Repair of O6-benzylguanine by the Escherichia coli Ada and Ogt and the human O6-alkylguanine-DNA alkyltransferases.

O6-Methylguanine is removed from DNA via the transfer of the methyl group to a cysteine acceptor site present in the DNA repair protein O6-alkylguanine-DNA alkyltransferase. The human alkyltransferase is inactivated by the free base O6-benzylguanine, raising the possibility that substantially larger alkyl groups could also be accepted as substrates. However, the Escherichia coli alkyltransferase, Ada-C, is not inactivated by O6-benzylguanine. The Ada-C protein was rendered capable of reaction by the incorporation of two site-directed mutations converting Ala316 to a proline (A316P) and Trp336 to alanine (W336A) or glycine (W336G). These changes increase the space at the active site of the protein where Cys321 is buried and thus permit access of the O6-benzylguanine inhibitor. Reaction of the mutant A316P/W336A-Ada-C with O6-benzylguanine was greatly stimulated by the presence of DNA, providing strong support for the concept that binding of DNA to the Ada-C protein activates the protein. The Ada-C protein was able to repair O6-benzylguanine in a 16-mer oligodeoxyribonucleotide. However, the rate of repair was very slow, whereas the E. coli Ogt, the human alkyltransferase, and the mutant A316P/W336A-Ada-C alkyltransferases reacted very rapidly with this 16-mer substrate and preferentially repaired it when incubated with a mixture of the methylated and benzylated 16-mers. These results show that benzyl groups are better substrates than methyl groups for alkyltransferases provided that steric factors do not prevent binding of the substrate in the correct orientation for alkyl group transfer.

The protein O 6 -alkylguanine-DNA alkyltransferase is an important component of the cellular resistance to the toxic and mutagenic effects of alkylating agents (1)(2)(3)(4). This protein has now been reported to be present in a wide range of organisms, and cDNA-derived amino acid sequences are available for at least 13 distinct proteins (2, 4 -6). The alkyltransferase repair proteins act by transferring alkyl groups from the O 6 -position of guanine onto a cysteine acceptor site located in the protein sequence. In all known alkyltransferases, this cysteine residue is contained within the sequence PCHR, and there are many other highly conserved residues. This provides strong evidence that the DNA reactions brought about by these proteins are the same irrespective of species. Escherichia coli contains two alkyltransferases, the products of the ogt gene and the ada gene. The content of Ogt protein remains constant, whereas the content of Ada is highly inducible in response to alkylating agents (1,3).
The E. coli Ada protein is the best characterized of all of the alkyltransferases, and a crystal structure at 2.1-Å resolution for the carboxyl-terminal domain of this protein (Ada-C), 1 which is fully active in repair of O 6 -methylguanine, has been published (6,7). This structure shows that the cysteine acceptor site is buried, and the authors suggest that the protein must undergo a significant conformational change upon binding a DNA substrate and that this alteration facilitates the transfer reaction.
There has been considerable interest in the development of inactivators of human alkyltransferase (AGT) because it is well established that the presence of this protein in tumor cells imparts resistance to killing by both therapeutic methylating agents such as dacarbazine and temozolomide and chloroethylating agents such as 1,3-bis(2-chloroethyl)-1-nitrosourea (8). One of the most promising compounds that has been used as an inactivator is O 6 -benzylguanine (9). The basis for this inactivation is well understood. The base analog is recognized as a substrate by the human alkyltransferase, and the benzyl group is transferred to the active-site cysteine, forming an S-benzylcysteine adduct and inactivating the protein (10). However, little is known of the means by which O 6 -benzylguanine binds to the active site, and despite the similarity in amino acid sequence, the Ada-C protein was not inactivated by O 6 -benzylguanine (10 -12). It was suggested that this difference is due to a restriction of the space available at the active site in the Ada protein (13). Such a steric effect could also account for published differences in the relative efficiencies with which alkyltransferases from different species repair substrates containing larger alkyl groups (2,4,14,15). Although methyl groups have been reported to be the best substrate for all of these repair proteins, larger alkyl groups are also repaired at slower rates, and the ability to act on these groups is quite species-specific.
