The Bacillus subtilis Counterpart of the Mammalian 3-Methyladenine DNA Glycosylase Has Hypoxanthine and 1, N 6 -Ethenoadenine as Preferred Substrates*

The AAG family of 3-methyladenine DNA glycosylases was initially thought to be limited to mammalian cells, but genome sequencing efforts have revealed the presence of homologous proteins in certain prokaryotic species as well. Here, we report the first molecular characterization of a functional prokaryotic AAG homologue, i.e. YxlJ, termed bAag, from Bacillus subtilis . The B. subtilis aag gene was expressed in Escherichia coli , and the protein was purified to homogeneity. As expected, B. subtilis Aag was found to be a DNA glycosylase, which releases 3-alkylated purines and hypoxanthine, as well as the cyclic etheno adduct 1, N 6 -ethenoadenine from DNA. However, kinetic analysis showed that bAag re-moved hypoxanthine much faster than human AAG with a 10-fold higher value for k cat , whereas the rate of exci- sion of 1, N 6 -ethenoadenine was found to be similar. In contrast, it was found that bAag removes 3-methylad-enine and 3-methylguanine (cid:1) 10–20 times more slowly than human AAG, and there was hardly any detectable excision of 7-methylguanine. It thus appears that bAag has a minor role in the repair of DNA alkylation damage and an important role in preventing

The AAG family of 3-methyladenine DNA glycosylases was initially thought to be limited to mammalian cells, but genome sequencing efforts have revealed the presence of homologous proteins in certain prokaryotic species as well. Here, we report the first molecular characterization of a functional prokaryotic AAG homologue, i.e. YxlJ, termed bAag, from Bacillus subtilis. The B. subtilis aag gene was expressed in Escherichia coli, and the protein was purified to homogeneity. As expected, B. subtilis Aag was found to be a DNA glycosylase, which releases 3-alkylated purines and hypoxanthine, as well as the cyclic etheno adduct 1,N 6 -ethenoadenine from DNA. However, kinetic analysis showed that bAag removed hypoxanthine much faster than human AAG with a 10-fold higher value for k cat , whereas the rate of excision of 1, N 6 -ethenoadenine was found to be similar. In contrast, it was found that bAag removes 3-methyladenine and 3-methylguanine ϳ10 -20 times more slowly than human AAG, and there was hardly any detectable excision of 7-methylguanine. It thus appears that bAag has a minor role in the repair of DNA alkylation damage and an important role in preventing the mutagenic effects of deaminated purines and cyclic etheno adducts in Bacillus subtilis.
Exposure of genomes to a variety of reactive intracellular metabolites and environmental agents results in chemical modifications of the DNA nucleobases, including oxidations, alkylations, and deaminations. Such DNA lesions are primarily repaired through the base excision repair pathway (BER). The first step of BER involves N-glycosylic cleavage of the basesugar bonds by damage-specific DNA glycosylases. The abasic site is cleaved by apurinic/apyrimidinic endonucleases or apurinic/apyrimidinic lyases, and repair is completed through a sequential action of a phosphodiesterase, a DNA polymerase, and a DNA ligase (reviewed in Refs. 1 and 2).
Alkylating agents represent one of the most abundant classes of mutagenic and genotoxic agents present in the environment. Repair of alkylation damage is initiated by DNA glycosylases, referred to as 3-methyladenine (3mA) 1 DNA gly-cosylases, because 3mA is a major substrate for these enzymes. Escherichia coli possesses two enzymes of this kind, the constitutively expressed Tag (3) and the alkylation-inducible AlkA (4). The Tag enzyme has a narrow specificity for removal of 3mA, 3-methylguanine (3mG), and 3-ethyladenine (5), whereas AlkA has a much broader specificity toward a wide range of modified base residues in DNA (4, 6 -10). The Tag and AlkA proteins share no significant sequence homology despite their functional similarity. The yeast Saccharomyces cerevisiae possesses one 3mA DNA glycosylase, MAG, which is homologous to E. coli AlkA (11). However, mouse, rat, and human cells possess none of these types of enzymes but another, structurally different protein termed AAG/ANPG/MPG (the term AAG will be used here) (12). Mammalian AAG recognizes and removes a broad range of damaged base residues, including deaminated adenine (hypoxanthine), cyclic etheno adducts, e.g. 1,N 6 -ethenoadenine, and a variety of different alkylated base residues, including 3mA, 3mG, and 7-methylguanine (7mG) (8,(12)(13)(14)(15)(16)(17)(18)(19).
