The abundant DNA adduct N7-methyl deoxyguanosine contributes to miscoding during replication by human DNA polymerase η

Aside from abasic sites and ribonucleotides, the DNA adduct N7-methyl deoxyguanosine (N7-CH3 dG) is one of the most abundant lesions in mammalian DNA. Because N7-CH3 dG is unstable, leading to deglycosylation and ring-opening, its miscoding potential is not well-understood. Here, we employed a 2′-fluoro isostere approach to synthesize an oligonucleotide containing an analog of this lesion (N7-CH3 2′-F dG) and examined its miscoding potential with four Y-family translesion synthesis DNA polymerases (pols): human pol (hpol) η, hpol κ, and hpol ι and Dpo4 from the archaeal thermophile Sulfolobus solfataricus. We found that hpol η and Dpo4 can bypass the N7-CH3 2′-F dG adduct, albeit with some stalling, but hpol κ is strongly blocked at this lesion site, whereas hpol ι showed no distinction with the lesion and the control templates. hpol η yielded the highest level of misincorporation opposite the adduct by inserting dATP or dTTP. Moreover, hpol η did not extend well past an N7-CH3 2′-F dG:dT mispair. MS-based sequence analysis confirmed that hpol η catalyzes mainly error-free incorporation of dC, with misincorporation of dA and dG in 5–10% of products. We conclude that N7-CH3 2′-F dG and, by inference, N7-CH3 dG have miscoding and mutagenic potential. The level of misincorporation arising from this abundant adduct can be considered as potentially mutagenic as a highly miscoding but rare lesion.

The N7 atom of deoxyguanosine is the most nucleophilic site in DNA and is susceptible to alkylation, forming various N 7alkyl deoxyguanosine adducts (13,16,17). These adducts include N 7 -CH 3 dG, N 7 -ethyl deoxyguanosine, and N 7 -benzyl deoxyguanosine (13,18,19). The deoxyguanosine adduct formed with the 8,9-exo-epoxide of the hepatocellular carcinogen aflatoxin B 1 is highly mutagenic, causing GC to TA transversion mutations (6,19). N 7 -CH 3 dG has been detected as the major DNA adduct formed by methylating agents and is the most abundant lesion in DNA aside from abasic sites (1,20,21) and ribonucleotides (22,23), present in lymphocytes at levels of 14 adducts/10 7 normal nucleotides for nonsmokers and 25 adducts/10 7 nucleotides in smokers (24). Endogenous methylation of DNA, apparently from SAM, has been identified as the primary source of N 7 -CH 3 dG adducts observed in the livers of untreated rats (25,26).
The miscoding and mutagenic potentials of the resulting depurination (i.e. abasic sites) and ring-opened (i.e. N 7 -CH 3 formamidopyrimidine (FAPY) dG) products of N 7 -CH 3 dG have been extensively studied (27)(28)(29)(30)(31). The miscoding potential of N 7 -CH 3 dG itself is not understood. Despite its abundance, N 7 -CH 3 dG has been largely ignored in favor of other alkylated bases due to its instability to depurination and base-catalyzed ring-opening. It has been assumed that N 7 -CH 3 dG is not mis-coding because it should not alter the canonical Watson-Crick hydrogen-bonding pattern (32). However, the techniques that were used to reach this conclusion were not very sensitive compared with modern methods, and only a few model DNA polymerases were considered (21,33).
In 1961, Lawley and Brookes (34) proposed that alkylation at the guanine N7 position might induce mispairing due to its lowering of the pK a of the N1 position from 9 to 7, favoring rare tautomers (17, 34 -36). Even a low level of misincorporation across a very abundant lesion would be similar in risk to a highly miscoding but rare lesion.
Koag et al. (37) employed an isosteric fluorine transitionstate destabilization approach to stabilize the glycosidic bond, to avoid depurination and mild deprotection conditions to prevent ring-opening to N 7 -CH 3 FAPY dG. Although the lesion inhibited catalysis by pol ␤, replication was reported to be highly accurate (i.e. dCTP was inserted opposite N 7 -CH 3 dG) (37). We used this 2Ј-fluoro analog, N 7 -CH 3 2Ј-F dG, and analyzed its miscoding potential with several Y-family translesion synthesis polymerases (human pols (hpols) , , and and Sulfolobus solfataricus Dpo4). N 7 -CH 3 2Ј-F dG caused miscoding with hpol and has mutagenic potential, which we infer is the case with N 7 -CH 3 dG.

