Reactivity and DNA Damage by Independently Generated 2′-Deoxycytidin-N4-yl Radical

Abstract

Oxidative stress produces a variety of radicals in DNA, including pyrimidine nucleobase radicals. The nitrogen-centered DNA radical 2′-deoxycytidin-N4-yl radical (dC·) plays a role in DNA damage mediated by one electron oxidants, such as HOCl and ionizing radiation. However, the reactivity of dC· is not well understood. To reduce this knowledge gap, we photochemically generated dC· from a nitrophenyl oxime nucleoside and within chemically synthesized oligonucleotides from the same precursor. dC· formation is confirmed by transient UV-absorption spectroscopy in laser flash photolysis (LFP) experiments. LFP and duplex DNA cleavage experiments indicate that dC· oxidizes dG. Transient formation of the dG radical cation (dG^+•) is observed in LFP experiments. Oxidation of the opposing dG in DNA results in hole transfer when the opposing dG is part of a dGGG sequence. The sequence dependence is attributed to a competition between rapid proton transfer from dG^+• to the opposing dC anion formed and hole transfer. Enhanced hole transfer when less acidic O6-methyl-2′-deoxyguanosine is opposite dC· supports this proposal. dC· produces tandem lesions in sequences containing thymidine at the 5′-position by abstracting a hydrogen atom from the thymine methyl group. The corresponding thymidine peroxyl radical completes tandem lesion formation by reacting with the 5′-adjacent nucleotide. As dC· is reduced to dC, its role in the process is traceless and is only detectable because of the ability to independently generate it from a stable precursor. These experiments reveal that dC· oxidizes neighboring nucleotides, resulting in deleterious tandem lesions and hole transfer in appropriate sequences.

Graphical Abstract

INTRODUCTION

DNA damage plays an important role in the development of genetic disease, including cancer for which it also provides the chemical foundation for many treatments. DNA damage is also increasingly recognized as the primary cause of aging.^1–3 Oxidative damage proceeds through radical and radical ion intermediates. A great deal has been learned about such processes that involve carbon-centered radicals via a variety of experimental approaches, including independent generation of individual reactive species and metastable products.^4–6 In addition to advancing basic knowledge, these discoveries have impacted biological chemistry in unexpected ways and in some instances have led to practical applications.^7–9 Although advances in our understanding of their reactivity have been reported, the roles of nitrogen-centered radicals in DNA damage are less well understood.^10–14 The reactivity of nitrogen radicals can be significantly different from those of carbon radicals, particularly in ways that are important in a biological polymer, such as DNA. For instance, nitrogen radicals react much more slowly with O₂ than do carbon radicals.^15–17 In addition, polarity and bond dissociation energy disparities can create large kinetic differences in the hydrogen atom abstraction reactions of carbon and nitrogen radicals. Herein, we report on the independent generation and reactivity of 2′-deoxycytidin-N4-yl radical (dC·, Scheme 1), a reactive species that has not been well-studied but is implicated in a variety of DNA damage processes.

Scheme 1.

Formation Pathways of dC·

Radiolytic generation of dC· from 2′-deoxycytidine (dC) in a matrix at low temperature and characterization by EPR revealed that the radical exists as the iminyl tautomer.¹⁸ Computational studies are consistent with the lower energy for this tautomer and that it may be the preferred isomer within DNA.¹⁸ dC· is produced rapidly via deprotonation of the radical cation, which is itself formed by direct ionization (Scheme 1A).^19,20 Although dC oxidation is several kcal/mol more unfavorable than that of 2′-deoxyguanosine (dG), the most readily oxidized native nucleotide, high energy radiation (e.g., X-rays, γ-radiolysis) ionizes DNA unselectively to form holes, including at dC (dC^•+).^2,21,22 The migration and ultimate trapping of holes in DNA at dG (dG·) have also been proposed to involve dC· (Scheme 1B).^23,24 Chloramine 1, which is formed from reaction of HOCl that is a product of inflammation, decomposes to generate dC· (Scheme 1A).^25,26 The photochemical generation of unusual photodamage products in which dC forms intrastrand cross-links with adjacent dC or dG has also been ascribed to the intermediacy of dC·.^27,28

RESULTS AND DISCUSSION

Independent Generation of dC· Nucleoside and Its Solution Phase Reactivity.