In the present work, we have investigated the ability of the Ada-C, Ogt, and AGT proteins to react with O 6 -benzylguanine as a free base and when incorporated into a 16-mer oligodeoxyribonucleotide. Site-directed mutation of the Ada-C protein to increase the size of the pocket surrounding the cysteine acceptor produced a protein that was able to react with free O 6 -benzylguanine, and the rate of this reaction was greatly increased when DNA was added. The mutated Ada-C protein and the Ogt were very effective in repairing O 6 -benzylguanine contained within an oligodeoxyribonucleotide, and both these proteins and the AGT preferentially removed benzyl groups from such substrates rather than methyl groups. These results provide further insight into the alkyltransferase reaction and into the design of therapeutically useful inactivators.
Expression of Ogt and Control and Mutant Ada-C Proteins-These proteins were expressed in E. coli JM109 cells using the pQE30 vector (Qiagen) for the expression of the protein with a small extension [Met-Arg-Gly-Ser-(His) 6 -Gly-Ser] at the amino terminus.
To express the control Ada-C, the region representing the cDNA coding for residues Gln 179 to Arg 354 of the Ada protein was amplified from pGEMADA (13) by polymerase chain reaction using primers introducing a BamHI site (underlined) at the 5Ј end (5Ј-GGCGAATTC-CGGGATCCCAATTCCGTCACG-3Ј) and a SacI site after the stop codon at the 3Ј end (5Ј-GGTACCGAGCTCCCCCATCGG-3Ј). Polymerase chain reaction was carried out with Pfu polymerase under the following conditions: initial denaturation at 92°C for 2 min, 20 cycles of denaturation (1 min at 92°C), annealing (1 min at 50°C), and extension (1 min at 72°C) followed by a final extension at 72°C for 5 min. The 581-bp product was gel-purified, digested with BamHI and SacI, and ligated into the pQE30 vector that had been digested with the same enzymes to form pQE30-Ada-C.
The double mutant pQE30-A316P/W336G-Ada-C was made by replacing the region between the BstEII and SacI sites in mutant pQE30-A316P/W336A-Ada-C with a fragment containing a mutation converting the codon for Trp 336 to Gly. This fragment was made using the polymerase chain reaction conditions given above with pGEMADA but replacing the 5Ј primer with a primer (5Ј-CGCTTTCAGGTTACCG-CGGGGGCGTGTCGC-3Ј) introducing the Trp to Gly mutation (doubleunderlined) and containing a sequence corresponding to the BstEII site in the Ada cDNA (underlined). The 110-bp product was digested with BstEII and SacI, and the 90-bp fragment was used to replace the corresponding sequence by ligation into the 3918-bp DNA piece isolated from pQE30-A316P/W336A-Ada-C digested with these enzymes.
The triple mutants pQE30-K314P/A316P/W336A-Ada-C and pQE30-K314P/A316P/W336G-Ada-C were produced by using the Chameleon TM double-stranded mutagenesis kit (Stratagene) according to the manufacturer's instructions, using the unique Xmn I site in the plasmid and the oligodeoxyribonucleotide 5Ј-GTGCCGCCAACCGCTG-CCTATCGT-TATAC-3Ј for the mutation of Lys 314 to Pro in pQE30-A316P/W336G-Ada-C and pQE30-A316P/W336A-Ada-C, respectively. (The mismatches used to produce the mutation of Lys 314 to Pro are shown by double-underlining.) The Ogt cDNA was amplified from pUCOgt plasmid DNA (19) by polymerase chain reaction using primers introducing a BamHI site (underlined) at the 5Ј end (5Ј-GCTGTAGGATCCATGCTGAGATTACT-TGAA-3Ј) and a KpnI site (underlined) after the stop codon (italic) at the 3Ј end (5Ј-GCATGCGGTACCTTACAGCAAAAGATAACC-3Ј) using the same conditions as described above for control Ada-C construct. The 531-bp product was digested with BamHI and KpnI enzymes and ligated into pQE30 vector digested with the same enzymes to form pQE30-Ogt. The ligation mixture was then electroporated into JM109 cells.
