Mechanistic Investigation on ROS Resistance of Phosphorothioated DNA

Phosphorothioated DNA (PT-DNA) exhibits a mild anti-oxidant property both in vivo and in vitro. It was found that 8-OHdG and ROS levels were significantly lower in dnd+ (i.e. S+) E. coli., compared to a dnd− (i.e. S−) strain. Furthermore, different from traditional antioxidants, phosphorothioate compound presents an unexpectedly high capacity to quench hydroxyl radical. Oxidative product analysis by liquid chromatography-mass spectrometry and quantum mechanistic computation supported its unique anti-oxidant characteristic of the hydroxyl selectivity: phosphorothioate donates an electron to either hydroxyl radical or guanine radical derived from hydroxyl radical, leading to a PS• radical; a complex of PS• radical and OH− (i.e. the reductive product of hydroxyl radical) releases a highly reductive HS• radical, which scavenges more equivalents of oxidants in the way to high-covalent sulphur compounds such as sulphur, sulphite and sulphate. The PS-PO conversion (PS and PO denote phosphorus-sulphur and phosphorus-oxygen compounds, respectively) made a switch of extremely oxidative OH• to highly reductive HS• species, endowing PT-DNA with the observed high capacity in hydroxyl-radical neutralization. This plausible mechanism provides partial rationale as to why bacteria develop the resource-demanding PT modification on guanine-neighboring phosphates in genome.

conflicting, mainly for lack of detailed mechanistic investigation in rationale on the promising hypothesis that the sparse PT sites meet anti-oxidant needs of a long nucleobase sequence.
Indeed, DNA oxidative damage is usually caused by highly reactive oxygen species (ROS), such as singlet oxygen, superoxide, hydrogen peroxide, and hydroxyl radical 17 . Many chemical modifications in DNA damage (e.g. DNA cleavage, DNA-protein cross links, purine oxidation, etc.) stem from the reactions with ROS, in particular with hydroxyl radical (HO • ). Hydroxyl radical is very rare in a living cell, but extremely reactive, scrambling electrons and hydrogen atoms from protein and metabolites in a diffusive manner, and even from stable biomolecules such as DNA and lipids. Both the heterocyclic base and sugar-phosphate backbone are vulnerable, in particular via electron abstraction on guanine base and hydrogen abstraction on ribose 5′ -C. If the nucleobase-repair systems are not able to immediately regenerate intact nucleobase, a mutation in genome would result from erroneous base pairing during replication or, consequently, DNA cleavage would occur consequently. It is also believed that metal ions such as iron and copper would enhance hydroxyl radical generation via Fenton chemistry 18 .
In short, DNA oxidative damage seems to begin with one-electron loss and formation of a radical cation "hole" at the encounter position 19 . The hole hops reversibly through electronic structure along the DNA helix, trapped at a guanine-rich place as a form of guanine radical cation (G •+ ) 20 . Hydrolysis of G •+ releases 8-hydroxy-2′ -deoxyguanosine , upon consequential DNA repairing in vivo. So 8-OHdG is an effective biomarker for DNA oxidative damage 21 .
Fischer et al. have also reported in vivo antioxidant capacity of 5′ -PT-dNTP in drug development for neurodegenerative disorders, but as attributing to metal-ion coordination and Fenton inhibition 22 . In contrast, low molecular-weight free thiols such as glutathione (GSH), cysteine (CYS), and homocystine can ameliorate oxidative damage in an alternative fashion, via an oxidative dimerization towards RSSR 23 .
In this work we measured the both the 8-OHdG and ROS intracellular levels of dnd+ (S + ) and dnd− (S − ) strains of E. coli without intervention of restriction genes dndFGHI to verify the anti-oxidant property of PT-DNA. Then we revealed its anti-oxidant mechanism with in vitro oxidation of pJTU1238 plasmid and phosphorothioate dinucleotide, using gel electrophoresis, circular dichroism spectroscopy, high performance liquid chromatography -mass spectroscopy, and quantum chemistry. Here we propose a unique anti-oxidant mechanism, different from metal-ion coordination 18 and sulfur-sulfur dimerization 23 in the literatures.

