INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Carcinogenesis is generally considered to be the result of a multistep process involving several genetic changes. During studies on carcinogenesis in mammalian systems, and in the process of trying to determine the molecular basis of neoplasia, members of the ras family of oncogenes (N-, K- and Ha-ras) have been implicated (1, 3-5, 7, 19). Mutations in these genes have been observed in various naturally occurring human tumors as well as in carcinogen-induced animal tumors (1, 3-5, 7, 19), and the transforming ras genes have been shown to be mutant alleles of cellular ras genes (1, 3-5, 7, 19). It also has been shown that the ability of mammalian ras genes to induce transformation is conferred by single point mutations within their coding regions. These mutations have been localized most frequently to codons 12, 13, and 61 (1, 3-5, 7, 19). In the three mammalian ras genes, the sequences in codons 12 and 13 contain GG doublets. Within the GG doublet in codon 12, the 3' G was the residue that was most frequently implicated in the mutational activation of the ras proto-oncogenes (1, 3-5, 7, 19).

Such sequence specificity of mutational activation of ras oncogenes has important implications for human cancer, in that the exact same pattern of mutational activation has been observed in a variety of solid human tumors. Mutant ras oncogenes have been identified in carcinomas of the pancreas, colon, lung, thyroid, skin, bladder, and kidney (3-5, 7). They have also been detected in melanomas and certain forms of myeloid leukemia (4). In these tumors, a significant percentage of all activating ras mutations occurs in the GG doublets of codons 12 and 13 of the ras genes, predominantly by GCAT transitions at the 3' G of these doublets (1, 5, 7, 10). Overall, GCAT transitions at the 3' G of the GG doublet in codon 12 of ras oncogenes represent the single most frequent ras mutation observed in solid human tumors, accounting for about 30% of all the observed mutations (15). In rare instances, GCAT transitions have been found to occur at the 5' G (rather than at the 3' G) of the GG doublet of codon 12 in solid human tumors (7, 35). Mutant ras genes have also been transfected into cells in culture, and it was seen that the mutant gene altered via transition at the 5' G of codon 12 had transforming ability approximately equal to that of the mutant gene altered at the 3' G of codon 12 (7, 27). Thus, in solid human tumors with activated ras oncogenes, the high frequency of GCAT transitions observed at the 3' G, when compared to those at the 5' G, of codon 12 may reflect differences in mutation frequency at those sites rather than a bias imposed by phenotypic differences (in terms of transforming potential) resulting from mutations at the two sites.

Our previous studies on mutations induced by thymidine (dT) in mammalian cells appear to be relevant to the sequence-specific nature of ras gene activation in human tumors, since we found that the sequence specificity of dT mutagenesis corresponds exactly to the sequence specificity of mutations in activated ras oncogenes in solid human tumors. The first step in mutagenesis by dT (or by the dT analog 5-bromodeoxyuridine) apparently is the inhibition of the ribonucleotide reductase-catalyzed reduction of CDP to dCDP, leading to an increase in the intracellular dTTP/dCTP ratio. Subsequently, dTTP, now the nucleotide in excess relative to dCTP, mispairs with template guanine, leading to a GCAT transition at the next round of replication (2, 14, 22). We have shown that such mutagenesis by dT (or 5-bromodeoxyuridine) exhibits strong sequence specificity in that it occurs preferentially at the 3' G of runs of two or more adjacent G residues, resulting in GCAT transitions at this position in the multiple guanine run (11, 16, 17, 33). In sequences with GG doublets, dT-induced mutations were found to occur at least 25 times more frequently at the 3' G residue of the GG doublet than at the 5' G residue of the doublet (17). This sequence specificity for dT mutagenesis is identical to that observed for activating mutations in ras proto-oncogenes in solid human tumors, in that such mutations also were found to occur preferentially at the 3' G of the GG doublet in codon 12 of these genes.

