Oncogene (2002) 21, 8886 – 8894 ª 2002 Nature Publishing Group All rights reserved 0950 – 9232/02 $25.00 www.nature.com/onc

Recent progress on the Ada response for inducible repair of DNA alkylation damage

Barbara Sedgwick1 and Tomas Lindahl*,1

1Cancer Research UK London Research Institute, Clare Hall Laboratories, South Mimms, Hertfordshire EN6 3LD, UK

Oncogene (2002) 21, 8886 – 8894. doi:10.1038/sj.onc. Methylating agents 1205998 Methylating agents can alkylate DNA at many sites producing a wide variety of base lesions (Figure 1) and Keywords: DNA repair; O6-methylguanine; DNA phosphotriesters. The relative proportions of the glycosylases different base lesions depend on the nature of the alkylating agent, its reaction mechanism and the secondary structure of the DNA target. Of the commonly used laboratory alkylating agents, methyl Introduction methanesulphonate (MMS), dimethylsulphate (DMS) and methyl iodide (MeI) react via a SN2 mechanism Methylating agents comprise a major class of DNA and methylate DNA almost exclusively at nitrogen damaging compounds that occur both endogenously moieties in the purine and pyrimidine rings. and in the environment. They are extremely cytotoxic In contrast, N-methyl-N’-nitro-N-nitrosoguanidine and frequently also mutagenic, and are employed for (MNNG) and N-methylnitrosourea (MNU), which chemotherapy of certain cancers. All organisms have are SN1 agents, alkylate both nitrogens and oxygens multiple DNA repair strategies to counteract the effects in the DNA bases and also oxygens in the sugar- of DNA alkylation. To defend against fluctuating phosphate backbone. The major lesions generated in levels of environmental alkylating agents, many double-stranded DNA are 7-methylguanine (7-meG), bacteria mount an inducible response that enhances 3-methyladenine (3-meA) and O6-methylguanine (O6- cellular resistance to these same agents. This adaptive meG). In single-stranded DNA, SN2 agents also response has been most extensively studied in E. coli,in efficiently induce formation of 1-methyladenine (1- which induced alkylation resistance results from meA) and 3-methylcytosine (3-meC) (Bodell and increased expression of four genes, ada, alkB, alkA Singer, 1979). These lesions are less readily formed in and aidB. The Ada, AlkA and AlkB proteins protect duplex DNA because the modification sites are against alkylation by repair of methylated bases in involved in base pairing, and are therefore shielded DNA using different mechanisms, whereas AidB may from alkylation. There are also some distinct minor directly destroy certain alkylating agents. These DNA sites of . Several of the base modifications repair activities are conserved from bacteria to man. block DNA replication and are cytotoxic, including 3- The Ada protein also serves as the positive regulator of meA, 3-methylguanine (3-meG), 3-meC and probably ‘the adaptive response to alkylation damage’, which is also 1-meA. O6-meG and O4-methylthymine (O4-meT) more concisely named ‘the Ada response’ by analogy mispair during DNA replication and are therefore with the SOS, SoxRS and OxyR responses. Various mutagenic. It is fortuitous that the most abundantly aspects of the Ada response have been the subject of formed lesion, 7-meG, is relatively innocuous (Singer previous detailed reviews (Friedberg et al., 1995; and Grunberger, 1983). In a different type of DNA Landini and Volkert, 2000; Lindahl et al., 1988; alkylation, environmental such as 1,2- Samson, 1992; Sedgwick and Vaughan, 1991; Seeberg dimethylhydrazine, tert-butylhydroperoxide and diazo- and Berdal, 1999). Here, we describe the properties of quinones generate methyl radicals that react with the inducible proteins with particular emphasis on guanine residues in DNA to form miscoding 8- recent findings, such as those from genome sequencing, methylguanine adducts (Hix et al., 1995). protein structural studies, analysis of substrate specifi- Conservation of the Ada response in many bacterial cities, and the newly discovered function of the AlkB species suggests the presence of direct alkylating agents protein. in their environment. Good candidate agents are those produced by microorganisms, but others may be formed by chemical reactions. Some Streptomyces sp. release alkylating antibiotics, such as streptozotocin (a derivative of MNU) and azaserine (O-diazoacyl-L- *Correspondence: T Lindahl; E-mail: [email protected] serine), into the soil creating an urgent need for an Repair of alkylated DNA B Sedgwick and T Lindahl 8887 It is of note that ada gene expression also increases in stationary phase to further the defence against alkylation; this enhanced expression is also dependent on RpoS (Taverna and Sedgwick, 1996). The preli- minary studies on the function of AidB need to be extended, including more precise definition of its eukaryotic homologues.

