Screening DNA Damage in the Rat Kidney and Liver by Untargeted DNA Adductomics

Abstract

Air pollution, tobacco smoke, and red meat are associated with renal cell cancer (RCC) risk in the United States and Western Europe; however, the chemicals that form DNA adducts and initiate RCC are mainly unknown. Aristolochia herbaceous plants are used for medicinal purposes in Asia and worldwide. They are a significant risk factor for upper tract urothelial carcinoma (UTUC) and RCC to a lesser extent. The aristolochic acid (AA) 8-methoxy-6-nitrophenanthro-[3,4-d]-1,3-dioxolo-5-carboxylic acid (AA-I), a component of Aristolochia herbs, contributes to UTUC in Asian cohorts, and in Croatia, where AA-I exposure occurs from ingesting contaminated wheat flour. The DNA adduct of AA-I, 7-(2′-deoxyadenosin-N⁶-yl)-aristolactam I, is often detected in patients with UTUC, and its characteristic A:T-to-T:A mutational signature occurs in oncogenes and tumor suppressor genes in AA-associated UTUC. Identifying DNA adducts in the renal parenchyma and pelvis caused by other chemicals is crucial to gaining insights into unknown RCC and UTUC etiologies. We employed untargeted screening with wide-selected ion monitoring-tandem mass spectrometry (wide-SIM/MS²) with nanoflow liquid chromatography/ Orbitrap mass spectrometry to detect DNA adducts formed in rat kidneys and liver from a mixture of 13 environmental, tobacco, and dietary carcinogens that may contribute to RCC. Twenty DNA adducts were detected. DNA adducts of 3-nitrobenzanthrone (3-NBA), an atmospheric pollutant, and AA-I were the most abundant. The nitrophenanthrene moieties of 3-NBA and AA-I undergo reduction to their N-hydroxy intermediates to form 2′-deoxyguanosine (dG) and 2′-deoxyadenosine (dA) adducts. We also discovered a 2′-deoxycytidine AA-I adduct and dA and dG adducts of 10-methoxy-6-nitro-phenanthro-[3,4-d]-1,3-dioxolo-5-carboxylic acid (AA-III), an AA-I isomer and minor component of the herbal extract assayed, signifying AA-III is a potent kidney DNA-damaging agent. The roles of AA-III, other nitrophenanthrenes, and nitroarenes in renal DNA damage and human RCC warrant further study. Wide-SIM/MS² is a powerful scanning technology in DNA adduct discovery and cancer etiology characterization.

Graphical Abstract

Introduction

The kidney performs many essential functions, including controlling volume status, maintaining electrolyte and acid-base balance, and releasing hormones that regulate blood pressure. The kidney also plays a pivotal role in eliminating exogenous drugs and toxicants; however, renal oxidases and reductases may also bioactivate carcinogens, leading to nephrotoxicity and cancer.^1,2 The National Toxicology Program (NTP) has conducted over 500 rodent bioassays on chemicals, of which 69 induced renal tubule cancer adenomas and carcinomas and a small number of transitional cell tumors of the renal pelvis.³ Some chemicals assayed include atmospheric pollutants, occur in the diet, or arise in tobacco smoke; others are pro-oxidants, which are linked to DNA damage and human RCC risk through epidemiology studies,^4–10 mechanistic studies in cell culture or animal models.^11,12 The pattern of renal tumor occurrence in rodents is similar to that observed in humans, where over 80% of diagnosed renal tumors originate from the epithelium of the renal tubule cells.¹³ Gender differences have been observed in RCC, with male rodents being often more susceptible. The metabolic processing of chemicals may cause some of these gender differences in susceptibilities.^14,15 RCC also occurs nearly twice as often in men as in women in the USA.¹³ Obesity and hypertension are also risk factors for RCC.¹⁶ Kidney cancer represents about 3.7% of all cancers in the United States; however, the chemicals that damage the human renal genome are mainly unknown.

DNA adducts occur through exposure to xenobiotics, their reactive metabolites, or endogenous electrophiles formed by oxidative stress. Some DNA adducts induce genetic mutations and potentially contribute to cancer development. Exposure to aristolochic acid I (AA-I) occurs by consumption of Arisolochia herbaceous plants, which are used for medicinal purposes in Asia and worldwide, or in the Balkans by consumption of wheat flour contaminated with comingling Aristolochia herbs or via contamination of the cultivation field soil.^17,18 The role of AA-I DNA adducts-induced mutations in cancer driver genes in UTUC and RCC is well documented.^6,19–25 There is limited data on other mutation-inducing DNA adducts present in the human kidney. 8-Oxo-2′-deoxyguanosine (8-oxo-dG) and 8-oxo-2′-deoxyadenosine (8-oxo-dA) are two markers of oxidative DNA damage; both lesions occur in the human kidney. However, the levels of these oxidative adducts were not significantly different between medulla-RCC and non-RCC patients or between pelvis-UTUC and pelvis-non-UTUC groups.²⁶ In the same study, the investigators reported a reduction in the global level of 5-hydroxymethyl-2′-deoxycytidine in the non-tumoral urinary tract mucosa of urothelial carcinoma patients, which may impact epigenetic events in renal cancer development.

There is a critical need to characterize the human kidney DNA adductome to identify the chemical agents that damage the genome and lead to nephrotoxicity and cancer development. Identifying renal DNA adducts can provide further insight into the etiologies of RCC and UTUC, especially in countries with low exposure to AA. Targeted and untargeted adductomics screening are powerful approaches to identify and quantify DNA adducts formed from various agents.^27–30 Targeted adductomics screening analyzes for known DNA adducts associated with a particular DNA-damaging agent or class of agents. This approach requires prior knowledge of the potential adducts formed by the chemicals of interest. Untargeted adductomics screening is a broader analysis that aims to identify and characterize known and unknown DNA adducts without prior knowledge of their structures.³¹ This approach utilizes advanced mass spectrometry-based methods to detect and measure a wide range of adducts in the DNA. Untargeted adductomics can provide valuable insights into the overall adduct profiles induced by DNA-damaging agents and help discover novel adducts that may be associated with cancer etiology.

In this study, we have applied our untargeted, data-independent (DIA) wide-selected ion monitoring/tandem mass spectrometry (wide-SIM/MS²) technology to screen for DNA adducts by Orbitrap followed by wSIM-City software for data analysis of DNA adduct formation in kidneys and liver of rats dosed with a mixture of 13 environmental, tobacco, and dietary carcinogens that might contribute to human renal and liver DNA damage. ^28,32–34 The chemical structures of 24 DNA adducts known to be formed by these carcinogens are reported in Table S1. Our study aims to determine the sensitivity of the wide-SIM/MS² scanning technology and the breadth of coverage of DNA adducts formed in rat kidneys and liver, where many procarcinogens undergo bioactivation. We also employed targeted MSⁿ scanning and constant neutral loss data-dependent acquisition (DDA-CNL/MS³) scanning to characterize DNA adducts further,^27,35–37 and investigated the “gas-phase fractionation” (GPF) approach with ion trap MS² combined with Orbitrap MS²/MS³.^28,34,38 These MS scanning technologies screen for the neutral loss of the 2′-deoxyribose (dR) moiety (116.0473 Da) from the adducted 2′-deoxynucleosides (dN) during collision-induced dissociation (CID) or higher-energy C-trap dissociation (HCD). The resultant aglycone ions ([M+2H-dR]⁺ = [BH₂]⁺) are the principal product ions of most chemically stable DNA adducts.^{27,28,31,39–41}

Materials and Methods.

Materials.

Calf thymus DNA (CT DNA), DNase I (Type IV, bovine pancreas), Benzonase ultrapure powder (Santa Cruz Biotechnology, Dallas, TX, 500 Ku, sc-391121C, reconstituted at 300 u/μL in 20 mM Tris HCl, pH 8.0, 2 mM MgCl₂, and 20 mM NaCl, 50% glycerol). Alkaline phosphatase (Escherichia coli), nuclease P1 (from Penicillium citrinum), RNase A (bovine pancreas), Rnase T1 (Aspergillus oryzae), proteinase K (Tritiachium album), ethanol for molecular biology (200 proof), Tris-HCl, bis-Tris, 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), EDTA, Na₂HPO₄, β-mercaptoethanol (βME), isoamyl alcohol, and chloroform were purchased from Sigma-Aldrich (St. Louis, MO). Bond Elut C18 SPE cartridges (100 mg/mL) were from Agilent (Santa Clara, CA). Phosphodiesterase I (Crotalus adamanteus venom) was purchased from Worthington Biochemicals Corp. (Newark, NJ). Acrylamide (99% pure), 8-methoxy-6-nitrophenanthro-(3,4-d)-1,3-dioxolo-5-carboxylic acid (aristolochic acid I, (AA-I)) was obtained from Sigma Aldrich. AA-I was provided as a powder extracted from the stem of Aristolochia contorta (Jilin Province, China, product number: A5512, batch number: WXBD0062V, and reported at 90% purity by HPLC/UV). 4-Aminobiphenyl (4-ABP, 98% purity), benzo[a]pyrene (B[a]P, 96% purity), dimethylnitrosamine (DMNA, 93% purity), 2-naphthylamine (2-NA, 95% purity), 2-nitrofluorene (2-NF, 98% purity), and o-toluidine (o-tol 98% purity) were purchased from Sigma Aldrich. 3-Nitrobenzanthrone (3-NBA, 95% purity) was from Finetechnology-ind., Hubei, China. 2-Amino-1-methyl-6-phenylimidazo[4–5-b]pyridine (PhIP, 99% purity), 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline (MeIQx, 98% purity), 2-amino-9H-pyrido[2,3-b]indole (AαC, 99% purity), and 2-amino-3-methyl-9H-pyrido[2,3-b]indole (MeAαC, 97% purity) were purchased from Toronto Research Chemicals (North York, ON, Canada). [¹³C₁₀]-dG (99% isotopic purity), [¹⁵N₅]-dG, and [¹⁵N₅]-dA (>98% isotopic purities) were purchased from Cambridge Isotope Laboratory (Tewksbury, MA). Optima^™ LC-MS grade H₂O, CH₃OH, CH₃CN, formic acid (FA), Dimethyl sulfoxide (DMSO), and HCl (1.0 N stock solution) were purchased from Fisher Chemical Co. (Pittsburgh, PA). Chromacol 03-FISV LC-MS vials were from Thermo Scientific (Waltham, MA). 9-Methoxy-6-nitro-phenanthro-[3,4-d]-1,3-dioxolo-5-carboxylic acid (AA-VII) and 10-methoxy-6-nitro-phenanthro-[3,4-d]-1,3-dioxolo-5-carboxylic acid (AA-III) were provided by Dr. Francis Johnson, Stony Brook University.⁴²

