABH2/mABH3 null mice

Human ABH2 and ABH3 have previously been characterized in vitro. Purified hABH2 and hABH3 repair both 1meA and 3meC in DNA, although with a different preference for ds versus ss substrates (see Introduction). In order to assess the properties and relevance of these repair enzymes in cellular extracts, the duplex oligonucleotides described above were incubated with nuclear extracts from liver, testis and kidney, cleaved with DpnII, and the reaction products analyzed by denaturing PAGE. MgCl₂ (10 mM) was added for optimal activity of mABH2. In addition, the amount of extract, DNA substrate and buffer conditions were titrated to obtain maximum repair activity and to minimize degradation of the DNA substrate (results not shown). The results indicate that mABH2 is the primary activity for repair of 1meA and 3meC in dsDNA in nuclear extracts (Figure 7 and Supplementary Figure 3). The repair activities at 1meA and 3meC lesions by mABH3-depleted cellular extracts are indistinguishable from the wild-type. Furthermore, extracts lacking both mABH2 and mABH3 have repair activities matching those of mABH2-depleted extracts. Similar results were observed with extracts from both liver and testis (Figure 7A and B and Supplementary Figure 3A).

To further investigate the repair of 1meA and 3meC lesions in extracts derived from gene-targeted mice, the amount of extract was varied between 0.5 and 10.0 g and repair activities were quantified (Figure 7B). These data support the hypothesis that mABH2 is the primary, or only, demethylase for repairing 1meA and 3meC in dsDNA. Furthermore, very low, but detectable, repair activity was observed in nuclear wild-type extracts from kidney, a tissue with low expression of mABH2 (Supplementary Figure 3C and results not shown).

We observed a low but reproducible and robust repair activity in extracts from mice having functional mABH2. To substantiate our findings somewhat, we attempted to up-concentrate the repair activity, taking advantage of the observation that recombinant hABH2 elutes as a sharp peak from a cation exchange column at pH 8.0 and 370 mM NaCl. Indeed, when nuclear extracts prepared from the livers of 10 wild-type mice were subjected to cation exchange chromatography, the repair activity eluted at the expected NaCl concentration, and the repair efficiency obtained with the peak fractions were several fold higher than in the case of the unfractionated extracts. Significant repair was obtained with as little as 0.2 g protein from peak fractions (Figure 7C, 7D–F). Again, the repair activity clearly appears to represent mABH2, as no activity was observed neither in the eluted fractions nor in the flow-through, when livers from mABH2 mice were used (Figure 7C and results not shown). The activity was dependent on the amount of protein added and incubation time (Figure 7E and F). This purification scheme was also followed with nuclear extracts prepared from kidney, a tissue with low expression of mABH2. As expected, the repair activity was significantly reduced compared to the liver (Supplementary Figure 3C). However, detectable activity could be obtained from fractionated wild-type nuclear extracts from kidney. Again, no activity was detected in mABH2-deficient extract. ss linear DNA is rapidly degraded in cellular extracts. In order to further elucidate the role of mABH3, we are currently attempting to establish repair assays by using exonuclease protected circular ssDNA containing 1meA and 3meC.

Discussion

Endogenous DNA methylation

Harmful, nonenzymatic, and possible enzymatic, DNA methylation and its repair is a topic that recently has attracted a great deal of interest, owing to the discovery of a novel, evolutionary conserved mechanism for repairing methylation damage. Both mammalian and bacterial AlkB proteins were shown to repair certain methylated bases in vitro, as well as to correct the repair deficiency of alkB mutant E. coli (Duncan et al, 2002; Aas et al, 2003). We demonstrate that mABH2, but not mABH3, null cells display a defect in removing methyl lesions from DNA and that spontaneously arising methyl lesions accumulate in the genome of the mABH2-targeted mice. This represents the first demonstration of accumulated, spontaneous methylation damage in repair-deficient cells, further substantiating the notion that repair of endogenously generated lesions is an important role for methylation repair. In accordance with this, the AlkB mechanism is well conserved between organisms that would be expected to experience very different loads of methylation from exogenous sources (Falnes and Rognes, 2003).