Synthesis of 2-F dG-and N 7 -CH 3 2-F dG-containing oligonucleotides
A fluorine analog of the lesion (N 7 -CH 3 2Ј-F dG) was prepared by modifying the approach of Lee et al. (32) (Scheme S1 and Figs. S1-S4). The 23-mer oligonucleotides were characterized by LC-ESI-MS (Figs. S5B and S6A). The N 7 -CH 3 2Ј-F dG-containing oligonucleotide was resistant to cleavage by FPG glycosylase, further confirming its identity as the intact lesion rather than the ring-opened N 7 -CH 3 FAPY 2Ј-F dG oligo-nucleotide, which is a substrate for this glycosylase (Fig. S6B). Following alkaline treatment to form the FAPY lesion, FPG glycosylase cleaved the lesion (Fig. S6B). We conclude that the desired N 7 -CH 3 2Ј-F dG lesion was present and that the two potential problems, depurination and ring-opening, had been avoided.

Primer extension past dG, 2-F dG, and N 7 -CH 3 2-F dG by Y-family DNA polymerases
Four Y-family DNA polymerases were studied (hpol , hpol , hpol , and Dpo4). hpol and Dpo4 were the most effective of these in producing full-length extension products, although they both stalled following the P ϩ 3 and P ϩ 2 products, respectively ( Fig. 1, A and D). In contrast, hpol stalled at the adduct (Fig. 1B), and hpol showed no distinction between the control templates and the adduct, stalling before the dG templates and the lesion (Fig. 1C).
To determine the insertions across the adduct, primer extension experiments were done with individual dNTPs. hpol and Dpo4 were highly error-prone with all three templates examined (Fig. 2, A and D). Dpo4 did not misincorporate dATP in the 2Ј-F dG control. In contrast with these two polymerases, hpol and hpol incorporated only the correct dCTP for all three templates under the same conditions (Fig. 2, B and C), and these two polymerases were not examined further. Rates for dCTP insertion by hpol were estimated (in single-nucleotide incorporation experiments) to be 1.9 min Ϫ1 for dG , 1.4 min Ϫ1 for 2Ј-F dG, and 0.9 min Ϫ1 for N 7 -CH 3 2Ј-F dG (i.e. there was a ϳ2-fold reduction in the rates when hpol encountered the lesion). For hpol , the rates of insertion of dCTP across the templates were 0.51 min Ϫ1 for dG, 0.56 min Ϫ1 for 2Ј-F dG, and 0.47 min Ϫ1 for N 7 -CH 3 2Ј-F dG. Thus, the rates when hpol encountered the lesion were comparable with the control templates.

Steady-state kinetics of individual dNTP insertion opposite dG, 2-F dG, and N 7 -CH 3 2-F dG by hpol and Dpo4
Steady-state kinetic analysis was performed for hpol and Dpo4 (Tables 1 and 2). The catalytic efficiencies and misincorporation frequencies for dG and 2Ј-F dG were comparable, as noted previously (38), suggesting that fluorine had little or no impact on polymerase recognition. With all three templates, hpol preferred to insert dCTP relative to other dNTPs (Table  1 and Fig. 3). However, there was a 2-fold lower efficiency for incorporation of dCTP at the N 7 -CH 3 2Ј-F dG lesion compared with dG and 2Ј-F dG. The efficiency for misinsertion of dATP was similar for all three templates, but the misinsertion frequency ϳ2-fold higher with N 7 -CH 3 2Ј-F dG (Table 1 and Figs. 3 and S7). The catalytic efficiency for dGTP misincorporation was ϳ3-fold lower for the N 7 -CH 3 2Ј-F dG lesion (Table 1 and Fig. 4). The efficiency for dTTP misincorporation was 3.5-fold greater for the N 7 -CH 3 2Ј-F dG lesion compared with the control templates, and the misincorporation frequency was 5-and 11-fold higher relative to dG and 2Ј-F dG (Table 1 and Fig. 4). Thus, the misincorporation frequency for N 7 -CH 3 2Ј-F dG was in the order dTTP Ͼ dATP Ͼ dGTP, ranging from 1 to 4% (Figs. S8 and S9).
Dpo4 preferentially inserted dCTP opposite N 7 -CH 3 2Ј-F dG ( Table 2). There was a 4-fold lower efficiency for insertion across the lesion compared with 2Ј-F dG (Figs. S8 and S9). The efficiencies for incorporating other dNTPs were in the order dATP Ͼ dTTP Ͼ dGTP ( Table 2).