The alkylated DNA adduct O⁶-methyl-2′-deoxyguanosine (O⁶-MedG) and its tri-deuterated homolog O⁶-[²H₃]-MedG was synthesized as reported.⁴³ The bulky aromatic DNA adducts N-(2′-deoxyguanosin-8-yl)-4-ABP (dG-C8–4-ABP), [¹³C₁₀]-dG-C8–4-ABP; N-(2′-deoxyguanosin-8-yl)-AαC (dG-C8-AαC), [¹³C₁₀]-dG-C8)-AαC; N-(2′-deoxyguanosin-8-yl)-MeIQx (dG-C8-MeIQx), [¹³C₁₀]-dG-C8-MeIQx; N-(2′-deoxyguanosin-8-yl)-PhIP (dG-C8-PhIP), [¹³C₁₀]-dG-C8)-PhIP; 10-(2′-deoxyguanosin-N²-yl)-7,8,9-trihydroxy-7,8,9,10-tetrahydrobenzo[a]pyrene (dG-N²-B[a]P), [¹³C₁₀]-dG-N²-B[a]P; 7-(2′-deoxyadenosin-N⁶-yl) aristolactam I (dA-N⁶-AL-I), [¹⁵N₅]-dA-N⁶-AL; 7-(2′-deoxyguanosin-N²-yl) aristolactam I (dG-N²-AL-I), [¹⁵N₅]-dG-N²-AL-I; 7-(2′-deoxyadenosin-N⁶-yl) aristolactam II (dA-N⁶-AL-II), [¹⁵N₃]-dA-N⁶-AL-II; 7-(2′-deoxyguanosin-N²-yl) aristolactam II (dG-N²-AL-II), and [¹⁵N₃]-dG-N²-AL-II were synthesized as described.^35,44–47 N-(2′-Deoxyguanosin-8-yl)-2-NA (dG-C8–2-NA), 1-(2′-deoxyguanosin-N²-yl)-2-NA (dG-N²-2-NA), 1-(2′-deoxyadenosin-N⁶-yl)-2-NA (dA-N⁶-2-NA), and [1,3,NH₂-¹⁵N₃]-dA-N⁶-2-NA were synthesized as reported.⁴⁸ 2-(2′-Deoxyguanosin-N²-yl)-3-aminobenzanthrone (dG-N²-3-ABA), N-(2′-deoxyguanosin-8-yl)-3-aminobenzanthrone (dG-C8-N3-ABA) and 2-(2′-deoxyguanosine-8-yl)-3-aminobenzanthrone (dG-C8-C2–3-ABA) were provided by Dr. Goncalo Gamboa da Costa, National Center for Toxicological Research, Jefferson, AR and Dr. Raji Singh, Leicester Cancer Research Centre, University of Leicester, UK.⁴⁹

Animal studies:

SAS Fischer 344 male rats were purchased from Charles River Lab (Wilmington, MA.) Animals were acclimated for one week under standard housing conditions on an AIN-76A diet. The animals were 10 weeks of age, and the average body weight at the start of the experiment was 185 ± 10 g. The animals were given an i.p. dose of a 13-carcinogen cocktail (0.465 ml DMSO containing each chemical at a concentration of 0.4 mg/mL), and the negative control animals received an i.p. injection of 0.465 mL DMSO. Five rats were used for DNA adductomics in the kidney of untreated and 13-carcinogen-dosed animals, and there were three untreated rats and four rats of the 13-carcinogen-dosed animals analyzed for liver DNA adducts. The list and structures of chemicals assayed are depicted in Scheme 1. The animals were given food and water ad libitum. The animals were euthanized with CO₂ at 24 h post-treatment. The liver and kidney were retrieved, rinsed three times in fresh, chilled phosphate-buffered saline (PBS) containing 10 mM BME, flash frozen on dry ice, and stored at −80 °C until DNA isolation.

Scheme 1.

Chemical structures of 13-carcinogen mixture dosed to the rats

Isolation and Enzymatic Digestion of DNA.

The tissues (~40 mg wet weight) were thawed on ice and homogenized in 2 mL of TE buffer (50 mM Tris-HCl pH 8.0 and 10 mM EDTA containing 10 mM βME) using a blade homogenizer (Pro Scientific, Oxford, CT). The homogenates were centrifuged at 3,000 g at 4 °C for 10 min. The nuclei pellet was digested with RNase A (150 μg) and RNase T1 (1.5 μg) for 1.5 h on a thermomixer at 37 °C, 900 rpm, followed by the addition of 0.1 vol of 10% SDS, and 0.4 mg proteinase K for 2 h. Protein was removed by adding chilled Puregene protein precipitation solution (250 μL) containing 10 mM βME (Qiagen, Germantown, MD), and residual lipids were removed from the supernatant by extraction with chloroform/isoamyl alcohol (24:1). DNA was precipitated by adding 0.1 vol. of 5 M NaCl followed by 1.5 volume of cold isopropanol and stored at −20 °C overnight. The DNA was retrieved by centrifugation, and the DNA pellets were washed twice in 70% ethanol/30% water. The DNA was briefly dried by vacuum centrifugation and reconstituted in LC-MS grade water. The DNA amount was estimated by UV spectra, assuming 1 unit of A₂₆₀ was equivalent to 50 μg of double-stranded DNA.

DNA (20 μg) in 5 mM Bis-Tris buffer (pH 7.1, 100 μL containing 10 mM MgCl₂) was spiked with isotopically labeled internal standards at 5 adducts per 10⁸ nucleotides (nts) levels before the digestion: [¹³C₁₀]-dG-C8–4-ABP, [¹³C₁₀]-dG-C8-AαC, dG-C8-[²H₃C]-MeIQx, [¹³C₁₀]-dG-C8-PhIP, [¹³C₁₀]-dG-N²-BPDE, [¹⁵N₅]-dG-N²-AL-I, [¹⁵N₅]-dA-N⁶-AL-I, [¹⁵N₃]-dG-N²-AL-II, and [¹⁵N₃]-dA-N⁶-AL-II. DNA samples were digested using a thermomixer at 37 °C, 900 rpm. Benzonase (300 U) and nuclease P1 (0.12 U) were added, and the incubation was done for 3.5 h, followed by adding phosphodiesterase 1 (3.2 mU) and alkaline phosphatase (40 mU) at 37 °C for 18 h. The digestion solutions were precipitated with 3 vol of ethanol and stored at −20 °C for 2 h. The precipitated enzymes were removed by centrifugation at 20,000 g at 4 °C for 10 min. The supernatant containing the unmodified nucleosides and DNA adducts was dried by vacuum centrifugation, reconstituted in 1:1 H₂O:DMSO (25 μL), sonicated, centrifuged at 21,000 g for 5 min, and transferred to silylated borosilicate glass inserts for LC-MS analysis. Online trapping was employed to remove non-modified 2′-deoxynucleosides.^28,50 Four μg of DNA was assayed by wide-SIM/MS² and CNL/MS³, and 2.4 μg DNA was assayed for targeted MS² and MS³. For O⁶-MedG analysis, DNA (20 μg) was spiked with O⁶-[²H₃]-MedG at 5.0 adducts per 10⁸ nts, employing the same digestion conditions reported above for the aromatic DNA adducts.⁵¹ Oasis HLB cartridges (30 mg) were conditioned with CH₃OH with 0.1% CH₃CO₂H, followed by LC-MS grade water containing 0.1% CH₃CO₂H (1 mL) before DNA digests were applied to the SPE cartridge. The SPE cartridges were washed with LC-MS grade water containing 0.1% CH₃CO₂H (1 mL), followed by 10% CH₃CN in water containing 0.1% CH₃CO₂H (1 mL), and O⁶-MedG was eluted with CH₃OH (1 mL). After drying by vacuum centrifugation, the SPE eluents were reconstituted in LC-MS grade water (25 μL), and 2.4 μg equivalent of DNA was assayed by targeted MS².

Mass Spectrometric and Nanoflow Liquid Chromatographic Analysis of DNA Adducts.

The non-polar aromatic DNA adduct analyses were performed on an UltiMate 3000 RSLC nano UHPLC System interfaced with an Orbitrap Fusion Tribrid MS (Thermo Fisher Scientific, Waltham, MA). Chromatography was performed using integrated column-emitters (75 μm ID × 200 mm, 10 μm orifice) from CoAnn Technologies, LLC (Richland, WA) and custom packed with Luna C18 stationary phase (5 μm particle size, 120 Å, Phenomenex Corp. Torrance, CA), and mounted in a Nanospray Flex ion source (Thermo Fisher Scientific, Waltham, MA). The LC solvents were (A) 0.05% HCO₂H in H₂O and (B) 0.05% HCO₂H in 95% CH₃CN. The DNA digests were injected on Thermo Acclaim PepMap trap cartridge RP C18 (0.3 × 5 mm, 5 μm particle size, 100 Å, Thermo Fisher Scientific, Waltham, MA) used to trap the aromatic DNA adducts present in the DNA digest. The trap column was washed with solvent A for 4 min at a 12 μL/min flow rate by the loading pump to remove the polar, non-modified nucleosides. After trapping, the adducts were back-flushed onto the analytical column at a flow rate of 0.6 μL/min for 1 min, then decreased to a flow rate of 0.3 μL/min, and a linear gradient commenced from 1% B and reached 99% B at 40 min. The column was held for 5 min at 99% B, followed by a 1 min gradient to 1% B over 1 min with a 6 min equilibration before commencing a new analysis. These conditions were employed for targeted and untargeted DNA adduct analyses of non-polar aromatic DNA adducts.