Results
In vivo Anti-oxidant Effect. Figure 1a shows DCF fluorescence of dnd+ (S + ) and dnd− (S − ) E. coli strains over time under oxidative stress mimicked with Fenton's reagent of 6 μ M FeSO 4 and 0.5 mM H 2 O 2 . Compared to the starting point, the intracellular ROS level represented by DCF fluorescence climbed up by 18.0 ± 8.4, 40.9 ± 10.0, 42.7 ± 4.7, 45.2 ± 7.8, 62.1 ± 11.7, and 94.8 ± 10.2% in the dnd− culture, at 1, 5, 10, 15, 20, and 25 min, respectively. By contrast, the intracellular ROS accumulation in E. coli (S + ) strain fluctuates at 3.8 ± 5.5, 22.5 ± 5.7, 14.9 ± 12.9, 27.3 ± 6.6, 40.5 ± 4.9, and 35.0 ± 12.7%, respectively. As shown in Fig. 1c, when the strains were charged with a moderate Fenton's reagent, the 8-OHdG level arose about 12.5 ± 14.7~58.3 ± 2.3% in the S + strain with increasing H 2 O 2 concentrations from 0.5 to 2.0 mM; conversely, the values in the S − strains rocketed up by 32.5 ± 3.3~180.0 ± 4.1%. In Fig. 1d In vitro Anti-oxidant Tests of Plasmid PT-DNA. Figure 2 showed the in vitro experimental results of plasmid oxidation with H 2 O 2 concentrations in a range of 0.0 to 1.5 μ M for 12 hours in the dark, with comparison to a ABTS pre-oxidized PT-DNA sample ([PT] ≈ 0.19 μ M). It is almost indistinguishable in the image of gel electrophoresis (Fig. 2a) and the circular dichroism (CD) spectra (Fig. 2b) owing to the technical limitation; so digital quantitation on the two in vitro experiments was carried out. Figure 2c show quantitation of the electrophoresis image by software image J 26 . Nicked circular DNA percentage in the total plasmid DNA (supercoil plus linear and nicked circular) was used to exhibit DNA oxidative damage 27 . The nicked-DNA percentage of PT plasmid slightly increased by 1.6 ± 1.6 and 2.6 ± 1.7% at 0.5 and 1.0 μ M H 2 O 2 , respectively. However, the values were measured as 4.0 ± 2.9 and 8.9 ± 1.7% in the dnd− plasmid, and − 0.4 ± 2.0 and 9.4 ± 1.6% in the pre-oxidized PT-DNA, respectively. All the three plasmid samples exhibited severe oxidative damage in the presence of 1.5 μ M H 2 O 2 , with the nicked-DNA percentages of 12.7 ± 1.4, 17.7 ± 1.7, and 24.5 ± 1.6% for the PT-, normal plasmid, and pre-oxidized PT-, respectively. Figure 2d shows quantitation of DNA conformational damage by oxidative stress, represented with a ratio of integrated areas under the positive band at 275 nm over the negative band at 245 nm in CD spectra. It has been reported that declining the positive ellipticity and inclining the negative ellipticity characterize oxidative damage on plasmid DNA 27 . For purpose of controlling errors from peak-wavelength shift, the integrated value in half-band width was computed and their ratios were used to denote DNA helicity. The ratios drop down by slopes of − 1.2 ± 0.6, − 3.1 ± 1.1, and − 7.1 ± 0.8 (× 10 −3 min −1 ) for the PT-, normal, and pre-oxidized PT-DNA plasmid, respectively. On the other hand, the absolute values of ratio in the order of dnd− > dnd+ > pre-oxidized PT-DNA plasmid indicated that DNA oxidation damaged the secondary structure significantly while DNA phorosphorothioation disturbed the B-helical conformation slightly 5 .

Comparison of PS, GSH, CYS, and vitC.