Nonenzymatic methylation of DNA is known to arise endogenously in bacteria, and DNA alkylation repair limits spontaneous mutations (Mackay et al, 1994; Taverna and Sedgwick, 1996). In mammalian cells, several possible sources of nonenzymatic methylation by methyl group donors exist, but only the universal methyl donor SAM has been investigated to any extent. More than 20 years ago, several studies addressed the relevance of endogenous DNA methylation by SAM (Barrows and Magee, 1982; Rydberg and Lindahl, 1982; Naslund et al, 1983) and concluded that SAM does methylate DNA nonenzymatically. Based on experiments with highly purified [³H methyl]SAM, Rydberg and Lindahl (1982) estimated that intracellular SAM could result in DNA methylation at a level similar to that expected to occur as a result of the continuous exposure of cells to 20 nM MMS (Rydberg and Lindahl, 1982).

In our MMS-exposure experiments, we measured the formation of approximately one 1meA lesion per 10⁶ adenines per millimolar MMS per hour, whereas spontaneous lesions accumulated in mABH2-targeted mice at an approximate rate of one lesion per 10⁷ adenines per year, that is at a rate roughly five orders of magnitude lower, and well comparable to that expected from exposure to nanomolar concentrations of MMS in the absence of repair (assuming a linear relationship between lesion formation and MMS dose; 1 mM/10⁵=10 nM). Thus, the intracellular concentration of methylating agents may well be close to the value estimated by Rydberg and Lindahl (1982).

The biological roles of mABH2 and mABH3

The mammalian AlkB homologs ABH2 and ABH3 both have the ability to demethylate 1meA and 3meC lesions in DNA, but there are also some important differences between these enzymes. Whereas hABH2 displayed a moderate, magnesium-dependent preference for dsDNA, hABH3 favored ssDNA (Falnes et al, 2002; Aas et al, 2003; Koivisto et al, 2004), and similar results have recently been obtained in the case of mABH2 and mABH3 (Lee et al, 2005). Moreover, the subcellular localization of hABH2 and hABH3 was found to be very different (Aas et al, 2003). hABH2 was exclusively nuclear and showed a cell-cycle-dependent distribution; it was found in replication foci during the S-phase, and accumulated in the nucleoli of non-S-phase cells. hABH3, on the other hand, was found both in the cytosol and in the nucleus, but was excluded from nucleoli. Interestingly, hABH3, but not hABH2, was also shown to efficiently demethylate lesions in RNA. Actually, the data presented here might strengthen the possible role of mABH3 as a specialized RNA repair enzyme, and experiments along these lines are underway. Our data indicate that mABH3 (and probably also hABH3) does not play a major role in removing 1meA and 3meC lesions from genomic DNA. High-resolution structure of substrate and product complexes of AlkB was recently published (Yu et al, 2006). The sequence divergence in the nucleotide-recognition lids between AlkB, ABH2 and ABH3 may be related to their significantly different affinity for different nucleic acid substrates observed in vitro (Aas et al, 2003; Falnes et al, 2004), and thus support our in vivo observations. However, our assays might lack the sensitivity required to identify repair at ss regions of DNA occurring transiently in vivo within transcription bubbles and replication forks. Furthermore, it is possible that mABH3 might have other, favored, DNA substrates.

Mammalian AlkB homologs

ABH1, the mammalian AlkB homolog with the strongest sequence similarity to the E. coli AlkB protein, was also the first one to be identified (Wei et al, 1996). However, no DNA repair activity has been demonstrated in the case of this protein, despite numerous efforts. Combination of biochemical and bioinformatics approaches led to the discovery of DNA repair by dioxygenases in bacteria and humans, and today the AlkB family in mammals consists of eight members, denoted as ABH1–ABH8 (Kurowski et al, 2003). However, only in the case of ABH2 and ABH3 has an in vitro activity been reported (Lee et al, 2005). It is tempting to speculate that mABH1 may carry out repair in vivo, although no activity has been ascribed to the purified protein so far. In order to evaluate this further, mABH2 knockout mice are currently being interbred with the mABH1-targeted mice (mABH1-targeted mice are viable—Nordstrand, Ringvoll and Klungland; unpublished data). The large number of lesions identified in DNA, as well as the comprehensive list of modified RNA nucleotides, suggest that the remaining ABH members may exhibit various exciting properties with regard to the repair and regulation of DNA and RNA species (Rozenski et al, 1999; Kurowski et al, 2003).