The samples for O⁶-MedG analysis were directly injected into a UChrom C18 column (0.75 μm ID X 150 mm 3 μm particle size) from Premier LCMS (Penn Valley, CA). The column was interfaced with an EASY-Spray^™ source (Thermo Fisher Scientific, Waltham, MA). The LC solvents were (A) 0.01% HCO₂H in H₂O and (B) 0.01% HCO₂H in 95% CH₃CN. The flow rate commenced at 0.30 μL per min and was held at 99% A for 4 min, then linearly increased to 12% B at 12 min, where the flow was decreased to 0.15 μL per min. The gradient increased linearly to 35% B at 33 min, then reached 99% B at 36 min, and held at 99% B for 4 min at a flow rate of 0.30 μL before returning to 99% A at 1 min and held for 6 min before commencing a new run.

General Orbitrap MS source parameters.

The source parameters were as follows: spray voltage, 2200 V (positive ion mode); ion transfer tube temperature, 275 °C; quadrupole isolation; isolation window wide-SIM (m/z 64), Orbitrap resolution, 120,000 (fwhm) at m/z 200; RF-lens, 40% (bulky aromatic and O⁶-MedG adducts). The data were acquired by Xcalibur^™ version 4.4 (Thermo Fisher Scientific, San Jose, CA).

Targeted Data MS² and MS³ Acquisition of DNA Adducts:

The DNA adduct levels were estimated by single-point calibration employing the aglcones’ areas produced by the targeted LC/MS² divided by the area of the isotopically labeled internal standards or surrogate internal standards. Adduct structures of the aglycone were characterized at the MS³ scan stage.^28,48 The MS² ion transitions [M+2H–dR]⁺ used to construct the extracted ion chromatograms (EICs) for adduct quantifications are: O⁶-MedG (m/z 282.1 → 166.0729) and O⁶-[²H₃]-MedG (m/z 285.1 → 169.0912); dG-C8–2-NA and dG-N²-2-NA (m/z 409.2 → 293.1145); dA-N⁶-2-NA (393.2 → 277.1196) and [1,3,NH₂-¹⁵N₃]-dA-N⁶-2-NA (m/z 396.2 → 280.1108); dA-C8–4-ABP (419.2 → 303.1353); dA-N⁶-2-AF (m/z 431.226 → 315.1353); dG-C8–4-ABP and dG-N²-N4-ABP (m/z 435.1 → 319.1302 ), [¹³C₁₀]-dG-C8–4-ABP (445.2 → 324.1470); dG-C8–2-AF and dG-N²-AF (m/z 447.2 → 331.1302); dG-C8-AαC (m/z 449.2 → 333.1207), [¹³C₁₀]-dG-C8-AαC (m/z 459.2 → 338.1375); N-(2′-deoxyguanosin-8-yl)-MeAαC (m/z 463.2 → 347.1363); dG-C8-MeIQx (m/z 479.2 → 363.1425), [¹³C₁₀]-dG-C8-MeIQx (m/z 489.2 → 368.1593); dG-C8-PhIP (m/z 490.2 → 374.1470), [¹³C₁₀]-dG-C8-PhIP (m/z 500.2 → 379.1640); dG-C8-N3-ABA/dG-N²-3-ABA (m/z 511.2 → 395.1251); 2-(2′-deoxyadenosin-N⁶-yl)-3-aminobenzanthrone (dA-N⁶-3-ABA, (m/z 495.2 → 379.1302); dG-C8-N3-ABA and dG-N²-3-ABA (511.1 → 395.1251); dA-N⁶-AL-I (m/z 543.2 → 427.1149); [¹⁵N₅]-dA-N⁶-AL-I (m/z 548.2 → 432.1001); dG-N²-AL-I (m/z 559.2 → 443.1098); dG-N²-B[a]PDE (m/z 570.2 → 257.0959, 285.0916, 454.1510), [¹³C₁₀]-dG-N²-B[a]PDE (m/z 580.2 → 257.0959, 285.0916, 459.1678). The normalized AGC was 200%, and the maximum injection time was 54 ms for the MS² and MS³ scan events. The HCD normalized CE was 20% for all adducts at the MS² scan stage, except for dG-N²-B[a]P which was set at 15% HCD. The HCD-normalized CE was 40% for all DNA adducts at MS³. The quadrupole isolation was set at 2 m/z for MS² and 3 m/z for MS³. The resolution was 30K for both scan events. All EICs were constructed with a 10 ppm mass tolerance window.

Untargeted DNA Adductomics Screening by wide-SIM/MS².

Wide-SIM/MS² DNA adductomics technique screens for putative DNA adducts by monitoring the co-elution of the modified nucleoside precursor ions ([M+H]⁺) in the wSIM scan mode and their corresponding aglycone ions ([BH₂]⁺) produced by the loss of [M+2H-dR]⁺ under HCD fragmentation at the MS² scan stage.²⁸ The wide-SIM/MS² method employed GPF and comprised of 10 scan events, where the odd-numbered events were wide-SIM scans, and the even-numbered events were MS². Each wide-SIM event had an isolation width of 64 m/z (including a 2 m/z overlap on both ends of the isolation ranges), screening a mass range of m/z 328–632. The sequential MS² scan events fragmented all ions in the same mass range with a product ion detection ranging from m/z 100 to 650 for MS² data acquisition. The quadrupole isolation mode was used. The normalized AGC was 3000%, the maximum injection time was 246 ms, and the resolution was 120K. For MS², the normalized AGC target was 2000%, and the maximum injection time was 118 ms. The HCD was 25% (normalized), except for the SIM window centered at m/z 600, which screens for dG-N²-B[a]P, where a stepped HCD of 10 and 15% was employed to maximize the signal for the aglycone of dG-N²-B[a]P.³⁴ The resolution was 60K, scanning from m/z 100 – 650. There was a substantial carryover for several hydrophobic DNA adducts of 3-NBA and AA of high abundance, which could result in cross-contamination of untreated control samples with the 13-carcinogen treated samples. Therefore, the samples were not randomized. The untreated DNA samples were assayed first, followed by the 13-carcinogen-treated samples. A blank injection of 1:1 H₂O:DMSO was injected before commencing the analysis of the untreated samples and before assaying the DNA of the dosed animals.

Generation of a Mass Inclusion List for DDA-CNL/MS³ screening.

The untreated (N = 5 for the kidney, N = 3 for the liver) and 13-carcinogen-dosed animals (N = 5 for the kidney and N = 4 for the liver ) were each pooled to have a single pooled control and treated sample per organ. These samples were then assayed by wide-SIM/MS² to generate a targeted mass inclusion list for DDA-CNL/MS³. The mass inclusion lists consisted of high-confidence score adducts identified in 13-carcinogen-treated samples with a minimum mean intensity of precursor ion of 5000 and a minimum aglycone intensity of 10,000 counts, observed in both pooled treated replicates. Putative DNA adducts were filtered to have scores > 0.95 and peak shape correlations > 0.6 (Pearson correlation coefficient of precursor and aglycone peak intensities).³⁴ The putative DNA adducts were further filtered and required a minimum number of scans > 4 in which the neutral loss was detected. The ratios of the MS²/MS¹ intensities were filtered for >0.5 and <10 ratios, and the putative adducts passing these criteria were then filtered further based on statistics (log2-fold change differences >1 over the untreated samples).

DDA-CNL/MS³ scanning methods.

Four different DDA-CNL/MS³ analyses were performed. Method 1 was conducted by untargeted scanning (no mass inclusion list). The method employed an exclusion list of background ions with retention times generated by injection of an enzyme digest without DNA using the “Background Exclusion” mode of the AcquireX feature of the operating software. Method 2 used two separate mass inclusion lists of putative DNA adducts obtained from pooled wide-SIM/MS² data at log2-fold changes >1 over the background adduct levels in untreated rats (p-values <0.05). Mass List 1 contained the 32 DNA adducts, including previously reported DNA adducts of the 13-carcinogen mixture listed in Table S1. Several additional DNA adducts of aromatic carcinogens (IQ, dG-C8-IQ; AA-II, dG-N²-AL-II, and dA-N⁶-AL-II) and isotopically labeled [¹⁵N₃ and ¹⁵N₅] internal standards of AA-I and AA-II were also screened. These additional DNA adducts were included to demonstrate the specificity of the DDA-CNL/MS³ screening. The internal standards were spiked at 5 adducts per 10⁸ nts and were positive controls. Mass List 2 was comprised of 60 (for kidney) and 58 (for liver) other putative DNA adducts tentatively identified by the pooled wide-SIM/MS² analysis, including five DNA adducts on Mass List 1 for cross-checking. DDA scanning priority was given to those putative adducts on Mass List 1. Method 3 employed untargeted scanning with GPF (GPF-DDA-CNL/MS³); no mass inclusion and background ion exclusion lists were used. Unlike Methods 1 and 2, the entire scan range was divided into 5 different GPF scan ranges to mimic the wide-SIM ranges used for wide-SIM/MS² analysis. The MS² detection was conducted with the ion trap, followed by criteria-based Orbitrap MS² (upon observing the neutral loss of dR −116.1 ± 0.2 Da in the ion trap MS²), followed by MS³. Method 4 employed the same GPF and the MS²/MS³ scheme of Method 3 and included the Mass List 1 from Method 2. These DDA-MS method parameters are summarized in Figure S1.

The DDA-CNL/MS³ methods 1 and 2 contained an MS scan window from m/z 297.5 to 602.5 followed by a data-dependent MS² (ddMS²) scan for ions with an intensity between 2.5 × 10³ and 5.0 × 10⁶ with a charge state of 1 or undetermined. The SIM scan mode setting (m/z 297.5 – 602.5) was employed because it permitted a higher AGC setting of 3000% to detect low abundant ions. In contrast, the full scan mode AGC setting maximum was 1250% and was less sensitive (unpublished observations, N. Ragi and Turesky). The resolution was 120K, and the maximum injection time was 246 ms. The MS² scan events employed a normalized AGC of 200%, and the maximum injection time was 100 ms. The MS² quadrupole isolation window was at 1.6 m/z. The HCD energy was 20% for all DNA adducts except for dG-N²-B[a]P, which was on mass list 1, and the HCD was 12%. The targeted loss inclusion was 116.0473 Da with a 15 ppm tolerance. A dynamic exclusion of 15 s was employed to eliminate the continuous fragmentation of the same adduct precursor if the MS² event was scanned 3 times within 30 s (with a mass tolerance of 15 ppm) and isotopes excluded. Only ions above the 2% threshold (relative intensity) were triggered for the MS³ scan event. For MS³, the AGC was 400%, the maximum injection time was 200 ms, and the MS³ isolation window was 3 m/z. The scan range m/z was 100 – 600, and the resolution was 30K for MS² and MS³ scan events. The data-dependent cycle time was 1.5 s, allowing for MS³ scans on the three most abundant ions detected on the mass list.