Owing to the μ M technical limitation of plasmid PT-DNA, diethyl phosphorothioate (PS) was used as a model compound of PT-DNA in the following in vitro comparative experiments with non-physiological oxidants ABTS • + and DPPH • radicals. GSH, CYS, and vitC appeared to possess a similar scavenging capacity with reference to 46.7 μ M ABTS • + in 30 min, with the measured IC50 values of 5.1 ± 0.2, 5.9 ± 0.1, and 7.1 ± 0.1 μ M, respectively. This value was 42.3 ± 1.2 μ M for PS, six times higher than those of traditional antioxidants. Similarly, the IC50 values towards 200 μ M DPPH in 30 min were determined as 36.5 ± 1.5, 31.2 ± 0.2, and 25.3 ± 0.4 μ M for GSH, CYS, and vitC, respectively; while the IC50 value for PS was measured to be 230.7 ± 8.1 μ M, also about 6 to 8 times higher than the others. Thus, in terms of IC50 value with reference to organic oxidants ABTS • + and DPPH • , phosphorothioate exhibited a mild anti-oxidant activity, in agreement with its weak reductive potential.
According to electrochemical measurement (Fig. 3b), the reduction potential of PS (0.74 V vs Ag/AgCl) is lower than that of 2′ -deoxyguanosine by about 0.28 V. The standard reduction potentials of guanine, adenine, thymidine, and cytosine were reported to be about 1.29, 1.42, 1.70, and 1.60 V 28,29 . So the standards reduction potential of phosphorothioate was estimated as 0.94 V after calibration, higher than that of Fe(III)/Fe(II) by 0.20 V, but lower than the G •+ /G potential. Therefore, the pro-oxidant effect for these antioxidants was determined to conform to the order of CYS > vitC > GSH ≫ PS (Figs S3 and S4). Similarly, oxidative damage of plasmid DNA in the DNA lesion experiment can be used to test the pro-oxidant effect in the presence of Fe(III)-modified Fenton's reagent 30 . PS gave no detectable pro-oxidativity up to 2.5 mM, while the other three antioxidants exhibited the DNA degradation at the relatively high concentrations (Figs 3c and S5). In silico Justification. We carried out intensive quantum chemical calculations for the reaction pathway from PS toward PH and PO (Table S1 and Fig. S8). Figure 5 shows the most plausible mechanism in which PH was generated from two oxygen-donations of H 2 O 2 . In the first step, one molecule of H 2 O 2 reacts with PS to form PSO anion (INT1) and release one molecule of water, with a calculated barrier of 23.2 kcal/mol in aqueous solution. Then, another molecule of H 2 O 2 reacts with PSO anion, releasing one molecule of sulphur dioxide (SO 2 ) and one water molecule and then leading to PH generation. The reaction barrier for the second step was computed to be 14.9 kcal/mol. By comparison to the PS-PH conversion, the most plausible mechanism for PS-PO conversion has a shortcut path in the presence of hydroxyl radical. Electron transfer from phosphorothioate anion to HO • or G •+ seemed barrier-free for the extremely high reactivity of hydroxyl radical -a pre-reactive complex of [PS---OH] −• was formed instantly. The radical anion complex transformed to a five-coordinate adduct (INT2) via phosphor hydroxylation, with a calculated barrier of 6.9 kcal/mol. Such an addition-elimination mechanism is similar to hydrolysis of organophosphorus pesticides 31 . The INT2 was higher in energy by 5.9 kcal/mol than the pre-reactive complex. INT2 underwent an elimination of hydrogen-sulphur radical (HS • ), forming the normal phosphate with a calculated barrier of 14.4 kcal/mol. The reaction of PS plus HO • towards PO plus HS • was exothermic by 27.8 kcal/mol. Therefore, the overall reaction barrier for the PS-PO conversion was estimated to be 14.4 kcal/mol (Δ G ≠ ) in an aqueous solution. As Fig. 6 shows, the frontier orbitals of GA and GsA were observed at the M06 (CPCM, solvent = water)/6-311 + G(d, p)//M06/6-31 G(d) level of theory.

