In search of a mutational hotspot

In vitro selection was used to define sequence contexts that significantly enhanced the mutagenic potential of 7,8-dihydro-8-oxoguanine (8-oxoG). Contexts that simultaneously reduced the efficiency of 8-oxoG cleavage by formamidopyrimidine DNA N-glycosylase and increased the efficiency of misincorporating A opposite the lesion by DNA polymerase were isolated from a pool of 48 random octanucleotide sequences. Kinetic analysis showed that the combined effects of poor repair and high miscoding resulted in 102to 103-fold increase in the mutagenic potential of 8-oxoG. Furthermore, the isolated sequence contexts correlated strongly with G3 T transversion hotspots in spontaneous mutational spectra reported for the Escherichia coli lacI and human p53 and factor IX genes. We present an example directly linking the interplay between DNA repair and replication to a ‘‘high risk sequence’’ for base substitution. The four base moieties in DNA are constant targets of damage by chemical processes such as depurination, deamination, and oxidation. DNA oxidation is caused by ionizing radiation, chemical agents (1), and, perhaps most importantly, normal metabolism (2). 8-Oxoguanine (8-oxoG) is one of the most prominent base oxidation products (3, 4) and has been implicated in mutagenesis, carcinogenesis, and aging (5). It has been shown to cause G 3 T transversions in vivo (6–9) and to be readily bypassed by DNA polymerases in vitro, coding for a nonmutagenic C and to a lesser extent a mutagenic A (10). The biological consequence of a DNA lesion is determined largely by the outcome of its interactions with the proteins responsible for maintaining the genetic integrity of the cell, namely the replicative and repair machineries. There are two prerequisites for a lesion to cause a mutation; repair enzymes must fail to remove it before the next round of replication takes place, and it must have coding properties different from those of its cognate base. To counter the mutagenic effect of 8-oxoG, Escherichia coli has evolved an elaborate repair system consisting of three genes, mutT, mutM, and mutY (7, 11). mutT codes for a sanitizing enzyme that converts 8-oxoGTP in the nucleotide pool to 8-oxoGMP, preventing the incorporation of 8-oxoG into DNA during replication (12). mutM or fpg codes for formamidopyrimidine DNA N-glycosylase (Fpg) (13), a protein with N-glycosylase and AP lyase activities (14, 15) that acts on both purine and pyrimidine oxidation products (16). However, action on 8-oxoG paired with C is believed to be the main function of Fpg in vivo. Fpg activity on 8-oxoG opposite A, a pair that results from replication errors, is greatly reduced (17) to prevent fixation of a G3T transversion. The third gene product in this repair system, MutY, is an N-glycosylase that removes A mismatched with 8-oxoG and to a lesser extent with G (18, 19). E. coli mutants deficient in any one of these three genes display a mutator phenotype and multiple mutations show synergistic effects (11). DNA damage-induced mutational spectra obtained from a wide range of species often display regions of highand low mutation frequency (hotand coldspots, respectively) which are lesionand gene-specific (20). These observations imply that a lesion’s mutagenic potential is partially determined by the sequence context in which it is presented. Sequence context effects on the efficiency of lesion recognition and removal have been reported for several DNA repair enzymes (21–23). Similarly, a strong influence of neighboring bases on both the efficiency of lesion bypass and the formation and extension of mispairs opposite several DNA lesions has been well documented both in vitro (24–27) and in vivo (28, 29). We, therefore, set out to define sequence contexts that modulate the mutagenicity of 8-oxoG by affecting the rate at which the lesion is repaired andyor the frequency at which the lesion miscodes for A during DNA synthesis.

The four base moieties in DNA are constant targets of damage by chemical processes such as depurination, deamination, and oxidation.DNA oxidation is caused by ionizing radiation, chemical agents (1), and, perhaps most importantly, normal metabolism (2).8-Oxoguanine (8-oxoG) is one of the most prominent base oxidation products (3,4) and has been implicated in mutagenesis, carcinogenesis, and aging (5).It has been shown to cause G 3 T transversions in vivo (6)(7)(8)(9) and to be readily bypassed by DNA polymerases in vitro, coding for a nonmutagenic C and to a lesser extent a mutagenic A (10).
