Oxidation of Thymine to 5-Formyluracil in DNA Promotes Misincorporation of dGMP and Subsequent Elongation of a Mismatched Primer Terminus by DNA Polymerase*

5-Formyluracil (fU) is a major oxidative thymine lesion generated by ionizing radiation and reactive oxygen species. In the present study, we have assessed the influence of fU on DNA replication to elucidate its genotoxic potential. Oligonucleotide templates containing fU at defined sites were replicated in vitro by Escherichia coli DNA polymerase I Klenow fragment deficient in 3'-5'-exonuclease. Gel electrophoretic analysis of the reaction products showed that fU constituted very weak replication blocks to DNA synthesis, suggesting a weak to negligible cytotoxic effect of this lesion. However, primer extension assays with a single dNTP revealed that fU directed incorporation of not only correct dAMP but also incorrect dGMP, although much less efficiently. No incorporation of dCMP and dTMP was observed. When fU was substituted for T in templates, the incorporation efficiency of dAMP (f(A) = V(max)/K(m)) decreased to (1/4) to (1/2), depending on the nearest neighbor base pair, and that of dGMP (f(G)) increased 1.1-5.6-fold. Thus, the increase in the replication error frequency (f(G)/f(A) for fU versus T) was 3.1-14.3-fold. The misincorporation rate of dGMP opposite fU (pK(a) = 8.6) but not T (pK(a) = 10.0) increased with pH (7.2-8.6) of the reaction mixture, indicating the participation of the ionized (or enolate) form of fU in the mispairing with G. The resulting mismatched fU:G primer terminus was more efficiently extended than the T:G terminus (8.2-11.3-fold). These results show that when T is oxidized to fU in DNA, fU promotes both misincorporation of dGMP at this site and subsequent elongation of the mismatched primer, hence potentially mutagenic.

4 is ring fragmentation products such as a urea residue and its analogues. The response of DNA polymerases to the first and second groups has been clarified fairly well by in vitro and in vivo studies (13-15, 22, 23). The third group includes 5-hydroxy-5methylhydantoin, a ring contraction product. The ability of this lesion to block DNA replication has been recently demonstrated by in vitro DNA polymerase reactions using a defined oligonucleotide template (24). The fourth group contains methyl oxidation products such as 5-hydroxymethyluracil and 5-formyluracil (fU) 1 . Several lines of evidence indicate that 5-hydroxymethyluracil is neither a replicative block nor mutagenic (25,26), hence being an innocuous lesion.
The genotoxic potential of fU belonging to the fourth group has been assessed in this (27,28) and other (29, 30) laboratories. In our previous approach, we synthesized 5-formyl-2'-deoxyuridine 5'-triphosphate (fdUTP) and studied its incorporation into DNA by DNA polymerases. fdUTP efficiently substituted for dTTP and to a much less extent for dCTP. Moreover, the pH-dependent variation of the substitution efficiency for dCTP suggested involvement of an ionized (or enolate) form of fU as a key intermediate responsible for the mispairing with template G. Such a mutation mechanism involving ionized bases (thymine and 5-bromouracil (BrU)) was originally suggested by Lawley and Brookes (31). Later, the pH-dependent variation 5 role in nucleotide selection by DNA polymerases (1,2), this process can also be affected by base pairing symmetry whether a X:Y base pair is formed from X (template):Y (dNTP) or X (dNTP):Y (template) (34)(35)(36). Secondly, the sequence context can affect the selection of dNTP opposite the template lesion (7). Thirdly, the base ionization mechanism somehow does not hold when fU is present in template DNA.
