Imbalanced Base Excision Repair Increases Spontaneous Mutation ...

JOURNAL OF BACTERIOLOGY, Nov. 1999, p. 6763–6771 0021-9193/99/$04.00⫹0 Copyright © 1999, American Society for Microbiology. All Rights Reserved.

Vol. 181, No. 21

Imbalanced Base Excision Repair Increases Spontaneous Mutation and Alkylation Sensitivity in Escherichia coli LAUREN M. POSNICK

AND

LEONA D. SAMSON*

Division of Toxicology, Department of Cancer Cell Biology, Harvard School of Public Health, Boston, Massachusetts 02115 Received 6 April 1999/Accepted 26 August 1999

Inappropriate expression of 3-methyladenine (3MeA) DNA glycosylases has been shown to have harmful effects on microbial and mammalian cells. To understand the underlying reasons for this phenomenon, we have determined how DNA glycosylase activity and substrate specificity modulate glycosylase effects in Escherichia coli. We compared the effects of two 3MeA DNA glycosylases with very different substrate ranges, namely, the Saccharomyces cerevisiae Mag1 and the E. coli Tag glycosylases. Both glycosylases increased spontaneous mutation, decreased cell viability, and sensitized E. coli to killing by the alkylating agent methyl methanesulfonate. However, Tag had much less harmful effects than Mag1. The difference between the two enzymes’ effects may be accounted for by the fact that Tag almost exclusively excises 3MeA lesions, whereas Mag1 excises a broad range of alkylated and other purines. We infer that the DNA lesions responsible for changes in spontaneous mutation, viability, and alkylation sensitivity are abasic sites and secondary lesions resulting from processing abasic sites via the base excision repair pathway. AlkA can also act on the simple alkyl lesions O2-methylcytosine and O2-methylthymine and the oxidized thymine products 5-hydroxymethyluracil and 5-formyluracil (3, 41). More recently, both Mag1 and AlkA have been shown to remove normal, undamaged bases from DNA in vitro (2). Cells deficient in 3MeA DNA glycosylase activity are very sensitive to killing by alkylating agents, demonstrating the important role these enzymes play in repairing lethal alkylation damage (12, 13). However, overexpression of 3MeA DNA glycosylases has adverse effects on cells, paradoxically increasing sensitivity to treatment with alkylating agents as well as increasing spontaneous mutation (2, 8, 15, 19, 21, 22, 49). Such effects have generally been attributed to the direct or indirect results of excessive abasic site formation (2, 8, 14, 15, 22). Other DNA glycosylases that participate in BER have been implicated in causing similar types of damage. Uracil DNA glycosylase activity increases spontaneous mutation in cells with high levels of uracil in DNA (e.g., in E. coli dut mutants) (reviewed in reference 14). Furthermore, expression of mutant uracil glycosylases that can remove C’s or T’s also confers a mutator phenotype (25). Glassner et al. previously found that overexpression of the S. cerevisiae Mag1 3MeA DNA glycosylase dramatically increased spontaneous mutation in S. cerevisiae and in AP endonuclease-deficient (xth nfo) E. coli (15). Here, we compared the abilities of the Mag1 and Tag glycosylases to increase spontaneous mutation, decrease cell viability, and increase alkylation sensitivity in both wild-type and xth nfo E. coli. We examined the spectra of mutations induced by each glycosylase and the ability of umuDC and recBC mutations to modulate glycosylase effects. We observed wide variations in the effects of the two glycosylases, and we infer that the differences are due to innate differences in their substrate specificities.

Alkylating agents transfer potentially dangerous alkyl groups to nucleophilic sites in DNA. Simple alkylating agents such as methyl methanesulfonate (MMS) or N-methyl-N-nitrosourea produce more than a dozen different lesions in DNA, including the exocyclic oxygen lesion O6-methylguanine (O6MeG) and the ring nitrogen lesions 3-methyladenine (3MeA) and 7-methylguanine (7MeG) (13). O6MeG is a mutagenic lesion that induces G:C-to-A:T transition mutations, while 3MeA is a lethal lesion that blocks DNA replication (13, 30). 7MeG is the most abundant lesion produced by simple alkylating agents but is generally believed to be innocuous (13, 31). In Escherichia coli, 3MeA lesions are repaired by the base excision repair (BER) pathway (reviewed in references 13, 41, and 48). The first step in this multistep process is the enzymatic cleavage by 3MeA DNA glycosylases of the glycosylic bond that connects bases to DNA. E. coli has two 3MeA DNA glycosylases; the constitutively expressed Tag glycosylase and the alkylation-inducible AlkA glycosylase (12, 13, 23). Repair of the apurinic or apyrimidinic (AP or abasic) site resulting from DNA glycosylase activity is usually initiated by an AP endonuclease that makes a nick 5⬘ to the AP site. The resulting 5⬘ blocking fragment is subsequently removed by deoxyribophosphodiesterase, and repair of the gapped DNA strand is completed by DNA polymerase I and DNA ligase. In E. coli, the major AP endonucleases are endonuclease IV (encoded by the nfo gene) and exonuclease III (xth) (13). The AlkA and Tag glycosylases have remarkably different substrate ranges. Tag acts almost exclusively on the lethal alkyl lesion 3MeA, although it can also act on the rarer lesion 3-methylguanine (3MeG), albeit inefficiently (4). In contrast, AlkA and many other 3MeA DNA glycosylases, such as the Saccharomyces cerevisiae Mag1 glycosylase, have very broad substrate ranges. Both AlkA and Mag1 can act on 7MeG, 7-methyladenine, 7-chloroethylguanine, 7-hydroxyethylguanine, hypoxanthine, and 1,N6-ethenoadenine (4, 5, 23, 33, 39, 40, 48).

