MutS Recognition of Exocyclic DNA Adducts That Are Endogenous Products of Lipid Oxidation*

The ability of the methyl-directed mismatch repair system to recognize and repair the exocyclic adducts propanodeoxyguanosine (PdG) and pyrimido[1,2-α]purin-10(3H)-one (M1G), the major adduct derived from the endogenous mutagen malondialdehyde, has been assessed both in vivo andin vitro. Both adducts were site-specifically incorporated into M13MB102 DNA, and the adducted genomes were electroporated into wild-type or mutS-deficient Escherichia colistrains. A decrease in mutations caused by both adducts was observed inmutS-deficient strains, suggesting that MutS was binding to the adducts and blocking repair by nucleotide excision repair. This hypothesis was supported by the differences in mutation frequency observed when hemimethylated genomes containing PdG on the (−)-strand were electroporated into a uvrA − strain. The ability of purified MutS to bind to PdG- or M1G-containing 31-mer duplexes in vitro was assessed using both surface plasmon resonance and gel shift assays. MutS bound to M1G:T-containing duplexes with similar affinity to a G:T mismatch but less strongly to M1G:C- and PdG-containing duplexes. Dissociation from each of the adduct-containing duplexes occurred at a faster rate than from a G:T mismatch. The present results indicate that MutS can bind to exocyclic adducts resulting from endogenous DNA damage and trigger their removal by mismatch repair or protect them from removal by nucleotide excision repair.

Malondialdehyde (MDA) 1 is produced endogenously through the processes of lipid peroxidation and eicosanoid biosynthesis (1,2). MDA is mutagenic in bacterial and mammalian cell assays and carcinogenic in rodents (3)(4)(5)(6)(7). It also induces p53independent cell cycle arrest at the G 1 /S and G 2 /M checkpoints (8). MDA reacts with DNA, forming a pyrimidopurinone adduct to deoxyguanosine (M 1 G) and an oxopropenyl adduct to deoxyadenosine (N 6 -[3-(1-oxopropenyl)]deoxyadenosine) (M 1 A) ( Fig.  1) (9 -12). M 1 G also can be formed by the reaction of the oxidative DNA damage product, base propenal, with deoxyguanosine (13). M 1 G is an abundant constituent of DNA from healthy human beings. It has been detected in several human tissues at levels from 2-150 per 10 8 bases (14 -17). M 1 G may account for a significant portion of the genotoxic and cell cycle regulatory activity of MDA. Site-specific mutagenesis experiments indicate that M 1 G induces mutations to T and A on replication in Escherichia coli and is a block to replication (18).
Given the high biological activity of M 1 G, it is important to identify the pathways by which it may be removed from DNA. We have shown that M 1 G and a structural analog, propanodeoxyguanosine (PdG), are removed in E. coli by nucleotide excision repair and that PdG is excised by both E. coli and mammalian nucleotide excision repair complexes in vitro (18,19). The experiments described in the present report were designed to test the hypothesis that M 1 G is recognized by the mismatch repair system. Mismatch repair exists to correct errors that arise during DNA replication, but recent studies indicate that it recognizes and acts on damaged DNA, including duplexes containing alkylated or platinated bases (20 -26). In both cases, the mismatch repair system attempts to remove the normal base opposite the adduct, which sets up a futile cycle of repair and replication that leads to cell toxicity. M 1 G and other exocyclic adducts are relatively small adducts that resemble normal DNA bases, so they may be substrates for removal by mismatch repair. Transfection of M13 genomes containing single M 1 G or PdG adducts into E. coli strains deficient in mismatch repair suggested that both adducts are recognized and repaired by MutS-dependent mismatch repair. This conclusion was supported by in vitro studies of purified MutS protein binding to M 1 G-and PdG-containing duplexes.
nucleotides were synthesized, purified, and characterized as described. Following purification, the M 1 G oligonucleotides were determined to be 99.7% pure by 20% polyacrylamide gel electrophoresis. For the surface plasmon resonance assays, duplexes were 5Ј-biotinylated by phosphoramidite chemistry on the nonadducted strand. Oligonucleotides used as hybridization probes were prepared using an Applied Biosystems (Foster City, CA) automated DNA synthesizer in the Vanderbilt University Center in Molecular Toxicology Molecular Genetics Core and purified using a SurePure TM oligonucleotide purification kit from Amersham Pharmacia Biotech.