Purification of Ogt and Control and Mutant Ada-C Proteins-JM109 cells transformed with pQE30-Ogt, pQE30-Ada-C, or one of the mutants were grown in a 1-liter culture inoculated with a grown-overnight culture at 1:50 dilution. At a cell density equivalent to an A 600 of 0.5, isopropyl ␤-D-thiogalactopyranoside was added to a final concentration of 0.3 mM, and the cells were harvested 4 h later. The pellet was suspended in 30 ml of 20 mM Tris-HCl, pH 8.0, 250 mM NaCl and disrupted using a French press. After centrifugation at 17,000 ϫ g for 45 min at 4°C to remove cell debris, the supernatant was applied to a 2-ml column of Talon IMAC resin (Clonetech) equilibrated with 20 mM Tris-HCl, pH 8.0, 250 mM NaCl, and the column was washed with this buffer containing 10 mM imidazole. The Ada-C protein was then eluted using 200 mM imidazole, and the fractions found to contain Ada-C by SDS-polyacrylamide gel electrophoresis were pooled and dialyzed immediately against 50 mM Tris-HCl, pH 7.6, 250 mM NaCl, 5 mM dithiothreitol, and 0.1 mM EDTA. The yield of protein was about 8 mg/liter of culture, and the purity was Ͼ90% as judged by SDS-polyacrylamide gel electrophoresis.
Purification of AGT-AGT was expressed and purified to homogeneity by ammonium sulfate precipitation, chromatography on Mono-S, and gel-filtration as described previously (10,20).
Assay of Dealkylation of Oligodeoxyribonucleotides Containing O 6 -Methylguanine or O 6 -Benzylguanine-The purified alkyltransferase proteins were incubated in 50 mM Tris-HCl, pH 7.5, 0. 5 mM dithiothreitol, and 0.1 mM EDTA with 5Ј-d(AACAGCCATATa 6 GGCCC)-3Ј in which a 6 G represents O 6 -benzylguanine or O 6 -methylguanine. The reaction was stopped by the addition of 1% SDS, and the mixture of oligodeoxyribonucleotides was separated by HPLC. The separation was carried out on a Beckman C18 5-m Ultrasphere ODS reverse-phase column (4.6 ϫ 250 mm) with a Brownlee RP-300 Aquapore 7-m precolumn (0.46 ϫ 300 mm) at 45°C using a flow rate of 2 ml/min and a linear gradient of increasing methanol at 0.23% per minute over 30 min, starting from 14% methanol in 50 mM sodium phosphate, pH 6.3. The retention times of the unalkylated, methylated, and benzylated oligodeoxyribonucleotides that were detected by absorbance at 254 nm were 9.5, 15, and 21 min, respectively.
Assay of Sensitivity to Inactivation by O 6 -Benzylguanine and Related Compounds--The purified Ada-C protein or mutants were incubated with O 6 -benzylguanine or other inhibitors in 0.5 ml of 50 mM Tris-HCl, pH 7.5, 0.1 mM EDTA, and 5.0 mM dithiothreitol in the presence or absence of calf thymus DNA as shown for 30 min at 37°C. The residual alkyltransferase activity was then determined by a 30-min incubation with a 3 H-methylated DNA substrate that had been methylated by reaction with N-[ 3 H]methyl-N-nitrosourea essentially as described (9,21).
