Comparative mutagenesis of the C8-guanine adducts of 1-nitropyrene and 1,6- and 1,8-dinitropyrene in a CpG repeat sequence. A slipped frameshift intermediate model for dinucleotide deletion.

In the Ames Salmonella typhimurium reversion assay 1,6- and 1,8-dinitropyrenes (1,6- and 1,8-DNPs) are much more potent mutagens than 1-nitropyrene (1-NP). Genetic experiments established that certain differences in the metabolism of the DNPs, which in turn result in increased DNA adduction, play a role. It remained unclear, however, if the DNP adducts, N-(guanin-8-yl)-1-amino-6 ()-nitropyrene (Gua-C8-1,6-ANP and Gua-C8-1,8-ANP), which contain a nitro group on the pyrene ring covalently linked to the guanine C8, are more mutagenic than the major 1-NP adduct, N-(guanin-8-yl)-1-aminopyrene (Gua-C8-AP). In order to address this, we have compared the mutation frequency of the three guanine C8 adducts, Gua-C8-AP, Gua-C8-1,6-ANP, and Gua-C8-1,8-ANP in a CGCG*CG sequence. Single-stranded M13mp7L2 vectors containing these adducts and a control were constructed and replicated in Escherichia coli. A remarkable difference in the induced CpG deletion frequency between these adducts was noted. In repair-competent cells the 1-NP adduct induced 1.7% CpG deletions without SOS, whereas the 1,6- and 1,8-DNP adducts induced 6.8 and 10.0% two-base deletions, respectively. With SOS, CpG deletions increased up to 1.9, 11.1, and 15.1% by 1-NP, 1,6-, and 1,8-DNP adducts, respectively. This result unequivocally established that DNP adducts are more mutagenic than the 1-NP adduct in the repetitive CpG sequence. In each case the mutation frequency was significantly increased in a mutS strain, which is impaired in methyl-directed mismatch repair, and a dnaQ strain, which carries a defect in proofreading activity of the DNA polymerase III. Modeling studies showed that the nitro group on the pyrene ring at the 8-position can provide additional stabilization to the two-nucleotide extrahelical loop in the promutagenic slipped frameshift intermediate through its added hydrogen-bonding capability. This could account for the increase in CpG deletions in the M13 vector with the nitro-containing adducts compared with the Gua-C8-AP adduct itself.

1-Nitropyrene (1-NP) 1 and the dinitropyrenes are common environmental pollutants (1)(2)(3). Most of the nitropyrenes are mutagenic (for a review see Ref. 4) and tumorigenic (5)(6)(7)(8), but their potency sometimes differs by more than an order of magnitude. Nitroreduction is a major pathway of bioactivation of all nitropyrenes, whereas O-esterification enzymes, in addition, play a crucial role in the mutagenicity of DNPs.
1-NP and the DNPs revert Salmonella typhimurium frameshift tester strains TA98 and TA1538 more efficiently than the strains TA100 and TA1535 that detect base pair substitutions (9). The most frequent mutation among the revertants in TA98 is a two-base deletion of a GpC or CpG pair within a CGCGCGCG hotspot sequence upstream of the hisD3052 mutation (10). The frequency of reversion induced by 1-NP in TA98 drops sharply in TA98NR that lacks the classical nitroreductase (11). By contrast, the frequency of reversion by 1,6-and 1,8-DNP is only slightly lower in TA98NR but is significantly reduced in TA98/1,8-DNP 6 , which is deficient in a specific arylhydroxylamine esterification enzyme (11,12). It appears that this enzyme is necessary for the expression of mutagenicity of the DNPs but not for that of 1-NP. The C8 guanine adducts of 1-NP and DNPs ( Fig. 1) have been thought to be responsible for a major fraction of their mutagenicity. Several site-specific studies from our laboratory showed that the 1-NP adduct N-(guanin-8-yl)-1-aminopyrene (Gua-C8-AP) is mutagenic in Escherichia coli (13)(14)(15). However, the type and frequency of mutations are dependent on DNA sequence context.