MATERIALS AND METHODS Plasmids and bacterial strains. The tag and MAG1 coding sequences were obtained from the plasmids pSL-Tag and pSL-MAG1 (15a) as BamHI-mung bean nuclease-HindIII fragments and cloned into XmaI-Klenow fragmenttreated pBAD24 (16) to create p24Tag and p24Mag (pBAD24 was kindly provided by L.-M. Guzman and J. Beckwith, Harvard Medical School, Boston, Mass.). For higher-level Tag expression, the tag coding sequence was subcloned

* Corresponding author. Mailing address: Department of Cancer Cell Biology, Harvard School of Public Health, 665 Huntington Ave., Boston, MA 02115. Phone: (617) 432-1085. Fax: (617) 432-0400. E-mail: [email protected]. 6763

6764

POSNICK AND SAMSON

J. BACTERIOL. TABLE 1. E. coli strains and plasmids used in this study

Strain or plasmid

Strains CSH101 CSH102 CSH103 CSH104 CSH105 CSH106 CSH106 umuC122::Tn5 TN1018 TN1028 TN1038 TN1048 TN1058 TN1068 TN1068 ⌬(umuDC)595::cat AB1157 GW2100 GW8017 GW8014 GW2771 MV1932 JCSS19 Plasmids pBAD24 p24Mag p24Tag pSL-MAG1 pSL-Tag pSE380 pSE-Tag pSU18-DC a

Genotype or description

Reference or source

ara ⌬(gpt-lac)5 F⬘ lac proA⫹B⫹; A:T to C:G As CSH101 but F⬘ G:C to A:T As CSH101 but F⬘ G:C to C:G As CSH101 but F⬘ G:C to T:A As CSH101 but F⬘ A:T to T:A As CSH101 but F⬘ A:T to G:C CSH106 umuC122::Tn5 CSH101 nfo1::kan ⌬(xth-pncA)90 zdh-201::Tn10 CSH102 nfo1::kan ⌬(xth-pncA)90 zdh-201::Tn10 CSH103 nfo1::kan ⌬(xth-pncA)90 zdh-201::Tn10 CSH104 nfo1::kan ⌬(xth-pncA)90 zdh-201::Tn10 CSH105 nfo1::kan ⌬(xth-pncA)90 zdh-201::Tn10 CSH106 nfo1::kan ⌬(xth-pncA)90 zdh-201::Tn10 TN1068 ⌬(umuDC)595::cat argE3 hisG4 leuB6 proA2 thr-1 ara-14 galK2 lacY1 mtl-1 xyl-1 thi-1 rpsL31 supE44 tsx-33 AB1157 umuC122::Tn5 AB1157 ⌬(umuDC595)::cat AB1157 recA441 lexA300(Def)::spc sfiA11 sfiC2 GW8014 recA⫹ lexA⫹ AB1157 ada alkA1a AB1157 recB21 C22

35 35 35 35 35 35 This work B. Demple B. Demple B. Demple B. Demple B. Demple B. Demple This work

Ampr, arabinose inducible MAG1 sequence in pBAD24 tag sequence in pBAD24 MAG1 sequence in pSL301 tag sequence in pSL301 IPTG inducible, Ampr tag sequence in pSE380 umuDC operon; Camr; pACYC184 ori

J. Beckwith (16) This work This work B. Glassner and L. Samson B. Glassner and L. Samson Invitrogen This work G. Walker

(27) (27) (27) (27) (27) (27)

G. Walker G. Walker G. Walker G. Walker M. Volkert M. Zaman

MV1932 was originally reported to be alkA1 tag (see references 12, 24, and 38). However, MV1932’s tag mutation was recently found to be an ada mutation 43.

as a BamHI-mung bean nuclease-HindIII fragment into NcoI-mung bean nuclease-HindIII-digested pSE380 (Invitrogen) to create pSE-Tag. Plasmid pSU18-DC, containing the umuDC operon in a pACYC184-based vector, was a kind gift of T. Opperman and G. Walker, Massachusetts Institute of Technology, Cambridge. The relevant genotypes of the strains and plasmids used in these experiments are listed in Table 1. E. coli CSH101 to CSH106 (also known as CC101 to CC106) (9) and TN1018 to TN1068, the nfo1::kan ⌬(xth-pncA)90 zdh-201::Tn10 derivatives of CSH101 to CSH106 (27), were obtained from Bruce Demple, Harvard School of Public Health, Boston, Mass. Strains GW8014, GW2100, GW2771, and GW8017 were obtained from G. Walker. JCSS19 was obtained from M. Zaman, Harvard School of Public Health. MV1932 was obtained from M. Volkert, University of Massachusetts Medical School, Worcester. P1 transductions. The umuC122::Tn5 allele was moved from strain GW2100 into CSH106, and the ⌬(umuDC)595::cat mutation was moved from GW8017 into TN1068 by standard techniques (35). The umuC122 transductants were screened for UV sensitivity by a UV gradient plate assay (described below), and the ⌬umuDC transductants were screened for the presence of the deletion allele by Southern blot analysis. Cell extracts. Overnight cultures of the alkA⫹ tag⫹ strain CSH106 or its derivatives were diluted 1:100 into maltose-ampicillin medium, i.e., minimal A medium (35) supplemented with 0.2% maltose, ampicillin (100 ␮g/ml), thiamine (0.0005%), and methionine (40 ␮g/ml). Cultures were grown to log phase at 37°C, induced with arabinose or isopropyl-␤-D-thiogalactopyranoside (IPTG), and harvested by centrifugation at 4°C after 3 h of additional growth. Pelleted cultures were frozen in liquid nitrogen and stored at ⫺80°C. Extracts were made by thawing frozen pellets on ice and sonicating them in chilled glycosylase reaction buffer (70 mM HEPES [pH 7.8], 1 mM dithiothreitol, and 5 mM EDTA) containing 5% glycerol. 3MeA DNA glycosylase assays. Glycosylase activity was measured by incubating cell extracts with a 3H-labeled alkylated DNA substrate and measuring release of 3H-labeled alkylated bases (14). The substrate was calf thymus DNA treated with N-[3H]methyl-N-nitrosourea (specific activity ⫽ 17.9 Ci/mmol; equivalent to 23.9 cpm/fmol). Reactions were carried out for 1 h at 37°C in glycosylase reaction buffer. After sodium chloride-ethanol precipitation of the DNA substrate and extract proteins, the supernatants were dried down under