The alkyltransferase samples were incubated with 7 g of 3 H-methylated DNA substrate containing 7,000 cpm of O 6 -[ 3 H]methylguanine and 0.36 mg of carrier calf thymus DNA in 50 mM Tris-HCl, pH 7.5, 5 mM dithiothreitol, and 0.1 mM EDTA in a total volume of 1 ml at 37°C for 30 min. The DNA was then precipitated with 0.25 N perchloric acid at 4°C and hydrolyzed by heating in 0.5 ml of 0.1 N HCl at 70°C for 30 min. This procedure was repeated, and the combined hydrolysates were taken, neutralized by the addition of 0.15 ml of 1 M Tris-HCl, and filtered through a circular 0.45-m filter (Millipore). The labeled 7-methylguanine and O 6 -methylguanine present were then separated by HPLC on a Beckman Ultrasphere ODS column (25 ϫ 0.46 cm) using isocratic elution at 37°C with 0.05 M ammonium formate, pH 4.5, containing 12% methanol. The eluate was monitored for radioactivity by mixing with 3.5 parts of Flow Scint III (Packard Instruments, Meriden, CT) and passing through a Radiomatic Flo-One/Beta A-140A radioactivity monitor (Packard Instruments). The counting efficiency was 35%.
The results were expressed as the percentage of the alkyltransferase remaining. Less than 5% of the activity was lost on incubation in the absence of the inhibitor. The graphs of activity remaining against inhibitor concentration were used to calculate an ED 50 value representing the amount of inhibitor needed to produce a 50% loss of activity.
Reaction of AGT with O 6 -Benzylguanine-Measurements of [8-3 H]guanine formation from O 6 -benzyl[8-3 H]guanine were carried out using an assay mixture consisting of 50 mM Tris-HCl, pH 7.5, 0.1 mM EDTA, and 5 mM dithiothreitol in a volume of 0.25 ml in the presence or absence of DNA as indicated. The formation of labeled product was stopped by the addition of 0.6 -0.8 ml of the same buffer containing 0.2 mM guanine and 0.2 mM O 6 -benzylguanine. Aliquots were then separated by reverse phase-HPLC on a Beckman Ultrasphere ODS column (25 cm ϫ 4.6 mm) using isocratic elution at a temperature of 36°C with a buffer of equal parts methanol and 0.05 M ammonium formate, pH 4.5. The eluate from the HPLC was monitored for radioactivity by mixing with 3.5 volumes of Flow Scint III and passing through a Radiomatic Flo-One Beta A-140A radioactivity monitor.

RESULTS
Experiments were carried out using the carboxyl-terminal domain of Ada-C, which is responsible for the activity to repair O 6 -methylguanine. The protein expressed contained residues 179 -354 of the full-length Ada protein. For comparison, all amino acid residues are indicated by number according to the full-length sequence. The key region of the protein containing the active-site cysteine residue that is located between Asn 313 and Lys 341 , which are conserved in all known alkyltransferases, is shown in Fig. 1.
Purified Ada-C was totally resistant to inactivation by O 6benzylguanine, and the resistance was not altered when calf thymus DNA was added (Fig. 2a). This confirms previous studies in which crude cell extracts of E. coli were used (10 -13) and rules out the possibility that the resistance is due to the rapid metabolism of the O 6 -benzylguanine by bacterial enzymes. The crystal structure of Ada-C suggests that the tryptophan residue at position 336 may limit access to the cysteine acceptor site (6,7). Studies with the human AGT have indicated that a proline residue located 5 residues to the amino side of this cysteine is necessary for optimal reaction with O 6 -benzylguanine (22,23). The equivalent position (316) in the Ada-C protein is occupied by an alanine. Therefore, a double mutant of the Ada-C protein was made in which Trp 336 was changed to Ala and Ala 316 was changed to Pro (13) (see Fig. 1). This protein was purified to homogeneity and found to be definitely sensitive to inactivation by O 6 -benzylguanine (Fig. 2a), but the inhibition was relatively weak with an ED 50 value of 280 M (Table I). This compares to a value of about 0.3 M for the inactivation of human AGT by O 6 -benzylguanine assayed under the same conditions (22). However, with the A316P/W336A-Ada-C mutant, there was a large stimulation of the rate of inactivation by O 6 -benzylguanine when DNA was present (Fig. 2a). The ED 50 value was reduced by 28-fold to 10 M in the presence of DNA.