Figure 1 Sites of methylation on the DNA bases. Thick arrows Ada, an O6-meG-DNA methyltransferase and indicate major sites of DNA methylation by SN1 agents. The curly arrow indicates an additional site alkylated by methyl radi- chemosensor for the adaptive response cals. Colours are used to indicate which damaged sites are re- paired by particular DNA repair activities: blue by Ada; green The multifunctional Ada protein repairs methylated by AlkA; red by AlkB; black, no known repair activity bases and also regulates the adaptive response (Figure 2). Ada is composed of two major domains that can function independently in DNA repair reactions. The adaptive response in other microorganisms. Further- 19 kDa C-terminal domain (C-Ada19) directly more, certain algae and fungi growing in saline demethylates the mutagenic bases, O6-meG and O4- environments generate MeCl as a product of chloride meT, and transfers the methyl groups on to its Cys-321 detoxification (Sedgwick and Vaughan, 1991). MeCl is residue (Demple et al., 1985). Ada therefore precludes probably the most abundant methylating agent in our mutation induction by this rapid but suicidal action. environment (Crutzen and Andreae, 1990). Chemically, The active site thiol of Cys-321 is buried in the 3-D direct acting alkylating agents may be formed by structure of C-Ada19. The protein was therefore nitrosations, in slightly acidic conditions, of amides, initially proposed to undergo a substantial conforma- amines, amino acids and peptides (Harrison et al., tional change on binding to its substrates to expose this 1999; Sedgwick, 1997; Sedgwick and Vaughan, 1991). active cysteine (Moore et al., 1994). An alternative These reactions could possibly occur in decaying model, however, used a predicted ‘helix – turn – helix – matter, in acidic soils or in putrid water. wing’ motif (HTH) to bind DNA and a E. coli cells that are unable to repair O6-meG or O4- flipping mechanism to align the alkylated guanine with meT residues have an increased frequency of sponta- the catalytic cysteine (Vora et al., 1998). The more neous mutagenesis when subjected to starvation/ recent crystal structure of the human O6-meG-DNA stationary conditions. Mutagenic alkylating agents must therefore arise in these cells, but their precise nature is unknown (Rebeck et al., 1989). S-Adeno- sylmethionine (SAM) weakly methylates DNA, but it acts by an SN2 mechanism and would not be expected to be an efficient . Thus, variation of the cellular SAM levels, over a 100-fold range, had no significant effect on spontaneous mutagenesis in E. coli (Posnick and Samson, 1999b). It is more likely that the major endogenous mutagen is a SN1 compound that alkylates oxygens in DNA. Bacterially catalyzed amine nitrosation increases in anaerobic and resting E. coli cells (Calmels et al., 1987; Kunisaki and Hayashi, 1979). A mutant deficient in this activity was less susceptible to spontaneous mutagenesis, suggesting that enzymatic nitrosation of amines or amides may be a source of endogenous mutagens (Harrison et al., 1999; Sedgwick, 1997; Taverna and Sedgwick, 1996). AidB protein shows some homology to the mammalian isovaleryl-coenzyme A dehydrogenases, and has been proposed to detoxify nitrosoguanidines (nitrosated amides) or their reaction intermediates (Landini et al., 1994). Expression of the aidB gene increases in anaerobic cells and is dependent on RpoS, the Figure 2 Diagrammatic representation of the E. coli Ada re- alternative sigma factor of RNA polymerase that sponse. The Ada protein is activated as a positive regulator by controls expression of many genes in stationary phase methylation of its Cys-38 residue in the amino-terminal half of (Volkert et al., 1994). A question of interest is whether the protein. This activation occurs by repair of methylphospho- triesters (PTE) in DNA or, less efficiently, by direct protein elevation of AidB activity occurs in the same methylation. The activated Ada protein induces expression of sev- conditions as spontaneous mutagenesis and is required eral genes resulting in increased DNA repair and probable to destroy endogenously generated nitroso-compounds. destruction of certain alkylating agents

Oncogene Repair of alkylated DNA B Sedgwick and T Lindahl 8888 methyltransferase (AGT) also implicates a HTH motif Self-methylation at Cys-38 converts Ada into a in DNA binding and an ‘arginine finger’ to extrude the transcriptional activator and dramatically increases its O6-alkG nucleotide from the duplex DNA. The sequence specific binding to the promoters of the ada- extrahelical base can then reach the recessed AGT alkB operon and the alkA and aidB genes (Landini and active site (Daniels and Tainer, 2000). These conserved Volkert, 2000; Sakumi and Sekiguchi, 1989; Teo et al., motifs in the E. coli C-Ada19 domain are shown in 1986) (Figure 2). The transferred methyl group does Figure 3. A nucleotide flipping mechanism has been not contact the promoter DNA, but may induce a documented for several other DNA methyltransferases conformational change in the protein that creates a and DNA glycosylases (Cheng and Roberts, 2001). new high affinity DNA binding face (Lin et al., 2001). The Ada protein also demethylates Sp-diastereo- A possible HTH motif and a basic segment of the isomers of methylphosphotriesters in DNA by methyl protein may have roles in specific DNA binding transfer on to the Cys-38 residue in the 20 kDa N- (Sakashita et al., 1995) (Figure 3). Self-methylated terminal domain (N-Ada20). This active cysteine was Ada binds to the ada and aidB promoters immediately previously considered to be Cys-69 (Sedgwick et al., upstream of the RNA polymerase binding sites, and C- 1988), but has now been more accurately identified as Ada19 may be required for direct interaction with Cys-38 in the recently solved 3-D structure of the RNA polymerase. In contrast, the self-methylated N- protein (GL Verdine et al., submitted). N-Ada20 Ada20 is sufficient to activate the alkA gene (Akimaru contains one Zn2+ tightly co-ordinated by 4 cysteine et al., 1990). The mechanisms by which self-methylated residues, Cys-38, Cys-42, Cys-69 and Cys-72 (Figure 3), Ada interacts with the promoters of the Ada regulon which are all essential for biological activity. The and RNA polymerase have been reviewed in detail bound metal is required for correct protein folding, (Landini and Volkert, 2000). and also activates the nucleophilicity of the methyl E. coli has a second O6-meG-DNA methyltransferase accepting Cys-38 residue. Methylation of Cys-38 does (Ogt) and a second 3-meA-DNA glycosylase (Tag) that not completely abolish Zn2+ binding (Myers and are expressed constitutively, and will repair some O6- Verdine, 1994). N-Ada20 has been a prototype for meG and 3-meA lesions in DNA during the vulnerable several other proteins that catalyze methyl group period in which the adaptive response is being induced transfer reactions using Zn2+ co-ordinated thiols as (Samson, 1992). Ingeniously, methylphosphotriesters, a nucleophiles, for example, enzymes involved in methio- relatively innocuous type of DNA damage, that are not nine synthesis and others in methanogenesis (Matthews repaired by Ogt or any other known constitutive and Goulding, 1997). Since C-Ada19 does not appear activity, are used as the inducing signal for the to contain a zinc residue to demethylate DNA by adaptive response. Repair of one of the two stereo- metallo-activated methyl transfer, the previously isomers of methylphosphotriesters therefore serves observed homology around residues Cys- solely as a sensor for changing levels of DNA 69 and Cys-321 may be coincidental. alkylation damage in bacteria (Lindahl et al., 1988).