Discussion
The in vivo experiments indicated that the model organisms (E. coli, as shown in Fig. 1) of this study led to the consistent results as wild type S. lividans (dnd+ , i.e. S + ) and the knocking-out mutant (dnd− , i.e. S − ) to oxidative stress reported by Dai et al. 16 . It should be acknowledged that the additional anti-oxidant capacity in the S + strain isn't major given that the decay time changed slightly for extracellular peroxide (12~13%) owing to coexistence of various anti-oxidant systems in the living organisms. Beside a slightly fast consumption of peroxide, it was first time to observe the intracellular ROS accumulation was suppressed by about 14~60% in the S + strain, albeit with a sink at 10-15 min upon anti-ROS response time-delay as reported in the previous literature 32 . Dai et al. reported no significant up-regulation of anti-oxidant genes based on RNA-seq experiment of wide type S. lividans upon oxidative stress 16 , in addition to the consistent observation on dndFGHI-negative E. coli in this work, clarifying the previous suspicion whether the dnd− knocking-out would disturb dndABCDE/FGHI restriction-response system ("sick cell"). However, the intercellular ROS response exhibited dramatical difference between the S + and S − cultures: the intracellular levels of 8-OHdG biomarker for DNA oxidative damage dropped significantly and those of GSSG biomarker for overall oxidation lowered as well; on the other hand, the GSH levels were similar in the two strains -consistent with no thiol peroxidase up-regulation based on RNA-seq experiment under oxidative stress 16 . All these results testified an effective anti-oxidant bypass in the S + strain to maintain a reductive atmosphere in cell, which may stem from PT-DNA itself.
pJTU1238 in the E. coli strains were used to solve the anti-oxidant biochemistry of PT-DNA. Plasmid DNA was extracted from the two E. coli strains and tested for anti-oxidant activity in vitro directly. A typical PT population was estimated as ~1/1000 nt in the S + strain based on the previous deep sequencing 1 , so the PT concentration in a stock solution of extracted dnd+ plasmid with optical density of 4.0 at 260 nm is quantified as ~200 μ g/mL in ds-DNA weight or ~0.6 μ M in PT equivalent. Owing to no further endogenous PT regeneration after the E. coli digestion, the anti-oxidant capacity of plasmid PT was limited at the micromolar (μ M) level, significantly lower than those in vivo oxidant levels in Fig. 1 15,33 . The plasmid experiment in Fig. 2d indicated that the oxidative   (Fig. 2c).
Although the anti-oxidant property of the extracted plasmid touched limit of determination at 1 μ M H 2 O 2 , it is worthwhile to note that the low-frequency PT in genome is in agreement with economical efficiency in biology, due to the low-frequency of DNA oxidative damage in reality, e.g. usually ~1 lesion per 10000 nt upon gamma radiolysis 34 .
Intriguingly, the antioxidant effect of phosphorothioate compound showed unexpected selectivity to hydroxyl radical in the anti-oxidant chemistry, compared to traditional antioxidants. It has been postulated that PT "chelator" might coordinate ferrous iron and shut down Fenton reaction 35 . But this seems unlikely in our scenario: first, the in vitro anti-oxidant activity of plasmid DNA was observed even without an iron catalyst (due to difficult manipulation of low concentration H 2 O 2 ); second, the PT consumption was observed at the same level of molar equivalent of H 2 O 2 in the extracted plasmid experiments, indicating that PT is consumable in the anti-oxidant process. There is no evidence that PT would be inert toward hydroxyl radical, irrespective of whether PT coordinates the metal ion or not. Moreover, the coordination of PS to Fe(II) was relatively weak -compared to Fe(II)-EDTA (IC50 = 1.27 ± 0.05 μ g/ml at 12.5 μ M [Fe 2+ ]), the absorbance of Fe(II)-ferrozine complex did not change in the presence of PS from 0.01 to 5.00 mM (Fig. S6). It is likely that even a strong chelator like EDTA (ethylene di-amine tetra-acetic acid) cannot inhibit oxidative stress mimicking with Fenton chemistry. Probably this is due the high sensitivity of the biological system to trace amount of oxidative damage on the DNA molecule, and so PT-metal coordination is likely to be unimportant in its anti-oxidant activity. It should also be mentioned that both GSH and CYS exhibited severe side-effect at concentrations above 250.0 μ M, that is, more ROS was generated (pro-oxidant effect). Indeed, hydroxyl radical -the dominant ROS product of Fenton reaction -is the most lethal to DNA molecule, responsible for the formation of G •+ radical cation and then resulting 8-OHdG 36,37 .