Although A316P/W336A-Ada-C is clearly sensitive to these inhibitors, it is still 2 orders of magnitude more resistant than the human AGT for which the ED 50 values in the presence of DNA are 0.1 M for O 6 -benzylguanine (Table I) and 0.04 M for 2,4-diamino-6-benzyloxy-5-nitrosopyrimidine and 2,4-diamino-6-benzyloxy-5-nitropyrimidine (16). This difference could not be reduced by changing Trp 336 to Gly, which occupies the equivalent position in AGT, or by including a third mutation in which Lys 314 was changed to Pro (Table I). [The rationale for introducing a second Pro at this position is that all of the O 6 -benzylguanine-sensitive mammalian AGT sequences including that of human AGT (Fig. 1) contain two prolines at positions equivalent to 314 and 316 of Ada-C (4, 10)]. In fact, as shown in Table I, A316P/W336G-Ada-C was actually significantly less susceptible than A316P/W336A-Ada-C, and the addition of a K314P mutation to either reduced sensitivity even further.
Previous studies with the human AGT have shown that O 6 -benzylguanine inactivates the protein by acting as a pseudosubstrate and that inactivation is accompanied by the The control (open symbols) and mutant A316P/W336A-Ada-C (filled symbols) were incubated for 30 min with the concentration of inhibitor shown in the presence (squares) or absence (circles) of 100 g of calf thymus DNA. The remaining activity was then determined and plotted as a percentage of that present when no inhibitor was added. a, results for O 6 -benzylguanine; b, results for 2,4-diamino-6-benzyloxy-5-nitrosopyrimidine; c, results for 2,4-diamino-6-benzyloxy-5-nitropyrimidine. formation of S-benzylcysteine at the active site with the liberation of guanine (10,24). The rate of formation of guanine can therefore be used to measure the ability of the alkyltransferase to react with O 6 -benzylguanine. The inability of Ada-C alkyltransferase to react with O 6 -benzylguanine was confirmed in this way because no [8-3 H]guanine was detected even when 300 g of protein was incubated for up to 3 h with O 6 -benzyl[8-3 H]guanine (Table I). Formation of [8-3 H]guanine by A316P/ W336A-Ada-C was easily measured, and the rate increased 24-fold in the presence of DNA (Table I). K314P/A316P/W336A-Ada-C was slightly less active in guanine production but was also stimulated 25-fold by DNA. The A316P/W336G-Ada-C and K314P/A316P/W336G-Ada-C mutants were also slightly less active, and guanine production was stimulated only 17-fold. These results agree well with the sensitivity to inactivation by O 6 -benzylguanine. They are also consistent with the ED 50 and the level of [8-3 H]guanine production by the human AGT. The latter is about 100 times greater than the formation from A316P/W336A-Ada-C (Table I).
The effect of DNA concentration on the stimulation of [8-3 H]guanine formation from O 6 -benzyl[8-3 H]guanine was measured using a 16-mer oligodeoxyribonucleotide (Fig. 3). Maximal stimulation required the addition of about 10 g (2 nmol) of the oligodeoxyribonucleotide and amounted to 25-fold and 17-fold, depending on whether residue Trp 336 was Ala or Gly (Fig. 3). When the human AGT was used, the maximal stimulation was only 6-fold (Fig. 3), in good agreement with earlier reports (25). Although the experiment shown in Fig. 3 was carried out with a small oligodeoxyribonucleotide, the maximal stimulation values obtained agree well with those found when high molecular weight calf thymus DNA was used ( Table I).