vacuum and resuspended in 0.1 N hydrochloric acid for descending paper chromatography in a 7:1:2 mixture of isopropanol, ammonium hydroxide, and water. 3MeA-containing spots were visualized by UV fluorescence of markers, cut from the paper, eluted in water, and counted by scintillation. Gradient plates and killing curves. MMS gradient plates were prepared by pouring agar containing 0.01% MMS into a square petri dish laid on a slant, placing the plate flat after solidification, and overlaying the slant with MMS-free agar. Overnight cultures were diluted into Luria-Bertani (LB)-ampicillin (100 ␮g/ml) medium containing inducer, grown to approximately 108 CFU/ml, and stamped across the gradient. Plates were scored after growth overnight at 37°C. For UV sensitivity assays, cultures were stamped on an LB plate and a UV step gradient was created by successively unshielding portions of the plate during UV exposure. For most killing curves, overnight cultures were diluted 104-fold into maltoseampicillin medium plus 0.2% arabinose or 1 mM IPTG, grown for 12 to 16 h, and treated with MMS at 37°C for 30 min. Survival was scored by plating dilutions on LB-ampicillin plates. For additional killing curves with strains containing pSE380-based vectors, overnight cultures were diluted 1:100 in maltose-ampicillin medium plus 1 mM IPTG, grown to log phase, treated with 0.3% MMS for the indicated times, and diluted and plated on LB-ampicillin. Mutation assays. For Rifr assays in CSH106-derived strains, overnight cultures were diluted 104-fold into maltose medium containing arabinose (for pBADbased vectors) or IPTG (for pSE380-based vectors), plus ampicillin and/or chloramphenicol (40 ␮g/ml). Ten 1-ml cultures were grown for 24 h at 37°C, concentrated, and plated on LB-rifampin (100 ␮g/ml) plates supplemented with ampicillin and chloramphenicol as appropriate. Rifr colonies were counted after 48 h. Prior to concentrating and plating, titers of three cultures were determined on LB plates containing ampicillin and/or chloramphenicol to determine colonyforming ability. Mutant frequencies were calculated by dividing median mutant number by CFU. Rifr assays with GW8014 and GW2771 were done as described above, with the following exceptions. Overnight cultures were diluted 106-fold into LB-ampicillin plus 1 mM IPTG, and 15 to 20 1-ml cultures were grown for 14 h at 37 or 42°C. For GW8014, 50- to 100-␮l aliquots of the cultures were plated on LB-rifampinampicillin plates. For lactose reversion assays, strains CSH101 to CSH106 (pBAD24 and

3MeA DNA GLYCOSYLASE OVEREXPRESSION IN E. COLI

VOL. 181, 1999

6765

FIG. 1. Glycosylase constructs increase 3MeA DNA glycosylase activity in E. coli cell extracts. Extracts were prepared as described in Materials and Methods. (A) 3MeA DNA glycosylase activity from p24Mag in the wild-type strain CSH106 as a function of arabinose concentration. The line with multiplication signs shows mean glycosylase activity; the solid triangles and open circles show the individual values for the two experiments at each arabinose concentration. (B) 3MeA DNA glycosylase activity from pSE-Tag or the control vector pSE380 in wild-type (WT) and xth nfo backgrounds after induction with 1 mM IPTG. The bars show mean glycosylase activity; the solid triangles and open circles show the individual values for the two experiments in each strain background.

p24Mag) and TN1018 to TN1068 (pSE380 and pSE-Tag) were grown and treated as described above, but the medium contained 0.02% arabinose (pBAD24 and p24Mag) or 1 mM IPTG (pSE380 and pSE-Tag), the cultures were plated on lactose-minimal A-ampicillin plates to measure Lac⫹ reversion, and colony-forming ability was determined on glucose-ampicillin plates. Also, for Tag assays, five 5-ml cultures were used instead of 10 1-ml cultures. Lac⫹ revertants were counted 3 days after plating.

RESULTS Modulation of 3MeA DNA glycosylase levels in E. coli. 3MeA DNA glycosylases normally play a beneficial role by repairing lethal alkylation damage, but imbalanced expression of these glycosylases relative to that of the other BER enzymes can have unexpected and injurious effects (2, 8, 15, 21). As part of our studies on endogenous DNA damage and spontaneous mutation, we set out to examine the effects on E. coli of 3MeA DNA glycosylase overexpression. In particular, we chose to study the effects of expressing the E. coli Tag and the S. cerevisiae Mag1 glycosylases. Mag1 was chosen because its inappropriate expression had previously been shown to cause a very strong spontaneous mutator phenotype in S. cerevisiae (15, 27, 49). Tag was chosen because it is native to E. coli and because its substrate range is very different from that of Mag. Tag’s substrate range is virtually limited to 3MeA, whereas that of Mag1 extends from 3MeA to include 3MeG, 7MeG, hypoxanthine, 1,N6-ethenoadenine, and even normal guanines (2, 4, 5, 23, 33, 39, 40, 48). To express the glycosylases in a tightly regulated manner, we cloned the MAG1 coding sequence under the arabinose promoter in the vector pBAD24. The tag sequence was cloned under the stronger, IPTG-regulated trc promoter in the vector pSE380 to maximize Tag expression. To confirm that the constructs produced active 3MeA DNA glycosylase, we tested whether extracts from induced cells could release the 3H-labeled alkylated base 3MeA from DNA in vitro. Figure 1A shows that arabinose induction of Mag1 caused substantial

increases in 3MeA DNA glycosylase activity, proportional to the arabinose concentration. Likewise, IPTG induction of Tag produced high levels of 3MeA DNA glycosylase activity (Fig. 1B). As further confirmation that the glycosylase constructs p24Mag and pSE-Tag produced active 3MeA DNA glycosylase, we tested their ability to protect a 3MeA DNA glycosylase repair-deficient strain from killing on MMS gradient plates. Both glycosylase constructs fully protected the alkA strain MV1932 from killing by a low (0.01%) concentration of MMS, indicating that the expressed proteins are biologically active (data not shown). Mag1 and Tag increase spontaneous mutation to rifampin resistance. We tested whether overexpression of 3MeA DNA glycosylases in E. coli might increase spontaneous mutation to Rifr. Induction of Mag1 in a wild-type strain of E. coli (CSH106) dramatically increased spontaneous mutation to Rifr, by more than 100-fold over control levels at the maximum arabinose dose (Fig. 2A), and the increase in spontaneous mutation was proportional to the increase in Mag1 glycosylase activity (Fig. 2B). Note that growth of E. coli containing the pBAD24 control vector in 0.2% arabinose did not increase mutation (Fig. 2A), confirming that arabinose is not mutagenic and that Mag1 induction was responsible for the increased mutation in the Mag1-expressing strain. Unlike Mag1, Tag did not cause an increase in Rifr mutant frequency in a wild-type strain (Fig. 3). We reasoned that an xth nfo strain, deficient in the ability to repair abasic sites, might be more sensitive to Tag’s effects. Indeed, Tag expression in an xth nfo background produced a mild 2.5-fold increase in mutant frequency (Fig. 3). The finding that Tag selectively promotes mutation in an xth nfo background suggests that Tag produces abasic sites but that only in xth nfo cells do these sites persist long enough to lead to mutations. Tag and Mag1 clearly have quantitatively different mutator

6766

POSNICK AND SAMSON

J. BACTERIOL.

FIG. 2. Mag1 expression increases spontaneous mutation. (A) Mutant frequency versus concentration of the inducer arabinose. The y axis shows Rifr mutant frequency in the alkA⫹ tag⫹ strain CSH106; the x axis shows arabinose concentration. The Mag1 strain contains p24Mag; the control strain contains pBAD24. The line with multiplication signs shows mean mutant frequency; the squares, diamonds, and circles show values for individual experiments. Mag1 values are shown by solid or shaded symbols; control values are shown by open symbols. For Mag1, n is 3, except for 0% arabinose, where n is 2. For the control strain, n is 2, except for 0% arabinose, where n is 1. (B) Mutant frequency versus Mag1 glycosylase activity. The data on mean glycosylase activity are from Fig. 1, and the data on mean mutant frequency are from panel A. Errors are as shown in those figures.

effects. The difference is unlikely to result from unequal activity for 3MeA removal, since Tag extracts had much higher 3MeA DNA glycosylase activity in vitro (Fig. 1). A more likely explanation lies in the different substrate preferences for each enzyme. Tag acts primarily on 3MeA, although a minor activity for 3MeG excision has been noted elsewhere (4). Mag1 has been shown to remove not only 3MeA, but also 7MeG, ethenoadenine, hypoxanthine, and normal guanines (2, 5, 39, 40). Thus, Mag1 may be capable of producing many more premutagenic abasic DNA lesions than Tag.