Regarding the issue of PT-DNA backbone cleavage during oxidation, we repeated LCMS analysis of GsA oxidation of our previous work 15 , using Fenton's reagent, instead. Oxidant ABTS • + resulted in the same product as the Fenton's reagent, but no G H A was detected in the presence of excessive amount of oxidants. The G H A formation observed in the previous work was due to peroxyacetic acid (PAA) usage or high concentration of H 2 O 2 15,16 . Importantly, the GsA oxidation led to only GA, no PSSP disulphide was observed as in the PS oxidation. In the absence of Fenton catalyst, no GA product was detected in the mixture of 200 μ M H 2 O 2 and 20 μ M GsA after 30 min of incubation in the Tris-HCl buffer (pH = 7.4) (Fig. S7). Thus, the PT-DNA oxidation by hydroxyl radical likely resulted in a clean conversion to PO-DNA, i.e. the normal form of phosphate.
Therefore, the oxidation of phosphorothioate by ROS generated by Fenton's reagent is different from free thiols. For example, CYS oxidation by H 2 O 2 involves in a two-step reaction, forming sulfenic acid (CSOH) intermediate and then disulfide (CSSC), sulphenic, sulphinic or even sulphonic acid with increasing ROS concentrations 38,39 . The oxidation of phosphorothioate dinucleotide could desulfurize to normal phosphate 40 , in particular for the dilute PT encountering hydroxyl radical in a weak alkaline circumstance.
Owing to the experimental limitation, we distinguished the question whether the hydrogen-phosphonate species (PH) is an intermediate in the PS-PO conversion or a branched by-product of H 2 O 2 -dependent oxidation by quantum-chemistry calculations. As shown in Fig. 5, the PH generation needs two molecules of H 2 O 2 with a higher reaction barrier, consistent with the HPLC-MS experimental observation in which the PH formation was highly dependent on the concentration of H 2 O 2 . The neural form of PH has two tautomeric structures, either as protonation on a phorsphorous center or as oxygen atom 41 . Hydrogen phosphonate would hydrolyze P-O bond into two parties at a rate of 3.2 × 10 5 M/s at 25 °C 42 . Thus, unlike phosphate, hydrogen phosphonate hydrolyzes quickly and causes DNA backbone cleavage, as the dnd+ phenotype in PT-DNA gel electrophoresis.
Based on the quantum calculations, hydrogen-sulfur radical HS • was generated in the PS-PO conversion (Fig. 5). Opposite to extremely oxidative hydroxyl radical, HS • is extremely reductive! HS • will degrade rapidly to sulphur (S 8   At this point we tackle the question of how ROS invasion can precisely hit the PT site in vivo and in vitro when a PT site has an average scatter of 100~10000 nt. We believe that radical G-hopping in DNA oxidative damage helps to explain why the guanine-neighbouring PT site captures DNA damage efficiently. As shown in Fig. 6, both the HOMO and LUMO of GA anion and radical coincided each other in population, indicating that electron donation occurred on the guanine moiety and phosphate group in the oxidation of normal DNA, which is in agreement with the previous studies 15 . By contrast, the HOMO of GsA anion was consistent with the HOMO of GA and the SOMO of GsA radical in population, indicating that the guanine moiety of GsA is able to donate electron as well; however, the LUMO of GsA radical was computed to be dominant for the sulfur atom in population, indicating the likely formation of a phosphorothioate radical, which, as a result, keeps the guanine moiety intact in the PT-DNA oxidation. Such an electron transfer was further confirmed by the distributions of Mulliken charge and spin density of the two forms of GsA and GA, where the GA radical presented a significant charge separation between the nucleobases (+ 0.41 and + 0.29) and phosphate (− 0.97) groups with an unpaired electron on the guanine moiety (0.922), while the GsA radial only loses the negative charge (− 1.296 → − 0.253) and casts most of spin density (1.021) on the phosphorothioate group. We conclude that the phosphorothioate radical is hydrolyzed as via the same TS3 and TS4 transition states as shown in Fig. 5, leading to the formation of normal phosphate linkage. This also explains why the GsA was more reactive in scavenging ROS than PS, since the guanine moiety played a role in bridging electron transfer, and the relatively higher pH environment benefitted the formation of the [PS-OH] •− pre-reaction state, which can also be accelerated in the presence of hydroxyl radical as an oxidant.