Figure 3 Alignment of the E. coli Ada and B. subtilis AdaA proteins with other putative bacterial homologues. Protein accession references are: Paracoccus pantotrophus, AAK61311; Mesorhizobium loti, NP_103342; E. coli, NP_416717; B. subtilis, NP_388062; Lactococcus lactis, NC_002662. Identical and conserved residues are highlighted in yellow, and similar residues in grey. A hinge region between the two major domains of the E. coli protein is indicated (Sedgwick et al., 1988). Red stars indicate the cysteine 2+ residues that ligand Zn (Cys-38, Cys-42, Cys-69 and Cys-72 in the E. coli protein). ‘CH3’?points to the methyl accepting cysteines, indicated in red for methylphosphotriester repair and green for O6-meG repair (corresponding to Cys-38 and Cys-321 in the E. coli protein). Putative HTH DNA binding motifs occur in both domains. Within the proposed HTH of the C-terminal domain is a conserved ‘arginine finger’ which is implicated in nucleotide flipping. An invariant Asn-hinge couples the recognition helix (second helix) to the active site cysteine in this domain (Daniels and Tainer, 2000)

Oncogene Repair of alkylated DNA B Sedgwick and T Lindahl 8889 Eukaryotic organisms do not appear to have a DNA very small defect in the reactivation of double-stranded methyltransferase that repairs methylphosphotriesters. DNA phage treated with SN2 agents (Dinglay et al., Presumably this is not required as eukaryotes do not 2000; Kataoka et al., 1983). Although 1-meA and 3- have an inducible response equivalent to that in meC are generated mainly in single-stranded DNA, bacteria. AlkB can repair these modifications not only in single- Cys-38 of the E. coli Ada protein can be directly strands, but also in double-stranded structures methylated by MeI (GL Verdine et al., submitted). obtained by reannealing. Indeed, direct methylation of this cysteine by SN2 The low level of MMS induced mutagenesis in E. agents may be an alternative but less effective mode of coli alkB mutants suggests that unrepaired 1-meA and Ada activation as a positive gene regulator (Figure 2) 3-meC have a low capacity for mispairing during DNA (Takahashi and Kawazoe, 1987). SN1 agents induce the replication, and instead block replication (Dinglay et response by efficiently forming methylphosphotriesters al., 2000). MMS treated poly(dC) and poly(dA) in DNA, whereas SN2 methylating agents may cause contain lesions that block DNA synthesis in vitro, induction by inefficiently generating a few phospho- and these are presumably 3-meC in poly(dC), and both triesters in a large chromosomal DNA target and then, 1-meA and 3-meA in poly(dA) (Boiteux and Laval, as the cellular level of Ada increases, by direct Ada 1982; Larson et al., 1985). methylation (Vaughan et al., 1993) (Figure 2). The Ada AlkB was recently shown to be an a-ketoglutarate- response conveys a defence against both types of Fe(II)-dependent dioxygenase, and repairs 1-meA and agents, against killing by SN2 agents and against both 3-meC in DNA by oxidative demethylation. Theore- mutagenicity and toxicity of SN1 agents. It therefore tical protein fold recognition predicted that AlkB was a seems logical that both types of agents should have a member of this superfamily of enzymes (Aravind and means of inducing the response. Koonin, 2001) (Figure 4), and this was verified by Once all the methylphosphotriesters in DNA have biochemical analyses (Falnes et al., 2002; Trewick et been repaired, unmethylated Ada protein will accumu- al., 2002). These non-heme iron enzymes require Fe(II) late in the cell. High concentrations of unmethylated as a cofactor, and a-ketoglutarate and O2 as co- protein (4200 molecules per cell) inhibit ada gene substrates. They catalyze the hydroxylation of an activation by the methylated form, and will conse- unactivated C-H bond in the substrate coupled to the quently switch off the adaptive response in the absence oxidative decarboxylation of a-ketoglutarate, from of further damage (Saget and Walker, 1994). which the products are succinate and CO2. One atom of oxygen from O2 is incorporated as a hydroxyl in the substrate and the other in the carboxylate group of AlkB, a 1-meA/3-meC DNA dioxygenase succinate (Figure 5a) (for reviews see Prescott and Lloyd, 2000; Ryle and Hausinger, 2002; Schofield and Independently of the discovery of the Ada response, E. Zhang, 1999). Using this mechanism, AlkB reverts 1- coli alkA and alkB mutants were isolated in screens for meA and 3-meC in DNA directly to adenine and MMS sensitive mutants (Clarke et al., 1984; Kataoka et al., 1983; Yamamoto et al., 1978). Although the function of the AlkA protein was soon elucidated, the role of AlkB was only recently resolved (Dinglay et al., 2000; Falnes et al., 2002; Trewick et al., 2002). Evidence that AlkB acts alone came from its ability to confer MMS resistance to human cells (Chen et al., 1994). Also, the observation that AlkB repairs methylated single-stranded DNA in vivo predicted that 1-meA and 3-meC were its substrates (Dinglay et al., 2000). Despite these findings, many attempts to develop in vitro assays for AlkB were unsuccessful until clues of the unusual cofactor requirement were revealed (Aravind and Koonin, 2001). The substrates of AlkB have now been identified as 1-meA and 3-meC in DNA (Trewick et al., 2002). These lesions (Figure 1) are predominantly generated in single-stranded DNA by SN2 methylating agents, MMS, DMS or MeI, and to a much lesser degree by Figure 4 Topological diagrams for three members of the a-keto- glutarate-Fe(II) dioxygenase family. The diagrams are based on SN1 agents, MNNG and MNU (Bodell and Singer, the experimentally determined structure for E. nidulans isopenicil- 1979; Singer and Grunberger, 1983). The repair of lin-N-synthase and structural models of prolyl-4-hydroxylase and these modifications by AlkB explains the phenotypic AlkB. Conserved amino acid residues of the postulated active site characteristics of alkB mutants: their greater sensitivity and Fe(II) binding site are shown (reproduced with permission from Aravind and Koonin, 2001). The C-terminal third of the to SN2 than to SN1 alkylating agents (Kataoka et al., AlkB protein contains sequences that are not present in other 1983), their inability to reactivate single-stranded DNA family members, and may be required for DNA binding and phage treated with SN2, but not SN1, agents, and their substrate recognition