Materials and methods

Construction of targeting vectors for mABH2 and mABH3

A specific murine probe of the mABH2 sequence was amplified from murine genomic DNA by polymerase chain reaction (PCR) (primers 1 and 2, Supplementary Table 1) and was used to screen a mouse 129/SvJ lambda genomic library (Stratagene). A genomic clone of 12.0 kilobases (kb) spanning the whole coding region of the mABH2 gene was used to produce the targeting construct. The correct recombination event will result in the deletion of 3.2 kb of the mABH2 gene, containing the exon 4 sequence, which includes the conserved 2-oxoglutarate and Fe(II)-dependent oxygenase domain as well as the AlkB domain (Figure 1A and B). The resulting construct, based on the vector pGT-N38 (New England Biolabs), possesses homologous arms of 3.7 and 3.2 kb, which flank the neo gene.

Primers 3 and 4 (Supplementary Table 1) were used to identify the mABH3 gene. A genomic clone of 13.8 kb spanning exons 5–8 of the mABH3 gene was used to produce the targeting construct, together with the vector pGT-N28 (New England Biolabs). The correct recombination event will result in the deletion of 5.6 kb of the mABH3 gene (Figure 2A and B). The deleted area, including exon 8 and part of exon 7, contains sections of the conserved 2-oxoglutarate- and Fe(II)-dependent oxygenase domain as well as the AlkB domain (Figure 2A). The resulting construct possesses homologous arms of 4.4 and 2.1 kb, which flank the neo gene. The accuracy of the targeting construct was verified as above. Verification of correct ES-cell targeting is outlined in Supplementary Figure 1.

Histopathology

The following tissues were collected from wild-type, mABH2, mABH3 and mABH2/mABH3 mice and fixed in 10% neutral-buffered formalin: the skeletal muscle, ovary, uterus, vagina (females) or testis, epididymis, accessory sex glands (males), skin, mesenteric lymph node, liver, small and large intestines, pancreas, lungs, spleen, kidney, heart, thymus, urinary bladder, adrenal gland, sternum, cervical, thoracic and lumbar spinal cord, pituitary gland, brain and eye. After fixation, tissues were trimmed, embedded in paraffin, sectioned at 5 m and stained with hematoxylin and eosin using standard methods previously described (Relyea et al, 2000; Brayton et al, 2001) All tissues were examined by a veterinary pathologist using light microscopy.

Preparation of genomic DNA from mouse liver

Approximately 100 g of mouse liver was incubated with 50 mM Tris (pH 8.0), 100 mM EDTA, 0.5% SDS and 250 g proteinase K at 55°C over night. After extraction of DNA with phenol/chloroform and addition of 0.1 volume of 3 M sodium acetate and 2 volumes of 100% ethanol, the DNA was spooled onto a fused-end Pasteur pipette. Finally, the DNA was rinsed in 70% ethanol and air dried for 15 min at room temperature and dissolved in 100 l dH₂O. The DNA concentration was determined using NanoDrop 3.0.1 and processed further as described below.

HPLC-MS/MS analysis

DNA samples (typically 50 g) or 1meA-containing oligonucleotide standards were enzymatically hydrolyzed to nucleosides by using nuclease P1, snake venom phosphodiesterase and alkaline phosphatase (Crain, 1990). This was followed by the addition of 3 volume of methanol and centrifugation at 16 000 g for 30 min. The supernatants were dried under vacuum, and the resulting residues dissolved in 50 l 0.1% formic acid/5% methanol in water (v/v), for 1-methyldeoxyadenosine (1me(dA)) analysis by HPLC-MS/MS. A portion of each sample was diluted to 1:25 000 for quantitation of the corresponding deoxyadenosine (dA).