The biological consequence of a DNA lesion is determined largely by the outcome of its interactions with the proteins responsible for maintaining the genetic integrity of the cell, namely the replicative and repair machineries.There are two prerequisites for a lesion to cause a mutation; repair enzymes must fail to remove it before the next round of replication takes place, and it must have coding properties different from those of its cognate base.To counter the mutagenic effect of 8-oxoG, Escherichia coli has evolved an elaborate repair system consisting of three genes, mutT, mutM, and mutY (7,11).mutT codes for a sanitizing enzyme that converts 8-oxoGTP in the nucleotide pool to 8-oxoGMP, preventing the incorporation of 8-oxoG into DNA during replication (12).mutM or fpg codes for formamidopyrimidine DNA N-glycosylase (Fpg) (13), a protein with N-glycosylase and AP lyase activities (14,15) that acts on both purine and pyrimidine oxidation products (16).However, action on 8-oxoG paired with C is believed to be the main function of Fpg in vivo.Fpg activity on 8-oxoG opposite A, a pair that results from replication errors, is greatly reduced (17) to prevent fixation of a G 3 T transversion.The third gene product in this repair system, MutY, is an N-glycosylase that removes A mismatched with 8-oxoG and to a lesser extent with G (18,19).E. coli mutants deficient in any one of these three genes display a mutator phenotype and multiple mutations show synergistic effects (11).
DNA damage-induced mutational spectra obtained from a wide range of species often display regions of high-and low mutation frequency (hot-and coldspots, respectively) which are lesion-and gene-specific (20).These observations imply that a lesion's mutagenic potential is partially determined by the sequence context in which it is presented.Sequence context effects on the efficiency of lesion recognition and removal have been reported for several DNA repair enzymes (21)(22)(23).Similarly, a strong influence of neighboring bases on both the efficiency of lesion bypass and the formation and extension of mispairs opposite several DNA lesions has been well documented both in vitro (24)(25)(26)(27) and in vivo (28,29).We, therefore, set out to define sequence contexts that modulate the mutagenicity of 8-oxoG by affecting the rate at which the lesion is repaired and͞or the frequency at which the lesion miscodes for A during DNA synthesis.
In Vitro Selection.A 52-nt oligo (Fig. 1) containing 8-oxoG surrounded by four randomized nucleotides on both the 3Ј and 5Ј sides was synthesized by standard phosphoramidite chemistry.The 52 mer (5 pmol) was primed with a 5Ј end-labeled 16 mer in 10 mM potassium phosphate, 16 mM Tris⅐HCl (pH 7.4), 65 mM NaCl, 1 mM 2-mercaptoethanol, 150 M dithiothreiotol, 1 mM EDTA, and 12% glycerol.T4 DNA polymerase holoenzyme complex then was assembled on the oligonucleotide in the absence of Mg 2ϩ as described (34).The reaction included 100 M dNTPs, 1 mM ATP (final concentrations), 140 pmol gp44͞62, and 100 pmol gp45 and 10 pmol of gp43.Reactions were initiated by addition of 10 mM MgCl 2 (final) and 2.5 nmol of heparin in a total volume of 65 l and were terminated by a 10-min incubation at 65°C.A molar excess of Fpg over the DNA then was added, and the reaction mix was incubated for 10 min at 37°C.The reaction was stopped by addition of formamide, and the 8-oxoG-containing strands that were refractory to Fpg cleavage were purified by 8 M urea PAGE.As depicted in Fig. 1, the lesion-containing strand was 52 nt long, which was readily separable from the complementary strand (46 nt) and Fpg cleavage product (20 mer).The selected molecules were PCR-amplified and subcloned into pBluescript KS ϩ by using a restriction endonuclease-free method (35), and the sequence of representative subclones was determined by using standard protocols.Data Analysis.A relational database computer program was used to analyze 76 in vitro selected samples.The sequences that represented the original randomized region were searched for all possible strings of two, three, or four nucleotides [both contiguous and interrupted by (n) nucleotides], and statistically under-and overrepresented sequences, as determined by a 2 test and P values Յ 0.05, were isolated.