The frequency with which a mutation occurs in a site-specific manner could represent a balance between the frequency with which an incorrect base is misincorporated at a particular site and the efficiency with which the mismatch at that site is eliminated by the repair mechanism of the cell. Recently, the role that repair might play in the pattern of UV-induced mutations in the p53 gene was studied, and the results suggested that preferential repair at certain sites might be involved in the sequence specificity of UV-induced mutations in the p53 gene (30). In contrast, the results presented below suggest that sequence-directed base mispairing can account for the preferential occurrence of mutations at the 3' G residue in the GG doublets of ras codon 12. This conclusion is based upon the analysis of the tendency for base mispairing at the 3' G versus the 5' G of the GG doublet in codon 12 of the three human ras genes, utilizing a modified in vitro DNA synthesis assay (6, 21). In order to permit an unequivocal analysis of sequence-directed base mispairing, as distinct from the effect of DNA sequence on mismatch repair, these experiments were carried out using exonuclease-free Klenow polymerase. Because of the use of this enzyme, we can conclude that the results described in this study are solely a reflection of the ability of certain sequences in the DNA molecule (namely, the GG doublet in codon 12 of human ras oncogenes) to direct base misincorporation, without the complication of any repair activity. Klenow polymerase has been used routinely by various other laboratories for studies on the sequence specificity of chemical mutagenesis (13, 24, 28, 29, 32).

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

PCR amplification and cloning of ras gene sequences. Regions (100 to 125 bp) of DNA around codon 12 of the human N-, Ki-, and Ha-ras genes were amplified by PCR from human genomic DNA containing wild-type ras sequences (Clontech Laboratories, Inc., Palo Alto, Calif.). Amplifications were carried out with a Perkin-Elmer Cetus DNA thermal cycler as previously described (26), with minor modifications. Briefly, amplification reactions contained 1 µg of template human genomic DNA; 500 ng of the purified oligonucleotide primer pairs specific for amplification of each ras gene sequence; 200 µM each of dATP, dGTP, dTTP, and dCTP; and 1 U of Taq polymerase obtained from Promega (Madison, Wis.). Ha- and N-ras-specific primers that amplify sequences around codon 12 were obtained from Clontech, while Ki-ras-specific primers were synthesized in our laboratory with an Applied Biosystems DNA synthesizer. Amplification of human genomic DNA was carried out by denaturation at 94°C for 10 min, followed by 35 cycles of 94°C for 1 min, 55°C for 1 min, and 72°C for 1 min. The PCR products were purified on 1% agarose gels and cloned into the pT7 Blue T-Vector from Novagen (Madison, Wis.) for use in the misincorporation experiments described below. Recombinant clones were identified, and correctly inserted ras gene sequences were identified by DNA sequencing with the Promega f-Mol DNA Sequencing System and an end-labeled primer complementary to a region upstream of the cloning site in the pT7 Blue T-Vector. The sequencing reactions were terminated with 10 µl of stop buffer and boiled to denature the DNA; the products were then electrophoresed on 8% polyacrylamide gels (18). The gels were dried and exposed to X-ray films for 18 h at 80°C.

In vitro misincorporation assay. A modified in vitro DNA synthesis assay was utilized to quantitate the extent of incorporation of an incorrect base (i.e., dT) opposite either the 3' G or the 5' G of the GG doublets in codon 12 of the human ras oncogenes. To provide templates for the misincorporation assays, DNA was isolated from pT7 Blue T-Vector recombinants, containing the 100- to 125-bp regions surrounding the GG doublet in codon 12 of the N-, Ki-, and Ha-ras genes. The DNA was purified from individual clones in the form of single-stranded closed circular molecules, as described by the manufacturer (Novagen).

Competition experiments. To prepare templates for the competition experiments, 43-base oligomers corresponding to sequences around the GG doublet in codon 12 of the Ki- and N-ras genes were synthesized by using an Applied Biosystems DNA synthesizer. As described above, the templates were annealed in separate reactions to their respective 3' G and 5' G misincorporation primers. Then 5 µl of the annealed DNA-primer mixture was added into tubes containing 0.33 µM [-³²P]dCTP and increasing concentrations of cold dTTP as competitor. This concentration of dCTP was used because it had been found to give maximum incorporation in the control reactions for the misincorporation assays. To each tube was added 0.6 µl of a 1:1 mixture of exonuclease-free Klenow polymerase (5 U/µl) and iPPase (5 U/µl). The reactions were incubated at 37°C for 30 min and then terminated. The total radioactivity in the extended product when [-³²P]dCTP alone was provided (representing correct incorporation) was assigned a value of 100%. Competition was then calculated as the measured percent decrease in radioactivity subtracted from 100%, when increasing concentrations of cold dTTP were added to the same concentration of labeled dCTP. The values for the 3' and 5' G residues were corrected as described in the Results section.