In view of the potential influences of base pairing asymmetry, the sequence context, and deviations from the base ionization mechanism mentioned above, we have prepared oligonucleotide templates containing site specific fU following the previously reported phosphoramidite method (33) and reexamined the base pairing capacity of template fU with the four possible nearest neighbor base pairs. The results show that fU in the template directs misincorporation of dGTP in a pH-dependent manner, supporting our previous results obtained by the analysis of fdUTP incorporation. by guest on July 9, 2020 http://www.jbc.org/ Downloaded from 7 incubated at 37 °C for 2 h. The sample was passed through a molecular weight cut off filter (Mw = 10000) and an aliquot of the filtrate was analyzed by HPLC equipped with a C18 WS-DNA column (4.6 x 150 mm, Wako). The sample was eluted by a gradient of methanol in 10 mM sodium phosphate buffer (pH 7.4) at a flow rate 0.8 ml/min. The concentration of methanol was 0% for a 0-5 min and 0-5% linear gradient for  min. The column temperature was maintained at 40 °C by a column oven and eluents were monitored at 280 nm.
Treatments with Repair Enzymes-----25F was 5'-end labeled as described for the primers (see below) and annealed to the complementary strand (25COM).
The template/primer used in the analysis was 24AF/13T, 24GF/13C, 24CF/13G and 24TF/13A that contained fU at the same position and different nearest base pairs next to fU (i.e. primer terminus base pairs). For control reactions, primer extension assays were also performed using template/primers (24AT/13T, 24GT/13C, 24CT/13G and 24TT/13A) that contained T in place of fU. Using these template/primers containing fU or T, kinetic parameters of nucleotide incorporation opposite fU and T were also determined. For dATP incorporation, the dATP concentration was 0.05-1 µM and the amount of Pol I Kf (exo -) was 0.002 unit, while for dGTP incorporation, those were 10-80 µM and 0.03 unit. The incubation time was 5 min for both dATP and dGTP incorporation assays. Under these conditions, the extent of primer elongation was essentially proportional to the reaction time and the unit of Pol I Kf (exo -) used. The initial velocity of the reaction (V) (average of two experiments) was calculated as the percent of the extended primer per min per 0.03 unit of Pol I Kf (exo -). K m and V max values were evaluated from a hyperbolic curve fitting program.
The pH effect on the incorporation of dATP and dGTP opposite template fU or T was determined by varying the pH (pH 7.2-8.6) of the polymerase reaction buffer as described above. The template/primer (24AF/13T and 24AT/13T , 0.5 pmol) was incubated with Pol I Kf (exo -) (0.03 unit) and dATP or dGTP (20 µM) at 25 °C for 5 min. For a wide pH range (pH 6.9-9.3), GTA buffer (buffering capacity pH 3.5-10) was used in place of the Tris buffer for the DNA polymerase reaction. The composition of GTA buffer was 3,3-dimethylglutaric acid, Tris, and 2-amino-2-methyl-1,3propanediol (17.3 mM each), and pH was adjusted by adding HCl or NaOH.
The parameters for mismatch extension with templates 24CT and 24CF were also determined in the same manner.

RESULTS
Nucleoside Composition of Oligonucleotides-----To ensure the validity of phosphoramidite method used in the present preparation of oligonucleotides containing fU, pilot oligonucleotides containing T (25T) and fU (25F) in the same sequence were prepared. 25T and 25F were digested by nuclease P1 and alkaline phosphatase and the nucleoside composition was analyzed by HPLC. Digestion of 25T resulted in the HPLC peaks of dC, dG, dT and dA with an expected molar ratio (8:6:4:7) (data not shown). In the HPLC analysis of the digested 25F, two extra peaks were observed in addition to the four normal nucleosides ( Fig. 1 A). The first peak eluted at 15.1 min was readily identified as 5-formyl-2'-deoxyuridine (fdU) by comparison with the retention time of authentic fdU. The peak at 32.4 min (indicated by *) was an unknown product. When authentic fdU was incubated under the same conditions as those used for oligonucleotide digestion, fdU was partially converted to this product ( Fig. 1 B).
The formyl group of fdU is fairly reactive and forms adducts with nucleophiles (40).
