Repair of O6-methylguanine and O4-methylthymine by the human and rat O6-methylguanine-DNA methyltransferases.

In order to compare the ability of the human and rat O6-methylguanine-DNA methyltransferases (transferases) to repair in vitro O6-methylguanine (O6-MeGua) and O4-methylthymine (O4-MeThy) residues, which are two mutagenic DNA adducts formed by alkylating agents, we have purified both proteins to homogeneity. Gel electrophoresis of the proteins shows that the O4-MeThy repair is due to the transfer of the methyl group from the alkylated base to the transferase molecules. However, both proteins repair with different efficiencies the O6-MeGua and O4-MeThy residues present in alkylated DNA, poly[d(G.C)], poly(dG.dC), or in alkylated poly[d(A.T)] and poly(dA.dT), respectively. Reaction of both proteins with either methylated residues follows a second-order kinetics. The rate constants are 1 x 10(9) M-1 min-1 for both proteins acting on O6-MeGua and 4.8 x 10(6) or 1.8 x 10(5) M-1 min-1 for the rat or human protein acting on O4-MeThy, respectively. The activity of the mammalian transferases on O4-MeThy present in a poly(dA.dT) substrate is inhibited by double-stranded DNA.

In order to compare the ability of the human and rat 06-methylguanine-DNA methyltransferases (transferases) to repair in vitro Og-methylguanine (06-MeGua) and O*-methylthymine ( 04-MeThy) residues, which are two mutagenic DNA adducts formed by alkylating agents, we have purified both proteins to homogeneity. Gel electrophoresis of the proteins shows that the 04-MeThy repair is due to the transfer of the methyl group from the alkylated base to the transferase molecules. considered as lethal lesions because they are a block to DNA replication (21, although these adducts are easily depurinated to form apurinic sites (3). 0 6 -M e t h y l~a n i n e (06-MeGua)' and 04-methylthymine (04-MeThy) residues cause base mispairing (4) and are believed to lead to mutagenesis and carcinogenesis (5). 06-MeGua and 04-MeThy are equally mutagenic in vitro, but it has been shown that 04-MeThy is strikingly more mutagenic than 06-MeGua in vivo (6).
In Escherichia coli, 06-MeGua and 04-MeThy residues are repaired by two DNA alkyltransferases (transferases), which transfer the alkyl group from the alkylated bases to one of their own cysteine residues, and are therefore suicide proteins (7).
The Ada protein, encoded by the ada gene, is a n inducible 39-kDa protein, which possesses two alkyl-acceptor cysteines * This work was supported by grants from the Institut National de la SantC? et de la Recherche MBdicale and from the Association pour la Recherche sur le Cancer. 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   However, the affinity of the bacterial and mammalian transferases for 04-MeThy residues was recently suggested by incubating crude cell extracts with a n oligonucleotide containing a single modified base (191, and the removal of 04-MeThy residues by a high amount of purified human transferase has also been reported (20).
The goal of our study was to determine whether the mammalian transferases were able to repair the 04-MeThy residues with the same efficiency as 06-MeGua. We have purified the transferases of human and rat origin and compared their activities on 06-MeGua and 04-MeThy residues present in various substrates. Our results show that the two eukaryotic proteins repair 04-MeThy residues in vitro and that the alkyl group is transferred to the acceptor protein. However, the rat and, especially, the human protein have a low affinity for these lesions.

Stmins
The pKT 100 plasmid carrying the human transferase cDNA (obtained from Dr. S. Mitra) was grown in E. coli J M 107 (11). The pcDNA I1 plasmid carrying the rat transferase cDNA (14) was from laboratory stocks and was grown in E. coli  This substrate (corresponding to 400 fmol of 04-MeThy) was incubated in 70 m~ Hepes, pH 7.6, 5 m~ EDTA (final volume 120 pl), with increasing amounts of protein, for 45 min at 37 "C. It was then hydrolyzed to nucleosides by sequential incubation with DNase I, snake venom phosphodiesterase, and alkaline phosphatase (23). After addition of authentic markers, the alkylation products (02-MeThy, N3-MeThy, 04-MeThy, and phosphotriesters) were separated by HPLC using a C18 pBondapack column eluting with 24% methanol/ H20 at a flow rate of 0.8 mVmin (24). They were quantified by scintillation spectroscopy.
Protein Gel Electrophoresis [SHIMNU-treated poly(dA.dT) (corresponding to 2400 cpm of 04-MeThy) was incubated (final volume 100 pl) with the E. coli, rat, or purified human transferases or with H4 cell extracts for 45 min at 37 "C. The protein separation was then done in 12.5% SDS-polyamylamide gels (25). The gels were fixed in methanollacetic acid, soaked in EN3HANCE (DuPont NEN), then exposed to x-ray films.