HPLC apparatus consisted of an HP1100 system (Agilent Technologies) with binary pump, autosampler and solvent degasser. Gradient chromatographic separation of 1me(dA) analytes was performed on a Zorbax SB-C18 2.1 150 mm² i.d. (3.5 m) reverse phase column equipped with an Eclipse XDB-C8 2.1 12.5 mm² i.d. (5 m) guard column (both from Agilent Technologies), at ambient temperature and a flow rate of 0.2 ml/min. The mobile phase consisted of A (0.1% formic acid in water) and B (0.1% formic acid in methanol), in a gradient elution starting with 95% A/5% B for 3.5 min, followed by a 7 min linear gradient of 5–50% B, and finally 25 min re-equilibration with the initial mobile phase conditions. Chromatographic analysis of dA was performed under isocratic conditions on a Supelcosil LC-C18-S 4.6 150 mm² i.d. (5 m) reverse phase column (Sigma-Aldrich). This was maintained at ambient temperature, with a mobile phase of 80% A/20% B at a flow rate of 1 ml/min.

Online mass spectrometry detection was performed using an Applied Biosystems/MDS Sciex 4000 hybrid triple quadrupole/linear ion trap mass spectrometer (Applied Biosystems Sciex) with TurboIonSpray probe operating in positive electrospray ionization mode. The analytes were monitored in the multiple-reaction monitoring mode, using the mass transition 266.2 150.1 for 1me(dA) and 252.2 136.1 for dA.

Repair of 1meA in MEF lines and preparation of genomic DNA

MEF cells at 80% confluence were treated with the indicated doses of MMS for 1 h at 37°C. Following two rinses with PBS and replacement of the media, cells were incubated for up to 36 h under the cell culture conditions described. Cells were trypsinized, re-suspended in DMEM, pelleted at 4000 g for 5 min at 4°C, washed in PBS and collected at 4500 g for 3 min at 4°C. DNA was isolated as described in the Wizard^®Genomic DNA Purification protocol (Promega) before HPLC-MS/MS identification of 1meA.

Furthermore, in mABH2 null cells, repair of 1meA was studied under replication arrest. Following MMS treatment for 1 h at 37°C and rinsing twice with PBS, 6 M aphidicolin (Calbiochem) was added to the fresh media. The cells were incubated for up to 36 h, and then harvested. DNA was isolated and finally separated using HPLC-MS/MS to identify 1meA as described above.

Cell culture and MMS exposure

Spontaneously immortalized clonal MEF cell lines were established from individual 13.5-day-old embryos (E13.5) obtained from heterozygous crossings (Hogan et al, 1994). All MEF lines were cultured at 37°C with 5% CO₂ in Dulbecco's modified Eagle's medium (DMEM) with high glucose, without L-glutamine (BioWhittaker^TM, BE12-614F), supplemented with 12.8% fetal bovine serum (PAA GmbH, A15-043), 0.3 mg/ml glutamine (BioWhittaker™, DE17-605E), 100 U/ml penicillin and 0.1 mg/ml streptomycin (BioWhittaker^TM, DE17-602E). The genotype of each cell line was determined by PCR as shown in Supplementary Figure 1F. The tetrazolium cytotoxicity assay (MTT) was performed basically as described (Espevik and Nissen-Meyer, 1986). The cells were seeded in 96-well plates with 2000 cells/well in 200 l medium. Cells were exposed to medium containing MMS in culture medium either continuously (200 l medium per well, 0–500 M) or for 1 h (100 l/well, 0–5 mM), 24 h after seeding. Cell numbers were measured using the MTT assay.

Protein purification

AlkB, hABH2 and hABH3 were purified as previously described (Aas et al, 2003).

Preparation of protein extracts

Organs were removed, quickly frozen on dry ice and stored at -80°C until use. Nuclear extracts were prepared as described previously (Klungland et al, 1999). A concentration of approximately 4 mg/ml should be expected. Fractionation of nuclear extracts was performed on a cation exchange Resource S 1 ml column (GE Healthcare) by FPLC ÄKTA (GE Healthcare). The nuclear extract prepared from 10 pooled livers/kidneys was centrifuged at 13 000 g for 5 min and loaded to the column at 0.5 ml/min. The proteins were eluted with an NaCl gradient (Buffer A: 50 mM NaCl/50 mM Tris pH 8.0/2 mM DTT; Buffer B: 2 M NaCl/50 mM Tris pH 8.0/2 mM DTT) increasing the NaCl concentration with 0–25% Buffer B from fraction 1 to 30 and collected in 0.5 ml fractions at 0.5 ml/min. Fractions 1–30 were tested for repair activity. The procedure was carried out at 4°C.