Steady-state kinetics of post-lesion incorporation of individual dNTPs opposite 2-F dG or N 7 -CH 3 2-F dG by hpol
Steady-state insertion kinetics provides information on dNTP insertion across a lesion but does not provide information about extension past the lesion. Steady-state kinetics were done for further extension after the correct bp (N 7 -CH 3 2Ј-F dG:dC) and a mispair (N 7 -CH 3 2Ј-F dG:dT). dT was used as the misincorporated base opposite the lesion because it showed the greatest misincorporation frequency in the steady-state insertion kinetics with hpol (Table 1). With the mispairs (2Ј-F dG:dT and N 7 -CH 3 2Ј-F dG:dT), only dATP was incorporated opposite the next residue (dT) (Fig. S10). The efficiency of hpol for incorporating dATP past the mispair was ϳ4-fold lower for the lesion N 7 -CH 3 2Ј-F dG than 2Ј-F dG, indicating some resistance to extension past the mispair (Table 3 and Fig. 5). On the other hand, the efficiency of hpol for inserting dATP past the N 7 -CH 3 2Ј-F dG:dC bp was 21-fold higher than the 2Ј-F dG:dC control, indicating that the correct pair was preferentially extended past the lesion. For the 2Ј-F dG control and lesion, dCTP had a similar efficiency of misincorporation past the correct pair. Finally, dTTP was misincorporated with a 6-fold higher efficiency for the lesion than the 2Ј-F dG control (Table 3 and Fig. 5).

LC-MS/MS sequence analysis of extension products formed by hpol and Dpo4
We introduced a dT:dU mismatch upstream of the site of dNTP addition to utilize uracil-DNA glycosylase (UDG) to cut the extension products for analysis by LC-MS/MS. Replication of the unmodified oligonucleotide gave only error-free products, as reported elsewhere (38). Replication across the lesion by Dpo4 also gave only error-free products, in support of the results of steady-state insertion kinetics (Table 4). hpol replicated through the lesion in both an error-free and an errorprone manner, resulting in three main products ( Table 5). The first product corresponded to error-free products (i.e. m/z 934.3: 5Ј-pTCATGA, m/z 1086.3: 5Ј-pTCATGAT, and m/z 613.2: 5Ј-pTCAT) Figs. S11 and S12. The second corresponded to misincorporation of dA (m/z 934.3: 5Ј-pTACTGA and m/z 1086.3: 5Ј-pTAGTCAT), and the third corresponded to misincorporation of dG (m/z 1086.3: 5Ј-pTGATCAT) (Figs. S9 and S10). The CID spectra of the products matched the predicted CID spectra of the sequences (Tables S1-S6).

Miscoding of N 7 -methyl deoxyguanosine
To confirm these assignments, mass spectra of commercial oligonucleotide standards with these sequences were compared with those of the observed products and were nearly identical. No products were observed containing the misincorporation of dT seen in the insertion kinetics experiments (Table  1). Relative areas were calculated for each product on the basis of the intensity of distinguishing CID ions (e.g. a 3 -B 3 ions distinguish the error-free product from the product with misincorporation of dA). The yields of the observed products were estimated to be 85% for error-free bypass, 10% for misincorporation of dA, and 5% for misincorporation of dG (Table 4).

Discussion
Alkylation of DNA was first described in 1960 (20,39), and the N7 atom of dG has long been known to be a major site of damage (34). The change in the pK a of the N1 atom (from 9 to 7) upon N 7 -methylation (34) was considered to be a potential reason for miscoding, evoking the original postulate of rare tautomer involvement in miscoding proposed by Watson and Crick (41). Due to this issue, one cannot consider an approach with 7-deaza dG for studying N 7 -alkyl dG miscoding, which would not reflect the electronic properties of the adduct. For discussion of the early studies on different alkylated bases and the development of a major role for O 6 -alkyl dG adducts in mutagenesis and carcinogenesis, see Lawley (39). Although O 6 -alkyl dG lesions are recognized to be important, the role of dG N 7 -alkylation has remained unclear. Some early studies concluded that N 7 -CH 3 dG was not miscoding (39,42), but the results of these studies are compromised by several issues, including the sensitivity of the assays in detecting miscoding, the lack of mammalian and microbial translesion DNA polymerases, and the lability of N 7 -CH 3 dG. In 2009, Boysen et al. (21) concluded that there was no evidence for miscoding by N 7 -CH 3 dG, although the authors suggested the 2Ј-F isostere approach we used here to address the issue. Lee and associates (37) used N 7 -CH 3 2Ј-F dG with pol ␤ and concluded that it was not miscoding but did not present limits of detection or utilize sensitive methods. N 7 -Alkyl dG adducts are of particular interest because of their high endogenous levels and also high levels following exposure to alkylating agents (21,39,43,44). N 7 -Alkyl dG Table 1 Steady-state kinetics of single nucleotide insertion opposite dG, 2-F dG, and N 7 -CH 3