Oncogene Repair of alkylated DNA B Sedgwick and T Lindahl 8890 deaminate methylamine, both releasing formaldehyde (Lizcano et al., 2000). AlkB repairs lesions generated in single-stranded DNA, and may act at DNA replication forks and transcription bubbles. More efficient binding of AlkB to single- than to double-stranded DNA may direct AlkB to these regions (Dinglay et al., 2000). The hypothesis that AlkB acts at DNA replication forks and actively transcribed regions explains why rapidly growing E. coli alkB cells are more sensitive to MMS than those in stationary phase (Dinglay et al., 2000). An association of AlkB with DNA synthesis is also seen in Caulobacter crescentus in which the alkB gene is co-expressed with activities required for DNA replica- tion. In C. crescentus, neither AlkB nor Ada are induced by MMS, thus an Ada response is not apparent in this bacterium (see below Figure 7) (Colombi and Gomes, 1997). Direct reversal by AlkB of 1-meA and 3-meC to adenine and cytosine in DNA indicates an accurate mode of DNA repair. Other precise mechanisms of Figure 5 (a) Basic enzymatic reaction catalyzed by a-ketogluta- direct reversal of DNA damage are those catalyzed by rate-Fe(II)-dependent dioxygenases. Iron-oxo intermediates oxi- suicidal O6-methylguanine-DNA methyltransferases, dize inert substrates coupled to the oxidative decarboxylation of a-ketoglutarate to succinate. (b) Schematic representation of the light requiring photolyases that monomerize cyclobutyl repair of 1-meA and 3-meC residues in DNA by AlkB. Oxidation pyrimidine dimers (Sancar, 1996) and the B. subtilis of 1-meA and 3-meC by AlkB requires Fe(II) as a cofactor, O2 spore SP lyase which reverses the unique spore and a-ketoglutarate as co-substrates, and generates CO2 and photoproduct (5-thyminyl-5,6-dihydrothymine dimer) succinate. Oxidized methyl groups are released as formaldehyde resulting in direct reversal of the lesions to the unmodified to two thymines. Similarly to AlkB, SP lyase was base residues. Proposed intermediates are shown. Note that the discovered 20 years ago but has been only recently alkylation positions in 1-meA and 3-meC are equivalent, although biochemically characterized (Rebeil and Nicholson, these sites in the pyrimidine rings of A and C are numbered dif- 2001). This enzyme is a member of the ‘radical SAM’ ferently by standard nomenclature superfamily that use an (Fe-S) centre and S-adeno- sylmethionine to generate adenosyl radicals to effect catalysis. Thus, both AlkB and SP lyase use metal cytosine with the release of the oxidized methyl group catalysis to carry out difficult chemical reactions in as formaldehyde (Trewick et al., 2002). The inter- DNA repair, free radical intermediates being required mediate products are probably 1- to attack the stable lesions. hydroxymethyladenine and 3-hydroxymethylcytosine Site specific mutagenesis and crystal structure that are expected to be unstable and decompose to determinations have identified a conserved motif of release formaldehyde (Figure 5b). Enzymes in the three non-adjacent amino acids, His-X-Asp/Glu-Xn- superfamily of a-ketoglutarate-Fe(II)-dependent dioxy- His as part of the Fe(II) ligand site in a-ketoglutarate- genases catalyze a variety of reactions including dependent dioxygenases. This triad is also conserved hydroxylations, desaturations and oxidative ring in the E. coli AlkB protein and several of its putative closures. For comparison with AlkB, another enzyme homologues (Figures 4 and 6). An alkB mutant of C. that oxidizes methyl groups is thymine hydroxylase. crescentus is sensitive to MMS (Colombi and Gomes, This dioxygenase, involved in the pyrimidine salvage 1997), verifying that the C. crescentus sequence is a pathway of the fungi Neurospora crassa and Rhodotor- true homologue. Curiously, the second His of the ula glutinis, oxidizes the methyl group of free thymine, proposed iron ligand site in this sequence is replaced and by three successive oxidations forms 5-hydroxy- by Ser. This is also apparent in putative Mesorhizo- methyluracil, 5-formyluracil and 5-carboxyuracil. To bium loti (Figure 6) and Sinorhizobium meliloti finally generate uracil, the C-C bond of 5-carboxyuracil homologues (data not shown). A different His residue is cleaved by a carboxylase. In vitro, thymine closer to the C-terminus may provide the third iron hydroxylase can also demethylate 1-methylthymine coordinating ligand in these proteins. Crystal struc- initially forming 1-hydroxymethylthymine which tures of a-ketoglutarate-dependent dioxygenases decomposes to thymine and formaldehyde (Thornburg suggest that a-ketoglutarate forms an electrostatic et al., 1993). This latter reaction appears to be similar interaction with a conserved Arg residue (Zhang et to the demethylation of 1-meA and 3-meC by AlkB. al., 2000). Multiple sequence alignments of these Furthermore, these oxidation reactions catalyzed by enzymes with AlkB (Aravind and Koonin, 2001) AlkB and thymine hydroxylase are similar to those of indicate that a conserved Arg towards the C-terminus cytochrome P450 which demethylates N-methylamines of the AlkB homologues may be involved in a- (Yang and Smith, 1996) and amine oxidases that ketoglutarate interactions (Figure 6).