In summary, due to the exceptional chemical nature of phosphorothioate, E. coli (S + ) possesses DNA-intramolecular PT antioxidant activity, which acted as a guard against oxidative stress via effectively scavenging hydroxyl radical (the most lethal species to genetic materials), decreased the intracellular ROS level, and therefore limited the 8-OHdG production and promoted the GSH:GSSG ratio, and thereby provided additional protection from DNA oxidative damage. The reductivity of phosphorothioate is powerful in protecting the most vulnerable moiety (guanine) of the DNA molecules, but not too intensive to induce a pro-oxidant side effect in cell.
In vitro hydrogen phosphonate species was a branched by-product at the high H 2 O 2 concentration. The PS-PO in vivo recycling does not require a hydrogen phosphonate intermediate, averting the dilemma of the PH-related DNA cleavage and genomic integrity discussed previously in the literatures. Figure 7 shows that the single-molecular PS-PO shortcut path led to the formation of highly reductive species, HS • , which as available for scavenging more surrounding ROS including O 2 −• and H 2 O 2 . The oxidation of HS • towards sulphur (S 8 ), sulfite (SO 3 2− ) and sulfate (SO 4 2− ) provided additional antioxidant capacity (see Fig. S9). Moreover, the computations clarified the reasons of why phosphorothioation frequently occurs at a guanine-rich location. Thus, at this stage, the biologically functional loop of anti-oxidative PT-DNA was closed with the dnd-gene-induced phosphate S-incorporation and the oxidative PS-PO recycling. It is also reasonable that eukaryote develops complicated karyotheca substructure tightly packing genetic materials away from mitochondria, while a prokaryote needs additional protector from oxidative metabolites. However, whether the PS-PO conversion cascades an in vivo ROS-signal to regulate PT-DNA expression in the ROS response needs further investigation. The novel mechanism of anti-ROS property of PT-DNA may be applicable in anti-tumor and anti-aging pharmaceutics. Methods General. E. coli DH10B harboring pJTU1238 was prepared as the E. coli (S + ) strain, and the one with dndC-disrupted pJTU1238 was used as the control in this work (namely the E. coli (S − ) strain). The E. coli (S + ) strain displayed the dnd phenotype in electrophoresis but the E. coli (S − ) strain did not. All chemicals were of analytical grade and used without further purification. 8-OHdG ELISA kit was obtained from the Shanghai Enzymelinked Biotechnology company. Plasmid Mini preparation kit was purchased from DingGuo Biotech company.

DCFH-DA determination.
In the in vivo ROS-level test, the 0.2 mg/ml DCFH-DA stock solution was added into the E. coli cultures to make the final concentration of 10 μ M. After 30 min incubation at 37 °C, the suspension was washed with Tris-HCl (pH 7.4) and diluted to 3 × 10 8 cfu. Then 150 μ l samples in 96-well plate were treated with Fenton's reagent. The intracellular ROS level was measured with the DCF fluorescence and the reaction condition was selected according to Fig. S1. 8-OHdG levels. The sample for the measurement of 8-OHdG levels was prepared in a four-hour incubation with Fenton's reagent, sonication in ice-cold Tris-HCl, and centrifugation at 8000 G to remove cellular debris. In the 8-OHdG test, the supernatant was mixed with the stop reagent of the ELISA kit, and the absorbance at 450 nm was measured to compare to a calibration curve constructed with the standard 8-OHdG solutions at 2.5 to 40 pg/ml. Pre-oxidation of plasmid PT-DNA. This procedure was developed based on the ABTS+ oxidation of GsA, in which GA was determined as the only oxidative product by HPLC and LCMS. Accordingly, the plasmid PT-DNA extracted from the dnd+ (S + ) E. coli was treated with 46.7 mM ABTS •+ and then purified as the pre-oxidized Ox(S+ ) plasmid DNA sample (Figs S2 and S7).