The GPF-DDA-CNL/MS³ methods 3 and 4 contained an MS scan from m/z 297.5 to 602.5 split into five scan segments with an isolation width of 64 m/z as done for wide-SIM/MS² (m/z 298–362, m/z 358–422, m/z 418–482, m/z 478–542, and m/z 538–602) with quadrupole filtering. The resolution was 120K, a maximum injection time of 250 ms, an AGC setting of 1000%, and a lock mass of 371.10124 m/z enabled. Ions within an intensity range between 2.5 × 10³ and 1.0 × 10⁶ were available for data-dependent MS² detection in the ion trap using the rapid scan rate with a quadrupole isolation of 1.6 m/z, HCD collision energy of 25%, a maximum injection time of 35 ms, and an AGC setting of 2000%. Dynamic exclusion was used with a 5 s duration, masses tolerances of ± 10 ppm, and isotopes excluded. MS² fragmentation with Orbitrap detection was performed for those ion trap MS² events for which a targeted loss inclusion (dR −116.1 ± 0.2 Da ) and relative intensity threshold of 5% criteria was met. MS² fragmentation in the Orbitrap was performed with the same parameters as the ion trap, except a maximum injection time of 50 ms was used, and the Orbitrap resolution was 15K. MS³ fragmentation was triggered upon observation of the accurate-mass neutral loss of 2′-deoxyribose (-dR: 116.0473 Da, 5 ppm) upon MS² fragmentation in the Orbitrap. MS³ fragmentation was performed with HCD with a normalized collision energy of 40%, maximum injection time of 50 ms, and Orbitrap detection at a resolution of 15K. The data-dependent cycle time for each of the five scan ranges was 0.6 s.

Analysis of DNA Adductomics Data Using the wSIM-City Algorithm.

The Thermo MS .raw data files were converted to the mzML and .MGF centroided data formats using Proteowizard v.3.0.19098 (2019–4-8) msconvert.exe raw file converter with the following command: “msconvert --filter ‘peakPicking true 1-’--simAsSpectra -- <rawfile name>”. The search algorithm for wSIM-City was provided with these mzML files as input. The statistical analysis and peak choosing processes are automated by the wSIM-City R library package.³⁴

DNA Adducts and Post-search Alignments, and DNA Adduct Scoring Criteria.

wSIM-City searches were completed using the following parameters: delta_search_mass = −116.0473, ppm_tol = 20, rt_tol = 0.1, mzmin = 135, mzmax = 634, and mzwid = 0.1.³⁴ The MS¹ and MS² m/z value intensity signals for each DNA adduct and its aglycone ([BH₂]⁺) were based on the observed maximum intensity peak detected for a given pair of Precursor – aglycone peak pairs detected within the neutral loss ppm tolerance window.³⁴ A custom R script was then applied to the peak intensity signals to assign the EIC peaks, using the matched filter algorithm from the R XCMS library,⁵² to detect the aglycone and precursor ion EICs within a 6-second retention time tolerance of each other. The MS² signals were employed in the remaining data alignment processes and statistical analysis since they typically have better signal-to-noise ratios than the precursor ions in wide-SIM scan mode.³⁴

Putative DNA adducts were scored using the calculated difference in m/z values between the precursor ions and the aglycones ([BH₂]⁺), accounting for the dR loss (−116.0473 Da, C₅H₈O₃). Molecular formula (MF) were calculated within a 5 ppm mass error for the m/z of both the precursor and the aglycone peaks and was used as filtering criteria for DNA adducts, keeping those precursor and aglycone peaks with an MF difference of -C₅H₈O₃ = 0. The refined list of adducts were then filtered by the observed ratio of the maximum MS² over MS¹ peak intensities (MS²/MS¹ ) >0.5 and < 10. Peaks detected across the twice double-injected pooled treated and untreated samples were aligned with the create_ref_table and join_align functions, using a retention time tolerance (rt_tol) = 0.5 and parts per million tolerance (ppm_tol) = 10.³⁴

Untargeted Data filtering and downstream statistics.

Statistics were completed using the R v4.3.1 programming language.⁵³ The wSIM-City algorithm was used to generate the mass inclusion list of putative DNA adducts from the pooled samples in the wide-SIM/MS² analysis described in Generation of a Mass Inclusion List for DDA-CNL/MS³ screening to select the precursor and aglycone ion pairs from the individual replicates of the kidney (N = 5) and liver (N = 4) from treated animals. Some MS² signals were missing across sample replicates. Therefore, a low random value near the minimum MS² detected intensity was used to impute any missing value and to calculate new p-values (Student’s t-test, one-sided, equal variances) and log2-fold changes (treatment/over control), principal components analysis (PCA), and the production of the volcano plots. The putative DNA adducts not present in at least 50% of the carcinogen-dosed group with a minimum mean signal intensity of 10,000 ion counts at MS² for either untreated or carcinogen-dosed animals were filtered out to reduce false positive data.

Results

Selection of chemicals.

The 13 chemicals selected for study (Scheme 1) are based upon findings from NTP long-term rodent bioassays, mechanistic studies in rodents,^11,12 or epidemiological studies linking genotoxicants in the environment, tobacco smoke, and dietary exposures to an increased risk of RCC, UTUC, or other urological cancers.^{4,5,8,10,20,25} The major DNA adducts formed by these carcinogens are shown in Table S1. The DNA adduct structures range from low molecular weight alkylated adducts formed with DMNA and acrylamide to bulky adducts formed with aromatic amines, heterocyclic aromatic amines (HAAs), polycyclic aromatic hydrocarbons (PAHs), nitro-PAHs, and AA-I. As described below, DNA adducts were screened by wide-SIM/MS² followed by targeted MS³ or CNL/MS³ scanning to obtain MS³ product ion spectra to support adduct identities.

Wide-SIM/MS² and DDA-CNL/MS³ adductomics scanning technologies and the wSIM-City Algorithm.

We employed nanoflow liquid chromatography and nanoelectrospray ionization coupled to Orbitrap MS using our untargeted DIA wide-SIM/MS² DNA adductomics scanning technology and wSIM-City algorithm to identify DNA adducts formed with the 13 carcinogen cocktail mixture.^28,34 Wide-SIM/MS² is a data-independent acquisition (DIA) scanning method. The wide-SIM scan event captures all precursor ions, and the ensuing MS² scan event covers all the fragment ions, including the aglycones. GPF was employed to increase the sensitivity of the analysis and simplify the spectra.^28,48 However, the MS¹ spectra remain a complex set of ions comprised of DNA adducts present at trace levels along with many other components from the background enzyme digest matrix, buffers, and chemicals co-purified with the DNA, which are present at much higher abundance than DNA adducts. The MS² spectra contain multiple fragment ions of these components in the wide-SIM/MS² scan event. Presumed known DNA adducts are visualized using extracted ion chromatograms (EIC) of the precursor ions [M+H]⁺ in wide-SIM MS scan window, and their corresponding aglycone ions [M+2H-116.0473 Da]⁺) in the MS² scan event. Wide-SIM/MS² allows for retrospective data mining for DNA adducts.^28,48 However, manual searching for all potential DNA adducts is labor intensive.

wSIM-City software processes mass spectral data acquired by wide-selected ion monitoring with GPF coupled to MS² fragmentation.^33,34 The wSIM-City algorithm is automated to detect an array of known and unknown putative DNA adducts analyzed by wide-SIM/MS².^34,54 The algorithm aligns the matching precursor ions in the wide-SIM scan event (64 m/z windows), and the co-eluting aglycone ions differing by −116.0473 Da in MS² scan event with high stringency for mass accuracy. The algorithm also examines matching retention times, relative ion abundances, and peak shape features of the precursor and aglycone ions to rank-score the putative adducts, providing an overall composite score (Figure 1).³⁴ Wide-SIM/MS² scanning showed superior coverage compared to CNL/MS³ scanning methods for detecting bulky, non-polar aromatic DNA adducts spiked-in calf thymus DNA.²⁸ Wide-SIM/MS² scanning also allows for semi-quantitation of adducts if stable isotopically labeled internal standards are present. However, unequivocal adduct identification requires an independent targeted MS³ or DDA-CNL/MS³ analysis, where unknown putative DNA adducts can also be further characterized by their MS³ spectra.

Figure 1.

Identification of DNA adducts using wide-SIM/MS² and wSIM-City. (A) Mechanism of 2′-deoxyribose loss of the modified DNA nucleoside under HCD. (B) Scanning method for detecting DNA adducts by high-resolution mass spectrometry (HRMS) showing multiple MS¹ and MS² isolation windows utilized for each duty cycle. (C) A scheme for the wSIM-City algorithm and (D) wSIM-City performing neutral loss search for 2′-deoxyribose loss and subsequent plotting of precursor [M+H]⁺ and aglycone [BH₂]⁺ ions.

The untargeted DDA-CNL/MS³ screening method in real-time selects the topmost abundant ions in the full scan spectra for fragmentation. Those precursor ions that have lost 116.0473 Da (± 15 ppm) to form the presumed aglycone ion undergo an MS³ scan event. The putative adduct is then placed on an exclusion list for a specific time to eliminate its continual fragmentation. DDA-CNL/MS³ scanning provides rich spectral data about the adduct structures with MS² and MS³ fragmentation data in a single run, an advantage over wide-SIM/MS² scanning. Employing a mass list of DNA adducts with prioritized scanning can improve the detection frequency of low-abundance DNA adducts by targeted CNL/MS³ (vide infra).