Oncogene Repair of alkylated DNA B Sedgwick and T Lindahl 8891 E. coli has constitutive and inducible O6-methyl- AlkA, a DNA glycosylase with broad substrate guanine-DNA methyltransferases (Ogt and Ada) and specificity 3-methyladenine-DNA glycosylases (Tag and AlkA) (Lindahl et al., 1988; Seeberg and Berdal, 1999). The 3-Methyladenine is an abundant alkylation lesion question of whether E. coli also has a constitutive generated by both SN1 and SN2 alkylating agents. It activity that repairs 1-meA and 3-meC is unresolved. is a strongly cytotoxic base because its methyl group The extreme sensitivity of AlkB mutants to alkylation protrudes into the minor groove of the DNA double in single-stranded DNA (Dinglay et al., 2000) implies helix, where it can block DNA replication and that a constitutive activity may not exist. Moreover, transcription. This lesion is repaired efficiently in all database searches of the E. coli genome have not living cells. E. coli possesses a constitutively expressed revealed any significant candidate homologue for such 3-methyladenine-DNA glycosylase, the product of the an activity (P Bates and B Sedgwick, unpublished tag gene, which appears highly specialized for the data). AlkB activity involves iron-oxo intermediates excision of 3-methyladenine (Lindahl et al., 1988; which may themselves be a threat to the cell due to Seeberg and Berdal, 1999). The E. coli alkA gene oxidative DNA and protein damage. Consequently, it encodes a second 3-methyladenine-DNA glycosylase, may be an evolutionary advantage to have high cellular which is inducible as part of the Ada response. The levels of AlkB only when it is absolutely essential for AlkA enzyme has a much broader substrate specificity survival, as occurs on exposure of E. coli to than Tag; it excises the minor lesion 3-methylguanine methylating agents or possibly during DNA replication relatively efficiently, whereas this is a poor substrate for in C. crescentus. The E. coli ada and alkB genes occur Tag. AlkA also catalyzes the release from DNA of the in a small operon regulated from the ada promoter. We can therefore predict that the AlkB protein will be present at a similar low level to the Ada protein in the uninduced cell, that is, approximately 2 to 4 molecules per cell (Potter et al., 1989; Rebeck et al., 1989; Vaughan et al., 1991). Although putative alkB sequence homologues are widespread from bacteria to humans, they are not universal, and are not found in the complete genome sequences of some microorganisms, for example, B. subtilis and Haemophilus influenzae, Archea or Saccharomyces cerevisiae. Curiously, three genes isolated from S. cerevisiae complement the sensitivity of an E. coli alkB, but not a recA, mutant to MMS. One of these genes encodes a protein kinase whereas Figure 7 Functional domain fusions and diverse gene organisa- tion in various bacterial species. Different coloured boxes repre- the others may be membrane glycoproteins. The sent different functional domains. Blue, AdaA; green, AdaB; explanation for this complementation remains yellow, AlkA; red, AlkB. The domains exist as separate proteins unclear (Wei et al., 1995). Several candidate human in B. subtilis, but are fused in different combinations in E. coli, AlkB homologues are found in the databases (P C. crescentus and Streptomyces coelicolor. Major lesions repaired by the different domains are indicated below the E. coli proteins. Bates and B Sedgwick, unpublished data), and it Small cross-hatched boxes represent Ada regulated promoters. In must be ascertained which of these have roles in C. crescentus, the alkB gene is co-expressed with genes involved in DNA repair. DNA replication

Figure 6 Alignment of the E. coli and C. crescentus AlkB proteins with other putative bacterial homologues. Protein accession references are: Vibrio cholerae, NP_23345; Pseudomonas aeruginosa, NP_251996; Xanthomonas campestris, NP_636023; E. coli, NP_416716; Brucella melitensis, NP_541717; C. crescentus, NP_418829; Mesorhizobium loti, NC_002678. Identical and conserved residues are highlighted in yellow, and similar residues in grey. Red stars indicate the Fe(II) ligand triad, HisXAspXnHis. The green star indicates an arginine residue that is possibly involved in a-ketoglutarate binding