Our goal was to identify a photochemical precursor to dC·, which could be irradiated at >300 nm (to minimize DNA damage), was sufficiently efficient so as to be useful in laser flash photolysis experiments, and would be suitable for DNA incorporation. dC· was previously generated photochemically from an oxime ester (2).²⁸ However, long irradiation periods were employed and the oxime ester was unstable in our hands to solid-phase oligonucleotide synthesis deprotection conditions. Other nitrogen-centered nucleoside radicals have been generated via tandem Norrish type I, β-fragmentation reactions.^14,16 dC·generation from a comparable precursor (3) was inefficient (data not shown). Consequently, we sought a precursor that, like 2 and 3, contains a weak nitrogen–oxygen bond but would meet the criteria mentioned above.

The nitro-substituted oximes (6a,b) were synthesized via nucleophilic aromatic substitution by 4²⁸ (Scheme 2), followed by desilylation. The bis-TBDMS protected oximes (5a,b) were readily purified, but carrying out desilylation on crude 5a,b also worked well. The deprotected oximes were present as a single tautomer in d₆-DMSO (Figure S1), whose identity is based upon those of related molecules (e.g., 2).^16,28 The nitrophenyl oxime nucleosides absorb strongly above 300 nm; 6a and 6b exhibit λ_max at 332 (ε = 2.1 × 10⁴ M⁻¹ cm⁻¹; DMSO/MeOH, 1:9 (v:v)) and 325 nm (ε = 2.0 × 10⁴ M⁻¹ cm⁻¹; DMSO/MeOH, 1:9 (v:v)), respectively (Figure S2).

Scheme 2. Nitro-Substituted Oxime Precursor Synthesis.

^aKey: (a) (i) NaH ii, (a) 1-fluoro-4-nitrobenzene; (b) 1-fluoro-2,4-dinitobenzene; (b) Et₃N·3HF.

The utility of 6a,b (0.1 mM) as photochemical sources of dC· was evaluated via quantitative HPLC product analysis, following broad band irradiation (λ_max = 350 nm) in a mixture of pH 7.2 phosphate (10 mM) buffered saline (100 mM), MeOH (10% v:v) and DMSO (1% v:v) (Table 1). Photolysis of the dinitro precursor (6b) under anaerobic conditions was essentially complete (98% conversion) within 10 min. Complete conversion was achieved by extending the irradiation to 30 min and resulted in quantitative yield (within experimental error) of dC. High yields of dC were also obtained from photolyses of 6a under degassed conditions, but complete conversion required more than 30 min of irradiation (Table S2). The presence of β-mercaptoethanol (BME, 10 mM) during anaerobic photolyses of 6a,b had no effect on the already high yield of dC. Although MeOH is presumed to be the hydrogen atom source, the exchangeable nature of the amino protons in dC precluded probing this using isotopic labeling.

Table 1.

dC Yields from Nitro-oxime Photolyses^a

substrate	O2	[BME] (mM)	% yield dCb	% yield dUb
6a	−	0	89 ± 9	ND
6a	−	10	79 ± 3	ND
6a	+	0	27 ± 4	15 ± 1
6a	+	0.1	25 ± 1	13 ± 1
6a	+	1	58 ± 3	8 ± 1
6a	+	10	83 ± 2	ND
6a	+	100	110 ± 4	ND
6b	−	0	100 ± 1	ND
6b	+	0	25 ± 2	12 ± 1

^aPhotolyses were carried out in pH 7.2 phosphate (10 mM) buffered saline (100 mM), MeOH (10% v:v) and DMSO (1% v:v). Irradiation time: 6a, 60 min; 6b, 30 min.

^bYields are the average ± standard deviation of three replicates and are based upon the initial concentration of 6a,b. ND, not detected.

Although the conversion efficiency was not markedly different, the yield of dC was significantly lower following aerobic photolysis of 6a or 6b. As O₂ was not expected to react with the nitrogen radical, the effect of the efficient carbon radical trap suggested more complicated underlying chemistry.^20,29 2′-Deoxyuridine (dU) was also observed, albeit in lower yields than dC (Table 1). This product was previously reported following dC oxidation under photoinduced single electron transfer conditions.³⁰ In addition, the analogous product 2′-deoxyinosine (dI) is formed from dA· under aerobic conditions.^15,16 Low yields of products corresponding to formal reaction between dC· and a molecule of oxime precursor (6a,b) (and/or dC·; see below) were detected by LC/MS (Figure S4). Candidates for these molecules (7, 8) are postulated to result from a common mechanism in which dC· adds to the C5–C6 bond of precursor 6a,b or dC· (Scheme S1). dC· is known to add to the π-bond of an adjacent dC in a dinucleotide to form intrastrand cross-links.²⁷ Although aryloxyl radical addition products to dC have not been reported, the corresponding C8-addition products to dG are known.^31,32