2-F dG by hpol
The oligonucleotides used were as follows, where X represents dG, 2Ј-F dG, or N 7 -CH 3 2Ј-F dG.

Table 2
Steady-state kinetics of single nucleotide insertion opposite dG, 2-F dG, and N 7 -CH 3

2-F dG by S. solfataricus Dpo4
The oligonucleotides used were as follows, where X represents dG, 2Ј-F dG, and N 7 -CH 3 2Ј-F dG.

Miscoding of N 7 -methyl deoxyguanosine
adducts are found at the highest levels not only after exposure to methylating agents but with other alkylating agents as well (17,39,44,45). Several examples of N 7 -alkyl dG adducts are found in laboratory animals and humans not knowingly exposed to exogenous agents, including N 7 -(2-hydroxy)ethyl dG, N 7 -(2-oxoethyl) dG, and N 7 -ethyl dG (44), but the origins of these adducts are not known. Although the levels of ribonucleotides and abasic sites have been reported to be higher than those of N 7 -CH 3 dG, they are rapidly repaired by multiple pathways (22,23), and the steady-state levels in cells are less than those of N 7 -CH 3 dG (43).
The base-catalyzed imidazole ring opening of guanyl N 7 -alkyl adducts has been recognized for many years. As pointed out by Gates et al. (17), N 7 -CH 3 dG is not unusually unstable, and at neutral pH, ring-opening is very slow; even at pH 8.9, the halflife is 9.8 h (46 -49). Although there was original uncertainty about the multiple forms of N 7 -CH 3 FAPY dG seen in chromatography, 15 N NMR studies demonstrated that the site of the formyl group did not change (46) and that the adduct exists in slowly equilibrating rotomeric forms. Studies with rat liver and bladder DNA reported that levels of N 7 -CH 3 dG decreased faster than those of the FAPY product, and levels of the two adducts were similar after 3-9 days (50, 51). However, Den Engelse et al. (49) reported only very low levels of the FAPY formed in rat liver following treatment with methylating agents. Some of the discrepancy may be due to the broadness of the N 7 -CH 3 FAPY dG peaks, affecting both the resolution and the sensitivity (46,49,50,52). In the report of Den Engelse et al. (49), no N 7 -CH 3 FAPY dG adducts were detected in rat liver (Ͻ0.5% of N 7 -CH 3 dG) up to 3 days after treatment with [ 14 C]dimethylnitrosamine. Even in the report of Kadlubar et al. (51), the level of N 7 -CH 3 FAPY dG did not reach the level of N 7 -CH 3 dG (in the rat bladder epithelium) until 9 days after treatment with [ 14 C]-methylnitrosourea. In considering all of this information, we conclude that the level of N 7 -CH 3 dG is considerable and that   Table 1. Table 3 Steady-state kinetics of single nucleotide extension past 2-F dG:dC, 2-F dG:dT and N 7 -CH 3 2-F dG:dC, and N 7 -CH 3

2-F dG:dT base pairs by hpol
The oligonucleotides used were as follows, where X represents 2Ј-F dG and N 7 -CH 3 2Ј-F dG.   , and N 7 -CH 3 2Ј-F dG (C and F) at position X in the sequences 5Ј-CGGGCTCGTAAGCGTCAT-3Ј and 3Ј-GCCCGAGCATTCGCAGTAXTACT-5Ј. Reactions were done at 37°C for 5-10 min by incubating 120 nM primer-template DNA complex with varying concentrations of hpol . For A, B, and C, 8 nmol of hpol was used, and the reaction was done for 10 min. In the case of D, we used 8 nmol of hpol was used, and the reaction was done for 5 min; for E, 10 nmol of hpol was used, and the reaction was conducted for 5 min; and for F, 5 nmol of hpol was used, and the reaction was conducted for 5 min, varying concentrations of dGTP and dTTP. Fitting was to a hyperbolic equation in GraphPad Prism version 8.0, and k cat and K m values are presented in Table 1.