Bacterial Strains and DNA Isolation-Single-stranded M13MB102 for the construction of M 1 G:C, M 1 G:T, or G:C genomes was isolated as described (27). Replicative form M13MB102 DNA was harvested using Qiagen columns (Chatsworth, CA). E. coli strain JM105 was used as the host bacterium for the production of M13MB102 methylated DNA and for the indicator plates. E. coli strain JM110 (dam Ϫ ) was used as the host bacterium for the production of unmethylated M13MB102 in hemimethylated genome studies.
Construction of Unadducted or Adducted M13MB102 Viral Genomes-Construction of gapped duplex M13MB102 DNA and ligation of adducted or unadducted 8-mers were as described previously (28). Briefly, double-stranded M13MB102 DNA was linearized with KspI and BssHII and then dialyzed with a 12-fold excess of single-stranded M13MB102 DNA in decreasing concentrations of formamide. In some experiments, the single strand contained uracil in place of thymine residues to minimize replication of the nonadducted strand. The resultant gapped duplex DNA was isolated by a 0.8% low melting point agarose gel run in 40 mM Tris acetate/1 mM EDTA buffer (pH 8). The gapped duplex band was excised, and the DNA recovered using the enzyme GELase TM , according to the procedure provided by Epicentre Technologies. From this point on, the use of Tris buffers was avoided because of observations that Tris can conjugate to M 1 G in basic and frozen conditions (29). M 1 G-, PdG-, and G-containing 8-mers (100 pmol) were phosphorylated prior to ligation using ATP (50 M) and T4 polynucleotide kinase in 50 mM MOPS buffer (pH 7.2). For the ligations, gapped duplex DNA was added to each of the phosphorylated G-, PdG-, and M 1 G-containing 8-mers along with 400 units of T4 DNA ligase and ATP (1 mM). The ligation reaction proceeded for 4 h at 16°C in 50 mM MOPS buffer. The reaction mixtures were then brought up to a volume of 100 l with water, and the DNA was purified by spinning through modified polyvinylidene difluoride membranes (Millipore). The ligation products were resolved on a 40 mM MOPS, 0.8% low melting point agarose gel. Doubly ligated DNA was excised from the gel and recovered using GELase TM except the supplied 50ϫ Bis-Tris-NaCl buffer was not used. No additional buffer was added. The amount of enzyme used for digestion was increased from 1 unit of enzyme/600 mg of gel slice to 1 unit/300 mg. Formation of G:T-and M 1 G:T M13MB102 DNA was the same as above except during the formamide dialysis, a 12-fold excess of single-stranded M13MB102 was used that contained a T at position 6256.
Hemimethylated DNA was prepared similarly except the singlestranded DNA did not contain uracil. Gapped duplex DNA with methyl groups on the (Ϫ)-strand was prepared by formamide dialysis of methylated linear M13MB102 and unmethylated single-stranded M13MB102 DNA. Gapped duplex DNA with methyl groups on the (ϩ)-strand was prepared by formamide dialysis of unmethylated linear M13MB102 and methylated single-stranded M13MB102 DNA.
Transformation of E. coli Cells and Determination of Mutation Frequency and Strand Utilization-Cells were SOS-induced and transformed by electroporation as described previously (18). Briefly, bacteria in logarithmic phase growth were SOS-induced with UV light before making them competent for transformation. The UV dose was determined by irradiating cells at increasing times from 0 to 3 min and then plating dilutions of the irradiated cells on LB plates. The optimal UV dose corresponded to roughly a 10% survival of the cells as compared with no exposure. For transformation, 3 l of DNA sample (ϳ25 ng/l) was added to 20 l of cells. The cell/DNA mixture was placed into a chilled microelectroporation cuvette (Life Technologies, Inc.), and the electroporations were performed at 1.5 kV/cm using a Life Technologies, Inc. Cell-Porator E. coli electroporation system. After electroporation, 1 ml of SOC medium was added, and the bacteria were plated on LB plates in the presence of competent bacteria and isopropyl-␤-Dthiogalactoside and allowed to grow overnight.