The lesion-containing strand was 5Ј end-labeled with 32 P, annealed to its complement, adjusted to a range of molar concentrations between 0.5 and 105 nM, and subjected to cleavage by Fpg (final concentration 0.25 nM) at 37°C in 10 mM Tris⅐HCl (pH 7.5), 1 mM EDTA, and 50 mM NaCl.Following product analysis by denaturing PAGE and quantification by molecular imaging, the reactions' initial rates were determined and plotted as a function of substrate concentration.The resulting curve was fit to a hyperbolic function by using the program SIGMAPLOT, and the values of K m and k cat were derived.
T4 DNA Polymerase Steady-State Kinetics.The 54 mers containing 8-oxoG at positions 1 and 2 and the 24 mer containing the lesion at position 4 (see above) were used as templates to determine the steady-state kinetics of nucleotide incorporation and extension opposite 8-oxoG by gp43 exo Ϫ .Each template was primed with one of three primers, terminating 1 nt 3Ј to the lesion (relative to the template strand), terminating with C or terminating with A opposite the lesion.The primed-template concentration was adjusted to 2.5 nM (final during the reaction) in 25 mM Tris⅐acetate (pH 7.5), 150 mM potassium acetate, and 10 mM 2-mercaptoethanol, and T4 gp 43 exo Ϫ was added (16 nM protein apparent final concentration) in the absence of Mg 2ϩ .Reactions were started by addition of 10 mM magnesium acetate and a single dNTP (1 nM to 40 M final concentration, in ranges appropriate for individual reactions) and were carried out for 5-60 s at 22°C.Single nucleotide incorporation rates were determined as a function of dNTP concentration, and the steady-state rate constants were derived as described for Fpg kinetics.The reported gp43 exo Ϫ final concentration was based on a protein assay and not on titration of the polymerase active site.The concentration of active molecules was likely lower than template concentration because the reaction's initial velocity could be measured in a time scale of 5 to 60 s even at the highest dNTP concentration used [under polymerase excess conditions, DNA synthesis across 8-oxoG has been shown to proceed at a rate of 10s of nucleotides per second (36), making it impossible to measure an initial rate within the time scale used in our experiments].

RESULTS
synthesized chemically and used as template for DNA synthesis by T4 DNA polymerase holoenzyme.The resulting doublestranded molecules then were subjected to exhaustive cleavage by Fpg (see Methods), and those representing poor substrates, i.e., molecules refractory to cleavage, were isolated by denaturing PAGE.Less than 1% of the total lesion-containing strands remained uncleaved after the Fpg treatment as determined by phosphorimager analysis (data not shown).The selected molecules were amplified by PCR and subcloned (see Fig. 1 and Methods), and the sequences of 76 independent representatives as well as controls that were not subjected to this selection were determined.A highly statistically significant nonrandom sequence distribution was observed readily in the pool of selected molecules (Table 1) but not in the control samples (data not shown).In four of the eight randomized positions (third, second, and first nucleotides to the 5Ј and fourth nucleotide to the 3Ј side of the lesion), the selected molecules showed a significant (P Ͻ 0.05) bias for purines, and in the remaining four positions (fourth nucleotide to the 5Ј side and first, second, and third nucleotides to the 3Ј side of the lesion), there was a significant bias for adenine.