Oncogene Repair of alkylated DNA B Sedgwick and T Lindahl 8892 minor products O2-methylcytosine and O2-methylthy- enzyme is comprised of three sub-domains and, as mine formed by SN1 agents, as well as the 8- found in several different DNA glycosylases, has a methylguanine lesion generated by methyl radicals helix – hairpin – helix motif for interaction with (Gasparutto et al., 2002). Furthermore, the AlkA damaged DNA positioned opposite a catalytic Asp enzyme liberates altered adenines such as ethenoade- residue in the active site. Similarly to most other nine generated as a consequence of exposure of DNA enzymes in the DNA glycosylase family, AlkA to lipid peroxidation products, hypoxanthine which is introduces a bend in its DNA substrate, followed by identical with hydrolytically deaminated adenine, and flipping-out of the damaged deoxynucleoside residue formyluracil and a ring fragmentation product of which is rotated towards the active site. Whereas other thymine generated by active oxygen (Privezentzev et DNA glycosylases with more precise and restricted al., 2000). These lesions are not generated by alkylating substrate specificity tend to place the substrate residue agents, so it is unclear why they should be repaired by in a deep and tightly fitting pocket, AlkA has a one of the enzymes of the Ada response. However, shallow groove or cleft, rich in aromatic and AlkA is present in cells at a higher constitutive level hydrophobic residues, which can accommodate the than Ada and AlkB, and may be sufficient to deal with diverse substrates of the enzyme. The prime considera- the small amounts of endogenously generated lesions. tion, however, should be the efficient excision of 3- The identification of 3-meA as a cytotoxic rather methyladenine from DNA. In this regard, it is than a mutagenic DNA lesion has been confirmed by interesting that a 3-methyladenine-DNA glycosylase, studies using the minor groove-specific alkylating agent induced in Helicobacter pylori on exposure to MNNG, Me-lex, a methyl sulphonate ester attached to a has a more restricted substrate specificity similar to dipeptide (Monti et al., 2002). This alkylating agent that of the E. coli Tag enzyme (O’Rourke et al., has a restricted target specificity; it forms 3-meA 2000). In spite of the detailed structural information efficiently, but does not generate adducts in the major on AlkA, it is still not entirely clear which features of groove of DNA, and binds preferentially to A/T rich the substrate are specifically recognized by the sequences. The glycosyl bond of 3-methyldeoxyadeno- enzyme. One possibility is that the altered charge sine moieties in DNA is relatively weak and has a half distribution in the key substrates, allowing for life of only 20 – 30 h at 378C and neutral pH. Whilst stronger interactions with electron-rich aromatic this rate of spontaneous hydrolytic cleavage is so slow amino acid residues in the active site, is a relevant that an active repair mechanism is required for removal recognition factor (Labahn et al., 1996). An alter- of the toxic 3-meA residues, a 104-fold acceleration of native model postulates that the decreased stability of the cleavage rate is sufficient to allow rapid DNA the glycosyl bond in the modified deoxynucleoside repair. This appears an easy task for enzymes, which residue is of crucial importance (Seeberg and Berdal, are often required to speed up reaction rates by 106 – 1999). 108-fold. In consequence, several reaction mechanisms have evolved for DNA glycosylases to remove 3-MeA from DNA in different organisms, giving the impres- Adaptive responses and diverse domain fusions in sion that 3-methyladenine-DNA glycosylase sequences different bacteria are poorly conserved. Thus, there is no obvious sequence homology between Tag, AlkA, and the A variable capacity for an Ada response occurs human enzyme although this latter protein has a among bacterial species. Such responses have been similarly structured active site to AlkA. observed by induction of cellular resistance to An intriguing consequence of the broad substrate alkylating agents, by enhanced DNA repair activities specificity of AlkA is that the enzyme can release the or by immunodetection of increased Ada protein normal bases guanine and adenine from DNA at a levels. Conserved protein homologues are now also very slow rate (Berdal et al., 1998). This is such a found in the sequenced genomes of many prokaryotes sluggish reaction that it is hardly relevant in vivo and eukaryotes. relative to the rate of spontaneous hydrolytic depur- The bacterial Ada and AlkA proteins provide an ination. At very high cellular levels of AlkA protein excellent example of the evolutionary fusion of protein this side reaction is troublesome, however, and may domains with related roles into ‘composite’ proteins. account for the surprisingly poor alkylation resistance Component domains that exist as separate proteins in of cells that greatly overexpress AlkA. By the same one species are fused in different combinations in reasoning, it would make sense to keep AlkA levels others (Marcotte et al., 1999). The 39 kDa Ada protein relatively low in normal proliferating cells, and only of E. coli is composed of two domains that account for induce the enzyme as part of the Ada response after a different DNA methyltransferase activities. Ada ortho- specific challenge by excessive DNA alkylation (Berdal logues of similar molecular weight are induced in et al., 1998; Posnick and Samson, 1999a). several other enterobacteria and Pseudomonas aerugi- The 3-D structure of AlkA is available and has nosa (Sedgwick and Vaughan, 1991), and database clarified the catalytic properties of the enzyme searches reveal homologues in other species, such as, (Labahn et al., 1996; Yamagata et al., 1996). A Mesorhizobium and Paracoccus (Figure 3). In B. comprehensive review (Hollis et al., 2000) describes subtilis, the two functional domains exist as separate the main structural features of AlkA. Briefly, the proteins, AdaA and AdaB, that demethylate methyl-