A possible process is illustrated for reaction between a radical and dC· (Scheme 3). The aryloxyl radical can recombine with dC· within the solvent cage to regenerate the corresponding oxime precursor (6a,b). Alternatively, addition to the C5,C6-pyrimidine π-bond (9) would give rise to an unusual diradical in which one spin is on carbon. Freely diffusing ArO· and dC· are also likely to react with dC· and/or 6a,6b (to form adducts analogous to 9). Such radical–radical reactions are often noncompetitive with reactions between radicals and closed shell molecules. This is discussed further below. Upon the basis of hydroxyl radical chemistry, addition to the more electron rich C5-position is expected to be preferred over C6.³³ Reversible addition (9) explains why high yields of dC are obtained under anaerobic but not aerobic conditions, as O₂ would rapidly and irreversibly trap the initial adduct. Iminyl radicals are known to add reversibly to aromatic systems.³⁴ The resulting peroxyl radicals (10) are analogous to those produced following reaction between hydroxyl radical and dC, whose reactivity is well characterized by Cadet and Wagner.^30,33 By analogy to the reactivity of peroxyl radicals formed from hydroxyl radical in the presence of O₂, 10 may form several products, including 7 and 8, which is consistent with the low mass balance of aerobic photolyses of 6a,b.^30,33

Scheme 3.

Nitro-Substituted Oxime Photolysis

Increasing the BME concentration to as little as 10 mM resulted in yields of dC that are comparable to those observed following anaerobic photolysis of 6a (Table 1). Even just 1 mM BME doubled the dC yield under aerobic conditions to 58%. The rate constant for BME trapping of a carbon radical (<10⁷ M⁻¹ s⁻¹) is typically more than 200 times slower than that for O₂ (2 × 10⁹ M⁻¹ s⁻¹). Hence, it is unlikely that the increased dC yield under aerobic conditions at 10 mM, let alone 1 mM BME, is due to competitive trapping of 9. Instead, we suggest that BME reduces the diffusible radicals (e.g., aryloxyl radicals, dC·) in competition with their reaction with dC· and/or 6a. Thiols rapidly react with aryloxyl radicals, and the yield of p-nitrophenol increases from 50 ± 8% to 88 ± 4% when aerobic photolysis of 6a is carried out in the presence of 10 mM BME.³⁵

Direct Monitoring of dC· Generation and Reactivity.

Laser flash photolysis (LFP) experiments enable direct observation of dC· and examination of its reactivity. Rich transient features are observed in the time-resolved spectra after nanosecond pulses (355 nm) of 6a or 6b (~50 μM) in aqueous buffer (pH 7.0)/acetonitrile (1:1, v:v) solutions. Following 355 nm excitation of 6a, transient absorption with a positive peak at 400 nm and a negative peak at 330 nm is observed immediately after the laser pulse (Figure 1A). The transient species of the positive peak at 400 nm is mainly ascribed to p-nitrophenoxyl radical (ArO·) based on comparison of the coincident decay lifetime (τ ~ 1 μs) and spectral feature with a previous report for ArO· (Figure S6).³⁶ The negative peak at 330 nm is ascribed to ground state depletion of 6a. The significant depletion in 6a indicates that the initial concentrations of ArO· and dC· are quite high. Their minimum concentrations are estimated to be ~4 μM, assuming that the radicals’ extinction coefficients at 330 nm are significantly weaker than 6a. The ArO· radical with the characteristic peak at 400 nm decays rapidly within ~2.5 μs, revealing a spectral feature red-shifted to 410 nm (Figure 1a). According to its λ_max and longer lifetime (τ ~ 53 μs), the transient species at 410 nm is ascribed to the N-centered radical dC·.¹⁹ These observations show that direct irradiation of 6a results in homolytic cleavage, giving rise to ArO· and dC· radicals simultaneously (Scheme 3). ArO· and dC· absorb at ~400 nm, but the former absorbs much more strongly, initially masking the dC· signal. Fortunately, these two radicals are differentiable by their distinct lifetimes. When the decay of ArO· is complete within ~2.5 μs, the weak spectral feature of dC· radical is revealed. Interestingly, as the time proceeds after 2.5 μs under Ar, the absorption intensity around 400 nm increases slowly until 100 μs and then barely decays over the course of a millisecond (ms) (Figure 1B–D). The transient signals that grow in originate from reaction of dC· with surrounding species, generating a new intermediate(s) that absorbs more strongly than dC· at ~400 nm and is discussed further below.

Figure 1.

