ABSTRACT
5-Formyluracil (fU) is one of the thymine lesions produced by reactive oxygen radicals in DNA and its constituents. In this work, 5-formyl-2'-deoxyuridine 5'-triphosphate (fdUTP) was chemically synthesized and extensively purified by HPLC. The electron withdrawing 5-formyl group facilitated ionization of fU. Thus, pKa of the base unit of fdUTP was 8.6, significantly lower than that of parent thymine (pKa = 10.0 as dTMP). fdUTP efficiently replaced dTTP during DNA replication catalyzed by Escherichia coli DNA polymerase I (Klenow fragment), T7 DNA polymerase (3'-5' exonuclease free) and Taq DNA polymerase. fU-specific cleavage of the replication products by piperidine revealed that when incorporated as T, incorporation of fU was virtually uniform, suggesting minor sequence context effects on the incorporation frequency of fdUTP. fdUTP also replaced dCTP, but with much lower efficiency than that for dTTP. The substitution efficiency for dCTP increased with increasing pH from 7.2 to 9.0. The parallel correlation between ionization of the base unit of fdUTP (pKa = 8.6) and the substitution efficiency for dCTP strongly suggests that the base-ionized form of fdUTP is involved in mispairing with template G. These data indicate that fU can be specifically introduced into DNA as unique lesions by in vitro DNA polymerase reactions. In addition, fU is potentially mutagenic since this lesion is much more prone to form mispairing with G than parent thymine.
Reactive oxygen species generated inside cells by various exogenous and endogenous agents attack DNA, thereby altering the genetic information carried by DNA. Thymine is the most susceptible target among the DNA bases toward ionizing radiation (1 ,2 ), a typical source of the reactive oxygen species such as hydroxyl radicals, hydrogen peroxide and superoxide anion radicals. The oxygen species, most likely hydroxyl radicals, react with thymine in DNA generating different classes of oxidative base modifications. These include 5,6-saturated products like thymine glycols, ring fragmentation products like urea residues, and oxidation products of the 5-methyl group (3 ). The genotoxic and biochemical effects of the 5,6-saturated and ring fragmentation lesions have been directly demonstrated using DNA (4 -9 and references therein) or deoxyribonucleoside 5'-triphosphates (10 -12 ) containing these lesions. In contrast, limited information is available on the effects of the 5-methyl oxidation products. Until now, 5-hydroxymethyluracil (hmU) and 5-formyluracil (fU) have been identified as stable 5-methyl oxidation products of thymine (13 -15 ). It appears that hmU is an innocuous lesion since likewise thymine, it pairs with adenine during DNA replication (16 , for some controversial data, see 17 ). The innocuous nature of hmU is also supported by the following observation. hmU is a natural DNA component of dinoflagellates (18 ) and Bacillus subtillis phages (19 ): it replaces 12-68% of thymine in dinoflagellates and even 100% in B.subtillis phages. In addition, when Tetrahymena is grown in a medium containing the deoxyribonucleoside form of hmU, up to 30% of thymine in DNA is replaced by hmU without apparent toxic effects (20 ).
fU is known to be produced by ionizing radiation (13 ,15 ) as well as photosensitized oxidation (21 ,22 ). Studies on the genotoxic effects of fU is quite limited (15 ), probably due to lack of the method to specifically introduce this lesion into DNA (23 ). Recent studies have revealed that fU is removed from DNA by 3-methyladenine DNA glycosylase II (AlkA protein) (24 ) and crude extracts from mammalian livers or human cells (25 ,26 ), implying some unfavorable genetic, structural or biochemical effects associated with this lesion. One possibility of such unfavorable effects is that the electron withdrawing property of the 5-formyl group in fU facilitates ionization of the base, thereby altering the base pairing specificity. This type of effect has been well demonstrated for 5-halogenated uracils. Ionization of 5-bromouracil (BrU), for example, promotes mispairing with G, generating a transition mutation via a noncanonical BrU:G base pair (27 ,28 and references therein). Another possibility is cross-link formation between the formyl group and amino groups of certain cellular constituents such as proteins (15 ,29 ,30 ). In light of these potential genotoxic effects of fU, it is appropriate to establish the base pairing specificity of fU and the method to prepare DNA containing fU as unique lesions.