To determine mutation frequencies, phage were eluted from the primary transformation plates, diluted, and then replated with JM105 on X-gal/isopropyl-␤-D-thiogalactoside indicator plates to give roughly 300 plaques/plate. The plaques on the secondary plates were then lifted using nitrocellulose membranes and probed for base pair substitution mutations at position 6256 by differential hybridization with 13-mer probes. Membranes from 12 modified phage plates and 12 unmodified phage plates were split evenly into four dishes. Each dish contained one of the four probes. Because there was only one lift/plate and not four identical lifts with one membrane being placed into each dish, the summation of mutations detected along with G hybridizations sometimes did not add up to 100%. The specificity of the probes for a 1-base change at position 6256 has been shown previously (28,30).
In Vitro Gel Mobility Shift Assays-(Ϫ)-Strand oligonucleotides were phosphorylated with 10 Ci of [␥-32 P]ATP using 10 units of T4 polynucleotide kinase and 5 l of kinase dilution buffer. The reaction was incubated at 37°C for 30 min and then heated to 70°C for 10 min to inactivate the enzyme. The labeled oligonucleotide was purified by elution from a G50 microspin (Amersham Pharmacia Biotech) column and annealed to a 5-fold excess of its unlabeled complementary strand. All binding reactions were carried out in 20-l incubations on ice. 10 l of 0 -400 nM purified MutS protein was added to 5 l (0.8 pmol/l) of labeled oligonucleotide and 5 l of reaction buffer (0.01 M MOPS-KOH, pH 7.4, 0.15 M NaCl, 8.4 mM MgCl 2 , 3.4 mM EDTA, 10 ng/l poly(dI⅐dC), 15% glycerol). Reactions were incubated on ice for 15 min, before loading onto a 4% polyacrylamide, 4% glycerol gel with a 40 mM MOPS, 1 mM MgCl 2 buffer run at constant voltage of 200 V for 2.5 h. Gels were then dried and analyzed using a 400E PhosphorImager (Molecular Dynamics, Sunnydale, CA).
Surface Plasmon Resonance Measurements-Surface plasmon resonance measurements were performed using a streptavidin chip (Amersham Pharmacia Biotech) and duplex DNA containing a 5Ј-biotin on the nonadducted strand. Homo-or heteroduplex 31-mers were prepared by heating to 80°C and cooling to room temperature over a 3-h period, in the presence of a 5-fold excess of the nonbiotinylated strand. All surface plasmon resonance assays were performed using a Biacore 2000 (Amersham Pharmacia Biotech) at 25°C in a running buffer of 0.01 M MOPS-KOH (pH 7.4), 0.15 M NaCl, 3.4 mM EDTA, 8.4 mM MgCl 2 , 0.008% Surfactant P20. Each streptavidin chip was washed with 20 l of 50 nM NaOH and each flow cell derivatized with a duplex oligonucleotide at a flow rate of 5 l/min, until a total binding of approximately 60 resonance units (RU) was achieved in each flow cell. MutS binding to the homo-or heteroduplex modified chips was assayed at various concentrations at a flow rate of 20 l/min, and the DNA derivatized chips were regenerated by washing with 20 l of 0.5% SDS. When included, ATP, ADP, or nonhydrolyzable analogs were added to the reaction mixture in 2-fold excess of MutS concentration.
Binding to each adduct was assayed at five different concentrations, and sensograms were prepared for kinetic analysis by subtraction of nonspecific binding to a G:C match. BIAEVAL 3.0 (Amersham Pharmacia Biotech) global analysis software was employed to simultaneously fit the association and dissociation data to a 1:1 Langmuir binding model as described previously (31).