Preparation of oligonucleotide substrates

A 49-mer oligonucleotide containing either N-1-methyl-dA (1meA substrate, Supplementary, Table 1) or N-3-methyl-dC (3meC substrate, Supplementary Table 1) was provided by ChemGenes (ChemGenes Corporation). The natural deoxynucleoside phosphoramidites were employed using standard procedures. The methylated deoxy CED-OP used were 5'-DMT-N1-Methyl deoxy Adenosine (n-fmoc) CED OP (Catalog No.: ANP-6121) and 5'-DMT-N3-Methyl deoxy Cytidine (n-bz) CED OP (Catalog No.: ANP-3851). After oligonucleotide synthesis, deprotection was performed using 28% ammonium hydroxide at 55°C for 10 h. Oligonucleotides were purified from the deprotection solution by using a Puri-Pak purification cartridge and 5'-DMT-ON protected crude oligonucleotides. The oligonucleotides were eluted as DMT-ON oligonucleotides, and then treated with 80% acetic acid for 15 min at 37°C before being repurified on a Puri-Pak cartridge and lyophilized. A control oligonucleotide (Supplementary Table 1) was provided by Invitrogen. All oligonucleotides included phosphorothioate linkages at their 5' and 3' ends to reduce exonucleolytic attack.

The oligonucleotides were end-labeled at the 5' terminus. The labeling reaction contained 28 pmol of the oligonucleotide, 1.48MBq/0.04 mCi [-³²P]ATP (Amersham), 20 U T4 polynucleotide kinase (New England Biolabs), TSE-buffer (20 mM Tris–HCl pH 9.5/1 mM spermidin HCl/0.1 mM EDTA) and appropriate kinase buffer in a total volume of 60 l (incubation for 45 min at 37°C). Equal molar amounts of the complementary oligonucleotide (Supplementary Table 1) containing either a thymine (T) or a guanine (G) residue opposite to the N-1-methyl-dA or N-3-methyl-dC, respectively, were then added to form duplex DNA substrate. This was incubated for 2 min at 90°C and then allowed to cool slowly to room temperature. End-labeled oligonucleotide substrates were purified following separation on a nondenaturing 20% PAGE gel.

DNA repair reactions

In order to avoid exonucleolytic degradation of the oligonucleotide substrates during incubation with cellular extracts, the optimal amount of substrate and nuclear extract were determined by titration (results not shown). A 50 l reaction mixture containing 50 mM Tris, pH 8.0, 2 mM ascorbic acid, 100 M 2-oxoglutarate, 40 M FeSO₄ and approximately 150 fmol ³²P-labeled ds oligonucleotide substrate (which roughly correspond to 16 000 c.p.m. for the freshly prepared 1meA substrate and 9000 c.p.m. for the freshly prepared 3meC substrate), containing either an N-1-methyl-dA or N-3-methyl-dC lesion located within a palindromic site for recognition by restriction endonucleases, was incubated at 37°C for 30 min with either 100 ng enzyme (AlkB, hABH2 or hABH3) or the amount of nuclear extracts indicated in the figures. In order to improve the repair, 10 mM MgCl₂ was added to reactions containing hABH2 and nuclear extracts. Reactions with purified enzymes were incubated at 50°C for 10 min to inactivate the enzymes. All other reactions were terminated by adding EDTA to 18 mM, SDS to 0.4% and proteinase K to 0.4 g/l, and incubating at 37°C for 30 min. The DNA was precipitated by adding 25 g tRNA (Roche), NaAc pH 5.5–0.3 M and 3 volume of ice-cold 100% ethanol, and placed at -20°C overnight. The precipitated DNA was washed in 70% ethanol and dissolved in 25 l dH₂O.

AlkB and its homologs oxidize the methyl group, resulting in direct reversal of the damage without cleaving the DNA strand. The restriction enzyme DpnII is methyl sensitive and was used to distinguish methylated and repaired oligonucleotides. The site of the damage is within the DpnII recognition sequence (5'-GATC-3'). An aliquot of 7 l from reactions with purified enzymes or the precipitated DNA from reactions with nuclear extract was cleaved by 20U DpnII (New England Biolabs) at 37°C for 1 h. Oligonucleotides were resolved by 20% denaturating PAGE gel. They were visualized and quantified using an ImageQuant (Typhoon 9410). Assays were repeated with organs from two (or three) different mice and gave consistent results.