In this work, we have synthesized 5-formyl-2'-deoxyuridine 5'-triphosphate (fdUTP) as a substrate for DNA polymerases. We show here that DNA polymerases utilize fdUTP as a substrate, thus making it possible to specifically introduce fU lesions into DNA by in vitro DNA polymerase reactions. Base pairing properties of fU are also studied by examining whether fdUTP can substitute for normal dNTPs during the replication of DNA templates in vitro.
5-formyluracil and 5-formyl-2'-deoxyuridine were synthesized according the reported procedures (31 ,32 ) and their structures were confirmed by UV and 1H, 13C NMR measurements (for UV and 1H NMR data, see 32 -34 ). DEAE-Sephadex A-25 and Dowex 50W-X2 were products of Pharmacia and Dow Chemical respectively. dATP, dGTP, dCTP and dTTP were purchased from Pharmacia and their purity was >99.3%. The primer complementary to the positions 6311-6327 in M13mp18 (PM13: 5'-GGTTTTCCCAGTCACGA) was synthesized on an automated DNA synthesizer and purified by reversed-phase HPLC (35 ). PCR primers (PF: 5'-TGTAAAACGACGGCCAGT, PR: 5'-CAGGAAACAGCTATGACC) to amplify the 194 bp region containing the multicloning site of pBluescript II SK+ were obtained from Japan Bioservice and used without purification. Escherichai coli DNA polymerase I (Klenow fragment) and T4 polynucleotide kinase were obtained from New England Biolabs. T7 DNA polymerase deficient in 3'-5' exonuclease (Sequenase Ver 2) and Taq DNA polymerase were from USB and Perkin Elmer respectively. [[gamma]-32P]ATP (185 TBq/mmol) was purchased from Amersham and [2,8-3H]dATP (0.71 TBq/mmol) from Moravek Biochemicals. M13mp18 single-stranded DNA and plasmid pBluescript II SK+ were prepared by standard procedures (36 ).
1H and 13C NMR spectra were recorded on a Varian Associates Gemini 200 spectrometer at 200 and 50.3 MHz respectively. 31P NMR spectra were obtained on a Bruker ARX-500 spectrometer at 202.5 MHz. UV spectra of 5-formyl-2'-deoxyuridine and its 5'-triphosphate were taken in 16 mM phosphate buffer (pH 4.5-11.5) on a Shimadzu UV-260 spectrophotometer at room temperature. Other UV spectra were recorded on a Beckman DU-68 spectrophotometer.
Reversed-phase HPLC was performed on a system consisting of Tosoh computer controlled CCPM pumps, a Tosoh UV-8000 photodiode array detector, and a System Instruments Chromatocorder 12. For an analytical purpose, samples were separated on a Wako WS-DNA C18 column (4.6 * 150 mm) using a gradient of acetonitrile in 0.1 M triethylammonium acetate (pH 7.0) with a flow rate 0.8 ml/min. The acetonitrile percent in the gradient was 1.5% (0-5 min), 1.5-10% (5-30 min, linear gradient), 10-1.5% (30-31 min). For preparative separation, a semipreparative column (8 * 300 mm) containing the same packing material as the analytical column was used. The flow rate was 1.8 ml/min and the following acetonitrile gradient was applied: 1.5% (0-10 min), 1.5-4% (10-40 min, linear gradient), 4-10% (40-41 min), 10% (41-51 min). Ion-exchange HPLC was performed using a system similar to that described for reversed-phase HPLC except that a Wako 5SAX column (4.6 * 250 mm) and solvents A (7 mM KH2PO4, pH 4.0) and B (0.25 M KH2PO4, pH 4.5, 0.6 M KCl) were used. Samples were eluted by a gradient of solvent B with a flow rate 1.0 ml/min. The percent of solvent B was 20-70% (0-25 min, linear gradient), 70% (25-30 min), 70-20% (30-31 min).