Identification of Known and the Discovery of Unknown Aristolochic Acid DNA Adducts by Wide-SIM/MS² employing the wSIM-City Algorithm.

AA-I is one of the 13 carcinogens dosed to rats; it is a potent DNA-damaging agent and a rodent renal carcinogen.⁵⁵ The power of wide-SIM/MS² scanning methodology in untargeted DNA adduct discovery is shown in the EIC profiles of the wide-SIM and MS² scan events of the 13 carcinogen-treated rats. Automated by wSIM-City extracts precursor and aglycone m/z peak intensities for each scan that showed a detected neutral loss of dR within 5 ppm mass tolerance immediately following the precursor ion scan (Figure 2). The EICs of the precursor ion peaks ([M+H]⁺) of dG-N²-AL-I at m/z 559.1572 (t_R 25.7 min) and dA-N⁶-AL-I at m/z 543.1623 (t_R 28.70 min) are shown in the wide-SIM scan and the corresponding co-eluting aglycone ions ([BH₂]⁺) for dG-N²-AL-I at m/z 443.1098 and dA-N⁶-AL-I at m/z 427.1149 are shown in the MS² scan (Figure 2). There are also two putative dG-AL-I isomers (t_R 25.2 and 26.6 min) and an abundant, presumed isomer of dA-AL-I (t_R 27.1 min), based on the accurate m/z values for precursor and aglycone ions. The targeted MS³ spectra of the dA-AL-I and the proposed dA-AL-I isomer have the two characteristic product ions attributed to the ionized AL-I moiety. The ion at m/z 292.0602 ([C₁₇H₁₀NO₄]⁺) is formed by inductive cleavage of the dA-N⁶-C7-AL bond, resulting in the neutral loss of adenine [C₅H₅N₅ 135.0545 Da] with the positive charge retained on the AL moiety. The second fragment ion at m/z 293.0677 occurs by homolytic cleavage of the dA-N⁶-C7-AL bond to form the AL radical cation ([C₁₇H₁₁NO₄]^•+) (Figure 3).⁴⁷ The targeted MS³ spectra of dG-N²-AL-I (t_R: 25.7 min) and the first proposed isomeric dG-AL-I adduct (t_R 25.0) display similar fragmentation patterns and also contain ions attributed to the charged AL moiety at m/z 292.0602 and 293.0677. The MS³ fragmentation pattern similarities among these modified nucleobases suggest that bond formation of these novel AA adducts occurs at the N²-dG and N⁶-dA atoms and the C7-AL atom; however, adduct formation at other sites of the AL moiety cannot be excluded (Figure 3).

Figure 2.

Wide-SIM/MS² and wSIM-City identification of known AA-I and novel AA DNA adducts in kidneys of untreated and 13-carcinogen-mixture treated rats 24 h after dosing. The precursor ions are depicted with black sticks, and the aglycones are depicted with red sticks. The arrows point towards the internal standards [¹⁵N₅]-dA-N⁶-AL-I and [¹⁵N₅]-dG-N²-AL-I. The top two rows represent dA-N⁶-AL-I and dG-N²-AL-I and their putative isomers, and the bottom two rows show [¹⁵N₅]-dA-N⁶-AL-I and [¹⁵N₅]-dG-N²-AL-I. The EIC data are from wSIM-City automated identification of the precursor and aglycone m/z peak intensities for each scan that showed a detected neutral loss of dR within a 5 ppm mass tolerance in the scan immediately following the precursor ion scan.

Figure 3.

MS³ spectra dA-N⁶-AL-I, dG-N²-AL-I, proposed novel dG-AL-I isomer and proposed dG- and dA-AA-III isomer adducts.

The second putative dG-AL-I isomer (t_R 26.6 min in Figure 2) was detected in rat kidneys but at very low levels in rat liver. This adduct undergoes less HCD fragmentation than the other AA-I DNA adducts, and the characteristic ions of the ionized AL moiety at m/z 292.0602 and m/z 293.0677 are not observed in the MS³ spectrum (Figure 3). The MS³ product ion base peak at m/z 428.0864 is attributed to the loss of the CH₃^• radical from the aglycone ([BH₂-CH₃]^•+), followed by ions at m/z 411.0841 attributed to the loss of CH₃OH [BH₂-CH₃OH]⁺; and m/z 401.0759 due to the losses of the CH₃^• and HCN ([BH₂-CH₃-HCN]^•+) (Figure 3). A recent study characterized the decomposition of N-sulfooxy-AL-I, an AA-I-activated form of AA-I,⁵⁶ by electron spin resonance, identifying a sulfate radical and two radicals centered at the N⁶ and C7 atoms of the AL-I moiety.⁵⁷ In the presence of dG, the homolytic cleavage of the N-sulfooxy-AL-I bond was predicted to produce multiple carbon and nitrogen-centered radicals of AA-I and dG, resulting in multiple dG-AA-I adducts.⁵⁷ Indeed, four novel dG-AL-I adducts were detected by the reaction of N-sulfooxy-AL-I with dG in vitro; their structures remain to be elucidated.⁵⁷ In contrast, the heterolysis of the N–OSO₃ bond of N-sulfooxy-AL-I forming the proposed reactive AL-I nitrenium/carbenium ion produces only dG-N²-AL-I and dA-N⁶-AL-I, which is most commonly accepted mechanism of AA-I DNA adduct formation.⁵⁸ Many activated aromatic amines, HAAs, and nitro-PAHs form DNA adducts predominantly at the C8 or N² positions of dG and, to a lesser extent, at the C8 and N⁶ positions of dA.⁵⁹ The structure of the putative dG-AL isomeric adduct (t_R 26.3 min in Figure 2) is uncertain. Notably, 3-NBA, a structurally related genotoxicant of AA-I, forms an uncommon adduct with dG, where bond formation occurs between C2 atom of the benzanthrone ring and the C8 atom of guanine to form dG-C8-C2–3-ABA.⁶⁰ Based on the MS³ fragmentation pattern of this novel dG-AL isomer (t_R 26.6 min in Figure 2, and mass spectrum in Figure 3), we hypothesize a stable carbon-carbon bond may have formed between dG and the C atom of the AL moiety of this novel dG-AL-I isomer, similar to that reported for the dG-C8-C2–3-ABA adduct.⁶⁰

Another putative AA adduct was detected by wSIM-City. The precursor ion occurred at m/z 519.1510, and its aglycone was at m/z 403.1037, with the EICs shown in Figure 4. The molecular weight of this adduct is consistent with the mass of an adduct formed between the activated AA-I and dC. The targeted CNL/MS³ product ion spectrum of the aglycone contains fragment ions at m/z values of 388.0794, 386.0770, and 371.0533, attributed to the losses of [BH₂-CH₃]^•+, [BH₂-NH₃]⁺, and [BH₂-NH₃-CH₃]^•+, respectively. The ion at m/z 292.0602 is assigned as the charged AL moiety, occurring by the neutral loss of cytosine from the aglycone [BH₂-C₄H₅N₃O]⁺ (Figure 4). The MS³ spectrum of this adduct is consistent with the proposed 7-(deoxycytidin-N⁴-yl) aristolactam I (dC-N⁴-AL-I) structure. A previous study reported the biomimetic formation of a dC-AL-I adduct upon reacting AA-I with zinc powder in the presence of calf thymus DNA.⁶¹ However, to our knowledge, dC-N⁴-AL-I has not been reported in rodents or human tissues. The mutational signature SBS22 ascribed to AA-I contains low C>G and C>T base pair mutations, which could implicate the dC-N⁴-AL-I adduct in AA-I mutagenesis and warrants study.²⁵

Figure 4.

wSIM-City EICs of wide-SIM/MS² analysis of precursor and aglycone ions of a proposed novel dC-N⁴-AL-I adduct in kidneys and liver of untreated and rats treated with the 13-carcinogen mixture 24 h following dosing. The precursor ions are shown in black, and the aglycone ions are depicted in red. The EIC data are from wSIM-City automated identification of the precursor and aglycone m/z peak intensities within a 5 ppm mass tolerance. The targeted CNL/MS³ spectrum and proposed fragments ions of dC-N⁴-AL-I are reported from the liver DNA of a 13-carcinogen-treated rat.

The proposed structural assignment dC-N⁴-AL-I and all other newly identified AA adducts and others are tentative. Large-scale syntheses of the adducts and characterization by NMR are required to elucidate their structures.

Identification of Aristolochic Acid Isomers in Commerical Aristolochia contorta Used For Dosing Rodents.

A commercial Aristolochia herbal extract (A. contorta, Jilin Province, China), not synthetic AA-I, was dosed to the animals. AA-I (8-methoxy) and its demethoxylated derivative 6-nitrophenathrene-[3,4-d]-1,3-dioxolo-5-carboxylic acid (AA-II) are often the primary genotoxic components in Arisolochia herbs.⁶² However, two other methoxy-substituted 6-nitrophenanthro-[3,4-d]-1,3-dioxolo-5-carboxylic acid moieties, AA-III and AA-VII, also occur in some Arisolochia plants.⁶³ The AA-I was extracted from the stem of this herb by a mixture of methanol and water and provided by the vendor as a powder. The AA-I purity by HPLC-UV was reported to be ~90%. We re-examined the AA-I herbal extract by HPLC using a UV/Vis diode array detector monitoring at 250, 320, and 390 nm (Figure 5). AA-I was the principal component (peak 2, t_R 19.31 min), accounting for ~85% of the UV/Vis absorbance at all wavelengths, followed by two minor components, each at about 5 – 8% relative abundance to AA-I. The peak at t_R 18.16 min co-eluted with synthetic AA-III and displayed an identical UV/Vis spectrum. AA-III occurs in some Aristolochia herbs, including A. argentina and A. chilensis.^63–65 The third peak at t_R 19.90 min is unknown; its retention time and UV/Vis spectrum are dissimilar to that of AA-VII (t_R 18.68 min, the peak of the reference chemical is not shown).

Figure 5.