Through genome sequencing, open reading frames with homology to the mammalian AAG enzymes have also been identified in prokaryotic species, for instance in Bacillus subtilis, Mycobacterium tuberculosis, and Borrelia burgdorferi. B. subtilis also possesses two AlkA homologues and a third DNA glycosylase protein with close similarity to AlkC(YhaZ), a recently identified repair gene in Bacillus cereus. 2 The high number of 3mA DNA glycosylases in Bacillus species probably reflects the heavy exposure of earth bacteria to environmental alkylating agents like methyl chloride. E. coli possesses an inducible repair response to changes in environmental exposure, termed the adaptive response (20,21). The Ada protein, which is a methyltransferase that becomes activated upon methylation exposure, regulates the adaptive response in E. coli and induces transcription of AlkA and other repair proteins involved in protection against alkylation damage. A similar response system is also shown to be present in B. subtilis (22).
The fact that AAG enzymes have been conserved in some but not all bacteria raises questions about the possible functional significance of this type of enzyme in prokaryotic cells. To elucidate the role of AAG-like proteins in prokaryotes, we have expressed the AAG homologue of B. subtilis, YxlJ, in E. coli, and characterized its enzymatic activities. We have shown that bAag indeed is a functional DNA glycosylase, excising many of the same modified base residues as its human counterpart. However, its preference for different modified residues is significantly different from that of the human enzyme, and it * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  seems to have a more important role in the removal of premutagenic residues induced by deamination and lipid peroxidation rather than for cytotoxic residues induced by alkylating agents.

MATERIALS AND METHODS
Bacterial Strains and Plasmids-DNA from the B. subtilis strain 168 was used for the cloning of the aag gene. E. coli strain BK2118 (tag alkA) was used in survival assays on methyl methanesulfonate (MMS)containing plates and for expression of Bacillus aag, whereas ER2566 (New England Biolabs) and BL21 codon plus (Stratagene) were used for expression only. E. coli expression vectors included pUC18 (New England Biolabs), pUC19 (New England Biolabs), pT7SCII (USB), and pQE30 (Qiagen).
Purification of B. subtilis DNA-10 ml of overnight culture was washed twice in sterile water, and the cells were resuspended in 0.2 ml of 2% Triton X-100, 1% SDS, 100 mM NaCl, 10 mM Tris, pH8, 1 mM EDTA, 0.2 ml of phenol/chloroform/isoamyl alcohol (24:24:1), and 0.3 g of glass beads (Sigma). The suspension was vortexed three times for 1 min and cooled on ice in between to break the cell walls. The DNA was precipitated with 96% ethanol, incubated for 5 min at 37°C with RNase A, and precipitated with ethanol.
Cloning of B. subtilis aag-Genomic B. subtilis DNA was used as template in a PCR reaction to amplify a DNA fragment containing the B. subtilis aag gene, termed bAag. The forward primer 5Ј-AAAACTG-CAGTGATCTGCTTCGG-3Ј (PstI restriction site underlined) was located 45 nucleotides upstream of the putative GCG start codon, and the reverse primer 5Ј-CGGATCCTCAAAGCGCCG-3Ј (BamHI restriction site underlined) was located 55 nucleotides downstream of the putative stop codon. The PCR product was cloned into the PstI and BamHI restriction sites of pUC18, pUC19, pQE30, and pT7SCII.
Alkylation Survival of BK2118 (tag alkA) Transformed with B. subtilis aag-Exponentially growing BK2118 transformed with pUC19/aag and pUC19 (control), were harvested, washed in PBS, and incubated for 30 min on ice, and appropriate dilutions were spread on plates containing 100 g/ml ampicillin and 0, 0.1, 0.5, and 1 mM MMS. Plates were incubated at 37°C for 2 days, and surviving colonies were counted.
Expression and Purification of B. subtilis and Human AAG Protein-Various B. subtilis aag constructs were used for expression analysis in BK2118 (pUC19, pT7SCII, and pQE30), ER2566 (pT7SCII), and BL21-RIL codon plus (pT7) at different temperatures (37,22, and 16°C), and the highest expression levels were obtained using BK2118/pUC19 growing at 37°C. 2 liters of cell culture from freshly transformed colonies were grown in LB medium with 100 g/ml ampicillin to A 600 of 1 and induced by 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside for 2 h. Cells were resuspended in 50 mM Tris, pH 7, 500 mM NaCl, and10 mM ␤-mercaptoethanol and sonicated three times for 30 s. (Vibtra Cell Sonicator, Sonics and Materials Inc.). Extracts were diluted with 4 volumes of 50 mM Tris, pH 7, and 10 mM mercaptoethanol and loaded onto a phosphocellulose column (10 ml) equilibrated with buffer A (50 mM Tris, pH 7, 50 mM NaCl, and10 mM mercaptoethanol). Active fractions were step-eluted by buffer B (50 mM Tris, pH 7, 1 M NaCl, and 10 mM mercaptoethanol), and alkyl base DNA glycosylase activity was used to monitor bAag purification. Because of the BK2118 tag alkA phenotype, the extracts have no background 3mA DNA glycosylase activity. Active fractions were pooled, dialyzed against buffer C (50 mM MES, pH 6, 50 mM NaCl, and10 mM mercaptoethanol) and applied to a MonoS column (HR 5/5, Amersham Biosciences). The column was eluted with a linear gradient of 50 -500 mM NaCl in buffer C, and fractions were tested for glycosylase activity and analyzed by denaturing SDS-PAGE. Human AAG was purified as described previously (14).