Images of autoradiograms were recorded on a Hewlett Packard ScanJet 3c scanner using the DeskScan II software. Intensity of the bands in the recorded images was analyzed by the NIH Image (Ver 1.55) using a built-in gel scan macro.
fdUMP was prepared following the reported method (37 ) with a slight modification. 2'-Deoxyuridine 5'-monophosphate (dUMP) was converted to 5-acetoxymercuri-2'-deoxyuridine 5'-monophosphate, then to 5(E)-styryl-2'-deoxyuridine 5'-monophosphate (SdUMP) (37 ,38 ). Oxidative cleavage of the 5-vinylic bond of SdUMP yielded fdUMP. In our modified procedure, sulfate present in the reaction mixture was removed from the reaction mixture by precipitation with 1 equivalent of BaCl2 prior to DEAE-Sephadex A-25 chromatography. Otherwise, a considerably large amount of sulfate was contaminated in fdUMP after the chromatographic purification. Triethylammonium salt of fdUMP was finally converted to the free acid by passing through a Dowex 50W-X2 (H+ form) column. The yield of fdUMP from dUMP was 20%: UV [lambda]max (10 mM phosphate buffer, pH 7.0) 278 nm, [lambda]min 249 nm; 1H NMR (D2O, TSP [delta] = 0 p.p.m.) [delta] 9.62 (s, 1H, CHO), 8.74 (s, 1H, H-6), 6.27 (t, 1H, J = 6.3 Hz, H-1'), 4.57 (m, 1H, H-3'), 4.28 (m, 1H, H-4'), 4.16 (m, 2H, H-5'a,b), 2.62-2.36 (m, 2H, H-2'a,b); 13C NMR (D2O, TSP [delta]= 0 p.p.m.) [delta] 192.5 (CHO), 165.2 (C-4), 155.9 (C-2), 153.2 (C-6), 114.6 (C-5), 90.2 (C-1'), 89.2 (d, C-4'), 73.5 (C-3'), 67.5 (d, C-5'), 42.9 (C-2'). These UV and 1H NMR data are comparable with those reported (39 ).
fdUMP was activated by N,N-carbonyldiimidazole (40 ) and converted to the corresponding 5'-triphosphate by condensation with pyrophosphate. Crude fdUTP was subjected to DEAE-Sephadex A-25 chromatography (2.5 * 17 cm) using a linear gradient of triethylammonium bicarbonate (0.01-0.5 M, 800 ml) followed by isocratic elution (0.5 M, 800 ml). Main peak fractions (monitored by UV absorption at 280 nm) eluted in the isocratic elution were combined and evaporated under vacuum. The residue was coevaporated with water three times to remove triethylammonium bicarbonate. This procedure yielded 3022 OD280 of fdUTP, which corresponded to 58% yield from fdUMP based on [epsilon]280 of fdUTP = 13 400. For DNA polymerase reactions, fdUTP was further purified by preparative reversed-phase HPLC as described above. fdUTP (triethylammonium salt): UV [lambda]max (10 mM phosphate buffer, pH 7.0) 278 nm, [lambda]min 249 nm; 1H NMR (D2O, TSP [delta] = 0 p.p.m.) [delta] 9.62 (s, 1H, CHO), 8.69 (s, 1H, H-6), 6.29 (t, 1H, J = 6.2 Hz, H-1'), 4.67 (q, 1H, H-3'), 4.31-4.22 (m, 3H, H-4'+H-5'a,b), 2.53-2.37 (m, 2H, H-2'a,b), triethylamine 3.27 (CH2) and 1.25 (CH3) p.p.m.; 13C NMR (D2O, TSP [delta] = 0 p.p.m.) [delta] 193.8 (CHO), 171.2 (C-4), 157.5 (C-2), 156.1 (C-6), 114.8 (C-5), 89.6 (C-1'), 88.6 (d, C-4'), 72.5 (C-3'), 67.6 (d, C-5'), 42.4 (C-2'), triethylamine 61.8 (CH2) and 10.4 (CH3) p.p.m.; 31P NMR (D2O, external 85% H3PO4 [delta] = 0 p.p.m.) -2.08 (d, [gamma]-P), -7.57 (d, [alpha]-P), -17.83 (m, [beta]-P).