FIG. 3. Tag expression increases spontaneous mutation in an xth nfo background. The wild-type strain is CSH106; the xth nfo strain is TN1068. Error bars represent standard errors of the means (n ⫽ 4).

Dependence of increased spontaneous mutation on UmuDC activity. In E. coli, the UmuD and UmuC proteins are required for efficient bypass replication of abasic sites (10, 32). If glycosylase-derived abasic sites are responsible for the increase in mutations associated with glycosylase overexpression, then the appearance of these mutations should be dependent on the umuDC genes. We therefore tested whether the 3MeA DNA glycosylases could induce spontaneous mutation in a umuDCdeficient background. Figure 4 shows that the umuC122::Tn5 null allele strongly suppressed the Mag1 mutator effect. It should be noted that Mag1 DNA glycosylase activity was not reduced in the umuC background (data not shown). To confirm that the Mag1-induced increase in spontaneous mutation was truly UmuDC dependent, we reintroduced functional UmuC protein by coexpressing the Mag1 vector with pSU18-DC containing the umuDC operon. The pSU18-DC plasmid not only restored the mutator phenotype associated with Mag1 overexpression but also enhanced Mag1’s ability to induce mutations in the wild-type background (data not shown). These results are consistent with the hypothesis that Mag1-induced mutations result from UmuDC-mediated DNA lesion bypass at abasic sites. To test whether Tag’s ability to increase spontaneous mutation in xth nfo cells was also umuDC dependent, we introduced the ⌬(umuDC)595::cat allele (47) into the xth nfo strain. Table 2 (top) shows that the ⌬umuDC allele did not suppress Tag’s ability to increase mutation in the xth nfo background, as would be expected for mutations resulting from abasic sites. This result was surprising, given that Tag selectively increased mutations in an xth nfo background deficient in abasic site repair, but not in a wild-type background (Fig. 3). Furthermore, we


VOL. 181, 1999

FIG. 4. A umuC mutation suppresses the Mag1-induced mutator effect. The strain background is CSH106, the umuC mutation is umuC122::Tn5, ⫹ Mag1 refers to p24Mag, and ⫹ Control refers to pBAD24. All experiments were conducted with 0.2% arabinose. Error bars show standard errors of the means (n ⫽ 3). WT, wild type.

also expressed Tag in a constitutively SOS-induced strain and found that Tag selectively increased mutations in the SOSinduced background but not in the non-SOS-induced parental background (Table 2, bottom). These results appear to contradict the results from the ⌬umuDC strain and suggest that Tag mutations do have a UmuDC-dependent component. An alternative possibility is that some Tag-induced mutations are UmuDC dependent (for example, in the xth nfo or constitutively SOS-induced background) but that Tag is able to induce mutations by a UmuDC-independent pathway in xth nfo ⌬umuDC triple mutants. Spectrum of base substitutions. Mutations to Rifr result primarily from base substitutions (20). To determine what types of base substitution are specifically increased by Mag1 and Tag expression, we used the lacZ strains created by Cupples and Miller for monitoring each of the six possible base substitution events (9). Figure 5A shows that induction of Mag1 with 0.02% arabinose increases most of the six base substitutions to some extent but that the fold increases are largest for G:C-to-C:G (27.5-fold), G:C-to-T:A (8.4-fold), and A:T-to-T:A (7.4-fold) transversions. (Note that the largest absolute number of mutants resulted from G:C-to-T:A transversions.) These results are consistent with a model in which alkylated or nonalkylated G’s or A’s are removed by Mag1 (2,

TABLE 2. The effect of Tag expression on mutation induction in ⌬umuDC xth nfo and recA441 lexA300(Def) E. colia Fold increase in mutation frequency (cells overexpressing Tag over control cells)

SEM

⌬umuDC xth nfo xth nfo

8.2 3.1

2.3 0.4

recA441 lexA(Def) recA⫹ lexA⫹

2.0 1.1

0.1 0.6

Strain background

a These experiments were performed as described in Materials and Methods, except that xth nfo and ⌬umuDC xth nfo cultures were grown from 24 to 48 h to allow cultures to reach saturation. n is 3 for all experiments, except for recA441 lexA(Def) strains, where n is 6. The wild-type (recA⫹ lexA⫹) strain is GW2771; the recA441 lexA(Def) strain is GW8014.