Despite the acknowledged role of metabolism and the ability of each nitropyrene to induce frameshifts, neither the frequency nor the genetic requirements for mutagenesis of the adducts formed by these carcinogens have ever been compared in the same organism. In order to address such structureactivity relationships, in the current work we have examined mutagenicity of the C8 guanine adducts formed by 1-NP and 1,6-and 1,8-DNP in a repetitive CpG sequence. We have constructed single-stranded M13 bacteriophage genomes in which an adduct was placed at the underscored dG of an inserted CGCGCG sequence in the lacZ ␣ fragment. Mutagenicity of the three adducts was examined in strains that are either repair-□ S The on-line version of this article (available at http://www.jbc.org) contains Fig. S1 and Tables S1-S3.
* competent or contained a defect in the dnaQ gene that encodes the ⑀-subunit of DNA polymerase III holoenzyme. Pol III from these strains has been shown to be defective in 3Ј-5Ј-proofreading exonuclease activity (16). In addition, we have determined the effects of this adduct in a strain with impaired postreplicative mismatch repair. Our results suggest formation of an extra-helical loop as the promutagenic intermediate, the stability of which may be linked to the mutation frequency. A modeling study was carried out to explore the structure of this hypothesized intermediate and to examine plausible reasons for added stability in the case of the nitro-containing adducts. Our findings suggest added possibilities for hydrogen bonding by the nitro group as likely sources of added stabilization.
Synthesis of d(CGCG*CG)-Fifty nmol of d(CGCGCG) was stirred at ambient temperature with 15 mol of 1-nitrosopyrene or 1-nitroso-6 (8)-nitropyrene and 3 mol of ascorbic acid in 100 l of N,N-dimethylformamide/sodium acetate/acetic acid buffer (0.1 M), pH 5.0 (1:9), under nitrogen and protected from light. An additional aliquot of 3 mol of ascorbic acid was added every 15 min during the 1st h, and the reaction was allowed to continue for 16 -20 h (19, 20). Noncovalently bound material was removed by extensive extraction with chloroform. The adducted and unadducted oligonucleotides were separated by reversed phase HPLC. Further purification of the oligonucleotides was carried out by denaturing PAGE. The oligonucleotides were desalted on a Sephadex G-10 column, dried, and stored at Ϫ20°C until further use.
Construction of Site-specifically Modified M13 Genomes-Bacteriophage M12mp7L2 (200 g) was digested with a large excess of EcoRI (2400 units) for 2 h at 25°C in 1 ml of 100 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 50 mM NaCl. Agarose gel electrophoresis indicated no visible band for the remaining circular DNA. 2-Fold molar excess of a scaffold 46-mer was annealed to the linear single-stranded DNA at a concentration of 100 ng/ml by heating at 67°C for 4 min followed by slow cooling to room temperature over a period of 3-4 h. The proportion of circularized vector as determined by agarose gel electrophoresis was 35-45%. Fifty-fold molar excess of the modified or unmodified hexanucleotide was ligated into the gap of this annealed DNA in the presence of 20 units of T4 DNA ligase (Invitrogen) in 40 mM Tris-HCl buffer, pH 7.8, 8 mM MgCl 2 , 16 mM dithiothreitol, and 1 mM ATP at 16°C overnight. A "mock ligation" was also carried out in which no oligonucleotide was included.
SOS Induction and Transformation in E. coli-E. coli cells were grown in 100-ml cultures to 1 ϫ 10 8 cells/ml and then harvested by centrifugation at 5,000 ϫ g for 15 min at 0°C. All E. coli cells were grown in Luria broth, except the dnaQ49 strain NR11446, which was grown in minimal medium. The cells were resuspended in an equal volume of ice-cold deionized water and recentrifuged at 5000 ϫ g for 30 min. This procedure was repeated except the cells were resuspended in 50 ml of water. The bacterial pellet was resuspended in 1 ml of glycerol/ water (10% v/v) and kept on ice until further use. To induce SOS, the following additional steps were introduced after the first centrifugation. The cells were resuspended in 50 ml of 10 mM MgSO 4 and treated with UV light (254 nm) (either 20 or 50 J/m 2 ) in 25-ml aliquots in 150 ϫ 50-mm plastic Petri dishes. The cultures were incubated in Luria broth at 37°C for 40 min in order to express SOS functions maximally. Following SOS induction, these cells were centrifuged, deionized, and resuspended in glycerol/water in a similar manner as described earlier (13) except all manipulations were carried out in subdued light.