Primer PM13 was 5'-end-labeled by [[gamma]-32P]ATP and T4 polynucleotide kinase as described (35 ). To examine if fdUTP could substitute for normal four dNTPs, -A, -G, -C, -T reactions (41 ), where one of four normal dNTPs was missing in the reaction mixture, were performed in the absence and presence of fdUTP. M13 DNA (0.2 pmol) primed with PM13 primer (template:primer molar ratio = 3:1) in Pol I reaction buffer (20 [mu]l) was replicated by Pol I Kf (0.1 U) in the presence of three normal dNTPs (50 [mu]M each) +- fdUTP (50 [mu]M) at 37oC and various pH for 30 min. Pol I reaction buffer consisted of 66 mM Tris-HCl (pH 7.2-9.2), 1.5 mM mercaptoethanol, 50 [mu]g/ml BSA and 6.6 mM MgCl2. For a control reaction, M13 DNA was replicated in the presence of four normal dNTPs (50 [mu]M each) under the same conditions (pH 7.8). The reaction was terminated by adding loading buffer containing 0.05% xylene cyanol, 0.05% bromophenol blue, 20 mM EDTA and 98% formamide. The sample was boiled and an aliquot (2 [mu]l) was subjected to 8% denaturing polyacrylamide gel electrophoresis. Electrophoresis was performed at 1800 V and the gel was autoradiographed at -80oC overnight. Reaction conditions for T7(exo-) were essentially the same as those for Pol I Kf except that 1 unit of the enzyme was used.
To estimate the substitution efficiency of fdUTP for dTTP or dCTP, -T and -C reactions were carried out in the presence of [3H]dATP. The extent of DNA synthesis was quantitated based on the amount of [3H]dAMP incorporated into newly synthesized products. M13 DNA template/PM13 primer (0.25 pmol) in Pol I reaction buffer (50 [mu]l) was incubated with Pol I Kf (2.5 U) in the presence of three normal dNTPs (50 [mu]M each) +- fdUTP (50 [mu]M) at 37oC and pH 7.8 or 9.0. The specific activity of [3H]dATP in the three dNTP mix was 16.6 Bq/pmol. At appropriate incubation time, an aliquot of the reaction mixture (8 [mu]l) was taken, mixed with 100 mM EDTA (8 [mu]l) and spotted onto a Whatman GF/C glass filter. The filter was washed with 5% trichloroacetic acid, ethanol, finally with ether and dried. The filter was soaked in a scintillation cocktail and radioactivity was counted on a liquid scintillation counter.
Polymerase chain reactions (PCR) were performed as follows. pBluescript II SK+ (3 pmol) was mixed with PF and PR primers (20 pmol each), appropriate dNTPs (250 [mu]M each) and Taq DNA polymerase (1 U) in reaction buffer (20 [mu]l) containing 50 mM KCl, 1.5 mM MgCl2 and 10 mM Tris-HCl, pH 8.3. The reaction was carried out on a Perkin Elmer GeneAmp PCR System Model 2400 in the presence of four normal dNTPs or three dNTPs (A, G, C) +- fdUTP for 30 cycles of 94oC for 1 min, 55oC for 15 s and 75oC for 2.5 min. The sample (10 [mu]l) was mixed with loading buffer containing 30% glycerol, 0.25% bromophenol blue, 0.25% xylene cyanol and separated on a 3% agarose gel (Agarose X, Wako). The gel was stained by ethydium bromide and photographed.