Mutual effects of adjacent G residues on dT misincorporation. For these experiments, three 43-base templates were synthesized based on the sequence around codon 12 of the human N-ras oncogene. One of them was identical to the wild-type N-ras sequence (5'-AGGT-3'). The other two differed from the wild-type sequence by a single base, through the substitution of either the 5' G of the GG doublet in codon 12 with a T (5'-ATGT-3') or the 3' G of the GG doublet with a T (5'-AGTT-3'). These three templates were annealed to their complementary primers in separate reactions. Misincorporation experiments were then carried out as previously described, providing dTTP as the only nucleotide. Control experiments, providing dCTP as the only nucleotide, were carried out simultaneously.

Effects of upstream flanking base on dT misincorporation opposite the 5' G of the GG doublet. To determine the effect of the upstream flanking base on dT misincorporation opposite the 5' G of the GG doublet in codon 12, three related 43-base oligonucleotides based on the sequences around codon 12 of the human Ki-ras gene were synthesized and used as templates. These templates were identical to each other except that they had either an A, a T, or a C immediately upstream of the 5' G. Of these three templates, the one with T immediately upstream of the 5' G of the GG doublet was identical to the wild-type Ki-ras gene (5'-TGGT-3'). To these three templates, the Ki-ras-specific 5' G misincorporation primers were annealed in separate reactions, and misincorporation assays were carried out to determine the effect of the upstream flanking base on the extent of formation of a G:T mispair at the 5' position of the GG doublet. Misincorporation assays were carried out as previously described, providing dTTP as the only nucleotide. Control experiments, providing dCTP as the only nucleotide, were carried out simultaneously to determine the extent of correct incorporation of dC opposite these 5' G residues.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

dT misincorporation in GG doublets in codon 12 of human ras genes. In vitro DNA synthesis assays were carried out to determine the efficiency of dT misincorporation opposite the 3' G versus that at the 5' G of GG doublets in different sequences representing the human ras genes. Single-stranded DNA molecules containing ras sequences around codon 12 of the N-, Ki-, and Ha-ras genes were used as templates in these misincorporation assays. The sequences around the codon 12 GG doublet in the N-, Ki-, and Ha-ras genes are shown in Table 1, along with the appropriate misincorporation primers for the three ras genes. A typical misincorporation assay with N-ras is shown in Fig. 1, in which dTTP was the only nucleotide triphosphate provided and which covers a 1,000-fold range in dTTP concentrations. It can be seen that both the 3' G and the 5' G in the codon 12 GG doublet were able to mispair with dT, leading to the formation of 24-mer and 25-mer extension products, respectively. (The 25-mer extension product is formed when analyzing dT misincorporation opposite the 5' G; this is due to the fact that the base upstream of the codon 12 GG doublet of N-ras is an A. Thus, once dT has mispaired with the 5' G leading to the formation of a 24-mer product, this molecule is immediately extended by one more base because of the subsequent, correct incorporation of dT opposite the upstream A, leading to the formation of the 25-mer. In contrast, the 24-mer extension product resulting from the misincorporation of dT opposite the 3' G does not significantly undergo further extension by subsequent mispairing of dT opposite the 5' G.) Over a range of dTTP concentrations from 2.5 to 250 µM, it was seen that misincorporation of dT was much higher opposite the 3' G than opposite the 5' G. The results of such assays were quantitated in terms of the percent extension (misincorporation) at each dTTP concentration, and the results are presented in Fig. 2A. It can be seen that there was a dramatic difference in the mispairing efficiency of dT opposite the 3' G versus that at the 5' G in the codon 12 GG doublet over the range of dTTP concentrations tested and that a much higher concentration of dTTP was required to generate a given level of misincorporation opposite the 5' G than opposite the 3' G. For example, for misincorporation opposite the 3' G in codon 12 of N-ras, approximately 15% of the primer was converted to the mispaired extension product at a dTTP concentration below 10 µM. In contrast, the same level of misincorporation opposite the 5' G was not attained at dTTP concentrations even as high as 250 µM. Thus, more than 25 times the concentration of dTTP was required to generate the same level of misincorporation opposite the 5' G compared to that opposite the 3' G of the GG doublet in codon 12 of N-ras. It can also be seen that the percentage of primer extended by mispairing was five to eight times higher for the 3' G than for the 5' G at all of the dTTP concentrations between 2.5 and 250 µM.