Raman spectra. Raman spectra were recorded on a Bayspec Raman spectroscopy. For each kinetic measurement, H 2 O 2 was added into the S + or S − E. coli strains washed by 10 mM PBS (E. coli OD 600 ~0.75, peroxide ~30 mM, in 10 mM PBS). Then the time-resolved spectra were collected in a rate of one spectrum per 7 seconds. The Raman intensity at the wavenumber of 876 cm −1 (O-O vibration) was used to monitor the peroxide concentration 45 .
CD spectra. CD spectra were measured on a Jasco J-815 spectropolarimeter. For each kinetic measurement, after running a blank spectrum with buffer (10 mM HEPES, pH 7.4), H 2 O 2 (final concentration was 1 μ M) was added into the plasmid DNA (final concentration was 63 μ M), and then 30 spectra were recorded in a rate of 1 spectrum per minute.
Gel electrophoresis. A typical gel electrophoresis was set on 1.5% agarose gels with 0.5 × TBE running buffer (45 mM Tris, 4.5 mM boric acid, 1 mM EDTA) at 100 V. Thereafter, the gels were stained with Geneview, visualized and digitally photographed by Alphalimager HP Gel Imaging system. Antioxidant assessment. Oxidant ABTS •+ was prepared by reacting ABTS with potassium persulfate according to the literature 46 . Using 46.7 μ M ABTS •+ solution, the scavenging rate (SR) was calculated in percentage of its feature absorbance at 734 nm after 30-min reacting with different amounts of the antioxidant. IC 50 value was estimated as the antioxidant concentration resulting in SR = 50%.
A similar method was used for the 0.1 mM DPPH oxidant. Instead, the feature absorbance at 520 nm was used to calculate the DPPH concentration 47 . IC 50 value was estimated as the antioxidant concentration resulting in SR = 50%.
The in vitro ROS level was measured by the modified TBARS assay 48 . The mixture of antioxidant and Fenton's reagent were vortexed and then settled for 30 min at 37 °C. TBA was dissolved in 10% trichloroacetic acid, with the final concentration of 0.67%. The solution was then heated in boiled water for 20 min. Finally the amount of TBARS formed in each sample was assessed with the absorbance at 532 nm.
The ferric-ferrous conversion was monitored with a modified colorimetric method, based on the absorbance of the Fe(II)-Ferrozine complex at wavelength of 562 nm.
Electrochemistry. Electrochemical measurement was performed with a CHI 660D electrochemical analyser, with a three-electrode system, i.e. a saturated AgCl electrode, a platinum wire, and a platinum disc electrode used as the reference, auxiliary, and working electrodes, respectively.
HPLC-MS analysis. The phosphorothioate oxidation was carried out in a mixed solution of 25 μ M PS or 20 μ M GsA-Rp and GsA-Sp with different concentrations of oxidants (H 2 O 2 or Fenton reagents), respectively. The mixtures were settled for 2 hours in the dark at room temperature before the HPLC-MS analysis. In the analysis, 10 μ l of reaction sample was loaded on Agilent's C18 reversed phase column (150 × 4.6 mm, 1.5 μ m), water and acetonitrile with 0.1% acetic acid were used as eluent A and eluent B in HPLC, respectively, and the flow rate was 0.3 ml/min with a gradient elution B from 5 to 95% in 10 min. Drying gas flow: 8 L/min; nebulizer pressure: 35 psi; drying gas temperature: 320 °C; capillary voltage: 3500 V (using the positive detection mode and scanning between 50 and 900 m/z for mass spectrometer analysis).
Quantum chemical calculation. All DFT (density functional theory) calculations were carried out with Gaussian 09 package 49 . Geometries were first optimized at the B3LYP/6-31 + G* level. In a typical calculation, the M06 and B3LYP levels of theory with the larger basis sets of 6-311 + G(d, p) were used to evaluate the geometric and energetic results in the gas phase. Solvent effect was considered with single point calculations of a continuum model (SMD) on reactants, transition states, intermediates, and products.