To determine the sequence context effects on the incorporation of fdUMP into DNA, primed M13 DNA was replicated by Pol I Kf in the presence of four normal dNTPs (50 [mu]M each) + fdUTP (2.5 [mu]M) at pH 7.8 as described above. The reaction was terminated by addition of EDTA (final concentration 10 mM). The reaction mixture was extracted by phenol, and DNA was recovered by ethanol precipitation. DNA was treated by 10% piperidine (100 [mu]l) in a tightly sealed test tube at 90oC for 30 min. Piperidine was evaporated under vacuum and the residue was washed by coevaporation with water (100 [mu]l) twice. To create reference sequence ladders, M13 DNA replicated with four normal dNTPs were subjected to standard Maxam-Gilbert reactions (42 ). Samples were dissolved in loading buffer and electrophoresed as described for minus dNTP reactions.
fdUTP was synthesized from dUMP in three steps. Spectrometric data including the UV, 1H, 13C, 31P NMR spectra were consistent with the structure of fdUTP. NMR signals at 9.62 p.p.m. (1H) and 193.8 p.p.m. (13C) of fdUTP clearly demonstrated the presence of the 5-formyl group. 31P NMR signals corresponding the [alpha]-, [beta]-, [gamma]-phosphates also confirmed the 5'-triphosphate group of fdUTP.
fdUTP obtained by DEAE-Sephadex A-25 chromatography was further purified by HPLC in the final purification step. Reverse-phase HPLC showed better separation than ion-exchange HPLC with respect to the modification at the 5 position. In a reversed-phase mode, dUTP (5-substituent = H), dTTP (CH3), fdUTP (CHO) were eluted in separate peaks (Fig. 1 A), whereas these triphosphates were not resolved in an ion-exchange mode (Fig. 1 D). Thus, fdUTP was purified by reversed-phase HPLC in the final step. The final fdUTP preparation was highly pure as can be seen from the analytical HPLC profile (Fig. 1 C). fdUTP was also treated by alkaline phosphatase and products were analyzed by reversed-phase HPLC. This treatment resulted in quantitative conversion of fdUTP to 5-formyl-2'-deoxyuridine (fdU), further confirming the structure and purity of fdUTP (data not shown).
pKa values of the base moiety of fdU and fdUTP were determined by pH titration of UV spectra (220-350 nm) of these compounds in 20 mM phosphate buffer. The pH titrated UV spectra displayed isosbestic points at 257 and 297 nm for fdU, and 254 and 292 nm for fdUTP. Figure 2 A shows the plots of the absorbance at 280 nm (A280) against pH for fdU and fdUTP. The shapes of the A280 versus pH plots fitted typical pH titration curves, from which pKa values of the base unit of fdU and fdUTP were determined by a graphical method. These data are listed in Table 0 along with those of 5-substituted uracil derivatives for comparison. The substitution of the formyl group for CH3 (thymidine) or H (uridine) at the position 5 reduced pKa by 1.1-1.5 units (Table 0 ). Similar results were obtained for nucleotides. The pKa values of the base unit of fdU (8 .2 ) and fdUTP (8 .6 ) were close to those of 5-bromouracil derivatives (Table 1 ).