Alkylbase DNA Glycosylase Assay-10,000 -40,000 dpm (glycosylase assay and HPLC, respectively) of calf thymus DNA alkylated with [ 3 H]methyl-N-nitrosourea (1.5 Ci/mmol; NET-408, PerkinElmer Life Sciences) was incubated with different amounts of cell extracts or purified protein for 30 min at 37°C as described previously (37). The DNA was precipitated with ethanol, and radioactivity in the supernatant was measured in a liquid scintillation counter (Tri-Carb 2900TR, Packard).
Assays for Enzyme Cleavage of Hypoxanthine, 1,N 6  P-5Ј-end-labeled duplex nucleotides in reaction buffer (70 mM MOPS, pH 7.5, 1 mM dithiothreitol, 1 mM EDTA, and 5% glycerol) were incubated for 30 min at 37°C and separated on a 20% denaturating polyacrylamide gel (Long Ranger, FMC BioProducts). 1 g/l purified E. coli Nfo enzyme was added to perform nicking of the sugar phosphate backbone after removal of the damaged base by the DNA glycosylases. The cleavage products were visualized with a PhosphorImager (Amersham Biosciences model 445SI).
The kinetic constants, K m and k cat were determined by incubating the respective proteins with increasing concentrations of oligonucleotides containing a single hypoxanthine or 1,N 6 -ethnoadenine. The initial velocities were calculated by Lineweaver/Burk plots.
DNA Sequencing and Sequence Analysis-Sequence analysis was performed using the GeneTingTang program (GeneTing, Lillestrøm, Norway). Homology searches were performed by the gapped BLAST alignment (30

An Open Reading Frame in Bacillus subtilis That Encodes a
Protein with Homology to the Mammalian 3-Methyladenine DNA Glycosylase-Mammalian 3mA DNA glycosylases have been characterized from mouse, man, and rat (13-15, 18, 23, 24) and are different in structure from those found in E. coli and most other bacteria. However certain bacterial species appear to possess protein homologues of the mammalian enzymes as judged from searches in the completed genome sequences. An alignment of putative AAG sequences from B. subtilis, M. tuberculosis, and B. burgdorferi shows 33, 39, and 37% identity to human AAG, respectively (Fig. 1). The yxlJ gene of B. subtilis, here renamed aag, translates into a polypeptide of 196 amino acids (GenBank TM identification number gi 3912954). Compared with the human AAG, this putative Bacillus protein contains all of the key residues involved in substrate recognition and catalysis, except for Tyr-162 and Tyr-165 which are replaced by His.
Cloning and Expression of B. subtilis Aag in E. coli-To characterize bAag, the coding region, including its promoter sequence, was cloned in the E. coli expression vector pUC19 (Fig. 2B). There are two putative start codons (ATG and GTG) located within the open reading frame, but upstream Shine-Dalgarno and promoter-like sequences makes the second putative start site, GTG, the most probable initiation codon (Fig.  2B). With this vector, bAag was expressed and purified to apparent physical homogeneity by a two-step procedure including phosphocellulose and MonoS chromatography (see "Materials and Methods"). The double mutant strain BK2118 (tag alkA) was used for the expression to avoid interference by possible contamination of endogenous 3mA DNA glycosylase activities during purification and characterization. The purification was monitored by standard assays for 3mA DNA glycosylase activity, and the final preparation was analyzed by SDS-PAGE (data not shown).