Before transformation, the constructed genome was subjected to another round of EcoRI treatment to digest any uncut or religated M13mp7L2 DNA. A 10-fold molar excess of a 46-mer that contained the DNA sequence complementary to the scaffold oligomer was added to each DNA solution, and the mixture was heated at 100°C for 2 min to remove the scaffold and rapidly cooled to 0°C. To monitor the extent of removal of the scaffold, a gapped genome was taken through the same protocol. An aliquot of each of these DNA solutions was subjected to agarose gel electrophoresis to ensure that the scaffold was quantitatively denatured. For each transformation, 40 l of the cell suspension was mixed with 4 l (500 ng) of DNA solution and transferred to the bottom of an ice-cold Bio-Rad Gene-Pulser cuvette (0.1-cm electrode gap). Electroporation of cells was carried out in a Bio-Rad Gene-Pulser apparatus at 25 microfarads and 1.8 kV with the pulse controller set at 200 ohms. Immediately after electroporation, 1 ml of SOC medium was added, and the mixture was transferred to a 1.5-ml microcentrifuge tube. Part of the cells was plated following a 15-min recovery at 37°C in the presence of the plating bacteria E. coli GW5100, IPTG, and X-gal to determine the number of independent transformants. The remainder of cells was centrifuged at 15000 ϫ g for 5 min to isolate the phagecontaining supernatant. Minus-two and plus-one mutant phages were detected directly from the progeny as blue plaques after 18 h of incubation at 37°C. All other colorless mutants were detected and isolated by oligonucleotide hybridization as described (14).
Modeling Studies-The molecular mechanics program DUPLEX (21,22) was employed to obtain energy minimized structures for the AP and 1,8-ANP adducts in slipped mutagenic intermediate structures (23) containing a two-base bulge. Optimized geometries for the AP and 1,8-ANP adducts were computed in modified nucleosides, using Hartree-Fock calculations with 6 -31 G* basis set, employing the program package Gaussian 98 (24). These provided the bond lengths, bond angles, and dihedral angles needed for the coordinate generator in DUPLEX. Gaussian 98 (24) was also employed to compute the partial charges for these adducts in modified nucleosides. CNDO calculations, which provide charges that are compatible with the rest of the DUPLEX partial charge set (21), were used for this purpose. These are given in Table S1 of the Supplemental Material. Other DUPLEX force field parameters for the AP and 1,8-ANP adducts are the same as those employed previously (25).

Construction and Characterization of M13 Genomes Containing a Single Adduct-Deoxyhexanucleotides, d(CGCG*CG)
containing the C8 guanine adduct of 1-NP, 1,6-DNP, or 1,8-DNP, were synthesized by reacting N-hydroxy derivative of 1-aminopyrene or 1-amino-6(8)-nitropyrene in N,N-dimethylformamide/water (1:9), pH 5.0 -5.5, for 16 -20 h at ambient temperature and purified by HPLC followed by PAGE (19,20). The adducted oligonucleotides were characterized by enzymatic digestion to nucleosides followed by HPLC analysis and by electrophoretic migration pattern of the fragments after piperidine cleavage (at 90°C for 1 h) in comparison with Maxam-Gilbert G reaction products of the unmodified hexamer (see Refs. 13 and 20 for details). They were also examined by electrospray ionization-mass spectrometry analysis. The negative ion electrospray ionization-mass spectra of the hexamer containing Gua-C8-AP, Gua-C8 -1,6-ANP, and Gua-C8 Da (theoretical 2052.34), respectively. It is noteworthy that in each synthesis the major product contained the C8 guanine adduct at the second G, which eluted 7-8 min after the unmodified peak and 6 -12 min before the other modified hexamers by reverse-phase HPLC (see Fig. S1 of the Supplemental Material for a typical chromatogram). Reinjection of the purified d(CGCG*CG) on a reverse-phase column showed no detectable level of either the unmodified or the other modified hexamers, and based on the chromatographic analysis, we estimated that the modified hexamers were in excess of 98% pure. Since further PAGE purification of each oligonucleotide was carried out, and the migration pattern of each adducted hexamer was different (data not shown), the purity of the hexamers used in this work was expected to be significantly higher than 98%. Fig. 2 shows a typical autoradiogram of the unmodified and the three adducted hexamers by denaturing PAGE.