Supplementary data

Supplementary data are available at The EMBO Journal Online.

Acknowledgements

We are grateful to Cecilie G Castellanos, Tara Catterall, Christine G Neurauter, Lene Uldal, Rune J Forstrøm and Ann S Langøy (GE Healthcare) for excellent technical assistance. We thank IngenKo, Australia, The Norwegian Transgenic Center (NTS) and the Center for Comparative studies at Rikshospitalet/Radiumhospitalet HF for the excellent service they provided. We also acknowledge the generous financial support provided by the National Program in Functional Genomics (FUGE) sponsored by the Norwegian Research Council, the Norwegian Cancer Society, Throne Holst Foundation and the European Union program 'DNA repair'. We also express our gratitude for the advices and enthusiasm that the late Erling Seeberg provided during the initial phase of the project.

References

Aas PA , Otterlei M , Falnes PO , Vagbo CB , Skorpen F , Akbari M , Sundheim O , Bjoras M , Slupphaug G , Seeberg E , Krokan HE (2003) Human and bacterial oxidative demethylases repair alkylation damage in both RNA and DNA. Nature 421: 859–863  | Article | PubMed | ISI | ChemPort |

Aravind L , Koonin EV (2001) The DNA-repair protein AlkB, EGL-9, and leprecan define new families of 2-oxoglutarate- and iron-dependent dioxygenases. Genome Biol 2, RESEARCH0007

Barrows LR , Magee PN (1982) Nonenzymatic methylation of DNA by S-adenosylmethionine in vitro. Carcinogenesis 3: 349–351  | PubMed | ISI | ChemPort |

Brayton C , Justice M , Montgomery CA (2001) Evaluating mutant mice: anatomic pathology. Vet Pathol 38: 1–19  | Article | PubMed | ISI | ChemPort |

Chen J , Derfler B , Maskati A , Samson L (1989) Cloning a eukaryotic DNA glycosylase repair gene by the suppression of a DNA repair defect in Escherichia coli. Proc Natl Acad Sci USA 86: 7961–7965  | PubMed | ChemPort |

Clarke ND , Kvaal M , Seeberg E (1984) Cloning of Escherichia coli genes encoding 3-methyladenine DNA glycosylases I and II. Mol Gen Genet 197: 368–372  | Article | PubMed | ISI | ChemPort |

Crain PF (1990) Preparation and enzymatic hydrolysis of DNA and RNA for mass spectrometry. Methods Enzymol 193: 782–790  | PubMed | ChemPort |

Delaney JC , Essigmann JM (2004) Mutagenesis, genotoxicity, and repair of 1-methyladenine, 3-alkylcytosines, 1-methylguanine, and 3-methylthymine in alkB Escherichia coli. Proc Natl Acad Sci USA 101: 14051–14056  | Article | PubMed | ChemPort |

Delaney JC , Smeester L , Wong C , Frick LE , Taghizadeh K , Wishnok JS , Drennan CL , Samson LD , Essigmann JM (2005) AlkB reverses etheno DNA lesions caused by lipid oxidation in vitro and in vivo. Nat Struct Mol Biol 12: 855–860  | Article | PubMed | ISI | ChemPort |

Dinglay S , Trewick SC , Lindahl T , Sedgwick B (2000) Defective processing of methylated single-stranded DNA by E. coli AlkB mutants. Genes Dev 14: 2097–2105  | PubMed | ISI | ChemPort |

Duncan T , Trewick SC , Koivisto P , Bates PA , Lindahl T , Sedgwick B (2002) Reversal of DNA alkylation damage by two human dioxygenases. Proc Natl Acad Sci USA 99: 16660–16665  | Article | PubMed | ChemPort |

Engelward BP , Dreslin A , Christensen J , Huszar D , Kurahara C , Samson L (1996) Repair-deficient 3-methyladenine DNA glycosylase homozygous mutant mouse cells have increased sensitivity to alkylation-induced chromosome damage and cell killing. EMBO J 15: 945–952  | PubMed | ISI | ChemPort |