The production of guanine from O 6 -benzylguanine in the absence of DNA by A316P/W336A-Ada-C was relatively insensitive to salt, with no decrease at 0.2 M NaCl and only a 22% reduction when 0.5 M NaCl was added (Fig. 4, insert). In contrast, stimulation of the activity by DNA was very salt-sensitive, with an 82% reduction at 0.2 M NaCl and complete abolition of the stimulation at 0.5 M (Fig. 4). This result is consistent with either a strong inhibition of DNA binding by salt or its interference with the activation of the protein structure caused by DNA binding.
The ability of the human and E. coli alkyltransferase preparations to react with O 6 -benzylguanine when it is incorporated into an oligodeoxyribonucleotide was examined by using a 16-mer oligodeoxyribonucleotide that contains O 6 -benzylguanine. The AGT, Ada-C, and A316P/W336A-Ada-C proteins were each incubated with this substrate for 10 min, and the production of the unalkylated form of the 16-mer oligodeoxyribonucleotide was measured after HPLC separation of substrate and product (Fig. 5a). All of the proteins were clearly able to remove benzyl groups from this substrate, but considerably more of the Ada-C protein was needed to achieve the same level of repair, and even with a considerable molar excess of the protein, repair was not complete. This suggests that repair is much slower by this protein than by AGT or A316P/W336A-Ada-C. This was confirmed by measuring the time course of repair of the benzylated 16-mer oligodeoxyribonucleotide (Fig.  5b). Repair by AGT or A316P/W336A-Ada-C was so rapid that the reaction had gone to completion within 1 min. In contrast, even with 5 times the amount of Ada-C protein, about 20 min were required for complete reaction (Fig. 5b).
To compare the abilities of the alkyltransferases to repair O 6 -benzylguanine with the repair of O 6 -methylguanine, the proteins were incubated with a mixture of the 16-mer oligodeoxyribonucleotide containing O 6 -benzylguanine and a 16-mer  oligodeoxyribonucleotide of identical sequence containing O 6methylguanine (Fig. 6). The Ada-C efficiently repaired the O 6methylguanine in this mixture, but repair of O 6 -benzylguanine was only apparent when all of the methylated substrate had been repaired. In contrast, the AGT and the A316P/W336A-Ada-C mutant proteins repaired both adducts, but these proteins preferentially acted on O 6 -benzylguanine because the peak corresponding to the 16-mer oligodeoxyribonucleotide containing O 6 -benzylguanine was reduced to a greater extent than that which represents the 16-mer oligodeoxyribonucleotide containing O 6 -methylguanine. Thus, benzyl groups are actually preferred substrates for both of these alkyltransferases. E. coli contains a second alkyltransferase gene (1,26,27). Its product, the Ogt protein, is a constitutive alkyltransferase. As shown in Fig. 5b, Ogt protein very rapidly repairs O 6 -benzylguanine contained in the 16-mer oligodeoxyribonucleotide. When tested by incubation with the mixture of the 16-mer oligodeoxyribonucleotides containing O 6 -benzylguanine and O 6 -methylguanine, the Ogt protein repaired both adducts at comparable rates (Fig. 6). Previous studies have shown that Ogt shows some sensitivity to inactivation by O 6 -benzylguanine but is considerably less sensitive than the human AGT (10,11). Thus, the Ogt protein resembles the A316P/W336A-Ada-C mutant in both a weak ability to react with O 6 -benzylguanine as a free base and a rapid reaction with this adduct when it is present in an oligodeoxyribonucleotide. As shown in Fig. 1, the Ogt protein has glycine in the position equivalent to Trp 336 of Ada-C and does contain a proline residue at the position equivalent to Lys 314 . DISCUSSION Preliminary results in model systems consisting of human tumors grown as xenografts in mice have suggested that O 6benzylguanine may be a useful drug to enhance chemotherapy by certain alkylating agents (4, 28 -30). Phase I clinical trials to develop this concept are currently underway. However, irrespective of the results of these trials, O 6 -benzylguanine may not be an ideal agent for this purpose. It is only sparingly soluble in water, and a polyethylene glycol-based vehicle is being used for administration to achieve plasma concentrations adequate to reduce the human AGT level adequately. The more active pyrimidine analogs of O 6 -benzylguanine such as 2,4diamino-6-benzyloxy-5-nitrosopyrimidine are very rapidly degraded and excreted (31,32). More potent water-soluble and metabolically stable derivatives of O 6 -benzylguanine may therefore offer advantages, but the lack of a clear understanding of how O 6 -benzylguanine binds to the active site of the protein and is recognized as a substrate is a handicap to such studies. Molecular modeling approaches would be a useful way to facilitate the design of better inhibitors, but at present the absence of a three-dimensional structure for the AGT protein prevents this approach. Because the crystallization and determination of the structure of Ada-C has already been accomplished (6,7), the identification of mutant forms of Ada-C that are able to react with O 6 -benzylguanine should facilitate such modeling studies.