HPLC/UV of AA-I (t_R 19.31 min) and AA-III (t_R 18.16 min) isomers in commercial Aristolochia contorta extract. The synthetic AA-VII has a t_R at 18.70 min (peak not shown) and a UV/VIS spectrum distinct from peak 3 at t_R 19.90 min in the extract.

The major dA-AL adduct in rat kidneys (t_R 27.1 min) and the first eluting dG-AL adduct (t_R 25.35) in rat kidneys (Figure 2) have the same retention times as the dA-AL-III and dG-AL-III adducts formed in kidney DNA of mice treated with AA-III and have matching MS³ spectra to those spectra in the rat kidney DNA (unpublished data, R. Turesky, V. Sidorenko, and T. Rosenquist). Assuming AA-I and AA-III have similar molar extinction coefficients and their corresponding isomeric dG-AL and dA-AL adducts have similar electrospray ionization and mass fragmentation efficiencies, the amount of the putative dA-AL-III isomer, based on peak area counts (MS² stage), is 4.6-fold greater than dA-AL-I, and the amounts of the novel dG-AL-III isomer at t_R 25.35 and the dG-AL isomer at 26.66 min, respectively, are 4.1 and 2.5-fold greater than dG-AL-I. Given that the amount of AA-III is ~10-fold lower than AA-I in the herbal extract, DNA adduct formation with AA-III in rat kidneys occurs at ~25 to 40-fold greater levels than those adducts formed with AA-I when normalized per mg of AA isomer dosed/kg body weight at 24 h post-dose treatment.⁵⁷

Identification of Other Aromatic DNA Adducts in Kidney and Liver by Wide-SIM/MS².

Wide-SIM/MS² and the wSIM-City algorithm screened for bulky aromatic DNA adducts of other carcinogens dosed, including 3-NBA, 4-ABP, 2-NA, 2-NF, o-Tol, PhIP, MeIQx, AαC, MeAαC, B[a]P. The dG-C8–4-ABP, dG-C8-MeIQx, dG-C8-AαC, dG-PhIP, and dG-N²-B[a]P adducts were identified in the kidney, liver, or both organs, based on the exact retention times to reference synthetic standards, or by their co-elution with stable isotopically labeled internal standards when available. The presumed dG-N²-N⁴-ABP isomer and dA-C8–4-ABP adduct were also detected.^66,67 Reference standards for these latter two DNA adducts were unavailable. In addition, several DNA adducts of 3-NBA,^49,60 2-NF,⁶⁸ two genotoxicants emitted in diesel exhaust, and renal carcinogens,^69,70 and MeAαC, a liver carcinogen, also emitted in diesel exhaust and present in tobacco smoke, were putatively identified.^71,72 The dG and dA adducts previously reported for 3-NBA,⁴⁹ dG-C8-AF,⁶⁸ and dG-C8-MeAαC⁷³ were detected in the kidneys or liver DNA of 13-carcinogen treated rats. Consistent with previous rodent studies, DNA adducts of o-Tol were not detected,^74,75 and only trace levels of 2-NA dG and 2-NA dA adducts were observed by targeted MS² (unpublished observations, N. Ragi and Turesky).⁷⁶

3-NBA, like AA-I, contains a nitrophenanthrene moiety. Several DNA adducts of 3-NBA were detected in the wide-SIM/MS² analysis, employing automated wSIM-City identification. A prominent adduct of dG-ABA (t_R 25.08 min) with two minor peaks at t_R 25.00 and 25.78 min were detected. (Figure 6). Synthetic dG-ABA reference standards enabled us to identify dG-N²-3-ABA (t_R 25.00 min) as the minor and dG-C8-N3-ABA (t_R 25.08 min) as the principal dG-ABA adducts, respectively (Figure 6). The targeted MS³ spectra are shown in Figure 7. The analysis shows that the synthetic dG-C8-C2–3-ABA elutes at t_R 25.78 min (Figure 6). The dG-C8-C2–3-ABA reportedly occurred at about 1% relative to dG-N²-3-ABA and dG-C8-N3-ABA in rats dosed with 3-ABA.⁴⁹ This proposed adduct appears as a shoulder peak at t_R 25.78 min. However, the low levels and the overlap with the predominating dG-C8-N3-ABA isomer prevented the acquisition of a well-resolved targeted MS³ spectrum to confirm its identity. The major dA-ABA adduct (t_R 27.02 min) and a presumed minor dA adduct (t_R 24.90 min) were detected. The MS³ spectrum of the principal dA-ABA adduct is consistent with the structure of the previously reported dA-N⁶-3-ABA adduct (Figure 7).⁴⁹ The MS³ spectrum of the previously unreported minor dA-ABA adduct contains the same fragment ion as dG-C8-N3-ABA at m/z 231.0800. This product ion is not seen for the other isomeric ABA adducts (Figure S2); this adduct may have formed at the C8 atom of dA and the N³ atom of 3-NBA.

Figure 6.

Wide-SIM/MS² EICs and wSIM-City identification of 3-NBA DNA adducts in the kidney of untreated and rats treated with the 13-carcinogen mixture 24 h post-dosing. The precursor ions are shown in black, and the aglycone ions are depicted in red. The arrows at t_R 25.00 and 25.78 min show where synthetic dG-N²-3-ABA and dG-C8-C2–3-ABA elute. The base peak at t_R 25.08 min is the retention time of synthetic dG-C8-N-3-ABA. The major peak is the proposed dA-N⁶-3-ABA adduct at t_R 27.02 min. The minor peak dA adduct at t_R 24.90 min is tentatively assigned as dA-C8-N-3-ABA. The EIC data are from wSIM-City automated identification of the precursor and aglycone m/z peak intensities within a 5 ppm mass tolerance.

Figure 7.

Targeted MS³ spectra of 3-ABA DNA adducts.

Representative untargeted wide-SIM/MS² EIC for DNA adducts of aromatic amines, HAAs, B[a]P-dosed carcinogens in kidney and liver DNA rats automated by wSIM-City are depicted in Figure 8, and the DDA-CNL/MS³ spectra are provided in the supporting information (Figure S2 – Figure S6). The supporting information also provides unfiltered EICs of these precursor and aglycone ions of DNA adduct detected by wide-SIM/MS² and manually extracted using FreeStyle version 1.8 (Figure S7 – Figure S8) or automated by wSIM-City (Figure S9).

Figure 8.

Wide-SIM/MS² and wSIM-City EICs of 4-ABP, HAA, B[a]P DNA adducts in kidney DNA of untreated and treated rats given a 13-carcinogen mixture 24 h following dosing The precursor ions are shown in black, and the aglycone ions are depicted in red. The EICs identified by wSIM-City are plotted for dG-C8–4-ABP, dG-C8-AαC, dG-C8-PhIP, dG-N²-B[a]P, and their aglycone m/z peak intensities within 5 ppm mass tolerance.

Estimated Levels of Nitroaromatic, Aromatic Amine, Heterocyclic Aromatic Amine, B[a]P, and O⁶-MedG Adducts in the Kidney and Liver by Targeted MS².

The liver is a major organ for the bioactivation of many aromatic amine and HAA procarcinogens by cytochrome P450s (P450 1A2 and P450 2E1);^77,78 the kidney also contains oxidases and reductases, including NAD(P)H:quinone oxidoreductase (NQO1), which bioactivate AA-I and 3-NBA.^1,58 Following untargeted adduct screening by wide-SIM/MS², the known and putative DNA adducts were analyzed by targeted MS² monitoring [M+H]⁺ → [M+2H-dR]⁺. A single-point calibration was employed using the internal standards of [¹³C₁₀]-dG-C8-HAAs, [¹³C₁₀]-dG-C8–4-ABP, [¹³C₁₀]-dG-N²-B[a]P, and [¹⁵N₅]-dA-N⁶-AL-I and [¹⁵N₅]-dG-N²-AL-I DNA adducts, the latter of which were used as surrogate internal standards for estimating isomer AA and isomeric ABA DNA adduct levels.

The dG adducts of 3-NBA (incompletely resolved dG-N²-3-ABA/dG-C8-N3-ABA/dG-C8-C2–3-ABA isomers), followed by dA-N⁶-ABA, the newly discovered dA-N⁶-AL-III isomer, dA-N⁶-AL-I, the novel dG-N²-AL-I isomer and O⁶-MedG were the most abundant DNA adducts formed in rat kidneys (Figure 9). Adduct level estimates were made on nineteen of the 21 DNA adducts detected (dG-C8–2-NA and dC-N⁴-AL-I were not measured, and the unresolved dG-N²-3-ABA and dG-C8–3N-ABA isomers were measured and combined as dG-3-ABA adducts. Thus, DNA adduct levels were determined for 18 DNA adducts. Striking differences in adduct levels of these carcinogens occurred between the kidney and liver (Figure 10). The levels of the proposed dA-N⁶-AL-III and dG-N²-AL-III adducts were significantly higher in the kidney than the liver at 24 h post-treatment. In contrast, the dA-N⁶-AL-I and dG-N²-AL-I levels were not significantly different between these organs. The proposed dC-N⁴-AL-I adduct was not part of the targeted analysis; the relative ion abundances of the precursor ion and aglycone signal dC-N⁴-AL-I in wide-SIM/MS² suggests that the adduct level was about 2-fold higher in the liver than kidney, assuming similar ionization efficiencies in both DNA samples (Figure 5).

Figure 9.

Estimates of DNA adducts in liver and kidney DNA of rats treated with 13-carcinogens 24 h post-dosing. [¹³C₁₀]-dG-C8–4-ABP was employed to estimate levels of dG-N²-N⁴-4-ABP and dA-C8–4-ABP; [¹³C₁₀]-dG-C8-AαC was used to estimate dG-C8-MeAαC and dG-C8-AF; [¹⁵N₅]-dA-N⁶-AL-I was employed to estimate dA-N⁶-AL-III and dA-ABA isomers, and [¹⁵N₅]-dG-N²-AL-I was used to estimate dG-ABA isomers which were estimated as a “single” adduct due to overlapping peaks.

Figure 10.