6767

15, 49) and A’s are preferentially inserted opposite the resulting abasic sites (10, 32, 42). To determine what base substitutions were increased by Tag overexpression, we expressed Tag in the xth nfo derivatives of the lacZ spectrum strains (9, 27) (Fig. 5B). As with Rifr, the magnitude of the changes is much less than that seen for Mag1 expression. More interestingly, Fig. 5B shows that Tag induces a very different profile of mutations than does Mag1. The most striking changes are a 0.4-fold decrease in A:T-to-C:G transversions and a 2.1-fold increase in A:T-to-T:A transversions. These results are consistent with heightened Tag activity at adenine residues. In contrast, for Mag1 overexpression, the largest increases (and the largest number of induced mutants) occurred at G’s. The effects of glycosylase overexpression on colony-forming ability. We suspected that high levels of 3MeA DNA glycosylase activity might be harmful to cells. Figure 6A confirms that the colony-forming ability of CSH106 cells expressing Mag1 decreases sharply as glycosylase induction increases. In comparison, the colony-forming ability of CSH106 cells was unaffected by Tag expression, and the colony-forming ability of AP endonuclease-deficient xth nfo TN1068 cells was only mildly affected by Tag expression (Fig. 6B). Although the Mag1-induced toxicity is expected to result directly or indirectly from abasic sites, the actual lesion or lesions causing the toxicity are uncertain. The toxic lesions could include the abasic sites themselves or single- and doublestrand breaks resulting from the further processing of abasic sites (7, 17, 28, 30, 34, 44). If double-strand breaks are truly a consequence of glycosylase activity, Mag1 expression should be particularly toxic in strains deficient in double-strand break repair (17, 28, 34). To test this hypothesis, we transferred the Mag1 plasmid into a recBC strain and its wild-type parent and then compared the effects of Mag1 expression in the two backgrounds. Figure 6C shows that the recBC strain is much more sensitive to Mag1 induction than is the wild-type strain, implicating Mag1 in the production of double-strand breaks. (Note that different strains and induction conditions were used in the experiments for Fig. 6C versus those for Fig. 6A; under these conditions, no Mag1-induced loss of viability was observed in the wild-type background.) In contrast to Mag1, high-level Tag expression (Fig. 6D) had no apparent effect on the growth of recBC E. coli. These findings suggest that Tag has a much lower ability than Mag1 to produce lesions capable of leading to double-strand breaks. 3MeA DNA glycosylase expression sensitizes E. coli to killing by MMS. In 1986, Kaasen et al. (21) reported that overproduction of AlkA could sensitize E. coli to killing by MMS. More recently, Kaina et al. (22) reported that the human 3-alkyladenine DNA glycosylase could sensitize Chinese hamster ovary cells to sister chromatid exchanges and chromosome aberrations. Likewise, our preliminary results from gradient plate assays suggested that Mag1 and Tag could sensitize cells to killing at high doses of MMS, even though they protected glycosylase-deficient alkA cells from killing by a low dose of MMS (0.01%). To more closely examine the sensitization process, we carried out detailed killing curve assays with the alkA⫹ tag⫹ strain CSH106. Figure 7A shows that Mag1 and Tag had profoundly different abilities to sensitize cells to MMS. Thus, a 30-min treatment with 0.08% MMS produced more than a 104-fold decrease in survival for Mag1-induced cells. By comparison, the same treatment produced about a 50% decrease in survival in Tag-induced cells. The large difference in killing between Tag and Mag1 is unlikely to result from differences in expression level or activity on 3MeA, since Tag extracts had 15-fold-

6768

POSNICK AND SAMSON

J. BACTERIOL.

FIG. 5. Mag1 and Tag preferentially increase different base substitutions. To monitor Mag1- and Tag-induced base substitutions, we expressed plasmid p24Mag in lacZ marker strains CSH101 to CSH106 and plasmid pSE-Tag in xth nfo lacZ marker strains TN1018 to TN1068 (Table 1). (A) lacZ mutant frequency in Mag1-expressing cells induced with 0.02% arabinose. (B) lacZ mutant frequency in Tag-expressing cells induced with 1 mM IPTG. For both graphs, gray bars show strains expressing glycosylase (Mag1 or Tag), and white bars show strains expressing control plasmid (pBAD24 or pSE380). The numbers above the gray bars show the fold change in mutant frequency induced by glycosylase expression. Where numbers are absent, fold changes could not be calculated because of zero values. Error bars show standard errors of the means (n ⫽ 3). WT, wild type.

higher 3MeA activity in vitro than Mag1 extracts (Fig. 1). Instead, this difference may reflect the substrate specificity of the glycosylases. As noted above, the only major MMS-induced DNA lesion that Tag acts on is 3MeA. The greater sensitization seen with Mag1-overexpressing cells may result from the ability of this glycosylase to remove the roughly 10-fold-moreabundant 7MeG lesion (5). However, as shown in Fig. 7B, Tag can sensitize cells more significantly at higher MMS exposures. These data suggest that robust removal of 3MeA alone can eventually create sufficient secondary DNA lesions to harm MMS-exposed, Tag-expressing cells. To test the possibility that glycosylase-induced sensitization is due to a nonspecific increase in sensitivity to DNA damage in general, we compared Mag1- and Tag-induced sensitization to MMS and UV killing on gradient plates. Despite sensitizing cells to killing by MMS, neither glycosylase sensitized cells to killing by UV (data not shown), suggesting that the sensitization effect is specific to alkylating agents. DISCUSSION The BER process is in a delicate balance. When BER activity is excessive, as in E. coli dut mutants (reviewed in reference 14), or when BER components are deliberately imbalanced, as in the case of Mag1 overexpression in S. cerevisiae (15, 49), a host of toxic and mutagenic consequences can ensue. Theoretically, these consequences should vary depending on which BER enzyme is out of balance. For example, an excess of glycosylase activity may lead to the increased formation of abasic sites and consequently increased mutation. However, an overabundance of both DNA glycosylase and AP endonuclease activity (relative to DNA polymerase or DNA

ligase) could lead to an accumulation of DNA strand breaks and could consequently influence viability (48). This study illustrates that the substrate range of DNA glycosylases can also have profound effects on the BER imbalance phenotype. We compared the effects of expressing two rather different 3MeA DNA glycosylases, the S. cerevisiae Mag1 and the E. coli Tag glycosylases. Overexpression of both enzymes produced a mutator phenotype in E. coli, although the intensity of their mutator effects differed greatly. Tag is a rather modest mutator (approximately twofold for Rifr), despite producing the highest 3MeA DNA glycosylase activity in vitro. Mag1 increased Rifr mutant frequencies by more than 100fold, despite having 15-fold-lower 3MeA activity than Tag. We suspect that the disparity between Tag and Mag1 can be attributed to substrate specificity. Mag1 acts on a wide range of substrates including such damaged bases as 3MeA, 7MeG, 3MeG, 7-methyladenine, hypoxanthine, 1,N6-ethenoadenine, 7-chloroethylguanine, and 7-hydroxyethylguanine (5, 33, 39– 41, 48). Moreover, Mag1 can also remove normal guanines (2) and has been reported to bind to abasic sites or abasic site analogs (11). In contrast, Tag’s activity appears to be limited primarily to 3MeA, having only minor activity for 3MeG and no detectable activity for the release of normal bases from DNA (2, 4, 13). These differences in substrate recognition may underlie the ability of Mag1 to create more mutagenic lesions than Tag. Several pieces of evidence suggest that Mag1-induced mutations are the result of error-prone bypass of abasic sites resulting from excessive Mag1 activity. First, Mag1-induced Rifr mutations are strongly suppressed by a umuC allele, and abasic site mutagenesis in E. coli is generally presumed to be UmuDC dependent (10, 32, 42). A similar result was recently