Miscoding of N 7 -methyl deoxyguanosine
any biological effects cannot be simply ascribed to abasic sites and N 7 -CH 3 FAPY dG. N 7 -CH 3 dG is a substrate for several glycosylases, in addition to removal due to nonenzymatic depurination (53,54), including 3-alkyladenine DNA glycosylase (AAG) in humans and the bacterial homologs 3-methyladenine glycosylase (AlkA), Bacillus cereus DNA glycosylase AlkD, and Streptomyces sahachiroi AlkZ (55)(56)(57). The chemical and biological half-lives of N 7 -CH 3 dG have been estimated to be in the range of 69 -192 h at 37°C and neutral pH (chemical) (17) and 29 -58 h (biphasic) in rat liver (presumably converting to an abasic site in the study cited, in that N 7 -CH 3 dG was not detected (49). N 7 -CH 3 FAPY dG is also a substrate for Escherichia coli FPG and other glycosylases (e.g. human OGG1, NTH1, and NEIL1) (58 -61). The point made here is that N 7 -CH 3 dG is persistent enough to be copied and miscoded, at least in tissues undergoing DNA replication.
In E. coli, N 7 -CH 3 FAPY dG was not highly mutagenic when bypassed (G to T transversion mutation frequency of Յ2%) (62). When N 7 -CH 3 FAPY dG was bypassed in a shuttle vector in simian kidney COS-7 cells, it readily produced G to T transversion mutations with 30% frequency (63). N 7 -CH 3 FAPY dG

Table 4 LC-ESI-MS/MS analysis of full-length extension products across N 7 -CH 3 2-F dG by hpol
The oligonucleotides used were as follows, where X represents 2Ј-F dG and N 7 -CH 3 2Ј-F dG. Products were cut at U, and the expected sequences began at the 3Ј T of the primer.