DNA adduct levels in kidneys and liver of rats treated with a 13-carcinogen mixture 24 h following dosing. [¹³C₁₀]-dG-C8–4-ABP was employed to estimate levels of dG-N²-N⁴-4-ABP and dA-C8–4-ABP; [¹³C₁₀]-dG-C8-AαC was used to estimate dG-C8-MeAαC and dG-C8-AF; [¹⁵N₅]-dA-N⁶-AL-I was employed to estimate dA-N⁶-AL-III and dA-3-ABA isomers, and [¹⁵N₅]-dG-N²-AL-I was used to estimate dG-3-ABA isomers. The statistical significance between adduct levels formed in the kidney and liver was done with Welch’s unpaired t-test using Prism 8.4.3 (GraphPad Software, La Jolla, CA). (*P < 0.05; **P < 0.01, ***P < 0.001, ****P < 0.0001)

The dG-ABA and dA-N⁶-ABA adducts were also significantly higher in the kidneys than in the liver. The dG- and dA adducts of 4-ABP, dG-C8-AαC, and dG-C8-MeAαC were higher in the liver than in the kidney, perhaps by efficient hepatic P4501A2 bioactivation.^78–80 In contrast, dG-C8-MeIQx and dG-N²-B[a]P levels were similar in both organs, while the dG-C8-PhIP level was significantly higher in the kidney than in the liver. In contrast to NBA and AA, which form DNA adducts at high levels in the kidney, the proposed dG-C8-AF adduct formed following nitro reduction of 2-NF to the reactive 2-hydroxyaminofluorene intermediate was detected in the liver only at 3.0 ± 0.3 adducts per 10⁹ nts. It was not found in the kidney (<5 adducts per 10¹⁰ nts).

DMNA was one of the 13 carcinogens dosed to rats. DMNA undergoes metabolism by P4502E1 to produce a reactive methylating agent to form O⁶-MedG as the major stable adduct of this procarcinogen.^77,81 We employed offline SPE to isolate O⁶-MedG because it is poorly retained by our online trapping method.⁴³ Quantitative targeted MS² analysis using the O⁶-[²H₃]-MedG internal standard showed that O⁶-MedG was 7.6-fold higher in the liver than in the kidney (10.7 ± 2.8 versus 1.4 ± 0.3 adducts per 10⁷ nts, p = 0.0016 ). O⁶-MedG is one of the predominant adducts formed in the liver and kidneys of rats treated with this 13-carcinogen mixture (Figure 9 and Figure 10). The relative estimates of these different adducts formed in the liver and kidney should be interpreted cautiously since the kinetics of DNA adduct formation and their peak levels were not determined.⁸² Furthermore, surrogate internal standards were employed for some adducts and could lead to less precise estimates.

DNA Adductomics Analysis by wSIM-City.

wSIM-City was used to profile non-polar and aromatic DNA adducts in the wide-SIM/MS² data of pooled kidney and liver samples of rats exposed to the 13-carcinogen cocktail and detected many other putative unknown DNA adducts (16678 and 14631 total ion pairs across pooled kidney and liver samples, respectively). Many of these signals were considered noise or false positives caused by poorly defined EICs, unrealistic MS²/MS¹ ratios, or artifacts in the data. Based on the mass lists of DNA adducts obtained from the pooled samples, we searched for similarly identified putative DNA adducts by wSIM-City in the individually treated rats (N = 5 for treated and untreated kidneys, and N = 4 for treated liver and 3 for untreated liver controls). These putative DNA adducts identified from the independent wSIM-City search results were extracted using a < 7 ppm mass error and a retention time window of ± 0.5 min. Data were subsequently filtered for missing data (≥ 50% present in the treatment groups), resulting in a table of intensities that included 3956 and 3381 adducts for the kidney and liver samples, respectively. These tables were used to perform principal components analysis and statistics for volcano plots.

The EICs of several known and newly identified DNA adducts, including dA-N⁶-AL-III, were extracted by automation with wSIM-City (Figures 2, 4, 6, 8). The PCA of the DNA adduct aglycone [M+2H-dR]⁺ intensities clustered the DNA samples into dosed and undosed (control) groups (Figure 11). Significance testing Student’s (t-test, p < 0.1 log2 FC <−1 or >1) detected 333 significantly altered putative DNA adducts in the kidney (148 are higher in the carcinogen-dosed group) and detected 356 significantly altered putative DNA adducts in the liver (205 were higher in the carcinogen-dosed group). The volcano plots show that many putative DNA adducts of the 13 carcinogens dosed were elevated with p < 0.10 in treated animals) (Figure 11). However, the pooled and individually-dosed animals were assayed by LC-MS at different times. Several DNA adducts initially identified in pooled wide-SIM/MS² samples were not detected by wSIM-City in the individually treated animals because of retention time differences (> 30 s) and failed to be selected for alignment. For example, low abundance dG-C8-MeAαC and the proposed dC-N⁴-AL-I adducts were missed in both the liver and kidney, and dG-N²-B[a]P was missed in the kidney. By relaxing the retention time search tolerance, these adducts can be selected for alignment. However, that may introduce considerably more ion pairs within the search tolerance window and potentially reduce the accuracy of the curation. Furthermore, the signals of several low-abundance DNA adducts were not significantly different from the background signals of the control animals in the wide-SIM/MS² as show in the volcano plots, such as dG-C8-MeIQx (Figure 11). This adduct is above the LOQ value in targeted MS² but occurs at low levels (1 – 2 adducts per 10⁸ nts). The untargeted wide-SIM/MS² scanning method is several fold less sensitive than targeted MS², and the dG-C8-MeIQx signal was only above 10,000 counts in one of the 5 kidney DNA samples but present in 3 of the 4 carcinogen-dosed liver samples. However, the imputation required for statistical analysis caused high variation, resulting in p-values > 0.1. Several other known DNA adducts occurring at low abundances (<1 adduct per 10⁸ nts) also failed to show significant differences from the background signals of the control animals when analyzed by wide-SIM/MS² analysis.

Figure 11.

Principal components analysis (PCA) and Volcano plots for untargeted DNA adductomic analysis from DNA of kidneys and livers from the control (CTRL) and rats treated (TRT) with a 13-carcinogen mixture 24 h following dosing. The PCA plots are the putative DNA aglycone adduct intensities of A) rat kidney and (B) liver. The data were clustered by PCA using the ropls R library. Black, red, and blue ellipses represent 95% CI ellipses for total, dosed, and control samples. The volcano plots of (C) kidney and (D) liver are reported as −1*log(p value) versus log2 fold change of MS² intensities for dosed over control groups for DNA adducts present in at least 50% of the dosed group and a mean intensity over 10,000 counts in either the control or carcinogen-dosed groups. Grey dots represent putative DNA adducts detected in the kidney or liver tissues. Red dots show wSIM-City identified and validated known DNA adducts detected. The blue lines mark DNA adduct intensities (as log 2 FC of −1 or 1) and Student’s t-test (p-value < 0.1 expressed as −1 * log (p-value), one-sided, equal variances). Note, the signal intensities for several known DNA adducts identified at low abundance by targeted MS² in the dosed animals were detected in <50% of the sample by wide-SIM/MS² and filtered out (p-value >0.1). The DNA adduct signals were manually added back to the volcano plot data for visualization purposes.

Targeted MS³, DDA-CNL/MS³ and GPF-DDA-CNL/MS³ Acquisition.

We employed criteria established in the metabolomics field to identify DNA adducts.⁸³ Confirming DNA adduct identity required co-elution with a synthetic reference standard for retention time and a matching MS³ spectrum of the modified nucleobase. There are many putative adducts of unknown structures. Targeted MS³ scanning provides the most robust and reliable mass spectral data, particularly for low abundance DNA adducts, where spectral averaging across the peak can improve the signal over the background noise and provide cleaner, high-quality MS³ spectra than those obtained with DDA scanning where only one up to several scans are acquired and co-isolated background signals cannot be removed. However, there is a limit to the number of putative DNA adducts that can be targeted for MS³ analysis, although a normalized retention time technique with scheduled MS³ scanning can increase DNA adduct spectral coverage.⁸⁴ We explored the untargeted DDA approach, which can identify more unknown adducts across the chromatogram than possible with targeted MS³ scanning.

DNA adduct formation resulting from exposure to two polar carcinogens, DMNA and acrylamide, and 11 carcinogens containing aromatic moieties (Scheme 1) was examined using our wide-SIM/MS² data acquisition with wSIM-City analysis to establish a mass list of putative DNA adducts to prioritize for targeted MS³ and DDA-CNL/MS³ scanning. DMNA forms O⁶-MedG and several DNA adducts of acrylamide and its reactive epoxide metabolite glycidamide (GA) were previously identified in mice and rats, with higher adduct levels formed for GA.^85,86 GA forms N7-(2-carbomyl-2-hydroxethyl)-guanine (N7-GA-Gua) and N3-(2-carbomoyl-2-hydroxyethal)-adenine (N3-GA-Ade). These adducts undergo deglycosylation and were measured as the modified nucleobases in previous studies.^85,86 Moreover, even if some portion of the adducts were stable towards our DNA isolation and nuclease digestion procedures, our online trapping method does not appear suitable for analyzing these adducts due to their high polarity. We decreased the starting m/z range to encompass putative acrylamide and GA DNA adducts. However, these adducts were not detected by wide-SIM/MS² or targeted DDA-CNL/MS³ (unpublished observations, N. Ragi and Turesky). The low molecular weight aromatic amine o-Tol forms several isomeric DNA adducts with dG and dA in vitro, but they were not detected in rodents.^74,87 Our untargeted and targeted LC-MS analyses also did not detect o-Tol DNA adducts in rat kidneys or liver. Of the 2-NA adducts, only trace levels of dG-C8–2-NA were detected by targeted MS², occurring in rat liver at levels of approaching 1 adduct per 10⁹ nts. Nineteen DNA adducts of the nine remaining aromatic carcinogens were identified by targeted MS³ (Table S2). The proposed dC-N⁴-AL-I was detected by wSIM-City after the targeted MS³ assays had been performed.