View larger version (12K): [in this window] [in a new window]   FIG. 1.   dT misincorporation in the GG doublet in codon 12 of N-ras. 32P-labeled 23-mer primers were extended to a 25-mer extension product through the misincorporation of dT opposite the 5' G of the GG doublet, with a further correct incorporation opposite the A upstream of the GG doublet (upper panel) or extended to a 24-mer extension product through the misincorporation of dT opposite the 3' G of the GG doublet, with no further extension (lower panel). The extension reactions were visualized with a blot analyzer as described in Materials and Methods.

View larger version (15K): [in this window] [in a new window]   FIG. 2.   Misincorporation assays for GG doublets in ras oncogenes. dTTP misincorporation was determined opposite the 3' G () and the 5' G () of the GG doublet in codon 12 of the human N-ras gene (A), the human Ki-ras gene (B), and the human Ha-ras gene (C). Extension of the appropriate 23-base misincorporation primers was determined as described in Materials and Methods, providing dTTP (or dCTP) as the only deoxynucleotide triphosphate.  and , dC incorporation opposite the 3' G and the 5' G, respectively. The underlined base represents the site of misincorporation, at either the 3' G or the 5' G residue of the GG doublet. Each graph represents the average of three separate experiments.

3

4

Competition experiments. The misincorporation experiments described above were carried out with only dT (the incorrect base) being provided. In these experiments, a major difference was observed in misincorporation at the 3' versus the 5' G residues of the GG doublets in codon 12 of the ras genes. As only the incorrect nucleotide was provided in these experiments, it was necessary to determine whether this 3'-G/5'-G differential would still exist if the correct nucleotide (dCTP) was provided along with the incorrect nucleotide (dTTP). Therefore, competition experiments were carried out to determine whether the previously observed difference in the misincorporation of dT opposite the 3' G and the 5' G of the GG doublet in codon 12 of the human N- and Ki-ras genes was maintained when the misincorporation assays were carried out in the presence of the correct nucleotide, dCTP.

View larger version (10K): [in this window] [in a new window]   FIG. 3.   Competition experiments. The incorporation of labeled dCTP in the presence of increasing concentrations of competing cold dTTP was determined opposite the 3' G () and 5' G () of the GG doublet in codon 12 of the human N-ras gene. Competition experiments were carried out with N-ras-specific misincorporation primers as described in Materials and Methods. The underlined base represents the site of competition. The graph represents the average of two separate experiments.

3

4

Mutual effects of adjacent G residues on dT misincorporation. In the previous misincorporation and competition experiments, dT had been shown consistently to misincorporate with higher efficiency opposite the 3' G than opposite the 5' G of the GG doublet in codon 12 of the human ras genes. A possible explanation for this difference could be that either one or both G residues in the GG doublet might have an effect on dT misincorporation opposite the other. For example, the 5' G residue of the GG doublet might, in some manner, have a stimulatory effect on dT misincorporation opposite the 3' G residue of the doublet or else the 3' G residue of the doublet might have an inhibitory effect on dT misincorporation opposite the 5' G. To determine whether there were such effects of one G residue on dT misincorporation opposite the other G residue in the doublet, misincorporation experiments were carried out involving modifications of the N-ras sequence 5'-AGGT-3'. As seen in Fig. 4, when the 5' G of the GG doublet was replaced by a T residue, there was a marked decrease in dT misincorporation opposite the adjacent G residue. For example, for misincorporation opposite the 3' G residue of the GG doublet, approximately 35% of the primer was converted to the mispaired extension product at a dTTP concentration of 25 µM. In contrast, for misincorporation opposite the same G residue when the 5' G of the GG doublet was replaced by a T residue in otherwise-identical templates, a concentration of 50 µM dTTP was required to attain approximately the same extent of misincorporation. Thus, approximately twofold-less dTTP was required to generate a given level of misincorporation when a G residue was flanked on its 5' site by another G residue (i.e., in a GG doublet) than when it was flanked by an upstream T residue. Similarly, at a given dTTP concentration of 25 µM, there was approximately 30% greater misincorporation opposite the 3' G residue of the GG doublet than opposite the G residue present at the same position but flanked by an upstream T residue. Control experiments were carried out for correct base incorporation in which dCTP was the only nucleotide provided. The results of these experiments showed that correct base incorporation opposite a G residue was not affected by the change of a G to a T residue for its upstream neighbor. These results indicate that a G residue flanked on its 5' side by another G residue does not have an inherently greater ability to undergo base pairing than a G flanked on its 5' side by a T residue. Thus, the results presented in Fig. 4 indicate that the 5' G residue of the GG doublet has a stimulatory effect (relative to that of a T residue) on misincorporation opposite the 3' G residue of the doublet.