Miscoding of N 7 -methyl deoxyguanosine
was a strong block to replicative polymerases (e.g. pol ␣ and pol ␦/proliferating cell nuclear antigen), but hpol , hpol , and the sequential action of hRev1/hpol and Dpo4 were able to bypass N 7 -CH 3 FAPY dG (29,30). With hpol , N 7 -CH 3 FAPY dG reduced the efficiency of dCTP insertion by an order of magnitude (29). Our previous work on the miscoding properties of N 7 -CH 3 FAPY dG (29,30) can be summarized and compared with the present work on N 7 -CH 3 dG. Steady-state kinetic experiments on misinsertion showed only a low frequency of miscoding with S. solfataricus Dpo4 (0.01-0.04) but higher frequencies (0.28 and 0.29 for dT and dG insertion, respectively) with E. coli DNA polymerase I Klenow fragment. LC-MS analysis showed only misincorporation of dA for both polymerases examined with levels of misincorporation (2-35%) but considerable Ϫ1 frameshifts (11-17%) (30). In a later study with mammalian translesion DNA polymerases (29), we observed 2-5% misincorporation at N 7 -CH 3 FAPY dG in steady-state kinetics and 11-29% misincorporation by LC-MS for extension products with hpol and . Thus, the extents of misinsertion of hpol (Tables 1 and 4) are similar in magnitude to those seen with N 7 -CH 3 FAPY dG (29), although the oligonucleotide sequence is not the same.
Although Dpo4 and hpol are sometimes considered homologs (64,65), they showed different abilities to replicate past N 7 -CH 3 2Ј-F dG (Fig. 1, B and D), with hpol strongly blocked at the adduct site. hpol has been shown to bypass DNA adducts formed with methyl methanesulfonate more efficiently than hpols and , and it also interacts directly with the ligase SHPRH to suppress methyl methanesulfonate-induced mutagenesis (66). hpol , which also inserted only dCTP, is also effective in inserting dNTPs across minor groove lesions, such as N 3 -methyl deoxyadenosine (67). Koag et al. (37) evaluated the kinetics of insertion of dCTP and dTTP across N 7 -CH 3 2Ј-F dG by pol ␤, a gap-filling X-family polymerase. The lesion decreased the rate of pol ␤ catalysis by ϳ300-fold, yet replication was accurate, and no misinsertion products were reported. The structures revealed Watson-Crick base pairing of N 7 -CH 3 2Ј-F dG with an incoming dCTP, but the metal ion coordination was not optimal for catalysis. When N 7 -CH 3 2Ј-F dG was crystallized with dTTP, an open conformation was found, with a staggered bp. The relatively low but finite level of misincorporation at the N 7 -CH 3 dG might seem unimportant. However, consideration needs to be given to the overall mutagenic load. In four different studies cited by Den Engelse et al. (49), the ratio of N 7 -CH 3 dG to O 6 -CH 3 dG adducts following treatment (of cells or rats) with dimethylnitrosamine or methylnitrosoureas was ϳ10:1. In our own studies with hpol (68), miscoding in the LC-MS assays was 77%, which may be compared with 15% here with N 7 -CH 3 2Ј-F dG ( Table 4). Multiplying the adduct level differences, 77 ϫ 0.1 ϭ 7.7 (O 6 -CH 3 dG), which can be compared with 15 ϫ 1 ϭ 15 (N 7 -CH 3 (2Ј-F) dG). Kunkel (43) has estimated a 200 -3,000-fold difference in endogenous cellular levels of N 7 -CH 3 dG over O 6 -CH 3 dG. In a more recent study with cultured human lymphoblastoid cells, Sharma et al. (69) reported a 12-fold higher level of N 7 -CH 3 dG adducts than O 6 -CH 3 dG after treatment with methylnitrosourea and a 900-fold higher level of N 7 -CH 3 dG in the untreated cells. Applying the difference in levels of miscoding to these levels of the adducts can therefore result in an even larger potential contribution of N 7 -CH 3 dG to miscoding and mutagenesis.
In summary, we have shown that hpol produces error-free bypass products in copying past N 7 -CH 3 2Ј-F dG and also misinserts dA and dG, differing from the products seen for N 7 -CH 3 FAPY dG, which inserted dT and produced a frameshift mutation (29). Our findings indicate that our results are not due to any contamination by the FAPY degradation product and also suggest N 7 -CH 3 dG contribution to mutagenicity in cells. Caveats need to be considered about comparing miscoding frequencies in different sequence contexts, the potential roles of DNA polymerases that were not included here, rates of enzymatic repair in different cells, and possibly other issues. Inserting plasmid vectors containing N 7 -CH 3 dG into cells to estimate mutation frequencies would be very problematic in terms of being sure that the lesion, even with the 2Ј-F group, was not modified before mutation occurred. In conclusion, the abundance of the adduct N 7 -CH 3 dG, coupled with the evidence for miscoding, argues that this lesion should no longer be considered innocuous.

LC-ESI-MS/MS analysis of full-length extension products across N 7 -CH 3 2-F dG by Dpo4
The oligonucleotides used were as follows, where X represents 2Ј-F dG and N 7 -CH 3 2Ј-F dG. Products were cut at U, and expected sequences began at 3Ј T.

Miscoding of N 7 -methyl deoxyguanosine
ing. Restriction endonucleases, UDG, FPG glycosylase, dNTPs, and T4 polynucleotide kinase were purchased from New England Biolabs (Ipswich, MA). Unmodified oligonucleotides and primers used for extension and steady-state kinetics were obtained from Integrated DNA Technologies (Coralville, IA) and were HPLC-purified. Primers used for LC-MS sequence analysis were also obtained from DNA Technologies (Coralville, IA) and were twice HPLC-purified. Human DNA polymerases hpol (catalytic core residues 1-432), hpol (catalytic core residues 1-420), and hpol (catalytic core residues 19 -526) and bacterial Dpo4 were expressed in E. coli and purified as described previously (70 -73). 1 H and 13 C NMR spectra were recorded on a 600-MHz Bruker NMR spectrometer; 31 P NMR spectra were recorded on a 500-MHz Bruker NMR spectrometer. Mass spectrometry was performed at the Vanderbilt Mass Spectrometry Research Core Facility using both Thermo low-resolution (LTQ) and highresolution (Orbitrap) spectrometers. Spectra of synthetic products (negative and positive ion modes) and modified oligonucleotides (negative ion mode) were obtained using a Waters Acquity UPLC instrument (Waters, Milford, MA) interfaced to a Thermo-Finnigan LTQ mass spectrometer (Thermo Scientific, San Jose, CA), also equipped with an electrospray source.