The hairpin region of a ϩ2 frameshift mutant clone of M13mp7 was digested with EcoRI (26). This was recircularized noncovalently by annealing a scaffold 46-mer, the two ends of which were complementary to the terminal 20 nucleotides of the linearized vector. The central segment of the oligomer was complementary to 5Ј-CGCGCG, which allowed the ligation of 5Ј-CGCG*CG to the ends of M13 vector by T4 DNA ligase. A control (i.e. unmodified) genome was constructed in a similar manner. A mock ligation was also carried out, in which no oligonucleotide was added.
In order to visualize the unmodified and modified DNA con-structs, a portion of each of these genomes was run on a 1% agarose gel in the presence of ethidium bromide. Densitometry analysis indicated that the efficiency of recircularization of the unmodified and the three modified vectors was 34 -35%. When these DNA constructs were treated with an excess of BssHII, Ͼ95% of the unmodified genome was digested to a linear material, whereas the presence of the adduct at the restriction site of the modified genome prevented it from being digested with this enzyme. This is expected because inhibition of restriction enzyme cleavage by the presence of DNA adducts including Gua-C8-AP has been shown in many earlier studies (13).
To remove the 46-mer scaffold from the M13 DNA, each DNA solution was heated at 100°C for 2 min and rapidly cooled to 0°C. Prior to heating, a 10-fold molar excess of a 46-mer that contained the DNA sequence complementary to the scaffold oligomer was added to the DNA solution to ensure that the scaffold, once denatured, did not reanneal on the M13 DNA. As shown in Fig. 3, subsequent to removal of the scaffold 46-mer, a band for the circular DNA construct was observed on an agarose gel when either a control or adducted hexamer was included in the ligation mix, but no circular DNA was detectable (Ͻ2%) in the mock construct.
Mutagenesis of the Three Adducts in Repair-proficient and -deficient Background-The constructed M13 vector used in this study is a ϩ2 derivative, which should generate colorless plaques in the presence of IPTG and X-gal. Either a Ϫ2 or a ϩ1 frameshift can restore the reading frame to a Lac ϩ phenotype generating blue plaques. Spontaneous mutagenesis of the unmodified vector as determined phenotypically by the reversion to blue plaques was 2 ϫ 10 Ϫ3 or less (Table I). In the repaircompetent strain DL7 the presence of Gua-C8-AP increased the MF to 1.7% as determined by an increased proportion of blue plaques (Table I and Fig. 4A). With SOS (50 J/m 2 ), the MF of Gua-C8-AP increased to 1.9%. The MF of the two DNP adducts, Gua-C8 -1,6-ANP and Gua-C8 -1,8-ANP, were 6.8 and 10.1% without SOS, which increased to 8.2 and 12.8%, respectively, with SOS (20 J/m 2 ). SOS induction with a higher dosage of UV (50 J/m 2 ) resulted in further increase in MF of the two DNP adducts, which was found to be 11.2 and 15.0%, respectively, for Gua-C8 -1,6-ANP and Gua-C8 -1,8-ANP (Table I and Fig.  4A). In another repair-competent strain, NR9295, the numbers were somewhat lower, although the trend was similar (Table I and Fig. 4B).
In a nucleotide excision repair-impaired strain (uvrA) the MF did not increase without SOS, but with SOS (20 J/m 2 ) there was a notable increase in MF, much more so than what we observed for the repair-proficient strains (Table I and Fig. 4A). Even though these bulky adducts are likely to be repaired by the nucleotide excision repair system, in single-stranded DNA it is unlikely to have a significant effect. We believe that the higher MF with SOS was due to a more pronounced SOS response, which in turn was a result of the persistence of the UV lesions in the E. coli DNA. For the mismatch repair-deficient strain (mutS), there was ϳ2-fold increase in MF of each of the three adducts compared with the repair-proficient strain. In the dnaQ strain, the increase in MF was most pronounced, and for Gua-C8-AP in the absence of SOS, MF was ϳ8-fold of the same in repair-competent strain, which increased to 11-fold with SOS (Table I and Fig. 4B). The extent of increase in MF for the two DNP adducts was much less pronounced in the dnaQ strain. Both with and without SOS, MF was ϳ3-fold relative to the repair-competent strain. A population of blue plaques from each transformation was subjected to DNA sequencing. In each case more than 90% of the blue plaques contained CpG deletions, although a small population of one-base additions was also detected.