Time-resolved UV–vis absorption spectra for 355 nm laser photoexcitation of 6a (50 μM) in aqueous buffer (pH 7.0)/acetonitrile (1:1, v:v) under anaerobic conditions at short (A) and long (B) time scales. Kinetics traces of the 400 nm band after 355 nm laser photoexcitation of 6a under anaerobic and aerobic conditions at short (C) and long (D) time scales.

Examining the Role of O₂ in dC· Reactivity.

Upon photolysis of 6a, the kinetic traces of the 400 nm band obtained respectively in Ar-saturated and in O₂-saturated conditions (Figure 1C) show that the fast decay within 4 μs and the slow increase within 100 μs are unaffected by the O₂ concentration, indicating that the decay of ArO· and the secondary reaction of dC· are independent of O₂. Such radicals are expected to not react with O₂.¹⁷ In addition, the initial intensity of the 400 nm band is independent of O₂, demonstrating that dC· formation is unaffected by O₂ (Figure 1C insert). As mentioned above, the initial band is produced within the nanosecond time scale of the excitation pulse, indicating that homolytic cleavage of 6a occurs from a short-lived singlet excited state rather than longer-lived triplet. O₂ is unable to quench the singlet excited state, due to its concentration and diffusional limitations, explaining why its presence does not affect the photocleavage of 6a.^37,38

Importantly, the decay of the intermediate generated from the secondary reaction of dC· was accelerated on the ms time scale under aerobic conditions (Figure 1D). Quenching of this species by O₂ implies that unlike dC· it is a carbon-centered radical (Scheme 3). For the photolysis of 6b alone, similar time-resolved spectral features and kinetic traces are observed (Figure S7), demonstrating the O₂-independent generation of dC· through the photoinduced homolytic cleavage of 6b (Scheme 3).

We next considered possible structures of the carbon-centered radical produced from the secondary reaction of dC·. The rate of formation of this species is independent of 6a,b concentration (data not shown), excluding the possibility of the reaction of dC· with radical precursor. Coproduced ArO· radical is also excluded because it decays (τ ~ 1 μs) much faster than the secondary intermediate is formed (τ ~ 53 μs). Interestingly, the fast decay of ArO· radical corresponds to its reaction with H₂O (Scheme S2), yielding monohydroxyphenoxyl radical (HO–ArO·), which may have a longer lifetime and thus might lead to the carbon-centered radical species (Scheme 3).³⁶ LC/MS analysis is also consistent with reaction between two molecules of dC·.

Reaction between two radicals is considered to be feasible because the concentration of the species noted above (≥4 μM) is almost one-tenth that of the substrate. If the rate constant for reaction between two radicals (k ~ 2 × 10⁹ M⁻¹ s⁻¹) is ≥200 times than that between a radical and closed shell molecule such as 6a (k ≤ 1 × 10⁷ M⁻¹ s⁻¹), the reaction between two radical molecules will dominate. Although the structure is unknown, we speculate that it results from addition to the π-bond of dC· (9, Scheme 3) (but addition to the C6-position is also possible). Reaction of dC· with the HO–ArO· radical from 6a is expected to be more rapid than that of the aryloxyl radical generated from 6b because the latter is more electrophilic. Indeed, the secondary reaction of dC· is faster when 6a is photolyzed than when 6b is photolyzed (Figure S8). The formation of 9 is reversible, but trapping by O₂ to produce 10 is irreversible. This mechanism provides an explanation for the lower yields of dC from 6a,b under aerobic conditions (Table 1). In addition, we examined the effect of laser pulse energy on the photolysis of 6a (Figure S9). We found that the yield of proposed intermediate(s) depended linearly on the fluence, arguing against the possibility of secondary reactions.

dC· Reactivity with Nucleosides.

To gain insight into the potential reactivity of dC· in DNA, we explored its reaction with four DNA nucleosides (dG, dA, dC, and T) (Figure 2). The initial fast decay of the 410 nm band is unaffected by dG, while its subsequent slow growth is accelerated relative to photolysis of 6a in the absence of the nucleoside (Figure 2E). The emergence of a new positive peak at ~500 nm and recovery of the negative peak at 330 nm accompanies the growth of the 410 nm peak (Figure 2A,B,E). These results indicate that dC· but not ArO· reacts with dG, forming a new species absorbing around 330, 410, and 500 nm (Figure 2C,D). Additional control experiments involving 355 nm LFP of p-nitrophenol also support that ArO· does not react with dG, since there is no change in the spectra and kinetics traces of ArO· in the presence of dG (Figure S6).

Figure 2.