Accordingly, this product is most likely an adduct between fdU and a nucleophilic molecule present in the reaction buffer or enzyme preparations. The long HPLC retention time of the product relative to fdU and retention of the UV absorption around 280 nm were also consistent with the adduct formation of the exocyclic formyl group of fdU. The fU moiety of fdU is known to be degraded by strong base and oxidizing reagents, giving rise to ring fragmentation products (27,41) and 5-carboxyuracil (42), respectively. If such products were formed during the preparation of 25F, they might be present as contaminated lesions in 25F. However, the retention of the UV absorption (around 280 nm) of the product was inconsistent with the ring fragmentation products bearing no chromophores. The retention time of the product (32.4 min) much longer than fdU (15.1 min) in the reversed phase HPLC column also contradicted the expected very short retention time of 5-carboxy-2'-deoxyuridine bearing a negative charge of a carboxylate ion. Thus, the product was not the ring fragmentation products or 5-carboxy-2'-deoxyuridine. Moreover, when the amount of authentic fdU converted to the putative adduct was taken into account, the corrected molar ratio of nucleosides in 25F agreed with the expected value (dC:dG:dT:dA:fdU = 8:6:3:7:1).
We also attempted to identify the structure of the unknown product by mass spectrometric analysis after isolating the product by HPLC. However, the attempt was unsuccessful because of the lack of the apparent molecular ion (M + ) in the spectrum. In an alternative approach, 25F (as a duplex) was digested by several DNA repair enzymes with different damage specificities. Consistent with the previous reports (28, 37), the treatment with AlkA followed by Endo IV resulted in incision of 25F at the fU site (data not shown). In contrast, neither Endo III, Fpg, nor Endo IV incised 25F 2 , supporting the absence of base damage other than fU in 25F. On the basis of the results from the composition analysis and the treatment with repair enzymes, we concluded that fU was successfully incorporated into oligonucleotides in the present procedure of synthesis.
Translesion DNA Synthesis at the fU Site-----To clarify whether fU present in template DNA constitutes a replication block, 24AF and 24AT containing fU and T, respectively, at the same site (4 nucleotides beyond the primer terminus) were primed by 32 P-labeled P10, and the templates were replicated by Pol I Kf (exo -) for up to 10 min. The resulting products were analyzed by denaturing PAGE (Fig. 2). After 3 min of incubation, the primer annealed to the undamaged template 24AT was almost completely extended to a fully replicated product (lane 2). Similarly, the primer annealed to 24AF containing fU was mostly extended to fully replicated and one nucleotide shorter products after 3 min of incubation (lane 6). In addition, very weak bands also appeared at and one nucleotide prior to the fU site, indicating a pause of DNA synthesis at these sites. Quantification of the arrested and bypassed products showed that 91% of the original primer was extended beyond the fU site at 3 min. This result indicates that fU in template DNA allows efficient translesion DNA synthesis.
Similar results were obtained with templates 24CF and 24CT (data not shown).
Nucleotides Incorporated opposite fU-----Since efficient translesion DNA synthesis occurred at the fU site, the nucleotide incorporated opposite this lesion was analyzed by a primer extension assay. In this assay, primers (13T, 13C, 13G, 13A) by guest on July 9, 2020 http://www.jbc.org/ Downloaded from that were one nucleotide shorter than the template fU site were annealed to appropriate templates (24AF, 24GF, 24CF, 24TF) and the primers were extended by Pol I Kf showing that the misincorporation frequency of dCMP and dTMP was below the detection limit under these conditions. Essentially similar results were obtained with other template/primers containing G:C (Fig. 3B), C:G (Fig. 3C), and T:A (Fig. 3D) as the nearest neighbor base pairs. Accordingly, fU directed incorporation of correct dAMP and to a less extent incorrect dGMP but not pyrimidine nucleotides (dCMP and dTMP).  Table II summarizes the parameters (V max and K m ) and efficiencies (f A = V max /K m ) of dAMP incorporation (average of two experiments). Although there were variations depending on the nearest neighbor base pair, the V max values for dAMP incorporation were consistently higher for T than fU. Conversely, the K m values for T were consistently lower than for fU.