Detection of M 1 G Mutagenicity in Wild-type and Mismatch
Repair Deficient Backgrounds-Duplex M13MB102 genomes containing G, PdG, or M 1 G at position 6256 were constructed by ligation of 8-mer oligonucleotides into gapped duplexes as described previously. The resulting vectors were electroporated into SOS-induced E. coli strains that were wild type or deficient in mismatch repair, and progeny phage were probed for mutations. Single base pair substitutions were detected by differential hybridization of plaque DNA with radiolabeled 13-mer probes specific for each type of base pair substitution at position 6526. It was anticipated that the involvement of mismatch repair in the removal of M 1 G or PdG would result in an increased mutation frequency in cells that were deficient in mismatch repair. Such an effect is seen following transformation of PdG-or M 1 G-containing genomes into strains deficient in nucleotide excision repair (18).
Surprisingly, transformation of either PdG-or M 1 G-containing genomes into a MutS-deficient strain led to a decrease in mutation frequency (Table I). Approximately a 3-fold decrease in total mutations was observed with genomes containing PdG. The percentage of mutations induced by M 1 G decreased to below the limit of detection of the assay.
We hypothesized that the decrease in mutations observed in mutS Ϫ cells is due to competitive binding of MutS and the uvrA 2 B complex to the adducts but repair only by the nucleotide excision complex (Fig. 2). MutS cannot initiate repair because both the (ϩ)-and (Ϫ)-strands of M13MB102 are meth-ylated. Thus, in a wild-type cell, MutS protects the adduct from repair by uvr(A)BC. In the (uvrA Ϫ ) strain, only MutS is present, so repair is not initiated, and the mutation frequency is increased 3-4-fold because of the prolonged half-life of the adduct. However, in the mutS Ϫ strain, uvr(A)BC is able to repair the adduct, and a decreased mutation frequency is observed. These observations and the hypothesis to explain them raise the question of whether the mismatch repair system is capable of removing PdG or M 1 G.
Mismatch Repair in Differentially Methylated DNA-To answer this question, we constructed hemimethylated plasmids containing PdG on the (Ϫ)-strand and methyl groups on either the (ϩ)-or (Ϫ)-strand. Methylated single-stranded or doublestranded M13MB102 was isolated from JM105 (damϩ), and unmethylated single-or double-stranded DNA was isolated from JM110 (dam Ϫ ). Hemimethylated DNA was prepared by formamide dialysis, and gapped duplex DNA was prepared with methyl groups placed selectively on the (ϩ)-or (Ϫ)-strand. Hemimethylated, adducted plasmids were transformed into E. coli, and the percentage of base pair substitutions at position 6526 was determined. The frequency of PdG 3 A transitions and PdG 3 T transversions were equivalent in all strains tested, so the total percentage of mutations is reported for simplicity. The percentage of base pair substitutions observed in JM105 (wt) was 1.5% when PdG was on the methylated strand and 1.0% and 1.1% when the strand opposite PdG was methylated or both strands were methylated (Table II)  mutation frequencies recorded in these experiments relative to those summarized in Table I resulted from the absence of uracil residues in the (ϩ)-strand in the experiments summarized in Table II. The presence of uracil residues in the (ϩ)-strand lowers the replication of the (ϩ)-strand and increases the detection of mutations induced by adducts on the (Ϫ)-strand.
To probe for PdG removal by mismatch repair, differentially methylated genomes was transformed into uvrA-deficient cells. This was necessary to eliminate competitive nucleotide excision repair. As expected from previous results, the percentage of mutations observed in the uvrA Ϫ background was 3-4-fold higher when methyl groups were present on the PdG-containing strand or on both strands. In contrast, the percentage of mutations was much lower when methyl groups were present on the strand opposite PdG. This is consistent with removal of PdG by the methyl-directed mismatch repair system.
When differentially methylated, PdG-containing genomes were transformed into mutS Ϫ cells, a decrease in mutations comparable with that reported in Table I was observed regardless of the methylation status of the genome. This suggests that removal of PdG by nucleotide excision repair is not sensitive to the presence of methyl groups on either the adducted or nonadducted strand.