VOL. 181, 1999


6769

FIG. 6. Effect of glycosylase induction on colony-forming ability. (A) Mag1 induction decreases colony-forming ability. CFU per milliliter were determined by measuring titers of cultures of CSH106 after 24 h of growth in the presence of the indicated dose of arabinose as described under “Mutation assays” in Materials and Methods. Error bars show standard errors of the means (n ⫽ 3). (B) Tag induction has only a mild effect on viability of wild-type CSH106 or xth nfo TN1068 cells. CFU per milliliter were determined as described for panel A. Error bars show standard errors of the means (n ⫽ 3). (C) A recBC strain is more sensitive than a wild-type (WT) strain to Mag1 induction. Note that experiments with the recBC allele were done with strains AB1157 and JCSS19, not CSH106 or its derivatives. Overnight cultures were diluted 1:100 (recBC) or 1:200 (wild type) in LB-ampicillin medium, grown to log phase, and treated with 0.2% arabinose. Aliquots were removed at the indicated times after addition of arabinose for titer determination on LB-ampicillin plates. This graph shows values representative of four experiments. (D) A recBC strain is not more sensitive than a wild-type (WT) strain to Tag induction. The protocol was as described for panel C, except that induction was with 1 mM IPTG. This graph shows values representative of three experiments. For panels C and D, colony-forming ability was measured at additional time points in several experiments. No significant changes in colony-forming ability were observed up to 8 h after induction.

reported for S. cerevisiae, where Mag1-induced mutations appear to be absolutely dependent on the Rev1-Rev3-Rev7 abasic site bypass system (15). In addition, Mag1-induced base pair substitutions observed in this study primarily involve transversions to adenine, a frequent mutagenic event at abasic sites in E. coli (10, 32, 42). For Tag, the evidence is less straightforward. The observation that Tag selectively increases mutation frequency in AP endonuclease-deficient or constitutively SOS-induced backgrounds is consistent with Tag-induced mutations resulting from UmuDC-driven bypass of abasic sites. On the other hand, expression of Tag in a ⌬umuDC xth nfo background also increased mutations, which argues that Tag-induced mutations are not UmuDC dependent. It may be that Tag-induced mutations can occur by both UmuDC-dependent and UmuDCindependent pathways and that Tag induces mutations solely by UmuDC-independent pathways in ⌬umuDC xth nfo cells. In recent years, several error-prone but non-UmuDC-dependent mutagenic pathways have been identified for E. coli (reviewed in reference 18), including the increase in spontaneous mutation mediated by DinB (26), frameshift mutagenesis by 2-acetylaminofluorene (36), and the UVM response (UV modulation of mutagenesis) (18). We have not yet tested the po-

tential role of alternate (UmuDC-independent) pathways for Tag-induced mutations in ⌬umuDC xth nfo cells. The spectra of mutations induced by Mag1 and Tag in lacZ marker strains are rather different from each other and appear to reflect differences in the two enzymes’ substrate specificities. Thus, Mag1 produced the strongest fold increases in G:C-toC:G, G:C-to-T:A, and A:T-to-T:A transversions; the largest absolute number of mutants arose from G:C-to-T:A transversions. This spectrum is consistent with Mag1’s ability to remove normal and endogenous methylated purines (2). Tag, known to act primarily on 3MeAs, produced substantially smaller changes than Mag1. The two largest changes were a 0.4-fold decrease in A:T-to-C:G transversions and a 2.1-fold increase in A:T-to-T:A transversions. Interestingly, these results are consistent with heightened Tag activity at adenine residues. Unlike Mag1, Tag was reported previously not to remove normal bases from DNA (2). Our findings suggest that Tag may have some low-level activity on A’s that was not detectable in the in vitro system (2) or that elevated Tag levels are removing endogenously produced 3MeA residues. In either case, removal of the A or 3MeA would create an abasic site; following the “A rule” for E. coli, insertion of A would be expected to create an A:T-to-T:A transversion (10, 32, 42). Removal of an endoge-

6770

POSNICK AND SAMSON

FIG. 7. 3MeA DNA glycosylase expression sensitizes wild-type E. coli to killing by MMS. (A) Overnight cultures of cells containing the Mag1 or Tag (closed symbols) or control plasmids (open symbols) were diluted 104-fold in maltose-minimal ampicillin medium plus 0.2% arabinose (Mag1) or IPTG (Tag), grown for approximately 12 to 16 h, and treated with the indicated dose of MMS for 30 min, and titers were determined on LB-ampicillin plates. Error bars show standard errors of the means (n ⫽ 3). (B) Overnight cultures of cells containing the Tag plasmid (closed symbols) or a control plasmid (open symbols) were diluted 100-fold in maltose-ampicillin medium plus 1 mM IPTG, grown to log phase, and treated with 0.3% MMS for the indicated time, and titers were determined on LB-ampicillin plates. Error bars show standard errors of the means (n ⫽ 3).

nous premutagenic lesion may likewise explain why we saw a decrease in A:T-to-C:G transversions with Tag expression. We observed several other differences between the effects of Mag1 and the effects of Tag that we think may be attributable to differences in substrate range. First, overexpression of Mag1 had much more deleterious effects on cell viability than did overexpression of Tag; we would argue that this effect is likely due to Mag1’s heightened ability to remove normal bases. Second, Mag1 sensitized cells to MMS much more strongly than did Tag. For MMS-induced lesions, the main difference between Mag1 and Tag activity is Mag1’s ability to remove 7MeG. Even though Mag1 removes 7MeG lesions less efficiently than 3MeAs, 7MeG is by far the most abundant MMSinduced DNA lesion. Unlike Tag, Mag1 is not a native E. coli enzyme. One could argue that Mag1 has more intense effects on mutation induction, viability, and MMS sensitization than Tag because it cannot interact with downstream components of the E. coli BER pathway (such as AP endonuclease) as effectively as Tag. We think that this explanation of Mag1’s profound effects is unlikely for the following reasons. First, Mag1 also has much stronger mutator and sensitization effects than Tag when these enzymes are overexpressed in S. cerevisiae, Mag1’s native cell (15). Second, from this study, Mag1 and Tag appear to produce different spectra of mutations, suggesting that they act on different lesions (or bases), rather than varying in their ability to complete repair after base removal. Finally, the E. coli AlkA glycosylase also sensitized cells to MMS more strongly than Tag (data not shown), even though AlkA is also native to E. coli. Our data also suggest a possible role for secondary (postabasic-site) lesions in some of the phenotypes observed during glycosylase overexpression. In particular, Mag1 overexpression was especially toxic in recBC cells, suggesting that Mag1 produces double-strand breaks (17, 28, 34). Possible sources of double-strand breaks in Mag1-overexpressing cells include DNA polymerase encounters with abasic sites or nicked abasic

