CO2 protects cells from iron-Fenton oxidative DNA damage in Escherichia coli and humans

Abstract

While hydroxyl radical is commonly named as the Fenton product responsible for DNA and RNA damage in cells, here we demonstrate that the cellular reaction generates carbonate radical anion due to physiological bicarbonate levels. In human and Escherichia coli models, their transcriptomes were analyzed by RNA direct nanopore sequencing of ribosomal RNA and chromatography coupled to electrochemical detection to quantify oxidation products in order to follow the bicarbonate dependency in H₂O₂-induced oxidation. These transcriptomic studies identified physiologically relevant levels of bicarbonate focused oxidation on the guanine base favorably yielding 8-oxo-7,8-dihydroguanine (OG). In human cells, the bicarbonate-dependent oxidation was further analyzed in the metabolome by mass spectrometry, and a glycosylase-dependent qPCR assay to quantify oxidation sites in telomeres. These analyses further identify guanine as the site of oxidation when bicarbonate is present upon H₂O₂ exposure. Labile iron as the catalyst for forming carbonate radical anion was demonstrated by repeating the bicarbonate-dependent oxidations in cells experiencing ferroptosis, which had a >fivefold increase in redox-active iron, to find enhanced overall guanine-specific oxidation when bicarbonate was present. The complete profiling of nucleic acid oxidation in the genome, transcriptome, and metabolome supports the conclusion that a cellular Fe(II)-carbonate complex redirects the Fenton reaction to yield carbonate radical anion. Focusing H₂O₂-induced oxidative modification on one pathway is consistent with the highly evolved base excision repair suite of enzymes to locate G-oxidation sites for repair and gene regulation in response to oxidative stress.

Keywords: DNA repair; DNA/RNA oxidation; Fenton reaction; bicarbonate.

Scheme 1.

Products of oxidative modification of nucleic acids by HO^• or CO₃^•− (hashed box), in which those in red were quantified in cellular RNA.

Fig. 1.

Changes in redox-active metabolite levels after adding H₂O₂ to the medium with dependency on the bicarbonate concentration. HPLC–MS/MS allowed quantification of metabolite levels in HEK293T cells preincubated in PBS medium with 0, 5, or 20 mM HCO₃⁻ for 1 h at 37 °C under atmospheric CO₂ before adding 100 μM H₂O₂. The reactions were allowed to progress for 15 min before quenching and analysis. Ratios for reduced vs. oxidized states of (A) GSH:GSSG and (B) NADH:NAD⁺ are reported, and the mass spectrometry (MS) intensities are provided for the nucleotide-monophosphates (C) AMP, (D) CMP, (E) GMP, and (F) UMP. The analyses were conducted with 3 to 6 replicates, and the levels of statistical significance are represented by *P < 0.05, **P < 0.01, and ***P < 0.001 calculated by a student’s t test.

Fig. 2.