Parameters of dAMP and dGMP Incorporation-----For
Consequently, the incorporation efficiency of dAMP (f A ) opposite fU was reduced to 1/4-1/2 of that opposite T. The efficiency difference between T and fU was not large but significant, showing that conversion of T to fU in template DNA slows down incorporation of the correct nucleotide dAMP. Table III summarizes Tables II (f A ) and III (f G ). The values of f RE for fU were consistently in a 10 -4 range, whereas those for T were in a 10 -5 range. Thus, the increase in f RE due to the substitution of fU for T was 3.1-to 14.3-fold (Fig. 4A). These increases arose from reduced f A (Table II) and increased f G (Table III). Comparison of the parameters (V max and K m ) in Tables II and III also indicated that discrimination of the nucleotide at the fU site originated from V max and K m . The averaged discrimination factors for V max and K m were 16 and 360, respectively, which were calculated from the ratio of the parameters for dAMP vs. dGMP incorporation. Therefore, the contribution of K m was much greater than that of V max .

pH Effects of dAMP and dGMP Incorporation opposite fU-----We have
previously reported pH-dependent misincorporation of fdUTP opposite template G by DNA polymerase and have pointed out the importance of an ionized (or enolate) form of fU in fU:G mispair formation (27). To ask whether this mispairing scheme also held for fU in template DNA, the pH effect on the dGMP misincorporation was analyzed.
Template/primers containing fU (24AF/13T) or T (24AT/13T) were incubated with Pol by guest on July 9, 2020 http://www.jbc.org/ Downloaded from I Kf (exo -) and a single dNTP (dGTP or dATP, 20 µM) at pH 7.2-8.6. The percentage of the extended primer resulting from incorporation of dAMP or dGMP was determined by PAGE analysis (Fig. 5). Incorporation of dAMP was virtually unaffected by the pH change and was less efficient for fU than T (Figs. 5A and 5C). In contrast, incorporation of dGMP opposite fU showed a clear pH-dependence and the amount of dGMP increased with increasing pH (Figs. 5B and 5D). Although dGMP incorporation opposite T was also pH-dependent, the increase with pH was extremely small (Figs. 5B and 5D). The pH effect on dGMP misincorporation opposite fU was further analyzed in a wider pH range (pH 6.9-9.3) using GTA buffer in place of Tris buffer (Fig. 5D inset). The plot of the efficiency of dGMP misincorporation against pH showed a sigmoidal curve reminiscent of a pH titration, though the efficiency was somewhat different between the GTA and Tris buffer systems. Since the pK a values of fU and T were 8.6 and 10.0, respectively, these results strongly suggest that an acidbase equilibrium of fU (Fig. 6A) is involved in the misincorporation of dGMP and the ionized (or enolate) form of fU forms a mispair with incoming dGTP during DNA synthesis (Fig. 6B left).  (Table II), whereas the corresponding V max values were several fold lower, presumably due to the difference in the incorporated nucleotide (C or A) or the sequence context. The extension of the mismatched primer termini containing fU:G and T:G were inefficient and the extension efficiency (f C ) was two or three orders of magnitude lower than that of the corresponding matched termini (Table IV). For both T and fU, discrimination of the matched and mismatched termini exclusively originated from K m .

Extension of Matched and Mismatched Primer
Interestingly, the mismatched primer termini containing the fU:G pair was extended with significantly higher efficiencies than those containing the T:G pair. The mismatch extension frequency (f EX = f C (mismatched terminus)/f C (matched terminus) for the same template) was in a 10 -2 range for fU, whereas that for T was in a 10 -3 range. The increase in f EX associated with the substitution of fU for T was 8.2-fold (24AF vs. 24AT) and 11.3-fold (24CF vs. 24CT) (Fig. 4B). Accordingly, conversion of T to fU in template DNA promotes not only misincorporation of dGMP (Fig. 4A) but also elongation of the resulting mismatched primer termini (Fig. 4B). by guest on July 9, 2020 http://www.jbc.org/ Downloaded from DISCUSSION fU is one of the major oxidative thymine lesions found in DNA and nucleoside that were exposed to ionizing radiation (44-46), Fenton type reactions (46, 47), photosensitized reactions (48,49), and peroxy radicals (50). The