Engelward BP , Weeda G , Wyatt MD , Broekhof JL , de WJ , Donker I , Allan JM , Gold B , Hoeijmakers JH , Samson LD (1997) Base excision repair deficient mice lacking the Aag alkyladenine DNA glycosylase. Proc Natl Acad Sci USA 94: 13087–13092  | Article | PubMed | ChemPort |

Espevik T , Nissen-Meyer J (1986) A highly sensitive cell line, WEHI 164 clone 13, for measuring cytotoxic factor/tumor necrosis factor from human monocytes. J Immunol Methods 95: 99–105  | Article | PubMed | ISI | ChemPort |

Evensen G , Seeberg E (1982) Adaptation to alkylation resistance involves the induction of a DNA glycosylase. Nature 296: 773–775  | Article | PubMed | ISI | ChemPort |

Falnes PO (2004) Repair of 3-methylthymine and 1-methylguanine lesions by bacterial and human AlkB proteins. Nucleic Acids Res 32: 6260–6267  | Article | PubMed | ISI | ChemPort |

Falnes PO , Bjoras M , Aas PA , Sundheim O , Seeberg E (2004) Substrate specificities of bacterial and human AlkB proteins. Nucleic Acids Res 32: 3456–3461  | Article | PubMed | ISI | ChemPort |

Falnes PO , Johansen RF , Seeberg E (2002) AlkB-mediated oxidative demethylation reverses DNA damage in Escherichia coli. Nature 419: 178–182  | Article | PubMed | ISI | ChemPort |

Falnes PO , Rognes T (2003) DNA repair by bacterial AlkB proteins. Res Microbiol 154: 531–538  | Article | PubMed | ISI | ChemPort |

Hogan B , Beddington R , Costantini F , Lacy E (1994) Manipulating the Mouse Embryo; A Laboratory Manual. Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press

Karran P , Hjelmgren T , Lindahl T (1982) Induction of a DNA glycosylase for N-methylated purines is part of the adaptive response to alkylating agents. Nature 296: 770–773  | Article | PubMed | ISI | ChemPort |

Karran P , Lindahl T , Griffin B (1979) Adaptive response to alkylating agents involves alteration in situ of O6-methylguanine residues in DNA. Nature 280: 76–77  | Article | PubMed | ISI | ChemPort |

Kataoka H , Sekiguchi M (1985) Molecular cloning and characterization of the alkB gene of Escherichia coli. Mol Gen Genet 198: 263–269  | Article | PubMed | ISI | ChemPort |

Kataoka H , Yamamoto Y , Sekiguchi M (1983) A new gene (alkB) of Escherichia coli that controls sensitivity to methyl methane sulfonate. J Bacteriol 153: 1301–1307  | PubMed | ISI | ChemPort |

Klungland A , Rosewell I , Hollenbach S , Larsen E , Daly G , Epe B , Seeberg E , Lindahl T , Barnes DE (1999) Accumulation of premutagenic DNA lesions in mice defective in removal of oxidative base damage. Proc Natl Acad Sci USA 96: 13300–13305  | Article | PubMed | ChemPort |

Koivisto P , Robins P , Lindahl T , Sedgwick B (2004) Demethylation of 3-methylthymine in DNA by bacterial and human DNA dioxygenases. J Biol Chem 279: 40470–40474  | Article | PubMed | ISI | ChemPort |

Kurowski MA , Bhagwat AS , Papaj G , Bujnicki JM (2003) Phylogenomic identification of five new human homologs of the DNA repair enzyme AlkB. BMC Genomics 4: 48  | Article | PubMed |

Laval J (1977) Two enzymes are required from strand incision in repair of alkylated DNA. Nature 269: 829–832  | Article | PubMed | ISI | ChemPort |

Lee DH , Jin SG , Cai S , Chen Y , Pfeifer GP , O'connor TR (2005) Repair of methylation damage in DNA and RNA by mammalian AlkB homologues. J Biol Chem 280: 39448–39458  | Article | PubMed | ISI | ChemPort |

Lindahl T (1993) Instability and decay of the primary structure of DNA. Nature 362: 709–715  | Article | PubMed | ISI | ChemPort |