Oncogene Repair of alkylated DNA B Sedgwick and T Lindahl 8893 phosphotriesters and O6-methylguanine in DNA Conclusions respectively (Figure 7). Induction of AdaA and AdaB proteins also occurs in B. cereus, but several other All four of the major chemically methylated bases, and Bacillus species show a limited or undetectable also four minor modifications, are repaired by the inducible response (Morohoshi and Munakata, 1995). versatile inducible enzyme activities of the Ada Nevertheless, putative AdaA homologues are found in response (Figures 1 and 2). It remains unknown the genomes of B. halodurans, B. anthracis and whether two very minor lesions, 3-methylthymine and Lactococcus lactis (Figure 3). Intriguingly, a further 1-methylguanine, are repaired at all, or whether they mode of domain fusion, that of AdaA with AlkA, slowly accumulate as persistent lesions. A striking occurs in Mycobacterium tuberculosis, Streptomyces feature of the inducible activities is the diverse coelicolor, Xanthomonas campestris and Vibrio cholerae mechanisms that are required to remove the different (Seeberg and Berdal, 1999; B Sedgwick, unpublished alkylated bases from DNA. The unstable glycosyl bond data) (Figure 7). Fusion of AdaA with AlkA in of 3-meA, 3-meG, O2-meT or O2-meC in DNA is different bacterial species seems to be almost as hydrolytically cleaved by the AlkA DNA glycosylase, common as AdaA with AdaB. but the more unusual strategies of direct demethylation Organization of the adaptive response genes also by the Ada DNA methyltransferase and the AlkB differs between different species (Figure 7). The ada DNA dioxygenase are required to repair more and alkB genes form an operon in E. coli and alkA is chemically stable adducts. Ada transfers the methyl located separately. In contrast, in B. subtililis, the adaA groups from the methylated oxygens of O6-meG and and adaB constitute an operon with the alkA in an O4-meT in DNA on to itself, whereas AlkB oxidises adjacent position, and no alkB gene has been detected. and labilizes the methyl groups in 1-meA and 3-meC. The methylated AdaA protein of B. subtilis regulates O6-meG-DNA methyltransferases and 3-meA-DNA the adaA – adaB operon and the divergent alkA gene by glycosylases from human cells have been intensively binding to a central region between their transcrip- studied, and human AlkB is under investigation. These tional start sites (Morohoshi et al., 1993). In M. human DNA repair enzymes may suppress oncogenic tuberculosis and S. coelicolor, the adaA – alkA gene is transformation by endogenous and environmental adjacent to an adaB homologue. An inducible 17 kDa alkylating agents. On the other hand, they may have O6-methylguanine-DNA methyltransferase reported in the unfortunate effect of conveying cellular resistance V. cholerae (Bhasin and Ghosh, 1995) could be the to many prospective anti-cancer drugs. AdaB protein.

References

AkimaruH,SakumiK,YoshikaiT,AnaiMandSekiguchi Falnes PO, Johansen RF and Seeberg E. (2002). Nature, 419, M. (1990). J. Mol. Biol., 216, 261 – 273. 178 – 181. Aravind L and Koonin EV. (2001). Genome Biol., 2, Friedberg E, Walker G and Siede W. (1995). DNA Repair and Research 0007. Mutagenesis. Washington, D.C.: ASM Press. Berdal KG, Johansen RF and Seeberg E. (1998). EMBO J., Gasparutto D, Dherin C, Boiteux B and Cadet J. (2002). 17, 363 – 367. DNA Repair, 1, 437 – 447. Bhasin N and Ghosh A. (1995). Mutat. Res., 336, 79 – 89. Harrison KL, Jukes R, Cooper DP and Shuker DEG. (1999). Bodell WJ and Singer B. (1979). Biochemistry, 18, 2860 – Chem. Res. Toxicol., 12, 106 – 111. 2863. Hix S, Morais MS and Augusto O. (1995). Free Radical Bio. Boiteux S and Laval J. (1982). Biochimie, 64, 637 – 641. Med., 19, 293 – 301. Calmels S, Ohshima H, Rosenkranz H, McCoy E and Hollis T, Lau A and Ellenberger T. (2000). Mutat. Res., 460, Bartsch H. (1987). Carcinogenesis, 8, 1085 – 1088. 201 – 210. Chen BJ, Carroll P and Samson L. (1994). J. Bacteriol., 176, Kataoka H, Yamamoto Y and Sekiguchi M. (1983). J. 6255 – 6261. Bacteriol., 153, 1301 – 1307. Cheng X and Roberts RJ. (2001). Nucleic Acids Res., 29, Kunisaki N and Hayashi M. (1979). Appl. Envir. Microbiol., 3784 – 3795. 37, 279 – 282. Clarke ND, Kvaal M and Seeberg E. (1984). Mol. Gen. Labahn J, Scharer OD, Long A, Ezaz-Nikpay K, Verdine GL Genet., 197, 368 – 372. and Ellenberger TE. (1996). Cell, 86, 321 – 329. Colombi D and Gomes SL. (1997). J. Bacteriol., 179, 3139 – Landini P, Hajec LI and Volkert MR. (1994). J. Bacteriol., 3145. 176, 6583 – 6589. Crutzen PJ and Andreae MO. (1990). Science, 250, 1669 – Landini P and Volkert MR. (2000). J. Bacteriol., 182, 6543 – 1678. 6549. Daniels DS and Tainer JA. (2000). Mutat. Res., 460, 151 – Larson K, Sahm J, Shenkar R and Strauss B. (1985). Mutat. 163. Res., 150, 77 – 84. Demple B, Sedgwick B, Robins P, Totty N, Waterfield MD Lin Y, Dotsch V, Wintner T, Peariso K, Myers LC, Penner- and Lindahl T. (1985). Proc. Natl. Acad. Sci. USA, 82, Hahn JE, Verdine GL and Wagner G. (2001). Biochem- 2688 – 2692. istry, 40, 4261 – 4271. Dinglay S, Trewick SC, Lindahl T and Sedgwick B. (2000). Lindahl T, Sedgwick B, Sekiguchi M and Nakabeppu Y. Genes Dev., 14, 2097 – 2105. (1988). Ann. Rev. Biochem., 57, 133 – 157.