To examine whether fdUTP could substitute for dTTP or the other dNTPs during DNA synthesis, the minus dNTP reaction developed by Topal et al. (41 ) was used. Primed M13 DNA were replicated by Pol I Kf in the presence of three normal dNTPs without or with fdUTP. Products were analyzed by denaturing polyacrylamide gel electrophoresis (Fig. 3 ). In -T reactions, elongation of the primer was terminated in a few nucleotides from the primer terminus due to lack of dTTP (Fig. 3 D, lanes with even numbers except for lane 16). However, the primer was efficiently extended in the presence of fdUTP (Fig. 3 D, lanes with odd numbers except for lane 1). The extended products in -T+fdUTP reactions were comparable with those obtained by a control reaction with four normal dNTPs (lane 16). These results indicate that fdUTP can efficiently substitute for dTTP during the replication of DNA. Comparison of -C reactions in the absence and presence of fdUTP revealed that fdUTP could also substitute for dCTP (Fig. 3 C). However, fdUTP replaced dCTP less efficiently than dTTP as can be seen from the primer elongation in -T and -C reactions with fdUTP (Fig. 3 C and D). The substitution efficiency of fdUTP for dCTP increased with increasing pH of the reaction mixture from 7.2 to 9.0, exhibiting a parallel correlation with the acid-base equilibrium of the base moiety of fdUTP (pKa = 8.6) shown in Figure 2 . Primer extension was partially inhibited at pH 9.2 in -C reactions. Such inhibition was also observed in -A, -G and -T reactions at pH 9.2. Presumably, the polymerase activity of Pol I Kf was partially inhibited at this pH. In contrast to -T and -C reactions, fdUTP did not replace dGTP and dATP so that extension of the primer in -G and -A reactions was not observed in the presence of fdUTP under these conditions (Fig. 3 A and B).
The substitution efficiency of fdUTP for dTTP was determined by comparing DNA synthesis in complete reactions (four dNTPs) and -T reactions with fdUTP. The extent of DNA synthesis was measured by incorporation of [3H]dAMP into newly synthesized products. The reactions were performed with Pol I Kf, and the radioactivity of incorporated [3H]dAMP is plotted against incubation time (Fig. 5 ). From the averaged slopes of the plots, it is roughly estimated that fdUTP replaces dTTP with 71 and 27% efficiencies at pH 7.8 and 9.0 respectively. We attempted to measure DNA synthesis in -C+fdUTP reactions at pH 7.8 and 9.0 to determine the substitution efficiency for dCTP. However, the amounts of the incorporated radioactivity in -C+fdUTP reactions were within the fluctuation of the background DNA synthesis in -C reactions both at pH 7.8 and 9.0 (data not shown). Thus, quantitative data on the substitution efficiency of fdUTP for dCTP were not obtained.
M13 DNA was replicated by Pol I Kf in the presence of four normal dNTPs (50 [mu]M each) + fdUTP (2.5 [mu]M) at pH 7.8 to produce DNA partially substituted by fdUMP. The products were isolated and treated by hot piperidine. Since fU is alkali labile (22 ,23 ,43 ), site-specific DNA cleavage occurs at fdUMP sites. Figure 6 shows the results of polyacrylamide gel analysis of the piperidine-treated products (lane 3) along with standard Maxam-Gilbert reference ladders (lanes 1 and 2). The bands arising from incorporated fdUMP appeared at all T sites. The intensities of the bands were essentially similar, and this was further confirmed by densitometric analysis of the autoradiogram (Fig. 6 B). These data indicate that fdUMP is incorporated primarily in place for dTMP under the present conditions. Furthermore, they suggest that, if not at all, the surrounding sequence contexts exert relatively minor effects on the incorporation frequency of fdUMP. Using reaction products prepared at pH 9.0, where fdUTP substituted for dCTP more efficiently than pH 7.8, an attempt was made to reveal the piperidine cleavage bands arising from incorporation of fdUMP at C sites as well as T sites. However, clear bands indicating fdUMP incorporation at C sites were not detected over the background. Probably, the incorporation frequency of fdUMP at C sites was too low to be detected by this method.
In the present study, fdUTP was chemically synthesized from dUMP and extensively purified by HPLC. The level of contaminating nucleotides was below the detection limit of the present HPLC analysis (Fig. 1 ). High purity of fdUTP is essential to study base pairing specificity since the minus reaction method used here is extremely sensitive to contaminating nucleotides. Only a few nucleotide extension of the primer over the background due to contaminating species could lead to erroneous conclusions on the base pairing specificity.