yield of fU in Fenton type reactions and γ-irradiation is comparable to those of 8-oxoG (47) and 5hydroxypyrimidines (45) that are known as major mutagenic oxidative base lesions (51)(52)(53)(54)(55). Bacterial (28, 37, 56) and mammalian (57, 58) cells contain repair enzyme or activity that excises fU from damaged DNA, implying potential genotoxic influences of this lesion in vivo. Direct incorporation of fdUTP into permeated E. coli cells resulted in a small but significant increase in chromosomal lacI mutation with G:C→A:T transitions being most preferred (59). fdU was also mutagenic to Salmonella typhimurium when added to the culture medium (44). In the present study, we have assessed the genotoxic potential of fU in template DNA by utilizing in vitro DNA replication reactions. The product analysis of translesion synthesis revealed that fU constituted very weak blocks to DNA synthesis (Fig. 2). Thus, unlike other thymine lesions such as thymine glycol, urea residues, and 5-hydroxy-5-methylhydantoin (13)(14)(15)24), fU will exert a weak to negligible cytotoxic effect due to inhibition of DNA replication in vivo. Conversely, fU was shown to be a potentially mutagenic lesion based on the following results. First, the substitution of T by fU promoted misincorporation of incorrect dGMP (1.1-5.6 times as f G ) and at the same time retarded incorporation of correct dAMP (1/4-1/2 as f A ), hence leading to 3.3-to 14.3-fold increases in the replication error frequency (f RE ) relative to T (Fig. 4A). Secondly, the resulting mismatched primer terminus containing an fU:G pair was more readily extended (8.2-11.3 times as f EX ) than that containing a T:G pair (Fig. 4B) (see also the discussion below on f EX ). This step will affect the probability that a genome DNA molecule is replicated to completion, and thereby scored as mutation. According to the present data, fU is moderately mutagenic, but for more quantitative estimation of the mutation frequency of fU, it is necessary to consider the influence of repair and the property of replicative DNA polymerases.
The mismatch extension frequency (f EX ) in the absence of proofreading depends explicitly on the binding constant of DNA polymerase to matched vs. mismatched template/primer DNA as well as on the concentrations of the template/primer DNA and next-correct dNTP (60). Accordingly, to evaluate the intrinsic mismatch extension frequency (f EX(int) ), possible differential binding of DNA polymerase to matched and mismatched primer termini need to be taken into account under standing start conditions, i.e. differential binding to T:A vs. T:G and fU:A vs. fU:G termini in this study. The relationship between f EX and f EX(int) has been formulated in equation (1) (60), where [D r ] and [D w ] are the concentrations of template/primer DNA having correctly and incorrectly paired primer termini (D r and D w ), respectively, and K r and K w are the equilibrium constants for dissociation of polymerase-D r and polymerase-D w complexes, respectively.
In equation (1), it is generally assumed that the affinity of DNA polymerase for a correctly paired terminus is similar to or higher than that for an incorrectly paired  Table IV approximately represents the intrinsic value. According to equation (1), the largest discrepancy between f EX and f EX(int) occurs when K w is much higher than [D w ] (K w >> 33 nM). In this case, equation (1) can be transformed into equation (2) by approximation.
Although the K r values of Pol I (exo -) are not known, those for Avian myeloblastosis reverse transcriptase (AMV RT) and Drosophila melanogaster DNA polymerase α have been estimated as 5 nM and 20-50 nM, respectively (60). Granted that the K r value of Pol I (exo -) is in a similar range (5-50 nM), the f EX values in Table IV are subjected to up to 7.6-fold reduction. However, the correction factor given by equation (2) is by guest on July 9, 2020 http://www.jbc.org/ Downloaded from presumably similar for the T and fU templates since the affinities (K r ) of Pol I (exo -) for the structurally resembling T:A and fU:A termini are likely comparable. These considerations suggest that the relative difference in f EX(int) for the T vs. fU templates remains similar to that shown in Fig. 4B, though the absolute value of f EX(int) may be lower than that obtained experimentally in this study (f EX ).