MutS Binding Detection by Surface Plasmon Resonance Assay-The results of the in vivo experiments suggested that MutS binds to PdG and M 1 G, triggering their removal by mismatch repair if the adduct-containing strand is not methylated. Therefore, we sought to confirm the ability of MutS to bind M 1 G or PdG in duplex DNA using an in vitro assay for protein-DNA interaction. Duplex 31-mers containing PdG or M 1 G at position 15 and C or T opposite the adduct were synthesized and bound to a streptavidin Biacore chip by the addition of a biotin residue to the 5Ј terminus of the unadducted oligonucleotide. A duplex oligonucleotide of identical sequence but containing a G:T mismatch at position 15 was used as a control.
MutS Binding to a G:T Mismatch-Two separate flow cells were derivatized individually with approximately 60 RU of either 5Ј-biotinylated G:T duplex or G:C duplex. The binding of solutions containing various concentrations of MutS were assessed as the MutS-containing solutions were passed across the derivatized chips (Fig. 3). After the binding of a 400 nM MutS solution had reached equilibrium, the binding to the G:T duplex produced an absolute change of 616 RU (100%) compared with a change of 61 RU (9.9%) in the G:C duplex. The latter indicated a weak nonspecific interaction between MutS and the G:C duplex. This nonspecific binding was subtracted out of all binding sensograms produced from G:T-or adduct-containing duplexes.
Binding to the G:T mismatch was assessed at various concentrations of MutS to produce binding isotherms that were subsequently analyzed for kinetic and thermodynamic param-eters. At MutS concentrations of 50 -400 nM, isotherms were globally fitted to a simple 1:1 Langmuir binding model. The BIAEVAL 3.0 program simultaneously determines kinetic association and dissociation constants from the rate of change of response with respect to time (slope of association and dissociation curves). The G:T mismatch had an apparent thermodynamic dissociation constant (K D ) of 18 nM. Kinetic measurements indicated a k d of 2.44 ϫ 10 Ϫ3 s Ϫ1 and k a of 1.36 ϫ 10 5 M Ϫ1 s Ϫ1 with a chi square value of 1.10, falling well within accepted statistical values for Biaeval 3.0 analysis. These thermodynamic constants compare favorably with a K D of 20 Ϯ 5 nM determined by DNA footprinting and gel shift analysis (32).
MutS Binding to PdG and M 1 G Adducts-The relative binding of MutS to PdG and M 1 G adducts was assessed in a manner similar to the G:T mismatch. Each adduct was probed at several different concentrations of MutS, nonspecific binding was subtracted out, and the sensograms were subjected to global kinetic analysis. As illustrated in Fig. 4, when all adductcontaining duplexes are probed with the same concentration of 200 nM MutS, the equilibrium level of binding of MutS to the M 1 G:T duplex is very close to that of a G:T mismatch (240 RU versus 219 RU). Binding to an M 1 G:C duplex gave a lower absolute response of 138 RU, and binding to both PdG:C and PdG:T duplexes was even less significant. PdG:C showed an

FIG. 3. Specificity of binding of MutS to G:T mismatch.
Homoor heteroduplexes containing a G:T mismatch were created by annealing a 5-fold excess of the nonbiotinylated strand to the 5Ј-biotinylated strand. DNA (60 RU) was immobilized on a streptavidin surface in separate flow cells of a Biacore chip. A, a saturating amount of MutS protein (50 l, 400 nM) was washed over each flow cell at a flow rate of 20 l/min before the protein was washed off with 20 l of 0.5% SDS (not shown). B, the bulk refractive index contribution from buffer solution was subtracted from sensograms to display the low level of nonspecific binding to the G:C homoduplex. equilibrium response of 85 RU, and PdG:T showed only 41 RU.