View larger version (14K): [in this window] [in a new window]   FIG. 4.   Mutual effects of adjacent G residues on dT misincorporation. Extension of the appropriate misincorporation primers was determined as described in Materials and Methods. Template sequences were based upon modification of the N-ras sequence. The underlined base represents the site of dT misincorporation. Symbols: , dT misincorporation opposite the 3' G in the sequence -AGGT-; , dT misincorporation opposite the 5' G in the sequence -AGGT-; , dT misincorporation opposite a G in the sequence -ATGT-, where the former 5' G is replaced by a T; , dT misincorporation opposite a G in the sequence -AGTT-, where the former 3' G is replaced by a T. The symbols , , , and  represent the corresponding dC controls. The graph represents the average of two separate experiments.

Effects of the upstream flanking base on dT misincorporation opposite the 5' G of the GG doublet. As described above, the relative difference in the extent of mispairing of dT opposite the 3' G versus that opposite the 5' G of the GG doublet in codon 12 varied among the three ras genes, being smallest for Ha-ras and largest for N-ras genes. The decreased differential in the case of Ha-ras appeared to arise primarily as the result of higher dT misincorporation opposite the 5' G in codon 12 of Ha-ras than at the 5' G residue in N-ras. The base immediately downstream of the 5' G in all three ras genes was the same, namely, the 3' G, but the upstream base was different in each of them: C in Ha-ras, T in Ki-ras, and A in N-ras. It was possible, therefore, that these different upstream bases might have an effect on the extent of dT misincorporation opposite the 5' G of the GG doublet in codon 12 of the three ras genes. To determine whether there was such an effect of the upstream flanking base on dT misincorporation opposite the 5' G of the GG doublet, misincorporation experiments were carried out involving modification of the Ki-ras sequence (5'-TGGT-3'). As shown in Fig. 5, the presence of different upstream flanking bases led to appreciable differences in the extent of dT misincorporation opposite the 5' G of the GG doublet in otherwise-identical templates. It can be seen that misincorporation of dT opposite the 5' G was highest when C was the upstream flanking base, lower when T was the upstream flanking base, and lowest when A was the upstream flanking base. For example, approximately 20% of the primer was converted to the mispaired extension products at a dTTP concentration of 50 µM when C was the flanking base immediately upstream of the 5' G. In contrast, 100 µM dTTP was required to attain approximately the same extent of misincorporation when the upstream flanking base was A. Thus, approximately twofold-less dTTP was required to generate a given level of misincorporation opposite the 5' G when C was the upstream flanking base than when A was the upstream flanking base. Also, for all dTTP concentrations between 10 and 100 µM, there was approximately 50 to 100% greater misincorporation opposite the 5' G of the GG doublet when C was the upstream flanking base than when A was the upstream flanking base.