Sequence Analysis of Selected Molecules.The 76 representative samples were analyzed for the presence of consensus sequences.We used a relational database computer program (MICROSOFT FOXPRO) to determine the frequency of every di-, tri-, and tetranucleotide sequence in the sample pool and calculated whether the observed values deviated significantly from what would be expected to arise by chance.Sequence gaps were accommodated to avoid biases toward contiguous sequence contexts (e.g., not only the frequency of the trinucleotide XYZ was determined, but also the frequency of XNYZ, XNNYZ, XYNZ, XYNNZ, etc., in every possible permutation).Statistically over-and underrepresented independent sequences are reported in Table 2.Many additional di-and trinucleotide sequences were overrepresented in the sample pool but might not be independent of the sequences listed in Table 2.For example, there were 11 samples with the sequence 5Ј-G(N)A (the nucleotide within the parentheses represents the original position of the lesion), a highly overrepresented number ( 2 ϭ 8.2) in a total sample size of 76.However, there were 45 samples with the sequence 5Ј-N(N)A, suggesting that 5Ј-G(N)A was not selected independently of 5Ј-N(N)A [11 G(N)A͞45 N(N)A, 2 ϭ 0.01].Because our selection targeted molecules that were not cleaved after exhaustive treatment with Fpg (Fig. 1), sequences observed in significantly high numbers were taken to represent contexts that reduced the efficiency of Fpg action on 8-oxoG.By inference, sequences observed in significantly low numbers are taken to represent contexts that promoted efficient cleavage of 8-oxoG.
Consensus Sequences that Modulated the Action of Fpg on 8-OxoG.The most striking consensus in sequences promoting inefficient cleavage by Fpg was the dominance of A as the 3Ј nearest neighbor of the lesion (59% of all samples).Next, two additional sequences were observed in significantly high numbers, a string of four purines on the 5Ј side of the lesion and ARAR on the 3Ј side of the lesion.Although samples containing the consensus 5Ј-(N)ARAR are a subset of those containing 5Ј-(N)A, their observed frequency in the pool of the latter sequence (12͞45) was much higher than expected by random distribution ( 2 ϭ 30), suggesting that they were selected independently.On the other hand, 5Ј-(N)A was still highly overrepresented in samples not containing 5Ј-(N)ARAR (33͞64, 2 ϭ 18.1).It is therefore probable that the presence of A as the 3Ј nearest neighbor exerted a negative effect on 8-oxoG removal by Fpg, whereas the presence of 5Ј-(N)ARAR enhanced this effect.Similarly, the presence of four purines on the 5Ј side of the lesion appeared to enhance the effect of the 3Ј-A [nine samples carrying 5Ј-RRRR(N)A of 45 with 5Ј-(N)A, 2 ϭ 13.6].However, there were eight samples with the sequence 5Ј-RRRR(N) in 31 samples not containing a 3Ј-A ( 2 ϭ 19.0), suggesting that the presence of four purines on the 5Ј side of the lesion exerted some negative effect on lesion removal independently of the 3Ј nearest neighbor.Furthermore, of 17 samples containing the consensus 5Ј-RRRR(N) and 12 samples containing 5Ј-(N)ARAR, only two samples contained both, suggesting that the two sequences were selected independently of each other and that either consensus was sufficient to effect poor lesion removal by Fpg.
Three dinucleotide sequences were underrepresented in the pool of selected samples (Table 2).These contexts were present in significantly low numbers even when the observed low frequency of pyrimidines in the total sample pool (Table 1) was taken into account.For example, the sequence 5Ј-(N)YC was not only significantly underrepresented in 76 total samples ( 2 ϭ 7.6, Table 2) but also in 24 samples containing the sequence 5Ј-(N)Y ( 2 ϭ 4.17) and 11 samples with the sequence 5Ј-(N)NC ( 2 ϭ 3.68).We take this to indicate that the underrepresented sequences are contexts that promoted efficient cleavage of the lesion by Fpg, more so than generally pyrimidine-rich sequences.Consensus Sequences that Modulated the Frequency of Misincorporating A Opposite 8-OxoG by DNA Polymerase.A very interesting observation came from a small subset of samples (17%) in which T rather than G was recovered at the original position of the lesion.As alluded to earlier, these samples may define a consensus sequence that promoted simultaneously inefficient cleavage of 8-oxoG by Fpg and a high frequency of misincorporating A opposite the lesion by DNA polymerase.Such a sequence context