Inefficient Excision of Methylated Bases by B. subtilis Aag-
Mammalian AAG has been shown to remove various methylated bases, including 3mA, 3mG, and 7mG (13,15,24). We examined the ability of the purified bAag to remove alkylated bases from calf thymus DNA treated with [ 3 H]methyl-N-nitrosourea (Fig. 3). The amounts of methylpurines present in such DNA were determined by HPLC to be 65% 7mG, 10% 3mA, and 0.7% 3mG (5). The relative activities for release of the different alkylated base residues were analyzed by treating methylated DNA with different amounts of enzyme and quantifying the radiolabeled excision products after separation by HPLC. Enzyme activity measurements were performed with purified human AAG as well as with the B. subtilis enzyme to compare efficiency and specificity of base excision by the two enzymes (Fig. 3). It appears that the rate of removal of 3mA and 3mG by bAag for the same amount of enzyme proceeds ϳ10 -20 times more slowly than for human AAG, whereas the activity toward 7mG is Ͻ1%. The difference between human and Bacillus Aag in the efficiency of repair of alkylated bases indicates that bAag may not be the major activity in B. subtilis for removal of alkylated base residues.
Efficient Removal of Deaminated Adenine and 1,N 6 -Ethnoadenine-Mammalian AAG has been reported to be active against a wide variety of premutagenic lesions, such as hypoxanthine (25), cyclic etheno adducts, e.g.1,N 6 -ethenoadenine and N 2 -ethenoguanine (26), and 8-oxoguanine (27). The activity of the bAag to remove these lesions was analyzed on oligonucleotides with a single lesion. bAag cleaved 1,N 6 -ethenoadenine with an efficiency similar to that of the human AAG, whereas hypoxanthine excision was much more efficient with bAag than with the human enzyme (Fig. 4). Further experiments were performed to compare the kinetics of hypoxanthine or 1,N 6ethenoadenine removal by B. subtilis and human AAG. There was no difference in the K m and k cat values observed between  sapiens; gi 12084553). Identical residues in all members of the group are highlighted in black, strongly similar in dark gray, weakly similar in light gray. The residues of human AAG essential in catalytic activity and "flipping" (12), are indicated with bold or bold with underlining, respectively. The ClustalW sequence analysis program from Pole Bio-Informatique was used to align the sequences.
AAG and bAag for excision of 1,N 6 -ethenoadenine (Table I). In contrast, bAag showed a 10-fold higher k cat value for hypoxanthine removal than did human AAG, whereas the K m values were found to be similar. This indicates that the difference in repair of hypoxanthine could be attributed to an increased turnover capacity of bAag compared with human AAG. No activity toward 8-oxoG, formamidopyrimidine (faPy), or abasic sites was found to be associated with bAag (data not shown).
B. subtilis Aag Can Only Partially Complement the Alkyla-tion Sensitivity of the 3mA DNA Glycosylase-deficient tag alkA Double Mutant of E. coli-The AlkA and the Tag DNA glycosylases of E. coli are both involved in the removal of 3mA, and mutants lacking both enzymes are several orders of magnitude more sensitive to methylating agents than the wild type. To monitor the ability of bAag to replace Tag/AlkA in the repair of alkylated DNA, E. coli strain BK2118 (tag alkA), transformed by the pUC-bAag plasmid, was plated on MMS-containing LB plates (Fig. 5). Surviving colonies were counted and compared with the survival of wild type and non-transformed BK2118.
The results showed that expression of bAag yielded ϳ80% complementation at 0.1 mM MMS and only 8% at 0.5 mM. Full rescue is obtained with plasmids expressing similar amounts of E. coli AlkA (data not shown). This suggests that bAag cannot fully replace Tag and AlkA in the alkylation repair of E. coli and is consistent with the biochemical characterization showing that bAag has limited capacity for the removal of 3mA. In contrast, it was shown previously that human AAG provides full complementation of alkylation survival of the tag alkA double mutant (15). DISCUSSION We have characterized the product of the yxlJ gene in B. subtilis, a homologue of the 3-mA DNA glycosylases found in mammalian cells. Like the mammalian enzymes, bAag is a DNA glycosylase removing several different types of damaged base residues from DNA. To our knowledge, this appears to be the first characterization of a prokaryotic enzyme of this type. AAG homologues are also found in several other (among them pathogenic) bacterial species. Surprisingly, HPLC analysis revealed that removal of the cytotoxic lesions 3mA and 3mG is 10 -15 times less efficient for bAag than for human AAG. In contrast, bAag is 10-fold more efficient than human AAG in the excision of the premutagenic adenine deamination product, hypoxanthine. The two enzymes are similar in their affinity toward the cyclic etheno adduct 1,N 6 -ethenoadenine. It thus appears that the primary function of the B. subtilis AAG enzyme would be to avoid mutagenesis and repair DNA containing deaminated adenine and cyclic etheno adducts rather than to prevent the cytotoxic effects of methylated base residues. This agrees with the presence in B. subtilis of several other 3mA DNA