Purification Protocol
The rat transferase was purified from the E. coli KT 233 expressing the rat transferase cDNA (14). Exponentially growing bacteria (1 liter of LB medium) were harvested by centrifugation, washed in SSC, and disrupted by sonication at 0 "C in the presence of proteases inhibitors. The purification steps were as described by Koike et al. (20) for the human transferase and were, successively: ammonium sulfate precipitation, DEAE-Sephacel chromatography, and Mono S column chroma-The human transferase was purified from E. coli JM 107 transfected with the pKT 100 plasmid (11). The purification steps were as described (20). tography.

RESULTS
Purification of the Rat Dansfemse-The proteins present after each purification step were detected by gel electrophoresis (Fig. 1). The purification (Table I) was about 870-fold with a 40% yield. Gel electrophoresis of the Mono S fractions containing the transferase activity showed a single band of protein ( Fig. 1) with a molecular mass of about 23 kDa, in agreement with the 22.2 kDa calculated from the cDNA sequence (14).
The purified human transferase showed also a single band in gel electrophoresis (data not shown).
Repair of 06-MeGua and OP-MeThy Residues by the Purified  R a t a n d Human Dansferases-The relative activity of the two eukaryotic transferases on 06-MeGua and 04-MeThy residues was measured by incubating the proteins with [3H]MNUtreated DNA or [3H]MNU-treated poly(dA-dT). To circumvent problems of interpretation owing to technical differences, [3H]MNU of the same specific activity was used to alkylate both substrates and the incubation mixtures contained the same amount of either 06-MeGua or 04-MeThy residues. Both proteins remove the 06-MeGua residues with the same efficiency from alkylated DNA, poly(dG.dC), or poly[d(G.C)] (Fig. 2,A and B , and Table 11). However, a much lower number of 04-MeThy residues was removed by the same amounts of proteins, particularly in the case of human transferase (Fig. 2) (e.g. after 45 min of incubation, about 3.5-and 20-fold less 04-MeThy than 06-MeGua residues were removed by 10 ng of rat or human protein, respectively). The removal of 04-MeThy from alkylated poly(dA-dT) or poly[d(A.T)l was identical (Table 11).
Using protein concentrations sufficient to repair about 50% of each alkylated base, the 06-MeGua removal from alkylated DNA or from alkylated poly(dG-dC) was completed in less than 2 min. However, the 04-MeThy repair by the human protein acting on poly(dA*dT) was much slower (Fig. 3). A plot of ln[S,][PJ/[P,][S,] as a function of time (Fig. 3, inset) was linear, showing that the reaction between the human protein and the methylated poly(dA.dT) follows a second-order kinetics. A similar kinetics was observed with the rat protein (not shown). The Activity-on 04-MeThy present in: a The activity (pmol of alkylated base repaireul mg of proteid45 min) was calculated from data of the linear part of the curves obtained by incubating increasing amounts of protein with the different substrates. It should be noted that, under our experimental conditions, the same amount of E. coli Ada protein (19-kDa fragment) removed 75 and 62% of the 06-MeGua or 04-MeThy residues present in alkylated DNA or poly(dA*dT), respectively. Both residues were removed at the same rate, and their total removal was performed in about 2 min, in agreement with previously reported data (8).