Concerning the mispairing mechanism of fU, we have previously suggested participation of the ionized (or enolate) form of fU based on the pH-dependent misincorporation of fdUTP opposite template G (27). Privat and Sowers also proposed a similar mispairing scheme on the basis of pK a measurement of fdU and related nucleosides (61). Consistent with this mechanism, the efficiency of dGMP misincorporation opposite template fU increased around the pK a value of fU (pK a = 8.6), whereas the corresponding increase for T (pK a = 10.0) was much smaller than that for fU (Figs. 5B and 5D). The result obtained for T also agrees with a small increase in the base substitution frequency of Pol I (exo -) in this pH range (62). Unlike thymine bearing an electron donating methyl group, fU has an electron withdrawing formyl group that promotes ionization of fU in an acid-base equilibrium (Fig. 6A).
According to the pK a values, the fraction of the ionized form of fU increases form 4% to 50% in the pH range of 7.2-8.6, but that of thymine is virtually negligible (0.2% to 4%). Ionized fU can form a base pair with incoming dGTP through two hydrogen bonds (Fig. 6B left). The base pair formed between ionized fU and dGTP essentially assume Watson-Crick geometry (or B form geometry) and can fits into the active site of DNA polymerase. Since the geometric recognition is key to discrimination of correct vs. incorrect nucleotides by DNA polymerases (1), this geometry probably promoted misincorporation of dGMP opposite fU. Participation of base ionization promoted by electron withdrawing substituents has been demonstrated in the mispairing of 5halogenated uracils (BrU and 5-fluorouracil) with G (32). Thus, fU and 5-halogenated uracils share a common mutation mechanism. Although participation of a rare enol tautomer of fU (Fig. 6A) in the mispairing with G can not be fully ruled out, recent NMR studies show that the tautomeric equilibrium between keto and enol forms of fU is not significantly affected by oxidation of the methyl group of T to the formyl group (63,64). Therefore, involvement of the enol form of fU is unlikely in the mispairing with G. It is assumed that after dGMP incorporation, the resulting fU(ionized):G pair in Watson-Crick geometry shifts to wobble geometry (fU(keto):G) due to the acid-base equilibrium (Fig. 6B right). However, a certain fraction of the base pair will still exist as an fU(ionized):G pair whose geometry can again promote incorporation of the next nucleotide. Probably this is the reason why the mismatched primer terminus containing an fU:G pair was more efficiently extended than that containing a T:G pair. Although there is no experimental evidence that directly shows the equilibrium between fU(keto):G and fU(ionized):G base pairs in duplex DNA, the presence of such a equilibrium has been demonstrated for BrU(keto):G and BrU(ionized):G pairs in a duplex oligonucleotide by the nuclear magnetic resonance (NMR) study (65).
According to the proposed mutation mechanism for fU, it is reasonable that 5hydroxymethyluracil, another methyl oxidation product of T, does not direct misincorporation (25,26) since the hydroxymethyl group has electron donating nature and can not promote ionization of the base. Although a mutation mechanism involving an altered acid-base equilibrium has been previously demonstrated for 5-halogenated uracils (32), to our knowledge, fU is the first example adapting to this mechanism among oxidative DNA base lesions.
To assess the sequence context effect on the base pairing property of fU, the 3'nearest neighbor base of template fU and the paired base (i.e. the primer terminus base pair) was systematically changed and the nucleotide incorporated opposite fU was analyzed. fU with the four possible nearest neighbor base pairs directed incorporation of dAMP and to a less extent dGMP but not dCMP and dTMP (Fig. 3) (Table V), suggesting a fundamental discrepancy in the experimental set up used in the present and their studies. We repeated this experiment using the same template/primer (i.e. template 1 and primer 3 shown in Table V). However, the reported result was not reproduced and misincorporation of only dGMP was detected (Table V). Thus, T in this sequence context was not particularly prone to incorporate dCMP. We also repeated another control experiment with Tth DNA polymerase under the reported conditions (29, 30).
Template 1 (see Table V for the sequence) was annealed to a 9-mer primer (5'-TGCAGGTCG) and primer extension assays were performed in the presence of a single dNTP at 74 °C for 5 or 10 min. Although they observed incorporation of dAMP opposite T under these conditions, we did not see incorporation of any nucleotides. We  and alkaline phosphatase, and subjected to HPLC analysis as described above. Note that fdU before incubation was eluted as a single peak (not shown) and no adduct (*) was observed.  Table III).