constitutes an ideal putative G 3 T transversion hotspot.Although it is clear that the transversions observed in our sample pool were generated by Taq DNA polymerase during PCR amplification, contexts that had directed the misincorporation of A opposite the lesion during second strand synthesis by T4 DNA polymerase already would have been selected during the Fpg cleavage step of the assay [8-oxoG paired with A is a poor substrate of Fpg (17)].In addition, we were interested in identifying contexts where Taq polymerase misinserted A opposite the lesion because such contexts could have similar effects on all DNA polymerases.Of 13 samples in which T was observed at the original position of the lesion, 10 had A as the 3Ј nearest neighbor ( 2 ϭ 14), 9 had G as the 5Ј nearest neighbor ( 2 ϭ 10.2), and 7 had both ( 2 ϭ 47.1; Table 3).No significant consensus was observed at other positions in the randomized portion of the 52 mer substrate, suggesting that the sequence 5Ј-G(N)A was sufficient to effect a higher rate of miscoding for A by 8-oxoG.Conversely, in 11 samples in which G and A were observed as, respectively, the 5Ј and 3Ј nearest neighbors (regardless of the nucleotide recovered at the position of the lesion), 7 samples contained T at the position of the lesion, again a significantly overrepresented number (Table 3 reports a 2 of 6.6 for 7 observed͞11 total based on equal probability for all four nucleotides, which is an underestimate in view of the fact that G would be expected to predominate).Because of the limited pool size (76 samples), we cannot exclude the presence of longer contexts with higher miscoding potential.
Verification of the in Vitro Selection Assay.To verify the effectiveness of the selection scheme, we determined the kinetic parameters for action of Fpg on 54 nucleotide oligomers containing 8-oxoG in the contexts 5Ј-CC(8-oxoG)CC, 5Ј-AA(8-oxoG)AA, and 5Ј-TG(8-oxoG)AC (Methods and Table 4).The first sequence was predicted to promote efficient lesion removal, whereas the latter two were predicted to promote the opposite effect (Table 2).A 24 mer containing the lesion in the context 5Ј-TG(8-oxoG)AT also was used.As indicated by the k cat ͞K m values (which provide an estimate of a reaction's efficiency), Fpg removed 8-oxoG from the 5Ј-CC(8-oxoG)CC context 9-fold more efficiently than from the 5Ј-AA(8-oxoG)AA context, in full agreement with our predictions.More interesting, the putative hotspot context 5Ј-G(8-oxoG)A lowered the efficiency of Fpg action by 33-fold compared with the CC(8-oxoG)CC context.The hotspot contexts promoted inefficient removal of the lesion predominantly by raising the K m of the reaction.
Similarly, we verified the elevated miscoding potential of the lesion in the 5Ј-G(8-oxoG)A contexts by measuring the kinetic parameters for T4 DNA polymerase synthesis past 8-oxoG in three of the four contexts described in Table 4. Parameters for incorporation of C and misincorporation of A opposite 8-oxoG as well as extension of the two resulting pairs were determined.(Values for a 3Ј 3 5Ј exonuclease-free polymerase are shown; similar observations were made with the proofreadingproficient polymerase).Confirming predictions made in Table 3, A was incorporated opposite the lesion 33-fold more efficiently in the context 5Ј-G(8-oxoG)A than the context 5Ј-CC(8-oxoG)CC and Ϸ4-fold more than in the context 5Ј-AA(8-oxoG)AA.It should be noted that the efficiency of incorporating C opposite the lesion was also higher in the 5Ј-G(8-oxoG)A context (14-and 4-fold, respectively) than 5Ј-CC(8-oxoG)CC and 5Ј-AA(8-oxoG)AA.As for the Fpg For translesion synthesis to be successful, efficient extension of the nucleotide incorporated opposite the lesion is required.Earlier work has shown that, for several lesions, including 8-oxoG and several DNA polymerases, incorporation of a mutagenic nucleotide is less efficient than the nonmutagenic nucleotide (A and C, respectively for 8-oxoG).But curiously, extension of the mutagenic pair is more efficient than the nonmutagenic one (10,37).Data presented in Table 4 confirm these observations.In all three sequence contexts, extension of the 8-oxoG:A pair was 4-to 50-fold more efficient than extension of the 8-oxoG:C pair.Of interest, although the extension of the A mispair was 50-fold more efficient than formation of the mispair in the 5Ј-CC(8-oxoG)CC context (V max ͞K m values of 4.5 and 0.07, respectively), the two steps were equally efficient in the context 5Ј-G(8-oxoG)A (values of 2.3 and 2.1).