View larger version (15K): [in this window] [in a new window]   FIG. 5.   Effect of upstream flanking bases on dT misincorporation opposite the 5' G of the GG doublet. Extension of the appropriate misincorporation primers was determined as described in Materials and Methods. Template sequences were based upon modification of the Ki-ras sequence. The underlined base represents the site of misincorporation. Symbols: , , and , dT misincorporation opposite the 5' G when C, T, or A, respectively, are the bases immediately upstream of the 5' G;  and , the dC incorporation controls when C or A are the bases immediately upstream of the 5' G. The graph represents the average of two separate experiments.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

In cultured mammalian cells, high levels of exogenously supplied dT dramatically increase the intracellular ratio of dTTP to dCTP available for DNA synthesis (2, 11, 14, 20). Subsequently, dTTP, the nucleotide in excess, mispairs with guanine residues in replicating DNA molecules, resulting in GCAT transitions (11, 16, 17, 33). The misincorporation studies presented here are based upon previous work done in our laboratory dealing with the sequence-specific mode of dT mutagenesis in mammalian cells (17). Those studies had shown that when cultured mouse cells carrying a single, chromosomally integrated copy of the bacterial gpt gene were subjected to high concentrations of dTTP, GCAT transitions occurred preferentially at the 3' G residue of GG doublets in the gpt gene (11, 33). The mutations that were observed in these experiments thus occurred under conditions of an imbalance in two specific nucleotide pools, namely, a high-dTTP pool and a low-dCTP pool. This makes it unlikely that these mutations arose as the result of other mutational mechanisms such as preferential alkylation of the C residue at this site but on the other DNA strand, deamination or repair, since the conditions of nucleotide pool imbalance that were employed in our studies are not known to involve such mechanisms in causing mutations. Preliminary analyses of base mispairing in vitro that utilized a DNA synthesis system comparable to that employed in this study (but with phage M13 DNA sequences that resembled the gpt gene sequences in the earlier dT mutagenesis experiments) suggested that dT misincorporation occurred more efficiently opposite the 3' G than opposite the 5' G of GG doublets in these M13 sequences (17a). These results raised the possibility that the observed sequence specificity of dT mutagenesis in mammalian cells could be due to the preferential formation of mismatched G:T base pairs at the 3' G residue of GG doublets. This would be of special interest in cancer research, as the sequence specificity of dT-induced mutations in mammalian cells corresponds exactly to the most common mode of activation of human ras proto-oncogenes in that the most frequently observed mutation in these genes in various solid human tumors is a single GCAT transition that occurs preferentially at the 3' G of the GG doublet in codon 12 of the ras genes. Therefore, the experiments presented here were designed to determine whether there is a preference for misincorporation of dT opposite the 3' G compared to that opposite the 5' G of GG doublets in codon 12 of the human ras genes and whether such a preference could help to explain the preponderance in human tumors of activating mutations at the 3' G residue of the GG doublets in codon 12 of these ras oncogenes.

In the present study, we utilized an in vitro misincorporation assay in a repair-free environment in order to examine the relative susceptibilities to dT mispairing of the 3' and 5' G residues of the GG doublet in codon 12 of the human N-, Ki-, and Ha-ras oncogenes. The data presented in Fig. 2 clearly demonstrate that there is a strong preference for mispairing of dT opposite the 3' G compared to that opposite the 5' G of the GG doublets in codon 12 of each of these three human oncogenes. Control assays showed no difference in the correct incorporation of dC opposite the 3' and 5' G residues of the GG doublets, indicating that the preferential misincorporation of dT opposite the 3' G was not due to an inherently increased ability of this residue to undergo base pairing. Furthermore, the greater preference for mispairing of dT opposite the 3' G versus the 5' G was exhibited even when the correct nucleotide, dCTP, was provided along with the incorrect nucleotide, dTTP, as seen in the competition experiments (Fig. 3). The competition experiments are of particular interest because the conditions in these experiments are more representative of the in vivo situation, where dT misincorporation would always have to compete with correct dC incorporation. If the preferential mispairing of dT opposite the 3' G residue in GG doublets occurred in vivo, this would predict a higher frequency of GCAT transitions at the 3' G versus the 5' G of the GG doublet in codon 12 of the human ras oncogenes. Indeed, this is consistent with the most common pattern of mutation observed in activated ras genes in a variety of solid human tumors in that it occurs primarily through GCAT transitions at the 3' G of the GG doublet in codon 12 of the ras oncogenes. For example, a majority of human pancreatic adenomas, colorectal carcinomas, and endometrial adenomas possess an activated Ki-ras oncogene with a GCAT transition at the 3' G of the GG doublet in codon 12, while Ha-ras is similarly mutated in human thyroid, urinary tract, and skin cancers (7), and N-ras is similarly mutated in hematopoietic cancers (1, 7).