Although the thermodynamic interactions between MutS and M 1 G:T or G:T duplexes are similar, Fig. 4 illustrates that there are differences in the rates of association and dissociation. Whereas MutS binds to PdG and M 1 G duplexes with a k a comparable with that of a G:T duplex, dissociation from PdG or M 1 G duplexes occurs significantly faster than dissociation from the G:T duplex. This observation is confirmed by the kinetic and thermodynamic constants that were extrapolated from sensogram data (Table III). The dissociation of MutS from M 1 G and PdG duplexes occurs approximately 5-fold faster than dissociation from the G:T duplex. Because the association rates are comparable, the lower K D for the MutS-G:T duplex reflects the lower rate of dissociation.
MutS Binding to M 1 G:T in the Presence of ATP or ADP-Adenine nucleotide-binding sites have been found to be highly conserved in both eukaryotic and prokaryotic MutS homologs (33). Several of the known MutS homologs fail to form complexes with mismatches in the presence of ATP by gel shift analysis. Therefore, we attempted to use adenine nucleotide modulation as a probe for MutS-adduct binding. The oligonucleotide derivatized chips were probed with different concentrations of MutS as above, but the buffer contained a 2-fold excess of either ADP, ATP, or a nonhydrolyzable analog of ATP (AMP-PNP or ATP-␥-S) with respect to MutS. As demonstrated in Fig. 5, the extent of binding of MutS to the M 1 G:T duplex is significantly reduced in the presence of ATP. In contrast, addition of ADP appeared to stabilize the interaction between MutS and the M 1 G:T duplex, denoted on the sensogram by a slower rate of dissociation of the MutS from the oligonucleotide surface.
MutS Binding to M 1 G and PdG Adducts by Gel Shift Analysis-We attempted to detect MutS binding to M 1 G and PdG duplexes by gel shift analysis. Radiolabeled 31-mer oligonucleotides containing mismatches or adducts were incubated with increasing amounts of purified MutS protein. MutS binding to the mismatch or adduct should retard the migration of the duplex through the gel. Fig. 6 depicts an average band shift experiment. Although MutS was incubated with duplexes in the presence of excess poly(dI⅐dC), a small amount of nonspecific binding was detected and was used as background for comparison with mismatch and adduct binding. Incubation of a G:T duplex with increasing amounts of MutS protein resulted in a strong gel shift, consistent with previous reports. The presence of MutS was required for detection of this shift. MutS incubation with a duplex containing M 1 G:T produced a shift above background levels. Significant binding was detected at concentrations as low as 200 nM. Incubation of MutS with M 1 G:C showed only a small amount of binding that was not far above background levels, and incubation with a PdG:C or PdG:T containing duplex did not yield any band shift. The results of these gel shift experiments are consistent with the results of the surface plasmon resonance experiments. DISCUSSION The present study provides in vivo and in vitro evidence that M 1 G and PdG are recognized by MutS and that they can be

TABLE III Kinetic and thermodynamic parameters of MutS binding to M 1 G-and
PdG-containing duplexes All data are the means Ϯ S.D. for three determinations of six different concentrations of MutS (0 -400 nM) using three different preparations of MutS. The resulting sensograms were subjected to global kinetic analysis using Biaeval 3.0 software, and association and dissociation data were simultaneously analyzed by a nonlinear curve fitting of the Langmuir binding isotherm. repaired by methyl-directed mismatch repair when the genome is not methylated on the adduct-containing strand. The interaction with MutS appears to be competitive with binding and repair of the adducts by nucleotide excision repair. Thus, when M 1 G and PdG are introduced on vectors methylated on both strands, they are not repaired by mismatch repair but are protected against nucleotide excision repair. Conversely, when mutS is deleted, the adducts are efficiently repaired by nucleotide excision repair resulting in a significant reduction in mutations relative to wild type. To probe for the binding of MutS to PdG and M 1 G in vitro, we employed a surface plasmon resonance assay and a gel mobility shift assay. The surface plasmon resonance assay provided kinetic and thermodynamic data for the protein-nucleic acid interaction. The dissociation constant for MutS binding to the G:T duplex was very similar to the value determined previously by gel shift analysis (32). Binding isotherms illustrated that the affinity for M 1 G:T was similar to that for G:T, but its affinity for M 1 G:C, PdG:C, and PdG:T was lower. Furthermore, the dissociation rates for MutS release from each of these adducts were approximately 10-fold faster than the dissociation rate from G:T. MutS binding to G:T and M 1 G:T duplexes also was detected by gel mobility shift analysis, but binding to M 1 G:C, PdG:T or PdG:C was not. We attribute this to the lower affinity of MutS for these duplexes and their higher dissociation rates as measured by surface plasmon resonance. These experiments illustrate the utility of surface plasmon resonance for detection of rapidly reversible protein-nucleic acid interactions. Although the kinetics of dissociation of MutS from M 1 G-and PdG-containing duplexes are rapid, the extent of binding is sufficient to protect M 1 G-or PdG-adducted genomes from nucleotide excision repair or to initiate methyl-directed mismatch repair as judged by the in vivo data. MutS binding to these genomes in vivo may be stabilized by the larger size of the genome relative to the 31-mer oligonucleotides used for the in vitro experiments or by binding to additional protein factors.