Our results provide strong support for the concept that the inability of certain alkyltransferases to react with O 6 -benzylguanine is due to steric factors blocking the access of the relatively bulky benzyl group to the cysteine acceptor site (13). Neither the W336A mutation as shown here nor the P316A mutation (13) alone are sufficient to impart significant sensitivity of Ada-C to O 6 -benzylguanine. This steric hindrance is therefore provided by: (a) the presence of a bulky tryptophan residue at position 336, which in the sensitive mammalian alkyltransferases is occupied by a glycine; and (b) the absence of a proline residue at the position-5 residues on the aminoterminal side of the active-site cysteine. All known mammalian alkyltransferases, which are readily inactivated by O 6 -benzylguanine, have this proline, and most of the microbial alkyltransferases lack it. The yeast alkyltransferase, which is totally refractory to O 6 -benzylguanine (10), lacks this proline and has a tryptophan residue at the equivalent position of tryptophan 336 in Ada-C. The E. coli Ogt protein, which is weakly reactive with O 6 -benzylguanine (10,11), lacks both the tryptophan and this proline but has a proline residue in the Ϫ7 position. All of the mammalian alkyltransferases have proline residues in both the Ϫ7 and Ϫ5 positions. Mutation of either of these residues to alanine imparts some resistance of the human alkyltransferase to O 6 -benzylguanine, but changing both has a much greater effect (22). It is therefore possible that the second proline at the Ϫ7 position in Ogt contributes to its ability to repair O 6 -benzylguanine.
The crystal structure of the Ada-C protein indicates that a change in the conformation of the protein must occur in response to DNA binding, and it was postulated that this activates the protein (6,7). Our results showing that the conversion of O 6 -benzylguanine to guanine by the mutant A316P/ W336A-Ada-C is stimulated about 25-fold by addition of DNA at low salt concentrations are consistent with this hypothesis. It is noteworthy that the stimulation of the formation of guanine from O 6 -benzylguanine by the mutant A316P/W336A-Ada-C by the 16-mer oligodeoxyribonucleotide (Fig. 2) was maximal with only about 2 nmol of the 16-mer oligodeoxyribonucleotide, even though the reaction contained 5 nmol of Ada-C protein. This indicates that more than 1 molecule of the protein can bind to 1 molecule of the 16-mer oligodeoxyribonucleotide and is consistent with a recent report based on sedimentation equilibrium analysis that up to 4 molecules of AGT can be accommodated on a 16-nucleotide length of DNA (33).