glycosylases like, for example, two AlkA homologues FIG. 2. Cloning of B. subtilis aag. A, organization of   (AlkA and YfjP) that could be primarily responsible for the repair of alkylation damage. The incomplete restoration of alkylation resistance in the E. coli double mutant alkA tag expressing bAag is consistent with this notion and suggests a less important role of this enzyme in the repair of alkylation damage. In contrast, AAG appears to be the only glycosylase present in human cells responsible for base excision repair of alkylation damage. Assuming that the AAG enzymes originate from prokaryotes already possessing other enzymes for alkylation repair, it can be inferred that Aag in mammals has evolved from being enzymes primarily involved in the removal of premutagenic lesions, e.g. cyclic etheno adducts and deaminated adenine, toward enzymes with broader substrate specificity, including cytotoxic alkylation products. Repair of hypoxanthine in E. coli can be initiated by two different enzymes, the functional homologue of AAG DNA glycosylase, AlkA, or a damaged base-specific endonuclease, Nfi, incising the DNA strand two nucleotides 3Ј of the lesion (28). Nfi appears to be the major activity for repair of hypoxanthine in E. coli as judged from the high nitrite-induced mutation frequency of the nfi mutant (29). Comparison of hypoxanthine removal by different 3mA DNA glycosylases shows that human and rat AAG are 100 -300-fold faster than AlkA from E. coli and Mag from S. cerevisiae (30). Our results indicate that the prokaryotic AAG 3mA DNA glycosylases are even more efficient enzymes in the removal of hypoxanthine than the mammalian AAG and, consequently, the structurally unrelated AlkA family of proteins also. Interestingly, the genome sequence of B. subtilis also reveals the presence of a gene that translates into a protein with homology to Nfi. Studies of this gene function are required to elucidate whether this represents an alternative mechanism for repair of hypoxanthine in B. subtilis.
Etheno bridged exocyclic DNA adducts, e.g. 1,N 6 -ethenoadenine and 3,N 4 -ethenocytidine, can arise spontaneously in the cell and will also be induced by mutagens that induce lipid peroxidation or directly by other mutagens such as vinyl chloride and urethane (31). Because AAG appears to be the major ethenoadenine DNA glycosylase in mammalian cells, it is possible that bAag also has a role in repair of premutagenic etheno adducts in B. subtilis.
Lau et al. (12,32) elucidated the molecular basis for recognition and catalytic activity of the human AAG enzyme. They performed structural and mutational analysis to identify amino acids essential for the enzyme's ability to differentiate between normal and damaged bases. Sequence alignment of the human and bAag shows that all amino acids of human AAG involved in recognition and catalysis are conserved in the B. subtilis sequence, except for Tyr in positions 162 and 165, which are both substituted by His in the Bacillus sequence. Site-specific replacement of Tyr-162 or Tyr-165 with Ala reduces efficiency in the repair of alkylated DNA (12). Substitution of these residues with His could thus also possibly explain the reduced activity of B. subtilis Aag for alkylated base residues. However, His residues are clearly closer to Tyr than Ala, and more site-specific mutagenesis experiments will be required to confirm this suggestion.
The B. subtilis aag gene is localized between a gene coding for a putative catalase and a gene of unknown function. The catalase and bAag are transcribed in opposite directions with putative Shine-Dalgarno AG-rich sequences and AT-rich promoter-like sequences identified upstream of their respective start codons, and their predicted promoter regions are separated by 20 nucleotides (Fig. 2). The close localization of the two genes and the fact that they are both probably involved in a regulated response to (DNA) damage raise the question of a common regulation of the two genes. Several B. subtilis genes are regulated through the B. subtilis variant of the E. coli SOS box, the so-called Din box (33,34); however, the upstream sequence of bAag (and the catalase gene as well) did not contain any sequence homologous to the B. subtilis Din box sequence. Nevertheless, it could still be speculated that common regulatory elements might be involved in the expression of aag and the catalase function in B. subtilis.
With all the different genome sequences now available, it becomes possible to make conclusions about the genetic complement of DNA repair in different organisms. Clearly, important differences exist also between different bacteria. The B. subtilis genome apparently has many DNA glycosylases for the repair of alkylation damage; however, the genome is apparently lacking an AlkB function, which was recently shown to be required for repair of 1mA and 3mC in E. coli (35,36). The presence of the mammalian type of 3mA DNA glycosylases in some but not in most bacteria also has a functional implication, which we have alluded to in this communication. It also appears important to realize that homologous genes do not have exactly the same function in all species but that the sequence variations might reflect important functional differences.