Although mismatch repair exists to remove errors made during DNA replication, the system does recognize and attempt to repair other DNA adducts. For example, MutS binds to O 6methylguanine and attempts to repair the strand opposite it (23). This leads to a futile cycle of removal and resynthesis that eventually causes cell death. A similar effect is observed with DNA containing G-G intrastrand cross-links induced by cisplatinum (35,36). The attempted repair of these lesions or protection from nucleotide excision repair provided by MutS binding is important to the therapeutic activity of methylating agents and cis-platinum. Tumor cells that have lost mismatch repair capability are more resistant to the cytotoxic action of both alkylating agents and cis-platinum (37,38). In fact, high levels of expression of the mismatch repair system appear to be an important determinant of the efficacy of cis-platinum against testicular tumors (39).
The affinity of MutS for M 1 G and PdG duplexes is comparable with that of MutS for the cis-platinum intrastrand G-G cross-link (23-107 nM versus 67 nM, respectively) (35). The latter value was measured by gel mobility shift analysis, which suggests that the dissociation rate of MutS from the cis-platinum adduct is lower than that from M 1 G or PdG. However, the affinity is strongly dependent on the base opposite the modified guanines. For example, no binding of human MutS to a cisplatinum adduct is observed when the base opposite each of the modified Gs is C or when the base opposite the 5Ј G is T (40). However, binding is optimal when the base opposite the 3Ј G is T. In our hands, MutS binds twice as strongly to M 1 G:T relative to M 1 G:C. This is further support for the observation that the base opposite the damaged DNA base is an important determinant of MutS binding.
An added feature of the binding of MutS to M 1 G is the chemical structure of the M 1 G as a function of the base opposite it. We have recently shown that M 1 G undergoes rapid and quantitative ring opening to N 2 -(3-oxopropenyl)-deoxyguanosine when present in a duplex DNA opposite C (Scheme 1) (41). Ring opening is reversed when the duplex is heatdenatured. No ring opening is detected when M 1 G is placed in duplexes opposite T. Thus, the reduced affinity of MutS for M 1 G:C duplexes may be a result of the presence of the ringopened form of M 1 G in this duplex.
This and previous studies establish that some overlap exists between the repair of small DNA adducts by global repair systems such as nucleotide excision repair and mismatch repair. Furthermore, binding of MutS to small lesions, even in the absence of subsequent repair, can protect these lesions from removal by nucleotide excision repair. M 1 G is among the most abundant exocyclic adducts found in the DNA of healthy humans (17). Its formation is linked to polyunsaturated fatty acid metabolism through lipid peroxidation or prostaglandin biosynthesis. Indeed, women consuming diets rich in polyunsaturated fatty acids exhibit a 10 -20-fold increase in the level of M 1 G in leukocyte DNA (34). It will be interesting to determine whether the levels of M 1 G in human tissue are sensitive to conditions that increase or decrease the levels of mismatch repair proteins (e.g. hereditary nonpolyposis colorectal cancer).