J. BACTERIOL.

sites (6, 17, 28, 30, 34) and nicking of single-stranded DNA opposite repair tracts or daughter-strand gaps (44, 45). In addition, a number of DNA glycosylases have recently been shown to bind abasic sites or modified abasic sites, including certain 3MeA DNA glycosylases (11, 29), human uracil DNA glycosylase (37), human thymine DNA glycosylase (46), and the E. coli mismatch-specific uracil DNA glycosylase (1). Such binding may normally serve to protect cells from the mutagenic and cytotoxic effects of AP sites (37, 46). However, it is possible that glycosylase binding to abasic sites may play a role in the glycosylase-induced mutator and decreased viability phenotypes, for example, by acting as a block to replication (46). In summary, we have examined the effects of overexpressing two different 3MeA DNA glycosylases in E. coli, namely, the Mag1 and Tag glycosylases. Both glycosylases increased spontaneous mutation and sensitized cells to MMS. However, Tag had much less pronounced effects than Mag1. We suggest that the differences in the effects of Mag1 and Tag may be due in part to differences in enzyme substrate specificities and binding activities. Further, we argue that the lesions responsible for the increase in spontaneous mutation and MMS sensitization and the decrease in colony-forming ability may be either abasic sites resulting from glycosylase activity or secondary lesions stemming from such abasic sites. Because the rates of repair of abasic sites and formation of secondary lesions will be affected by the balance of DNA glycosylase activity and downstream BER activities and because the balance of such activities might vary in different cell types, the effects of glycosylase overexpression are likely to differ from cell to cell or organism to organism. ACKNOWLEDGMENTS We thank G. Walker, J. Beckwith, M. Zaman, and B. Demple for strains and plasmids. This work was supported by grants from the National Institute of Environmental Health Sciences (P01-E03926) and the National Cancer Institute (R01-55042) to L.D.S., who is a Burroughs Wellcome Toxicology Scholar. L.M.P. was supported by a fellowship from the Pharmaceutical Research and Manufacturers of America Foundation and by a Training Program in Environmental Health Sciences grant from the National Institute of Environmental Health Sciences (5 T32 ES07155). REFERENCES 1. Barrett, T. E., R. Savva, G. Panayotou, T. Barlow, T. Brown, J. Jiricny, and L. H. Pearl. 1998. Crystal structure of a G:T/U mismatch-specific DNA glycosylase: mismatch recognition by complementary-strand interactions. Cell 92:117–129. 2. Berdal, K. G., R. F. Johansen, and E. Seeberg. 1998. Release of normal bases from intact DNA by a native DNA repair enzyme. EMBO J. 17:363–367. 3. Bjelland, S., N.-K. Birkeland, T. Benneche, G. Volden, and E. Seeberg. 1994. DNA glycosylase activities for thymine residues oxidized in the methyl group are functions of the AlkA enzyme in Escherichia coli. J. Biol. Chem. 269: 30489–30495. 4. Bjelland, S., M. Bjoras, and E. Seeberg. 1993. Excision of 3-methylguanine from alkylated DNA by 3-methyladenine DNA glycosylase I of Escherichia coli. Nucleic Acids Res. 21:2045–2049. 5. Bjoras, M., A. Klungland, R. F. Johansen, and E. Seeberg. 1995. Purification and properties of the alkylation repair DNA glycosylase encoded by the MAG gene from Saccharomyces cerevisiae. Biochemistry 34:4577–4582. 6. Cao, Y., and T. Kogoma. 1995. The mechanism of recA polA lethality: suppression by RecA-independent recombination repair activated by the lexA(Def) mutation in Escherichia coli. Genetics 139:1483–1494. 7. Chaudhry, M. A., and M. Weinfeld. 1997. Reactivity of human apurinic/ apyrimidinic endonuclease and Escherichia coli exonuclease III with bistranded abasic sites in DNA. J. Biol. Chem. 272:15650–15655. 8. Coquerelle, T., J. Dosch, and B. Kaina. 1995. Overexpression of N-methylpurine-DNA glycosylase in Chinese hamster ovary cells renders them more sensitive to the production of chromosomal aberrations by methylating agents—a case of imbalanced DNA repair. Mutat. Res. 336:9–17. 9. Cupples, C. G., and J. H. Miller. 1989. A set of lacZ mutations in Escherichia coli that allow rapid detection of each of the six base substitutions. Proc.

VOL. 181, 1999


Natl. Acad. Sci. USA 86:5345–5349. 10. Dogliotti, E., P. Fortini, and B. Pascucci. 1997. Mutagenesis of abasic sites, p. 81–101. In I. D. Hickson (ed.), Base excision repair of DNA damage. Landes Bioscience, Austin, Tex. 11. Eide, L., K. G. Berdal, M. Bjoras, and E. Seeberg. 1998. Interactions of 3-methyladenine DNA glycosylases with AP-sites in the DNA, abstr. 24, p. 106. In DNA repair: bacteria to humans, program and abstracts volume. Genetics Society of America, Bethesda, Md. 12. Evensen, G., and E. Seeberg. 1982. Adaptation to alkylation resistance involves the induction of a DNA glycosylase. Nature 296:773–775. 13. Friedberg, E. C., G. C. Walker, and W. Siede. 1995. DNA repair and mutagenesis. American Society for Microbiology, Washington, D.C. 14. Glassner, B. J., L. M. Posnick, and L. D. Samson. 1998. The influence of DNA glycosylases in spontaneous mutation. Mutat. Res. 400:33–44. 15. Glassner, B. J., L. J. Rasmussen, M. T. Najarian, L. M. Posnick, and L. D. Samson. 1998. Generation of a strong mutator phenotype in yeast by imbalanced base excision repair. Proc. Natl. Acad. Sci. USA 95:9997–10002. 15a.Glassner, B. J., and L. D. Samson. Unpublished data. 16. Guzman, L.-M., D. Belin, M. J. Carson, and J. Beckwith. 1995. Tight regulation, modulation, and high-level expression by vectors containing the arabinose PBAD promoter. J. Bacteriol. 177:4121–4130. 17. Horiuchi, T., and Y. Fujimura. 1995. Recombinational rescue of the stalled DNA replication fork: a model based on analysis of an Escherichia coli strain with a chromosome region difficult to replicate. J. Bacteriol. 177:783–791. 18. Humayun, M. Z. 1998. SOS and Mayday: multiple inducible mutagenic pathways in Escherichia coli. Mol. Microbiol. 30:905–910. 19. Ibeanu, G., B. Hartenstein, W. C. Dunn, L.-Y. Chang, E. Hofmann, T. Coquerelle, S. Mitra, and B. Kaina. 1992. Overexpression of human DNA repair protein N-methylpurine-DNA glycosylase results in the increased removal of N-methylpurines in DNA without a concomitant increase in resistance to alkylating agents in Chinese hamster ovary cells. Carcinogenesis 13:1989–1995. 20. Jin, D. J., and C. A. Gross. 1988. Mapping and sequencing of mutations in the Escherichia coli rpoB gene that lead to rifampicin resistance. J. Mol. Biol. 202:45–58. 21. Kaasen, I., G. Evensen, and E. Seeberg. 1986. Amplified expression of the tag⫹ and alkA⫹ genes in Escherichia coli: identification of gene products and effects on alkylation resistance. J. Bacteriol. 168:642–647. 22. Kaina, B., G. Fritz, and T. Coquerelle. 1993. Contribution of O6-alkylguanine and N-alkylpurines to the formation of sister chromatid exchanges, chromosomal aberrations, and gene mutations. Environ. Mol. Mutagen. 22:283–292. 23. Karran, P., T. Hjelmgren, and T. Lindahl. 1982. Induction of a DNA glycosylase for N-methylated purines is part of the adaptive response to alkylating agents. Nature 296:770–773. 24. Karran, P., T. Lindahl, I. Ofsteng, G. B. Evensen, and E. Seeberg. 1980. Escherichia coli mutants deficient in 3-methyladenine-DNA glycosylase. J. Mol. Biol. 140:101–127. 25. Kavli, B., G. Slupphaug, C. D. Mol, A. S. Arvai, S. B. Petersen, J. A. Tainer, and H. E. Krokan. 1996. Excision of cytosine and thymine from DNA by mutants of human uracil-DNA glycosylase. EMBO J. 15:3442–3447. 26. Kim, S. R., G. Maenhaut-Michel, M. Yamada, Y. Yamamoto, K. Matsui, T. Sofuni, T. Nohmi, and H. Ohmori. 1997. Multiple pathways for SOS-induced mutagenesis in Escherichia coli: an overexpression of dinB/dinP results in strongly enhancing mutagenesis in the absence of any exogenous treatment to damage DNA. Proc. Natl. Acad. Sci. USA 94:13792–13797. 27. Kunz, B. A., E. S. Henson, H. Roche, D. Ramotar, T. Nunoshiba, and B. Demple. 1994. Specificity of the mutator caused by deletion of the yeast