Armstrong et al. (44 ,45 ) synthesized 5-formyluridine 5'-triphosphate (fUTP), a ribonucleotide counter part of fdUTP, using an approach different from this study. In their method, 5-formyl- 2',3'-isopropylideneuridine was phosphorylated by [beta]-cyanoethylphosphate. Subsequently, the product was treated by 4N NaOH-methanol followed by 50% acetic acid at 100oC for 2 h to remove protecting groups. However, such harsh treatments can not be used for the deoxy analog since anomerization and hydrolysis of the N-glycosidic bond catalyzed by base and acid respectively, become serious problems. The authors further showed that fUTP could replace UTP in the replication of poly(dA-dT) by RNA polymerase. However, the deoxy counter part, fdUTP, was not synthesized so that the substrate properties of fdUTP for DNA polymerases was not tested in their study. We attempted to synthesize fdUTP using 5-formyl-2'-deoxyuridine (fdU) as a starting material (39 ). For an unknown reason, the yield of fdUMP by phosphorylation of 5'-OH with 2,2,2-tribromoethyl phosphoromorpholinochloridate was too low to be used in the subsequent reactions. Recently, preparation of oligonucleotides containing fdU based on the phosphoramidite method was reported (23 ). They attempted to demonstrate the presence of fdU after enzymatic digestion of the oligonucleotide. However, the results were somewhat puzzling and intact fdU was not detected without prior derivatization of fdU in the oligonucleotide. Unambiguous demonstration of totally intact fdU in the oligonucleotide remains to be seen.
In this work, we have determined pKa of the base unit of fdU and fdUTP since we thought that pKa was a key parameter for understanding the base pairing property of fU, as was indeed the case for 5-bromouracil (BrU) (27 ,28 ). Towards the end of this study, pKa of the base unit of fdU was reported by Privat and Sowers (46 ). The reported values (pKa = 8.12) agrees well with that obtained in this work (pKa = 8.2). The pKa values of fdU and fdUTP were lower than those of dT and dTMP by 1.4-1.5 units, and comparable with BrU derivatives (Table 1 ). Therefore, substitution of the electron withdrawing formyl group (fU) for the electron donating methyl group (thymine) clearly alters the electronic structure of the pyrimidine ring, thereby facilitating ionization of the base moiety. The fraction of the base-ionized form of fdUTP calculated from pKa is 4% under the physiological pH (7.2), whereas that of dTTP is 0.16% (calculated as dTMP). Accordingly, these data suggest that both neutral (keto) and ionized forms of fU need to be taken into account when studying the genetic or biochemical effects of this lesion.
We have examined the ability of fdUTP to serve as a substrate for DNA polymerases with different enzymatic properties. For example, Pol I Kf has an associated 3'-5' exonuclease activity for proofreading, but T7(exo-) and Taq DNA polymerase lack this activity. T7(exo-) and Taq DNA polymerase are highly processive, while Pol I (Kf) is a distributive enzyme (47 ,48 ). Despite such differences in the enzymatic properties, all DNA polymerases tested here utilized fdUTP in place for dTTP during DNA replication (Figs 3 and 4 ). When fdUMP was incorporated in place for dTMP by Pol I Kf, the surrounding sequence context exerted only minor effects on the incorporation frequency of fdUMP (Fig. 6 ). These data suggest that hydrogen bonding and stacking interactions involved in an fU:A base pair are essentially similar to those of the canonical T:A base pair, and that the fU:A pair in a DNA helix does not introduce significant perturbations into surrounding DNA. When incorporated opposite template A, incoming fdUTP is likely to be stabilized through the interactions similar to dTTP and fits into the Watson-Crick geometry in the polymerase active site. The decrease in the substitution efficiency for dTTP at pH 9.0 relative to pH 7.8 (Fig. 5 ) is probably due to ionization of the base moiety of fdUTP. The negative charge on the base unit could inhibit proper binding of incoming fdUTP to the DNA-polymerase binary complex. This may also explain, at least to some extent, why the base-ionized form of fdUTP can replace dCTP only with a limited efficiency (see below).