Oncogene Repair of alkylated DNA B Sedgwick and T Lindahl 8894 LizcanoMJ,UnzetaMandTiptonKF.(2000).Anal. Schofield CJ and Zhang Z. (1999). Curr. Opin. Struct. Biol., Biochem., 286, 75 – 79. 9, 722 – 731. Marcotte EM, Pellegrini M, Ng H-L, Rice DW, Yeates TO Sedgwick B. (1997). Carcinogenesis, 18, 1561 – 1567. and Eisenberg D. (1999). Science, 285, 751 – 753. Sedgwick B, Robins P, Totty N and Lindahl T. (1988). J. Matthews RG and Goulding CW. (1997). Curr. Opin. Chem. Biol. Chem., 263, 4430 – 4433. Biol., 1, 332 – 339. Sedgwick B and Vaughan P. (1991). Mutat. Res., 250, 211 – Monti P, Campomenosi P, Ciribilli Y, Iannone R, Inga A, 221. Shah D, Scott G, Burns PA, Menichini P, Abbondandolo Seeberg E and Berdal KG. (1999). Base Excision Repair of A, Gold B and Fronza G. (2002). J. Biol. Chem., 277, DNA Damage: Repair of Alkylation Damage to DNA. 28663 – 28668. Hickson I (ed). Austin, TX: Landes Bioscience, pp 151 – Moore MH, Gulbis JM, Dobson EJ, Demple B and Moody 168. PCE. (1994). EMBO J., 13, 1495 – 1501. Singer B and Grunberger D. (1983). Molecular Biology of Morohoshi F, Hayashi K and Munakata N. (1993). J. Mutagens and Carcinogens: Reactions of Directly Acting Bacteriol., 175, 6010 – 6017. Agents with Nucleic Acids. New York: Plenum Press. Morohoshi F and Munakata N. (1995). Mutat. Res., 337, Takahashi K and Kawazoe Y. (1987). Mutat. Res., 180, 97 – 110. 163 – 169. Myers LC and Verdine GL. (1994). Curr. Opin. Struct. Biol., Taverna P and Sedgwick B. (1996). J. Bacteriol., 178, 5105 – 4, 51 – 59. 5111. O’Rourke EJ, Chevalier C, Boiteux S, Labigne A, Ielp L and Teo I, Sedgwick B, Kilpatrick MW, McCarthy TV and Radicella JP. (2000). J. Biol. Chem., 275, 20077 – 20083. Lindahl T. (1986). Cell, 45, 315 – 324. Posnick LM and Samson LD. (1999a). J. Bacteriol., 181, Thornburg LD, Lai M-T, Wishnok JS and Stubbe J. (1993). 6763 – 6771. Biochemistry, 32, 14023 – 14033. Posnick LM and Samson LD. (1999b). J. Bacteriol., 181, Trewick SC, Henshaw TF, Hausinger RP, Lindahl T and 6756 – 6762. Sedgwick B. (2002). Nature, 419, 174 – 178. Potter PM, Kleibl K, Cawkwell L and Margison GP. (1989). Vaughan P, Lindahl T and Sedgwick B. (1993). Mutat. Res., Nucleic Acids Res., 17, 8047 – 8060. 293, 249 – 257. Prescott AG and Lloyd MD. (2000). Nat. Prod. Rep., 17, VaughanP,SedgwickB,HallJ,GannonJandLindahlT. 367 – 383. (1991). Carcinogenesis, 12, 263 – 268. Privezentzev CV, Saparbaev M, Sambandam A, Greenberg Volkert MR, Hajec LI, Matijasevic Z, Fang FC and Prince MM and Laval J. (2000). Biochemistry, 39, 14263 – 14268. R. (1994). J. Bacteriol., 176, 7638 – 7645. Rebeck GW, Smith CM, Goad DL and Samson L. (1989). J. Vora RA, Pegg AE and Ealick SE. (1998). Proteins: Struct. Bacteriol., 171, 4563 – 4568. Funct. Genet., 32, 3–6. Rebeil R and Nicholson WL. (2001). Proc. Natl. Acad. Sci. Wei Y-F, Chen BJ and Samson L. (1995). J. Bacteriol., 177, USA, 98, 9038 – 9043. 5009 – 5015. Ryle MJ and Hausinger RP. (2002). Curr. Opin. Chem. Biol., Yamagata Y, Kato M, Odawara K, Tokuno Y, Nakashima 6, 193 – 201. Y, Matsushima N, Yasumura K, Tomita K, Ihara K, Fujii Saget BM and Walker GC. (1994). Proc. Natl. Acad. Sci. Y, Nakabeppu Y, Sekiguchi M and Fujii S. (1996). Cell, USA, 91, 9730 – 9734. 86, 311 – 319. Sakashita H, Sakuma T, Akitomo Y, Ohkubo T, Kainosho Yamamoto Y, Katsuki M, Sekiguchi M and Otsuji N. (1978). M, Sekiguchi M and Morikawa K. (1995). Proc. Japan J. Bacteriol., 135, 144 – 152. Acad., 71, 198 – 201. Yang CS and Smith TJ. (1996). Biological Reactive Sakumi K and Sekiguchi M. (1989). J. Mol. Biol., 205, 373 – Intermediates V. Snyder R (ed). New York: Plenum Press, 385. pp 385 – 394. Samson L. (1992). Mol. Microbiol., 6, 825 – 831. Zhang Z, Ren J, Stammers D, Baldwin JE, Harlos K and Sancar A. (1996). Annu.Rev.Biochem.,65, 43 – 81. Schofield CJ. (2000). Nature Structural Biol., 7, 127 – 133.

Oncogene