Profile of RNA oxidation products or sites upon adding H₂O₂ to E. coli or human cells showing dependency on the bicarbonate concentration. The redox-active products from RNA base oxidation ho⁵C, OA, ho⁵U, and OG were profiled in (A) E. coli or (B) HEK293T total RNA via nuclease/phosphatase digestion of the polymers to nucleosides and HPLC-UV-ECD quantification (SI Appendix, Fig. S2 and Table S2). RNA direct nanopore sequencing of the LSU and SSU rRNA from (C) E. coli, (D) HEK293T cytosolic rRNA, and (E) HEK293T mitochondrial rRNA were profiled from the oxidized cells to measure changes in base miscalls with ELIGOS2 that report on modification sites in the strands (SI Appendix, Figs. S3 and S4) (21). The base miscall analysis for the E. coli and mitochondrial rRNAs was conducted by comparison of the cellular RNAs against a synthetic RNA of the same sequence without modification made by in vitro transcription (IVT). This approach permitted a profile of those RNAs in cells not exposed to oxidant [i.e., background (bkgd)]. In contrast, the human cytosolic rRNAs are too G/C rich to allow synthesis of the RNAs without modifications via IVT; therefore, the comparison for the oxidized samples was against the rRNA from the nonoxidized cells resulting in no background being reported. The background for the E. coli was obtained from cells grown for 24 h at 37 °C in LB Miller medium under atmospheric CO₂ levels, and the HEK293T cells were grown to ~80% confluency in DMEM in a humidified incubator with 5% CO₂ at 37 °C. The cells were placed in PBS with 0, 5, or 20 mM bicarbonate for 1 h (HEK293T) or 2 h (E. coli) before adding 100 μM H₂O₂. The oxidations proceeded for 15 min before quenching and harvesting the total RNA for analysis. The analyses were conducted in triplicate for HPLC-UV-ECD and duplicate trial for direct RNA nanopore sequencing with levels of statistical significance represented by *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 calculated by a student’s t test. *The OA values from the HPLC–UV–ECD analysis are an overestimation resulting from overlap in the elution of A and OA in the HPLC.

Fig. 3.

Bicarbonate dependency in telomere oxidation sites upon addition of H₂O₂ assayed by qPCR telomere length measurements. (A) Scheme to illustrate telomere DNA oxidation types analyzed via glycosylase removal of the damaged site. Direct measurement of the telomere length before and after oxidation reports on frank strand breaks that occur upon oxidation. Pyrimidine oxidation sites are revealed by EndoIII, an enzyme for which the preferred substrate is Tg, while the glycosylase can also remove 5hoC and 5hoU from DNA. Sites of sugar oxidation and abasic sites are substrates for EndoIV to yield strand breaks that are quantified by telomere-specific qPCR. Purine oxidation sites are found by Fpg, a glycosylase that favorably removes OG and Fapy-G from DNA, while the enzyme can also remove OA and Fapy-A. The bicarbonate dependency in H₂O₂-mediated oxidation of the telomeres was followed by qPCR to quantify (B) strand breaks, (C) EndoIII-sensitive sites, (D) EndoIV-sensitive sites, and (E) Fpg-sensitive sites (SI Appendix, Fig. S5 and Table S3). The oxidations were conducted by adding 100 μM H₂O₂ to HEK293T cells pre-equilibrated for 1 h in PBS medium with 0, 5, or 20 mM bicarbonate at 37 °C under atmospheric CO₂ levels, followed by reaction quenching and harvesting of the gDNA. The background (bkgd) measurements were obtained from cells not exposed to oxidant. The analyses were conducted in triplicate with levels of statistical significance represented by **P < 0.01 and ***P < 0.001 calculated by a student’s t test.

Fig. 4.

Iron dependency in oxidatively derived damage to HEK293T nucleic acids upon H₂O₂ addition. (A) Colorimetric assay determination of labile iron relative concentration before and after a 16 h treatment with the ferroptosis-inducing compound erastin. (B) Change in OG nucleoside levels in total RNA measured by HPLC–UV–ECD before and after erastin treatment, as a function of added H₂O₂ and bicarbonate to the culture medium. (C) Telomere lengths were measured by qPCR before and after erastin treatment and as a function of bicarbonate concentration during H₂O₂-mediated oxidation. qPCR quantification in telomeres of (D) EndoIII-, (E) EndoIV-, or (F) Fpg-sensitive sites in the high vs. basal iron level cells before and bicarbonate-dependent H₂O₂ oxidation. In all cases, background (bkgd) refers to cells grown under a 5% CO₂ atmosphere in DMEM without exposure to H₂O₂. In the oxidation studies, the cells were equilibrated with 0, 5, or 20 mM bicarbonate in PBS for 1 h before adding 100 μM H₂O₂ for 15 min at 37 °C under atmospheric CO₂ followed by quenching the reaction and harvesting the nucleic acids to be analyzed. The analyses were conducted in triplicate.