Time-resolved UV–vis absorption spectra for 355 nm laser photoexcitation of 6a (50 μM) + dG (3 mM) in aqueous buffer (pH 7.0)/acetonitrile (1:1, v:v) under anaerobic condition at short (A) and long (B) time scale. Difference transient spectra of pure 6a from those of 6a + dG after 355 nm laser photoexcitation at short (C) and long (D) time scales. (E) Kinetics traces of the 400 nm band after 355 nm laser photoexcitation of 6a with and without dG under anaerobic condition. (F) Concentration dependence for the rate constant obtained from the slow growth of the 410 nm band after 355 nm laser photoexcitation of 6a + dG under anaerobic condition. The red line is the fit.

By subtracting the spectra of 6a from those of 6a + dG for each time slice, we obtain the spectra for the new species formed by the reaction of dC· radical with dG. The difference spectra obtained from 3.5–40 μs (Figure 2C) and 40–180 μs (Figure 2D) show the spectral characteristics for dG^•+/dG·, as indicated by absorption bands at 310, 390, and 500 nm.^39–41 To confirm the spectral assignment, we measured the spectra of dG^•+ and dG· (Figure S10) from the reported photochemical generation of SO₄^•−, which oxidizes dG.^39,42 It fully agrees with the spectra obtained from the dC· + dG (Figure 2). Moreover, it is well recognized that both the dG^•+ and dG· exhibit three resolved absorption bands at 310, 390, and 500 nm, as well as a flat and weak absorption above 600 nm. Following the one-electron oxidation to give rise to dG^•+ (pK_a = 3.9), rapid deprotonation within ~50 ns produces dG·.^39,42 Hence, dG^•+/dG· are observed following reaction of dC· + dG. These data provide strong evidence that dC· oxidizes dG to produce dG^•+/dG·. The signal intensity at 410 nm from 6a + dG plateaus within 40 μs but continues increasing for more than 100 μs in the absence of dG due to slower product formation (Figure 2E). Therefore, the difference spectra would have a decrease for the absorption intensity after 40 μs. In other words, the decrease of dG·intensity between 40 and 180 μs (Figure 2D) is an artifact of this difference analysis and is not a measure of the true dG·decay.

Interestingly, for the photolysis of 6a in the presence of the other three nucleosides (dA, dC, and T), there is no obvious change in the time-resolved spectra and kinetics traces compared to 6a alone (Figure S11), indicating that there is no obvious reaction between dC· radical and either of dA, dC, and T. Considering that dG has the lowest redox potential (1.29 V vs NHE) among all the DNA nucleosides, these results suggest that the reaction of dC· radical with dG to generate dG^•+/dG· occurs through the one-electron oxidation pathway.^21,43,44 Furthermore, we measured the reaction rate of dC· + dG by following the growth kinetics at 410 nm and assessed the concentration dependence varying the dG concentration from 1.0 to 6.0 mM (Figure 2F). From the slope, the second-order reaction rate constant of dC· radical with dG is deduced to be 5.9 × 10⁶ M⁻¹ s⁻¹.

2′-Deoxycytidin-N4-yl Radical (dC·) Generation and Reactivity in DNA.

Solid-phase synthesis of oligonucleotides was carried out using the corresponding phosphoramidite of 6a because the dinitro oxime (6b) was slightly unstable to concentrated aqueous ammonium hydroxide, the standard deprotection condition. (Note that for convenience, the same identifier is used for a molecule, whether it is a nucleoside or within an oligonucleotide.) Consequently, oligonucleotides were prepared using commercially available fast-deprotecting phosphoramidites and the corresponding β-cyanoethyl phosphoramidite of 6a.⁴⁵ Standard synthesis conditions were used with the exception that the oxime phosphoramidite was coupled for 5 min and 5% phenoxyacetic anhydride was substituted for acetic anhydride as capping agent. Oligonucleotides containing 6a were deprotected using concentrated aqueous ammonium hydroxide (25 °C, 8 h), purified by denaturing polyacrylamide gel electrophoresis, and characterized by mass spectrometry. Dodecameric DNA (11) containing 6a opposite dG is destabilized (T_M = 42.5 ± 0.5 °C) compared to when dC is paired with the purine (13, T_M = 50.4 ± 1.0 °C) by ~8 °C. We hypothesize that the large destabilization is at least partially attributable to 6a existing as the imine tautomer when opposite dG, which disrupts hydrogen bonding. The slightly higher average T_M when 6a is opposite dA (12, T_M = 36.9 ± 0.7 °C) than when dC is present (14, T_M = 34.7 ± 2.3 °C) in an otherwise identical dodecamer is consistent with this hypothesis.