The differential susceptibility to dTTP misincorporation of the individual G residues in the GG doublets studied might reflect their differential structural and reactive properties, as well as the stereochemical constraints on their abilities to form mismatches. In this context, it should be noted that the experiments presented above suggest that the adjacent G residues in GG doublets interact with each other in some way to influence the extent of misincorporation of dT opposite each other. As seen in Fig. 4, a G residue present upstream of another G, thus effectively creating a GG doublet, was found to stimulate dT misincorporation opposite the downstream G. On the other hand, when a G was present downstream of another guanine residue, it was shown to inhibit dT misincorporation opposite the upstream G. Presumably, the mutual interactions between the adjacent G residues in GG doublets, one residue stimulating and one residue inhibiting dT misincorporation opposite the other, would be additive to generate the differential in dT misincorporation observed between the 3' G and the 5' G residues in the doublets.

In addition to contributing to an understanding of the molecular basis of site-specific mutations in human oncogenes, the results presented here also demonstrate the abilities of very short sequences to affect mutagenic potential. Within these sequences the mutability of a given base, or at least its potential for mispairing, is dramatically affected by the base immediately upstream and the base immediately downstream of the target base. In these sequences, the effect of the adjacent base can be either stimulatory or inhibitory. In the case of the ras oncogenes, the mutual interactions between the two adjacent G residues in codon 12 would appear to create a "hot spot" for mutations caused by dT mispairing at the 3' G of the GG doublet and a "cold spot" for such mutations at the 5' G of the GG doublet. This observation could help explain the overwhelming preponderance of mutations that have been observed at the 3' G versus the 5' G of the GG doublet in codon 12 of the human ras oncogenes in various solid human tumors.

The above discussion focused on the interactions between the two G residues in the GG doublet in codon 12 of ras oncogenes. However, it appears that the GG doublet, with its strong, mutual interaction between the adjacent G residues, is a special case of a general rule for flanking base effects on dT misincorporation opposite G residues. Figure 4 showed the effect of changing the upstream or downstream G residue to a T residue, and we also have tested the effects of having A or C residues upstream or downstream of a G residue (17b). In general, any change in the base either upstream or downstream of a G residue led to a change in the ability of that G residue to mispair with dT. Thus, the GG doublet can be seen as a special case within the spectrum of flanking base effects. In terms of the effects of an upstream base, however, the GG doublet represents not only a special case but also the extreme case: among all the possible bases flanking a G residue on its 5' side, a G residue as the 5' neighbor (creating a GG doublet) had the maximal effect in stimulating dT misincorporation opposite the 3' G residue. Because of the maximal stimulatory effect on base mispairing of a G residue 5' to another G residue, the special case of the GG doublet in codon 12 of ras oncogenes can be seen as a "homing signal" for mismatch mutagenesis at the 3' G residue of the GG doublet.

Another aspect of flanking base effects was seen in experiments on the effects of changing the base immediately upstream of a GG doublet. The data presented in Fig. 5 show that the base immediately upstream of the GG doublet influenced the extent of mispairing of dT opposite the 5' G of the GG doublet. dT misincorporation opposite the 5' G was shown to be highest when C was the base immediately upstream of the GG doublet, intermediate when T was immediately upstream of the doublet, and lowest when A was the base immediately upstream of the GG doublet.

The effects of neighboring bases on mutagenesis and base mispairing also have been studied in other systems. In vitro studies on the effects of neighboring bases on alkylation by SN1 alkylating agents such as N-methyl-N-nitrosourea and N-methyl-N'-nitro-nitrosoguanidine have shown that alkylation of G residues is dependent on flanking bases (9). However, it should be noted that the mutations observed in our mutagenesis experiments occurred under conditions of a specific imbalance in two nucleotide pools, namely, a high-dTTP and a low-dCTP pool. Such pool perturbation is not known to cause site-specific alkylation of DNA. Thus, sequence-specific alkylation damage within GG doublets, as seen for mutagenesis by alkylating agents, is not relevant to our studies since it presumably is not operative under conditions of a nucleotide pool imbalance. The same may be said of other cellular phenomena that may exhibit or lead to sequence bias such as deamination or DNA repair.