To investigate phenotypically undetectable base substitutions and other types of frameshifts, we used oligonucleotide hybridization with a 17-mer probe complementary to the region of M13 where the 6-mer was inserted. The probe was designed to bind only to the non-mutant plaques. Therefore, all nonhybridized or weakly hybridized plaques were considered putative mutants and subjected to DNA sequencing. In the progeny from the control construct with SOS ϳ1% progeny (13/ 1357) were mutants, which showed one-base deletions in various sites in the CGCGCG insert. The MF of clear plaques did not increase in the adducted vectors, although both base substitutions and one-base deletions were detected. We conclude that in this repetitive CpG sequence, the predominant mutation induced by the three adducts is Ϫ2 frameshifts.
Modeling Studies-To investigate possible structural reasons for the enhancement in two-base deletions when the adducts contained a nitro group on the pyrene ring, we have carried out a computer modeling study. Specifically, we wished to obtain structures of the slipped mutagenic intermediate, which has been proposed to cause two-base frameshifts in NarI-type sequences (23). We investigated the following Sequence 1, with G4* modified by AP or 1,8-ANP. An extensive conformational search for slipped mutagenic intermediate structures of this type with modification at guanine C8 by another aromatic amine, N-acetyl-2-aminofluorene (AAF), provided a reasonable starting model (27). AAF was replaced by AP or 1,8-ANP; the base sequence was adjusted to match that in the present work (C1 replacing a G), and the structure was energy-minimized. A recently improved version of the molecular mechanics program DUPLEX (21), which accurately reproduces detailed sequencedependent conformational features of B-DNA duplexes (22), was employed in this work. In this improved DUPLEX, solvent water is treated implicitly with a new steep, sigmoidal distancedependent dielectric function that levels off to the dielectric constant of water at an interatomic distance equivalent to about one solvation shell. High level quantum mechanical calculations were used to obtain geometries of the AP and 1,8-

1-Nitropyrene and Dinitropyrene Mutagenesis in E. coli
ANP. These revealed that the oxygens of the NO 2 are out of plane with the AP ring system, with a dihedral angle O1-N-C9 -C10 (see Fig. 1) of Ϫ31.6 degrees. An in-plane position of these oxygens produces steric crowding with hydrogens in the adjacent ring system. The DUPLEX molecular mechanics calculations employed this quantum mechanically derived orientation of the NO 2 . Stereo views of the modeled structures are shown in Fig. 5. In both cases, a base-displaced intercalated conformation is adopted with the aromatic ring system stacked on the G6-C9 base pair. The modified, unpartnered G4* has a syn-glycosidic torsion angle and is displaced into the major groove, with the covalently linked pyrene ring system directed toward the minor groove; the bulged out, unpartnered C5 is positioned on the minor groove side and interacts on one face with the edge of the pyrene ring system. These structures share features with the NMR solution structure of the AP adduct in a DNA duplex with normal partner C in adopting a syn-guanine base-displaced intercalated conformation, with the modified G displaced into the major groove (28). In the duplex structure the looped out  Table S2 of the Supplemental Material).
Even though the slippage structures for the AP and 1,8-ANP are very similar, a key feature distinguishes them that can account for greater stability in the case of the 8-nitro derivative. Specifically, the nitro group oxygens are placed in a position to permit water-mediated hydrogen bonds between the carbonyl oxygen of the looped out C5 and one of the nitro oxygens; furthermore, this same nitro-oxygen can also form a second water-mediated hydrogen bond to the sugar O4Ј of G6. Table S3 of the Supplemental Material gives geometric features of these hydrogen bonds. In Fig. 5C we have modeled in these waters explicitly. This network of water-mediated hydrogen bonds is a feature only of the nitro-containing Gua-C8 -1,8-ANP and would serve to stabilize its slippage structure, including especially the looped out C; the latter would be more mobile in the AP slippage structure, as is the looped out partner C to the AP-modified G in the NMR solution structure of the duplex. DISCUSSION This study unequivocally demonstrated that the guanine C8 adducts of 1,6-and 1,8-DNP are significantly more mutagenic in E. coli cells than the 1-NP adduct. With SOS, MF of each adduct increases. The trend is similar in both repair-competent and repair-impaired cells, even though the magnitude of frameshift is highly dependent on the type of repair defect.