Due to the complications introduced by generating dC· from the nitrophenyl oximes under aerobic conditions, photolysis experiments involving oligonucleotides were carried out under degassed conditions. The concentration of ³²P-labeled DNA was ≤10 nM in these experiments. Even if 99.9% of O₂ was removed, its concentration (~200 nM) would be far greater than that of the radical(s), yet low enough that it would not trap (k_O2[O₂] < 4 × 10² s⁻¹) reversibly formed 7a competitively with unimolecular DNA processes.

Tandem Lesion Formation.

Tandem lesions consist of two or more contiguously damaged nucleotides. Several nucleobase radicals produce tandem lesions by either reacting directly with an adjacent nucleotide or via the corresponding peroxyl radical.⁴⁶ Tandem lesions are significant because they are more difficult to repair than isolated damage sites.^47–50 The precedent that is most similar to dC· is dA·, which initiates traceless tandem lesion formation by abstracting a hydrogen atom from the C5-methyl group of a 5′-adjacent thymidine.¹¹ The radical center in dC· is also positioned within the major groove and is positioned to react preferentially with the methyl group of the 5′-adjacent thymidine (Figure 3). The right-handed helical twist positions the methyl group of a 5′-thymidine ~1.4 Å closer to the N4-cytosine atom than that of a 3′-adjacent thymidine. Duplex 15 was designed to probe for tandem lesion formation and DNA damage via hole transfer.^10,11 The latter would result in lesions predominantly at dG₉ and dG₁₀, whereas damage at T₁₂ and dG₁₁ is indicative of a tandem lesion.^51–54 Purine damage is selectively revealed as a strand break by treating the photolysate with the base excision repair enzyme, formamidopyrimidine glycosylase (Fpg).⁵⁵ Treating photolyzed 5′-³²P-15 with Fpg produces (background corrected) cleavage at dG₁₁ (18.1 ± 1.1%) but not dG₉ or dG₁₀ (Figure 4, Figure S12). Piperidine treatment is used to convert alkali-labile lesions at dG and/or T into strand breaks. After correcting for background cleavage, photolysis of 5′-³²P-15 produces alkali-labile lesions at T₁₂ and dG₁₁ (Figure 4, Figure S12). The absence of damage at dG₉ or dG₁₀ suggests that hole transfer from dC· does not occur and is discussed further below.

Figure 3.

Molecular model of dC· flanked by thymidine. Distance shown is from N4 of dC· to the hydrogen atom of the thymine methyl group.

Figure 4.

Tandem lesion formation: strand cleavage at indicated nucleotides following photolysis by Fpg or piperidine. BME (0.7 M) is present during piperidine treatment.

Tandem lesion formation was validated using 5′-³²P-15 and characterized further using other duplex substrates. Piperidine treatment in the presence of BME, which suppresses cleavage at 8-OxodGuo, results in a reduction in cleavage at dG₁₁ and an increase at T₁₂ in 5′-³²P-15 (Figure 4, Figure S12).⁵⁶ This is the expected observation if a tandem lesion comprises damage at dG₁₁ and T₁₂ is produced in 5′-³²P-labeled substrate.¹¹ NaBH₄ treatment of the photolysate reduces piperidine-induced cleavage at T₁₂ by 3-fold (Figure S13). This is consistent with formation of fdU, which is piperidine labile, as opposed to hmdU, which is not, at T₁₂ upon photolysis. A similar cleavage pattern was observed following piperidine treatment of photolyzed 5′-³²P-18 (Figure S14), albeit at lower levels. Fpg cleavage at dA₁₃ of 5′-³²P-18 (4.6 ± 0.1%) is more than 3-fold lower than at dG₁₁ in 5′-³²P-15 (Figure 4). This is consistent with the lower rate of oxidation of the less electron rich adenine than guanine and the effect of local sequence on the purine oxidation potential.^21,52 Consistent with the greater distance between the dC· radical center and the thymine methyl hydrogen of the 3′-adjacent thymidine (Figure 3), alkali-labile lesions are not detected in 5′-³²P-17 (Figure S15). Support for the intermediacy of T· (Scheme 4) in the process was gleaned from photolysis of 5′-³²P-16 in which 2′-deoxyuridine (dU) was substituted for T₁₂. Alkali-labile strand breaks are reduced to 1.1 ± 0.1% at dG₁₁ in 5′-³²P-16, an approximately 16-fold reduction from 5′-³²P-15 (Figure 4, Figure S16). Piperidine-induced strand cleavage is 1.4 ± 0.1% at dU₁₂ in 5′-³²P-16. The source of this low level of damage is unknown, and we cannot rule out addition of dC· to the 5′-adjacent π-bond.²⁷

Scheme 4.