Base pairing properties of fU with C, A and G were studied using minus dNTP reactions catalyzed by Pol I Kf and T7(exo-). Although the substitution efficiency was much lower than dTTP, fdUTP partially replaced dCTP in -C reactions (Figs 3 and 4 ). These results indicate that incoming fdUTP can potentially form a base pair with template G in the polymerase active site, thus allowing misincorporation of fdUMP opposite G. Moreover, as shown in the product analysis (Figs 3 C and 4 C), the primer terminus containing a putative fU:G mispair was extended in the subsequent polymerization reaction. The substitution efficiency for dCTP increased with increasing pH of the reaction mixture from 7.2 to 9.0 (Fig. 3 C), showing a parallel correlation with the increasing fraction of the base-ionized form of fdUTP (pKa = 8.6). The pH-dependent substitution for dCTP was also observed for T7(exo-). Thymine is known to form a T:G mismatch, but pH-dependent elongation of the primer was not observed in -C reactions without fdUTP under the present conditions. The parallel correlation between ionization of the base unit of fdUTP and the substitution efficiency for dCTP strongly suggests that the ionized form of fU is responsible for the mispairing with G, although possible participation of the enol tautomer (49 ) or the wobble base pair (50 ,51 ) can not be fully ruled out. The involvement of ionized fU is also supported by the energetic consideration: the free energy of base ionization is lower than tautomerization for 5-substituted uracils, and the energy difference between these two forms increases as pKa of 5-substituted uracils decreases (46 ,52 ). Since pKa of fU is between those of thymine and BrU, the ionized form of fU is favored over the enol form by 1.98 (thymine) -3.4 (BrU) kcal/mol. The geometry of the fU(ionized):G base pair mimics that of the C:G pair. However, unlike the canonical C:G pair involving three hydrogen bondings, it is likely that the fU(ionized):G pair forms only two hydrogen bondings (Fig. 2 B). Thus, ionization of the fU moiety appears to have two consequences: it increases the chance for mispairing with G through two hydrogen bondings, but also inhibits proper binding of incoming fdUTP to DNA polymerases. The balance of these two effects determines the substitution efficiency of fdUTP for dCTP.
To generate base substitution mutations, DNA polymerases must insert an incorrect nucleotide in the first step of the reaction, then the resulting mispaired primer terminus need to be extended in the subsequent reaction. The present data show that fU is much more prone to form a mispair with G than thymine, and the primer terminus containing an fU:G mismatch is extendable, at least to a certain extent, by DNA polymerases. Based on the pKa values shown in Table 0 , it is estimated that the fractions of ionized BrU, fU and T likely responsible for mispairing with G are 7.9, 4 and 0.16% respectively, under physiological conditions (pH 7.2). A simple extrapolation of these data suggests that the mispairing frequencies for BrU:G : fU:G : T:G are 49:25:1. Fluctuations to slightly higher pH further promote the formation of the fU:G mismatch (and BrU:G) over T:G since the fraction of ionized fU (and BrU) dramatically increases around this pH region. In this context, whether resulting from oxidation of thymine in dTTP or DNA, fU is expected to be potentially mutagenic. The potential genotoxicity associated with fU may also explain why repair enzymes removing this lesion from damaged DNA exist in cells (24 ,25 ,38 ).
We thank Kenji Mitiue and Katsuhiro Matsui for the assistance in the preliminary work of this study, and Kenji Kanaori for obtaining NMR spectra. This work was supported by Grants-in-Aid from the Ministry of Education, Science and Culture, Japan (H.I.).
REFERENCES
Return