Synthesis, purification, and characterization of 2-F dG and N 7 -CH 3 2-F dG-containing DNA oligonucleotides
Modified oligonucleotides bearing 2Ј-fluorines were synthesized with Expedite reagents (Glen Research, Sterling, VA) on a 1-mol scale utilizing a Perspective Biosystems model 8909 DNA synthesizer and a standard synthetic protocol (75). We chose the ␤-anomer for the 2Ј-fluoro analogs because this configuration has been shown not to alter sugar puckering in DNA; this is the typical configuration for the 2Ј-deoxynucleotides (76 -78). The coupling of N 7 -CH 3 2Ј-F dG phosphoramidite was performed off-line for 2 h. The remainder of the synthesis was done online using standard procedures. Modified oligonucleotides were cleaved from the solid support, and exocyclic groups were deprotected in a single step using anhydrous methanolic K 2 CO 3 (50 mM), stirring at room temperature for 8 h. CH 3 OH was removed by sweeping with a stream of N 2 gas. Oligonucleotides were purified by reversed-phase HPLC with a Phenominex Alumina RP octadecylsilane (C 18 ) column (250 mm ϫ 4.6 mm, 5 m). The solvents used were aqueous 100 mM triethylammonium acetate (mobile phase A) and 100 mM triethylammonium acetate in H 2 O/CH 3 CN (1:1, v/v) (mobile phase B). The flow rate was 1.5 ml/min with the following gradient: initial 20% B, increased to 25% B over 5 min, held at 25% for 15 min, increased to 40% at 20 min, held for 5 min, then 100% at 25 min, and held until 30 min and 5% B at 31 min and re-equilibrated to 0% B for 5 min (all v/v). The UV detector was set at 240 nm. The collected fractions were lyophilized to dryness, redissolved in water, and desalted using ZipTip U-C18 columns prior to characterization.
The identity of the N 7 -CH 3 2Ј-F dG-containing oligonucleotide was further confirmed by subjecting it to FPG glycosylase. The N 7 -CH 3 2Ј-F dG-containing oligonucleotide was 32 P-labeled at the 5Ј-end using T4 polynucleotide kinase (New England Biolabs) and annealed to its complementary strand by heating at 95°C for 5 min and then allowing it to cool to room temperature overnight. A second portion of the N 7 -CH 3 2Ј-F dG-oligonucleotide was treated with NaOH and stirred for 12 h at room temperature to create a hydrolyzed N 7 -CH 3 FAPY-2Ј-F dG oligonucleotide. It was also 5Ј-end-labeled ( 32 P-label and T4 polynucleotide kinase) and then annealed with its complementary strand. Both oligonucleotides were subjected to treat-ment with FPG glycosylase for 1 h at 37°C. Reactions were quenched with 9 l of quenching dye (20 mM EDTA, (pH 9.0) in 95% formamide, v/v) and the products were separated on a 20% acrylamide (w/v) electrophoresis gel. Results were visualized using a phosphorimaging system (Bio-Rad, Molecular Imager FX) and analyzed by Quantity One software as described previously (38).

LC-MS analysis of full-length extension products by hpol and Dpo4
An 18-mer primer bearing a 2Ј-deoxyuridine (5Ј-FAM/ CGGGCTCGTAAGCGTC(dU)T-3Ј) was annealed to the 23mer oligomer used above, in a molar ratio of 1:1. Full-length extension reactions were done using similar conditions as in the steady-state experiments, with the exception of primer-Miscoding of N 7 -methyl deoxyguanosine template complex (2.5 M), hpol (150 nM), Dpo4 (300 nM), and dNTPs (500 M). Reactions were incubated at 37°C for 1 h. Reactions were quenched by spin column separation to remove Mg 2ϩ and dNTPs, and the extension product was treated with 25 units of UDG at 37°C for 4 h and then with 0.25 M piperidine, heating at 95°C for 1 h. H 2 O was added to the reaction mixture, which was lyophilized and then redissolved in H 2 O (70). Products were analyzed by LC-MS/MS, performed using a Waters Acquity UPLC system linked to a Thermo-Finnigan LTQ mass spectrometer with electrospray ionization in the negative ion mode. Separation by chromatography was done using an Acquity UPLC system BEH octadecylsilane (C 18 ) column (1.7 m, 2.1 mm ϫ 50 mm) with UPLC conditions as described previously (40).