Proposed Mechanism for Tandem Lesion Formation from dC·

These cleavage studies are consistent with an O₂ dependent tandem lesion formation process reminiscent of that involving dA·.¹¹ The yields of tandem lesions are commensurate with the level of labile lesion at dG₁₁ in 15 (~18–20%, Figure 4). Following dC· hydrogen atom abstraction from the 5′-thymidine methyl group (Scheme 4, step 1), O₂ trapping (step 2) of T· produces a peroxyl radical (Tp·) that adds to the π-bond of the 5′-adjacent purine (step 3), which undergoes further oxidation to yield the ultimate lesion via a series of steps (step 4).¹¹ More structural information was obtained by UPLC–MS/MS analysis of photolyzed 11 (Figures S17–S19). A tandem lesion comprising 5′-(8-oxodGuo)-(fdU)-(dC) (19) was identified. The identity of 19 was supported by analysis of photolysate reduced by NaBH₄, which generated a tandem lesion that is consistent with 20. The tandem lesion (19) is replaced by 21 when 11 is photolyzed in the presence of BME (0.2 mM). This is consistent with the intermediacy of the peroxyl radical (Tp·, Scheme 4). Furthermore, the presence of dC is consistent with hydrogen atom abstraction by dC·. A rigorous study on a trinucleotide in which Tp· was generated adjacent to dG produced tandem lesions containing Fapy·dG in greater yield than the respective product(s) comprising 8-oxodGuo.⁵⁷ However, Fapy·dG was not detected in these experiments.

2′-Deoxyguanosine Oxidation by 2′-Deoxycytidin-N4-yl radical (dC·) in DNA.

dC· has been proposed to play a role in the deprotonation of dG^•+, which is the step that traps guanine oxidation at a particular nucleotide during DNA hole transfer (Scheme 1B).²³ During this process, dC· is proposed to formally abstract a hydrogen atom from dG to produce dG·. It is not known whether this is a hydrogen atom abstraction or proton coupled electron transfer process. The LFP experiments using 6a,b indicate that dC· oxidizes dG via a single electron pathway (Figure 2), resulting in formation of dG^•+. As briefly mentioned above, 5′-³²P-15 was designed to probe for and distinguish between hole transfer and tandem lesion formation.^53,54 Alkali-labile cleavage at dG₉ and dG₁₀, a signature of hole transfer in this sequence, was not detected when photolysis was carried out at pH 7.2, or 5.1 where a greater concentration of dG^•+ was expected.

dC· oxidation of dG produces dC⁻, a strong base, which would rapidly deprotonate an opposing dG^•+, trapping the hole at that nucleotide (Scheme 5). Hole transfer from the initially formed dG^•+ must compete with deprotonation. The rate constant for hole transfer involving dG^•+ can span several orders of magnitude depending upon the flanking sequence.^58,59 If dC· oxidizes the opposing dG as proposed by Anderson, we speculated that the competitiveness of hole transfer with deprotonation would improve significantly if dG^•+ were part of a dGGG sequence (22). 5′-³²P-Labeling of the strand containing 6a in 22 reveals qualitatively similar results as 5′-³²P-15, although the cleavage yield at the 3′-terminal dG in the trinucleotide sequence (dG₉) is lower (Pip, 11.0 ± 0.5%; Fpg, 8.3 ± 1.9%). Moreover, Fpg cleavage at the nucleotide opposite dC· in 22, dG₃₂ = 0.4 ± 0.1%, is 1.2 ± 0.2% and 1.3 ± 0.3% at dG₃₁ and dG₃₀, respectively (Figure 5, Figure S21). Strand damage is not detected in 24 subjected to the same irradiation conditions (Figure S24). In comparison, Fpg cleavage at the dG opposing dC· in 15 is 0.4 ± 0.1% (Figure S22). The absolute yields of cleavage at the dGs are low, but the conversion efficiency of dG^•+ into a labile lesion is ~10%.⁵³ Furthermore, the low level of strand scission within dG₃₀-dG₃₂ on the opposing strand in 22 provides an explanation for why hole transfer is not detected at dG₉ and dG₁₀ of 15. The intervening T-A base pair reduces the amount of electron transfer ~10-fold, which is below the detection limit in this experiment.⁵³

Scheme 5.

Competition between Proton Transfer and Hole Transfer Following dG Oxidation by dC·

Figure 5.

Opposing purine oxidation by dC· and trapping of dG^•+.