When the kinetics of the combined incorporation͞extension reactions are compared, the 5Ј-G(8-oxoG)A context promoted nonmutagenic translesion synthesis 80-fold more efficiently than the 5Ј-CC(8-oxoG)CC context and promoted mutagenic translesion synthesis 15-fold more efficiently.By further combining the effects on Fpg and DNA polymerase, the sequence context 5Ј-G(8-oxoG)A was shown to have the potential of generating G 3 T transversions 450-fold more efficiently than the context 5Ј-CC(8-oxoG)CC.
Correlation of the in Vitro Derived Consensus Sequences with in Vivo Mutational Spectra.The authenticity of our in vitro-selected putative hot-and coldspots could be further tested by searching in vivo mutational spectra.We searched a database of mutations scored in the E. coli lacI gene (38), which is sufficiently large in total number of mutations and represents a relatively large sequence diversity, for all spontaneous or free radical-induced G 3 T transversions and determined the frequency of the 16 possible sequence contexts 5Ј N(mutation)N (reference to G 3 T transversions throughout this report is restricted to mutations generated spontaneously or by free radical-producing agents).As can be seen in Table 5, a striking 100͞147 total mutations had A as the 3Ј nearest neighbor ( 2 ϭ 109), and 42 of the latter had G as the 5Ј nearest neighbor ( 2 ϭ 117).Although it was clear that the sequence 5Ј-G(G 3 T)A was overrepresented in the context 5Ј-N(G 3 T)A (42͞100, 2 ϭ 11.6), the presence of 3ЈA was significant even at sites where the 5Ј nearest neighbor was not G (58͞105, 2 ϭ 38.4).Because these mutations were scored based on a phenotypic change, the data can be skewed by the presence of a large number of mutations at one site.We therefore searched for the frequency of G 3 T transversions as a function of sequence context without considering the number of mutations at each site.Of 37 different sites where G 3 Ts were observed, 19 had A as the 3Ј nearest neighbor and 7 fell in the sequence 5Ј-G(G 3 T)A, both highly overrepresented frequencies.These observations are in striking agreement with the predictions of our in vitro selection.Furthermore, only 11 mutations at 2 sites were scored where C was the nearest neighbor on both sides of the mutations, and 2 mutations at one site were surrounded by Ts (representing random distribution in the first case and significant underrepresentation in the second).Because the lacI gene encompasses a limited sequence diversity, it was possible that the observed biases toward G 3 T transversions in the context 5Ј-G(G)A simply reflected a higher incidence of this sequence in the gene.The lower portion of Table 5 rules out this possibility.There are 362 Gs in lacI, and only 68 have A as the 3Ј nearest neighbor, significantly fewer than expected by random distribution ( 2 ϭ 5.6).Similarly, 19 Gs fall in the context 5Ј-G(G)A, reflecting random distribution.On the other hand, 38 Gs are in the context 5Ј C(G)C, significantly more than expected by chance ( 2 ϭ 10.45).
Although we used prokaryotic proteins in our selection, we were interested to see whether this putative mutational hotspot is significantly overrepresented in spectra of human genes.In the tumor suppresser p53 gene, G 3 T transversions are found within the context 5Ј-G(G)A in 60͞672 total mutations at 12͞104 sites (38) (both numbers significantly overrepresented, 2 ϭ 7.71, 4.65, respectively).Similarly, G 3 T transversions in the human factor IX gene (antihemophilic factor B) have a significantly overrepresented 10͞71 mutations in the putative hotspot context scored at 8͞47 mutable sites ( 2 ϭ 6.97, 8.72, respectively).