Dansfer of the Methyl Group from 04-MeThy to the Acceptor Proteins-This transfer was checked by gel electrophoresis of
the proteins (Fig. 4). A band of about 23 kDa was observed when the purified rat or human transferases were incubated with the alkylated poly(dA.dT) (Fig. 4, lanes b and c). As a control, the polymer was incubated with the E. coli Ada protein (19-kDa fragment) (Fig. 4, lane a). However, when the same amount of substrate was incubated with crude H4 cell extract (Fig. 4, lane d), no radioactive protein was detected in the gel, although the transfer of the methyl group from 06-MeGua could be detected when a lower amount of cell extract was incubated with [3HlMNU-treated DNA (26).
It should be noted that no significant repair of the 02-MeThy or of the phosphotriesters was observed with the rat or the human protein under our experimental conditions (data not shown).
Inhibition of the 04-MeThy Repair by Double-stranded DNA "Crude extracts from rat (H4 cells) (Fig. 4)  obtain experimental conditions similar to those previously used with the cell extracts, the activity of the purified rat transferase was measured in the presence of crude cell extracts or of purified cellular DNA. The removal of 04-MeThy residues decreased to less than 10% in the presence of cellular extracts (200 pg of proteins) and dramatically decreased when cellular DNA was added in the incubation medium, whereas the activity on 06-MeGua was not significantly modified under these conditions using as substrate either MNU-treated DNA or poly(dG.dC) (Fig. 5). A similar decreased transferase activity on 04-MeThy in the presence of DNA was observed with the purified human protein (data not shown). It should be stressed that the DNA concentrations used in the present experiments are in the range of the amounts of DNA contained in the cellular extracts previously used to examine the 04-MeThy repair ( Fig. 4) (18). DISCUSSION The purified transferases from human and rat origin are able to repair in vitro the 04-MeThy residues present in an alkylated poly(dA-dT). However, the two proteins have a low affinity for these residues. The rate constant for the human and rat transferases acting on 06-MeGua residues at 37 "C was 1 x log M -~ min-l, in good agreement with the value reported by Chan et al. (27), who determined this value by extrapolation of data obtained at lower temperatures. However the rate constants for the rat and, especially, the human transferases acting on 04-

x IO5
min-l, respectively. Few results have provided evidence that 04-MeThy was repaired in mammalian cells. The removal of 04-MeThy from a poly(dA-dT) substrate by monkey liver extracts was suggested by Becker and Montesano (28), but it was not shown whether this repair was due to a transferase activity. More recently the transfer of the methyl group from 04-MeThy to the human transferase has been shown (20). No comparative data were available, however, on the relative repair of 06-MeGua and @-MeThy by the eukaryotic transferases. Therefore our results, which show that the repair of 04-MeThy is due to the actual transfer of the methyl group to the acceptor human or rat protein, are the first to compare the relative activities of these two purified proteins on 06-MeGua and 04-MeThy residues and to determine the rate of the reactions.
The 04-MeThy repair was measured using as substrates alkylated poly(dA.dT) and poly[d(A.T)], because alkylation of DNA produces a very low amount of 04-MeThy, representing 0.06% of the total alkylation (29). However, it is known that a B-conformation is a general structure to alternating pyrimidine-purine polymers (30). Furthermore, the two proteins remove 06-MeGua residues with the same efficiency from alkylated DNA, polyfdG-dC), or poly[d(G.C)l, and, in the same experimental conditions, the 19-kDa fragment of the E. coli Ada protein removes 04-MeThy residues from poly(dA.dT) as efficiently as 06-MeGua from DNA. Therefore the two eukaryotic transferases and the E. coli protein have different abilities to repair 04-MeThy, although they share the same active site (16). There is also a high degree of homology between the rat and the human transferases (14); nevertheless, these two proteins have different activities on 04-MeThy residues. However, inhibition of the transferase activity on poly(dA.dT) by doublestranded DNA may suggest that the protein does not bind well to polymers consisting of only A and T residues, and that the activity for 04-MeThy in normal DNA could be higher than in p o l~~* d T ) . This i~b i t i o n by DNA can also explain why many experiments failed to demonstrate W-MeThy repair by cell extracts containing large amounts of cellular DNA (18). If we assume that these results can be extrapolated t o the in vivo situation, the higher activity of the human protein on 06-MeGua suggests that 04-MeThy could be seldom repaired, because the transferase present in the cells might be preferentially used to repair 06-MeGua residues. The different activities of the human and rat proteins on 06-MeGua and 04-MeThy, and the higher transferase inducibility in rat cells by DNA damage (311, would also suggest that the 04-MeThy residues are less efficiently repaired in human cells. However, human cells of Mer+ phenotype possess a higher number of transferase molecules than rodent cells (32), and the overall repair capacity for @-MeThy might be similar in the cells of different origin. The different activity of the rat transferase on 06-MeGua and 04-MeThy residues could in part explain the different rates of removal of these two adducts, which have been measured in rat hepatocytes in vivo (33), although we cannot exclude the possibility that another repair mechanism exists in vivo for the removal of 04-MeThy adducts.