28. 29. 30. 31. 32. 33.

34. 35. 36. 37.

38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49.

6771

structural gene (APN1) for the major apurinic endonuclease. Proc. Natl. Acad. Sci. USA 91:8165–8169. Kuzminov, A. 1995. Collapse and repair of replication forks in Escherichia coli. Mol. Microbiol. 16:373–384. Labahn, J., O. D. Scharer, A. Long, K. Ezaz-Nikpay, G. L. Verdine, and T. E. Ellenberger. 1996. Structural basis for the excision repair of alkylationdamaged DNA. Cell 86:321–329. Larson, K., J. Sahm, R. Shenkar, and B. Strauss. 1985. Methylation-induced blocks to in vitro DNA replication. Mutat. Res. 150:77–84. Lindahl, T. 1993. Instability and decay of the primary structure of DNA. Nature 362:709–715. Loeb, L. A. 1985. Apurinic sites as mutagenic intermediates. Cell 40:483–484. Matijasevic, Z., M. Boosalis, W. Mackay, L. Samson, and D. B. Ludlum. 1993. Protection against chloroethylnitrosourea cytotoxicity by eukaryotic 3-methyladenine DNA glycosylase. Proc. Natl. Acad. Sci. USA 90:11855– 11859. Michel, B., S. D. Ehrlich, and M. Uzest. 1997. DNA double-strand breaks caused by replication arrest. EMBO J. 16:430–438. Miller, J. H. 1992. A short course in bacterial genetics: a laboratory manual and handbook for Escherichia coli and related bacteria. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y. Napolitano, R. L., I. B. Lambert, and R. P. P. Fuchs. 1997. SOS factors involved in translesion synthesis. Proc. Natl. Acad. Sci. USA 94:5733–5738. Parikh, S. S., C. D. Mol, G. Slupphaug, S. Bharati, H. E. Krokan, and J. A. Tainer. 1998. Base excision repair initiation revealed by crystal structures and binding kinetics of human uracil-DNA glycosylase with DNA. EMBO J. 17:5214–5226. Santerre, A., and A. B. Britt. 1994. Cloning of a 3-methyladenine-DNA glycosylase from Arabidopsis thaliana. Proc. Natl. Acad. Sci. USA 91:2240– 2244. Saparbaev, M., K. Kleibl, and J. Laval. 1995. Escherichia coli, Saccharomyces cerevisiae, rat and human 3-methyladenine DNA glycosylases repair 1,N6ethenoadenine when present in DNA. Nucleic Acids Res. 23:3750–3755. Saparbaev, M., and J. Laval. 1994. Excision of hypoxanthine from DNA containing dIMP residues by the Escherichia coli, yeast, rat, and human alkylpurine DNA glycosylases. Proc. Natl. Acad. Sci. USA 91:5873–5877. Seeberg, E., and K. G. Berdal. 1997. Repair of alkylation damage to DNA, p. 151–168. In I. D. Hickson (ed.), Base excision repair of DNA damage. Landes Bioscience, Austin, Tex. Strauss, B. S. 1991. The ‘A rule’ of mutagen specificity: a consequence of DNA polymerase bypass of non-instructional lesions? Bioessays 13:79–84. Volkert, M. 1999. Personal communication. Wang, T.-C., and K. C. Smith. 1986. Postreplicational formation and repair of DNA double-strand breaks in UV-irradiated Escherichia coli uvrB cells. Mutat. Res. 165:39–44. Wang, T. C., and H. Y. Chang. 1991. Effect of rec mutations on viability and processing of DNA damaged by methylmethane sulfonate in xth nth nfo cells of Escherichia coli. Biochem. Biophys. Res. Commun. 180:774–781. Waters, T. R., P. Gallinari, J. Jiricny, and P. F. Swann. 1999. Human thymine DNA glycosylase binds to apurinic sites in DNA but is displaced by human apurinic endonuclease 1. J. Biol. Chem. 274:67–74. Woodgate, R. 1992. Construction of a umuDC operon substitution mutation in Escherichia coli. Mutat. Res. 281:221–225. Wyatt, M. D., J. M. Allan, A. Y. Lau, T. E. Ellenberger, and L. D. Samson. 1999. 3-Methyladenine DNA glycosylases: structure, function, and biological importance. Bioessays 21:668–676. Xiao, W., and L. Samson. 1993. In vivo evidence for endogenous DNA alkylation damage as a source of spontaneous mutation in eukaryotic cells. Proc. Natl. Acad. Sci. USA 90:2117–2121.