DISCUSSION
Based on the biological observation that mutations cluster in hot and coldspots, we initiated this project on the premise that sequence context strongly modulates the mutagenic outcome of an important base oxidation product, 8-oxoG.We went on to isolate such sequences in vitro and to verify their competence by kinetic analysis.Finally, we were able to go full circle and find very strong correlation between our data and available in vivo mutational spectra, demonstrating the effectiveness of the in vitro selection technique.
As for most other DNA N-glycosylases, Fpg action is limited to removal of lesions in double-stranded DNA (39).The randomized portion in the oligonucleotide substrate used in the selection assay necessitated the use of DNA polymerase to generate a perfect complement for the chemically synthesized single strand containing 8-oxoG (Fig. 1).An interesting and advantageous outcome of this step was the generation of The 5Ј-G(8-oxoG)A putative hotspot not only supported higher efficiency of misincorporating A opposite the lesion by T4 DNA polymerase relative to other contexts (see Results and Table 4) but also higher efficiency of incorporating C, suggesting that this context had the general effect of minimizing lesion-induced distortion in the DNA structure.In addition, T4 DNA polymerase incorporated A opposite 8-oxoG and extended the resulting primer terminus with almost identical efficiency (Table 4) in this sequence context, simulating natural DNA synthesis (the rates of successive polymerization steps remain relatively constant during DNA synthesis on undamaged templates).In contrast, large variations in the efficiency of these two steps were observed during the incorporation of C opposite the lesion in the 5Ј-G(8-oxoG)A context and the incorporation of A opposite the lesion in the 5Ј-CC(8-oxoG)CC context.The latter observation suggested that the putative hotspot reduced lesion-induced distortions in DNA structure more efficiently when A was paired with the lesion.
The strong correlation between predictions made by the in vitro selection and bacterial in vivo spectra on the one hand and human in vivo spectra on the other suggested that the sequence context effect was mediated by altering lesion presentation independently of bound proteins (repair glycosylases or DNA polymerases).On the other hand, the stronger correlation between the in vitro data and the bacterial mutational spectrum (relative to the human spectra) can be interpreted to reflect differences in the prokaryotic and eukaryotic enzymes involved in processing the lesion.Although it is probable that both notions are correct, further work is needed to resolve the relative effect of this sequence context on different polymerases and glycosylases.In addition, correlation with in vivo data must take into account the influence of the putative hotspot sequence context on MutY.
We are currently in the process of further verifying our results by introducing plasmids carrying 8-oxoG in the putative hot-and coldspots into bacterial and mammalian cells and measuring the rate of G 3 T transversions in the different contexts.We are also using purine analogs as nearest neighbors of 8-oxoG to define the exocyclic groups involved in effecting the higher mutagenic potential of the lesion.Finally, sequence context effects on modulating the mutagenic potential of other oxidative lesions such as 5-hydroxycytosine will be addressed using a similar in vitro selection scheme.

Table 1 .
Observed nucleotide frequency at randomized positions in 76 independent samples of molecules cleaved inefficiently by Fpg 2 and P values are for deviation from a random distribution of 0.25 for each of the four nucleotides. *

Table 2 .
In vitro-selected sequence contexts for inefficient and efficient cleavage of 8-oxoG by Fpg Listed sequences are for the strand that originally contained the lesion (N).Following convention, R ϭ purine, Y ϭ pyrimidine, and K ϭ G or T.

Table 3 .
Statistical analysis of the appearance of the sequence 5ЈG(T)A in the pool of in vitro-selected samples Reported sequences are for the template strand and the original position of the lesion is indicated by parentheses.

Table 4 .
Kinetics of Fpg and T4 DNA polymerase processing of 8-oxoG within putative mutational hot-and coldspots

Table 5 .
Nearest neighbors frequencies at sites of spontaneous, ionizing radiation-, or oxidation-induced G 3 T transversions in the lacI gene of E. coliNearest neighbors distribution at G 3 T transversion sites in E. coli lacI