We hypothesize that the mechanism of two-base deletion involves generation of a promutagenic slipped frameshift inter-mediate whose structure is similar in the different adducts, but whose stability is greater in the more mutagenic adduct. Whereas these intermediates may form spontaneously in repetitive sequences (29), certain lesions such as the guanine C8 adducts of nitropyrenes and AAF are particularly efficient in generating such intermediates in high frequency (13,30). Scheme I shows our working model and also suggests how the slippage and elongation might be linked to proofreading and mismatch repair activity. To investigate if this may be the case, we have performed a computer modeling study comparing Gua-C8-AP with Gua-C8 -1,8-ANP. This study supports our hypothesis and shows how the 8-nitro substituent can stabilize the looped structure in solution.
High fidelity of DNA replication is maintained by at least three major steps (16). A high level of efficiency of the DNA polymerase in incorporating correct nucleotides is certainly very important. Nevertheless errors do occur, and in a large fraction of such cases the 3Ј-5Ј-exonuclease activity associated with the DNA polymerase can remove the incorrectly incorporated nucleotides. In addition, a DNA mismatch repair system detects and corrects mismatched nucleotides shortly after replication. It has been estimated that nucleotide selection discriminates against errors by 200,000 -2,000,000-fold, proofreading by 40 -200-fold, and mismatch repair by 20 -400-fold, each depending on the type of error (31). However, the ability of the DNA polymerase to incorporate the correct nucleotide opposite many carcinogen-DNA adducts is impaired, and the types and frequencies of misincorporation are often dependent on the DNA sequence surrounding the lesion. It has been suggested that when the replicative polymerase, pol III, encounters a replication-blocking lesion, it detaches from the replication terminus and translesional polymerases are recruited (32). Other polymerases such as pol II, pol IV, and pol SCHEME I. A model to account for the CpG deletion mutagenesis in the (CG) 3 sequence and the effects of proofreading and mismatch repair.

1-Nitropyrene and Dinitropyrene Mutagenesis in E. coli
V have been reported to be involved at this stage for translesion synthesis (32)(33)(34)(35). Some of these bypass polymerases are errorprone. The current work indicates the roles of two complementary repair systems in correcting the promutagenic slipped frameshift intermediate. As we have suggested in Scheme I, when either repair is impaired, there would be an increase in CpG deletion events. The guanine C8 adduct formed by AAF, a potent frameshift mutagenic lesion, which also intercalates with base displacement (36) and promotes slipped frameshift intermediates, shares some of the properties of the nitropyrene-DNA adducts (30,34,35,37,38). Our modeling studies of this slipped frameshift intermediate suggest that the bulged structure is more stable in the case of the 1,8-ANP compared with the AP adduct by virtue of its added hydrogen-bonding potential.
Current investigations in E. coli indicate that mutagenic bypass of bulky lesions can be carried out by pol II, pol III, pol IV, or pol V depending on the specific lesion and the base sequence context (35,39). Relaxed steric constraints in the active site appear to be a feature common to Y family bypass DNA polymerases, such as pol IV (40,41). Our modeling efforts suggest that specific hydrogen bonding capabilities of the nitro group, through water-mediated hydrogen bonds within the bulge, are plausible structural features that would stabilize the slipped mutagenic intermediate more in the 1,8-ANP adduct compared with the AP adduct itself. It is also conceivable that there may exist stabilizing water-mediated or direct hydrogenbonding interactions of the nitro group with specific amino acid residues within the polymerase. Since the 1,6-ANP adduct would also allow for such unique hydrogen-bonding interactions, stabilization of a slipped intermediate through an analogous structural mechanism could be envisioned.
Recent studies (32)(33)(34)(35), including the current work, suggest three major issues relating to translesion synthesis. First, it is important to analyze the structure, conformation, and stability of the promutagenic intermediate induced by the lesion, such as the slipped frameshift intermediate described here. Second, the DNA polymerases that can bypass the damage either with or without help from accessory proteins play crucial roles in both error-free and error-prone bypass events. This area of research has lately generated some intriguing data (32,42). Finally, the roles of DNA repair proteins that are involved either before or after the replication process must also be taken into account. It is fascinating how the different branches of chemistry and genetics are merging to decipher these puzzles of replication and mutagenesis.