Mackay WJ , Han S , Samson LD (1994) DNA alkylation repair limits spontaneous base substitution mutations in Escherichia coli. J Bacteriol 176: 3224–3230  | PubMed | ISI | ChemPort |

Mukherji M , Chien W , Kershaw NJ , Clifton IJ , Schofield CJ , Wierzbicki AS , Lloyd MD (2001) Structure-function analysis of phytanoyl-CoA 2-hydroxylase mutations causing Refsum's disease. Hum Mol Genet 10: 1971–1982  | Article | PubMed | ISI | ChemPort |

Naslund M , Segerback D , Kolman A (1983) S-adenosylmethionine, an endogenous alkylating agent. Mutat Res 119: 229–232  | Article | PubMed | ISI | ChemPort |

Ougland R , Zhang CM , Liiv A , Johansen RF , Seeberg E , Hou YM , Remme J , Falnes PO (2004) AlkB restores the biological function of mRNA and tRNA inactivated by chemical methylation. Mol Cell 16: 107–116  | Article | PubMed | ISI | ChemPort |

Relyea MJ , Miller J , Boggess D , Sundberg J (2000) Necropsy methods for laboratory mice: biological characterization of a new mutation. In Systematic Approach to Evaluation of Mouse Mutations, Sundberg P, Boggess D (eds), pp 57–89. Boca Raton: CRC Press

Rozenski J , Crain PF , McCloskey JA (1999) The RNA Modification Database: 1999 update. Nucleic Acids Res 27: 196–197  | Article | PubMed | ISI | ChemPort |

Rydberg B , Lindahl T (1982) Nonenzymatic methylation of DNA by the intracellular methyl group donor S-adenosyl-L-methionine is a potentially mutagenic reaction. EMBO J 1: 211–216  | PubMed | ISI | ChemPort |

Saintigny Y , Delacote F , Vares G , Petitot F , Lambert S , Averbeck D , Lopez BS (2001) Characterization of homologous recombination induced by replication inhibition in mammalian cells. EMBO J 20: 3861–3870  | Article | PubMed | ISI | ChemPort |

Sedgwick B (2004) Repairing DNA-methylation damage. Nat Rev Mol Cell Biol 5: 148–157  | Article | PubMed | ISI | ChemPort |

Tano K , Shiota S , Collier J , Foote RS , Mitra S (1990) Isolation and structural characterization of a cDNA clone encoding the human DNA repair protein for O6-alkylguanine. Proc Natl Acad Sci USA 87: 686–690  | PubMed | ChemPort |

Taverna P , Sedgwick B (1996) Generation of an endogenous DNA-methylating agent by nitrosation in Escherichia coli. J Bacteriol 178: 5105–5111  | PubMed | ISI | ChemPort |

Trewick SC , Henshaw TF , Hausinger RP , Lindahl T , Sedgwick B (2002) Oxidative demethylation by Escherichia coli AlkB directly reverts DNA base damage. Nature 419: 174–178  | Article | PubMed | ISI | ChemPort |

Valegard K , Terwisscha van Scheltinga AC , Dubus A , Ranghino G , Oster LM , Hajdu J , Andersson I (2004) The structural basis of cephalosporin formation in a mononuclear ferrous enzyme. Nat Struct Mol Biol 11: 95–101  | Article | PubMed | ISI | ChemPort |

Wei YF , Carter KC , Wang RP , Shell BK (1996) Molecular cloning and functional analysis of a human cDNA encoding an Escherichia coli AlkB homolog, a protein involved in DNA alkylation damage repair. Nucleic Acids Res 24: 931–937  | Article | PubMed | ISI | ChemPort |

Wu RS , Hurst-Calderone S , Kohn KW (1987) Measurement of O6-alkylguanine-DNA alkyltransferase activity in human cells and tumor tissues by restriction endonuclease inhibition. Cancer Res 47: 6229–6235  | PubMed | ISI | ChemPort |

Yu B , Edstrom WC , Benach J , Hamuro Y , Weber PC , Gibney BR , Hunt JF (2006) Crystal structures of catalytic complexes of the oxidative DNA/RNA repair enzyme AlkB. Nature 439: 879–884  | Article | PubMed | ISI | ChemPort |