To increase the competitiveness of hole transfer with proton transfer, O6-methyl-2′-deoxyguanosine (MeG, 23) was substituted for dG opposite dC·. We anticipated that the corresponding radical cation would be a much poorer proton donor to dC⁻ due to the absence of an N1-hydrogen, resulting in a longer lifetime for the radical cation and therefore greater hole transfer. The highest occupied orbitals (HOMOs) of dG and MeG were calculated based on density functional theory at the level of PCM/B3LYP/6-31++G(d,p). The calculations indicate that the MeG HOMO is 3.1 kJ/mol higher in energy than that of dG, which also will assist oxidation (Figure S35). In addition to MeG being more readily oxidized, a duplex (25) containing this modification opposite 6a has a higher T_M (39.7 ± 2.3 °C) than when it is opposite dC (26, 31.1 ± 1.3 °C). We postulate that this is due to the favorable Watson–Crick base pairing with the favored tautomer of 6a and suggests that MeG perturbs the duplex structure less than when dG is present. We do not know the chemical product from MeG oxidation or how this nonnative nucleotide affects the HOMO distribution over the trinucleotide sequence. However, cleavage at the corresponding trinucleotide sequence in 23 is 0.6 ± 0.1%, 5.3 ± 1.1%, and 0.6 ± 0.1% at dG₃₀, dG₃₁, and MeG, respectively, which is significantly greater than in 22 (Figure 5). We propose that this is indicative of hole migration initiated from dC· oxidation of the opposing MeG.

Additional evidence for MeG oxidation was obtained by comparing strand damage in 27 to that in 15. Duplex 27 differs from 15 only in that the nucleotide opposite the radical precursor (6a) is MeG instead of dG. Photolysis of 5′-³²P-27 in which the precursor-containing strand is labeled yields an alkali-labile lesion cleavage pattern at G₁₁ and T₁₂ (Figure S26) that is consistent with the tandem lesions observed in irradiated 15 (Figure 4, Scheme 4). 5′-³²P-Labeling of the complementary strand (Figure S25) produces 0.9 ± 0.1% alkali-labile cleavage at MeG upon photolysis and comparable or even greater piperidine-induced cleavage at dA₃₆ (0.8 ± 0.1%) and T₃₇ (2.3 ± 0.5%). In contrast, strand damage is not detected at A₃₆ or T₃₇ when the complementary strand of 15 is labeled (Figure S24), and cleavage at dG₃₈ is only 0.4 ± 0.1%. We propose that damage at T₃₇ and A₃₆ is a result of one electron oxidation of MeG. One speculative explanation involves deprotonation of MeG^•+ to produce 28 (Scheme 6). Peroxyl radical 28 is poised in the major groove to abstract a hydrogen atom from the methyl group of the 5′-adjacent thymidine (T₃₇). T· is then transformed into an alkali-labile lesion via Tp·, which propagates damage in the 5′-direction per Scheme 4. The reactivity is comparable to that observed for dC·, dA·, and other nucleobase peroxyl radicals.^{11,12,46,60,61} There is also precedent for tandem lesion formation from a cyclopropyl substituted nucleotide probe designed to detect purine radical cations produced during DNA hole transfer.⁶²

Scheme 6.

Proposed Mechanism for Tandem Lesion Formation from dC· Oxidation of MeG

CONCLUSION

Although 2′-deoxycytidin-N4-yl radical (dC·) is involved in several DNA damage processes, its reactivity is not well understood. As a nitrogen radical, the lack of reaction with O₂ has a significant effect on the chemical processes that dC·partakes in. We studied the reactivity of dC· using a photochemical precursor. Similar to other nucleobase radicals in which spin is in the major groove, we determined that dC· produces tandem lesions by reacting with the 5′-adjacent nucleotide.^61,63 Its reactivity in this regard is most similar to that of dA·, another nitrogen-centered radical that is positioned in the major groove.^10,11 Like dA·, tandem lesion formation from dC· is traceless because it is reduced to dC during the process. Unlike dA·, an isolated dC·, which is a stronger oxidant, directly oxidizes dG.²¹ In this regard, dC· is more similar to a nucleoside alkyl aminyl radical.⁶⁴ Hole transfer from the dG^•+ formed must compete with deprotonation by dC⁻, a strong base, and is detected when the local sequence is dGGG, which provides a conducive pathway. Additional evidence for oxidation of the opposing dG was obtained using MeG, which has not been employed previously as a probe for hole transfer. The chemistry observed in DNA containing MeG is fully consistent with hole transfer. These data are consistent with the proposed role that dC plays in trapping a DNA hole on a particular dG (Scheme 1B).²³ However, dC· formation is the rate-determining step in this process, and its feasibility remains in question.¹⁸ This issue and other aspects of dC· reactivity will be the focus of future endeavors.