The reaction of the mutant A316P/W336A-Ada-C and of the Ogt alkyltransferase with O 6 -benzylguanine when this substrate was contained in an oligodeoxyribonucleotide was very much faster than the reaction with the free base, even when the latter reaction rate was enhanced by the addition of unalkylated DNA. The wild-type Ada-C reacts with O 6 -benzylguanine only when it is incorporated into an oligodeoxyribonucleotide. This indicates that the binding of the DNA positions the adduct in the correct orientation to facilitate repair. There are now numerous examples of enzymes that repair or methylate DNA that act via base-flipping mechanisms (34 -36). These include HhaI and HaeIII methyltransferases (37,38), T4 endonuclease V (39,40), uracil DNA glycosylase (41,42), 3-methyladenine DNA glycosylase II (43,44), and DNA photolyase (45). Several authors have suggested that alkyltransferase follows a similar general mechanism (34,36,46,47), and the crystal structure of the Ada-C protein would be consistent with such a model, in which the methylated base is flipped out of the DNA structure and rotated into a binding pocket in the protein (6,7). Our results show that as long as this pocket in the alkyltransferase can accommodate the benzyl group, this substrate is preferred over methyl. Such a preference could be related to the more facile displacement of benzyl groups over methyl in bimolecular displacement reactions or to the presence of a hydrophobic region at the active site that binds benzyl better than methyl.
Although it is well established that alkyl adducts on the O 6 -position of guanine that are slightly larger than methyl such as ethyl and chloroethyl are effectively removed from DNA by alkyltransferases from mammalian sources and bacteria (1,2,8,14), there is little information showing repair of more bulky groups. Our results indicate that AGT would be expected to remove benzyl and related groups very effectively and thus could protect against some environmental carcinogens such as N-nitrosomethylbenzylamine and 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (48,49). For example, O 6benzyl groups produced by agents such as the esophageal carcinogen N-nitrosomethylbenzylamine (50) would be repaired very effectively, and our results support preliminary indications that mammalian alkyltransferases remove O 6 -pyridyloxobutyl adducts produced by the principal tobacco carcinogen, 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (51,52). Such repair is potentially very important for two reasons: (a) the alkyltransferase protein may be an important means of protection against the toxic effects of such adducts, and (b) both N-nitrosomethylbenzylamine and 4-(methylnitrosamino)-1-(3pyridyl)-1-butanone are also metabolized to methylating agents that form O 6 -methylguanine. Because there is a limited supply of alkyltransferase molecules that are not regenerated after use, the preferential repair of benzyl groups may lead to increased persistence of O 6 -methylguanine. This, in turn, may lead to the occurrence of G to A transitions in the Ki-ras gene that are well documented in tumors arising from 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone administration (53).
It is not known with certainty what is the source of DNA damage that provides the natural substrate for DNA repair by the alkyltransferase family of proteins. The ubiquitous nature of these proteins suggests that there is a widespread endogenous alkylating agent that forms adducts at the O 6 position of guanine. It has been suggested that this agent is either Sadenosylmethionine (54) or a nitroso compound formed by re-action with an endogenous amine (55). Lipid peroxidation products are also possible sources (56). The possibility that DNA methylation by one of these mechanisms has necessitated the production of a specific repair protein to deal with it would explain why all of the known alkyltransferase repair proteins act very efficiently on O 6 -methyl groups but vary dramatically in their abilities to accommodate larger adducts. However, the increased spontaneous mutations occurring in E. coli strains deficient in both Ada and Ogt alkyltransferases are not all the G to A and T to C transitions that would be expected from miscoding of the O 6 -methylguanine and O 4 -methylthymine substrates for alkyltransferase (57). Our results showing that benzyl adducts are not only substrates for the human AGT and E. coli Ogt reaction but are actually preferred substrates suggest that there may well be other endogenous alkylating agents. The noninducible Ogt protein is more likely than Ada to be responsible for reducing the rate of spontaneous mutations in E. coli (19). Although this Ogt protein reacts only weakly with O 6 -benzylguanine as a free base (10,11), as shown in Figs. 5b and 6, it reacts much more readily with O 6 -benzylguanine when this is present in DNA, and such adducts are therefore highly likely to be targets of repair in vivo.