Sequence-specific mutations caused by perturbations of nucleotide pools also have been studied in other systems. In a study of mutations induced in the aprt gene of CHO cells by high levels of dT, an effect of adjacent bases was considered to be evidence for a "next-nucleotide" effect (25). However, this next-nucleotide effect was not exhibited by mutations that were induced by deoxyribonucleotide pool perturbation in the hgprt gene in human cells (20). In addition, results in our laboratory on mutations in the gpt gene induced by nucleotide pool perturbation could not be explained by this effect (33).

In vitro DNA synthesizing systems have been used previously to study the effects of the local DNA sequences on base mispairing. The effects of different combinations of flanking bases on the formation of mismatches were studied, and it was seen that the frequency of misincorporation at a given site was affected by the upstream and downstream bases (21, 23). For example, in experiments involving DNA polymerase alpha, a T residue downstream of a G residue was shown to cause greater base mispairing opposite that G than did a downstream A (23). These observations are in agreement with results obtained in our laboratory (17b). However, the other studies did not attempt to address the relationship between flanking base effects observed in cell-free systems and sequence-specific mutagenesis in living cells. (The other studies did, however, indicate that all polymerases need not have the same specificity in terms of misinsertion at a given site [21].) In contrast, the results concerning GG doublets in the present study demonstrate a direct correlation between the effects of adjacent bases on misincorporation in a cell-free system and the occurrence of sequence-specific mutations in the corresponding gene sequences in cultured cells and in solid human tumors.

The preference for G:T mismatch formation at the 3' G of the GG doublet in codon 12 of the three human ras proto-oncogenes that we have demonstrated in our experiments is consistent with the most frequently observed pattern of activating mutations in ras oncogenes in solid human tumors. Nevertheless, the relevance of our results for tumorigenesis needs to be demonstrated directly. The lowest dTTP concentrations at which significant differences in the extent of misincorporation opposite the 3' and the 5' G residues in the GG doublets in codon 12 were detected (approximately 3 µM) are greater than the intracellular concentrations of dTTP that have been reported (31, 34). However, it is conceivable that during replication in vivo compartmentalization of intracellular dTTP could occur in the immediate proximity of the replication fork and thus create high local concentrations of dTTP. Such localized dTTP concentrations might be high enough to cause aberrant incorporation in a sequence-specific manner as described here. However, such transient, localized, elevated levels of intracellular dTTP would not be detected by measurements that rely on cellular extracts. In addition, information regarding intracellular nucleotide pool perturbation in human and animal systems is as yet scarce. It is also conceivable that not just dT but any agent that increases the intracellular dTTP/dCTP ratio could mimic the effect of dT and trigger sequence-specific mutagenesis as described above. It should also be noted that the concentrations of dTTP that were used in the present study were those that generated large levels of misincorporation as well as high levels of mutagenesis. However, tumor induction by mutation in vivo is presumably a very low frequency event. Thus, events occurring at low frequency in vivo, such as a rare misincorporation event in a sequence-specific manner, could be sufficient to generate the activating mutations in ras oncogenes observed in human tumors. For such rare misincorporation events to occur, it is conceivable that elevated dTTP concentrations and unbalanced dTTP/dCTP ratios may not even be necessary. Instead, the mispairing potential of G residues within GG doublets, as described in this study, may be strong enough to target mispairing to the 3' G residue of the GG doublet in codon 12 of ras oncogenes even in the absence of a dNTP pool imbalance.

In living systems, sequence-specific mutagenesis might be the end result of the interaction between two separate biological mechanisms, namely, the sequence-directed misincorporation of a noncomplementary base during replication as described here and the subsequent repair of such a mispair as influenced by the local DNA sequence. Indeed, a relationship between mutagenesis and sequence-specific repair of UV-induced cyclobutane pyrimidine dimers in the human p53 gene has been reported (30). Slow repair was seen in a majority of positions that were frequently mutated in skin cancer, suggesting that sequence-specific repair efficiency may contribute to mutations leading to cancer. On the other hand, our results suggest that sequence-directed mispairing can account for a major portion of sequence-specific mutagenesis in ras oncogenes in human tumors.