We determined the efficacy of detecting these 19 DNA adducts (Table S2) and screening other putative unknown DNA adducts employing DDA-CNL/MS³ and GPF-DDA-CNL/MS³ scanning methods compared to wideSIM/MS². Some adducts are present in DNA at ~1 adduct per 10⁸ nts or lower. The DDA-CNL/MS³ product ion spectra of representative DNA adducts are reported in Figures S2 – S6. Four DDA-CNL/MS³ scanning strategies were evaluated (see DDA-CNL/MS³ scanning methods in Materials and Methods, and Table S2). With the untargeted DDA-CNL/MS³ analysis (Method 1), 16,663 MS² and 137 MS³ spectra were acquired on kidney DNA, and 15,926 MS² and 183 MS³ spectra were obtained from the liver DNA, of which 4 DNA and 5 DNA adducts, respectively, in the kidney and liver, were previously detected by wide-SIM/MS². High-quality DDA-CNL/MS³ spectra were acquired on dA-N⁶-AL-I, dG-N²-3-ABA and dG-C8-N-3-ABA, dA-N⁶-ABA, and dG-C8–4-ABP. With the targeted DDA-CNL/MS³ analysis containing the inclusion mass lists (Method 2), 1640 MS² and 338 MS³ spectra were acquired from the two mass lists in the kidney, and 1525 MS² and 504 MS³ spectra were acquired in the liver sample mass lists. Twelve DNA adducts in the kidney and 17 from the liver DNA previously detected by wide-SIM/MS² displayed high-quality DDA-CNL/MS³ spectra. The untargeted GPF-DDA-CNL/MS³ analysis (Method 3) detected five adducts among the most abundant DNA adducts previously detected by wide-SIM/MS² in the liver and six in the kidney samples. With the GPF-DDA-CNL/MS³ analysis containing the inclusion mass lists (Method 4), 13 DNA adducts were detected in both the liver and kidney samples. Thus, DDA-CNL/MS³ screening was not as sensitive and comprehensive as the wide-SIM/MS² with the wSIM-City algorithm, even when employing a background exclusion list.

The DDA-CNL/MS³ detected other putative DNA adducts. Several putative lipid peroxidation DNA adducts were identified;⁸⁸ however, the precursor ions of many other putative adducts are not reported in our or another comprehensive database.^34,89 Many of the DDA-CNL-MS³ spectra were of poor quality; other spectra showed protonated guanine or adenine as fragment ions, but further interpretation is required to propose plausible DNA adduct structures.

GPF and Ion Trap MS² (GPF-DDA CNL/MS³): A Novel and Promising Approach for DNA Adductomics.

Two significant factors limit the data-dependent screening sensitivity of our DDA-CNL/MS³ approach and, more generally, other trap-based data-dependent screening assays. The first is the trapping capacity of the Orbitrap, which determines the full scan limit of detection of the precursor ions. The second is the rate at which the instrument can scan, which determines the number of detected precursor ions that can be sampled by MS² fragmentation. Either of these events can be the limiting factor in determining the sensitivity of an assay. In one case, all detected precursor ions can be fragmented, but there may be analytes of interest below the precursor ion limit of detection. In the other case, all precursor ions of interest are detected, but the instrument is not scanning fast enough to sample all the observed precursor ions. It is unclear which of these scenarios is the limiting factor for our standard DDA-CNL/MS³ approach with the samples analyzed in the experiment. Our working hypothesis is that both limiting factors must be addressed to make the methodology capable of screening for all relevant DNA adducts in the in vivo samples.

We sought to improve the full scan sensitivity of our DDA-CNL/MS³ approach by incorporating GPF using the fractionated m/z scan ranges of the SIM window of the wide-SIM/MS² approach. GPF reduces the mass range of ions entering the Orbitrap, which should result in more detectable ions within the narrower mass range; however, this also means more instrument time is spent acquiring full scan data, and therefore, less time is available for MS² fragmentation. To offset this aspect, we experimented with using the ion trap for more rapid MS² data collection and the ability to perform MS² fragmentation while the Orbitrap is acquiring full scan data. Accurate mass MS² and MS³ spectra were acquired by performing data-dependent Orbitrap MS² fragmentation upon observation of the nominal mass loss (± 0.5 Da) of dR (−116 Da) in the ion trap MS² fragmentation spectra, followed by Orbitrap MS³ fragmentation upon observation of the accurate mass loss (5 ppm) of dR (−116.0473 Da). A comparison of this GPF-DDA CNL/MS³ approach with that of the standard DDA-CNL/MS³ approach is shown in Table S2, and the precursor, CNL/MS² and MS³ scans across the peaks are shown in supporting information (Figure S10). The GPF-DDA-CNL/MS³ did not outperform the standard approach; however, there were significant additional differences in the method, precluding a direct comparison of the results. Utilizing GPF and/or ion trap fragmentation requires further development and optimization and will be investigated in future efforts.

DDA-CNL/MS³ Data Analysis using the Mass Spectral Database.

The data acquired during DDA-CNL/MS³ analysis is well-suited for DNA adduct identification through our recently completed DNA adduct mass spectral database.⁹⁰ Our database is available for manual examination and downloading in the .MSP (NIST format)91 and .db (mzVault format, Thermo Scientific)92 formats at the DNA Adduct Portal Web site (https://sites.google.com/umn.edu/dnaadductportal) Our database is available for manual examination and downloading in various formats, including .MSP (NIST format),⁹¹ .csv (comma-separated values), and .db (mzVault format, Thermo Scientific)⁹² at the DNA Adduct Portal website (https://sites.google.com/umn.edu/dnaadductportal). The .MSP format is highly flexible and can be used directly with open-source tools such as NIST MS Search and MS-DIAL.93 and Global Natural Product Social Molecular Networking (GNPS) with molecular spectrum networking to aid in identifying unknowns.⁹³ Also, the .MSP files can be converted to the NIST proprietary format with the free NIST Lib2NIST program (https://chemdata.nist.gov/mass-spc/ms-search/Library_conversion_tool.html) for use in any software using this format. For example, our DDA-CNL/MS³ spectra from liver of the 13-carcinogen-dosed rats were searched against the spectral library produced search results (using NIST MS Search score) found dG-C8-PhIP (score 93.8), dG-C8-MeIQx (score 98.0), dG-C8-AαC (score 83.9), dG-C8–4-ABP (score 87.8) (Figures S2, S4 and S5).

Discussion

Matsuda’s laboratory in 2006 was the first to screen for multiple DNA adducts in human tissues.^94,95 They showed that the triple quadrupole (QqQ-MS) could screen multiple DNA adducts by monitoring a wide range of SRM transitions ([M + H]+ > [M + 2H −116]⁺) with single integer m/z increments covering the entire mass range of putative adducts. Many putative DNA adducts were detected in the human esophagus and lung, and some adducts appeared to occur at different levels between the two tissues. It is not known whether these precursor ions are DNA adducts or simply other compounds that underwent the loss of 116 Da in tandem MS. This untargeted technique was also employed in other DNA adductomics applications, including lipid peroxidation DNA adducts in human gastric mucosa.⁹⁶

Our laboratory extended the field of DNA adductomics by employing the linear ion trap (LIT) MS, which allows for multistage MSⁿ scanning and complete spectral characterization of the DNA adducts.³⁵ The CNL/MS³ scan was performed from the DDA-MS² acquisitions using an ion count threshold. The loss of dR (116 ± 0.5 Da) was set as the criteria to trigger the subsequent MS³ scan of the top 5 to 10 most abundant ions to obtain the fragmentation spectra. We applied this scanning technology to screen for DNA adducts in human hepatocytes treated with the tobacco carcinogen 4-ABP and in the liver of rats treated with cooked-meat carcinogen MeIQx. This study demonstrated that the DDA-CNL/MS³ scanning mode can be performed with an LIT. Ion trap multistage MSⁿ scanning is a significant improvement over the CNL experiments conducted by QqQ, where the only information attained is that an ion of a given mass underwent the neutral loss of 116 Da in tandem MS. There are several key advantages of using the MS³ over MS² scan stage for screening DNA adducts. The MS³ scanning provides superior specificity than the MS² scan stage and, hence, a better signal-to-noise ratio, which is required for measuring low levels of adducts in cellular and tissue DNA.^36,41,97 Furthermore, the ability to obtain high-quality MS³ spectra of adducted nucleobases provides robust mass spectral data for identifying DNA adducts in rodents and human tissues.^35,36 The MS³ scanning technology has significantly advanced the DNA adductomics field of study. More recently, ion trap and quadrupole (Q)-hybrid instruments, which include Q-trap, Q-time-of-fight (TOF), and Q-Orbitrap-tribrid instruments, have emerged for targeted and untargeted DNA adductomics screening.^29,98,99 The advent of the HRAMS LIT-Orbitrap-hybrid MS has further improved the robustness of untargeted DNA adduct screening, and the accurate mass scanning for the neutral loss of dR (116.0473 Da) decreases the number of false positives observed for DDA-CNL/MS³ with the low-resolution LIT.³⁷ HRAMS LIT-Orbitrap-hybrid MS has also been employed in untargeted DDA-CNL/MS³ screening of DNA interstrand crosslinks formed with chemotherapeutic agents,^100,101 and most recently DNA-RNA crosslinks.¹⁰²

The focus of our paper was to determine the sensitivity and breadth of DNA adduct coverage of the wide-SIM/MS² scanning technology to detect putative DNA adducts formed in rat kidneys and liver from a variety of genotoxicants and to characterize their structures by targeted MS³ and CNL/MS³ scanning technologies.^28,34,38 Wide-SIM/MS² is more comprehensive than the untargeted DDA-CNL/MS³ with lower detection limits for bulky aromatic DNA adducts (Table 2S).^28,34,54 Our previous studies using calf thymus DNA spiked with bulky aromatic DNA adducts also showed that wide-SIM/MS² was more sensitive than DDA-CNL/MS³,^28,34 and our wide-SIM/MS² findings in this animal study support those results. Although future improvements of the GPF and Ion Trap MS² approach (GPF-DDA-CNL/MS³) may improve DNA adduct coverage by untargeted DDA-CNL/MS³. The wide-SIM/MS² method and SIM-City algorithm detected 19 DNA adducts in rat liver and kidney treated with the 13-carcinogen mixture. The targeted MS² analyses showed that DNA adduct levels occurred over a 2,000-fold range: the DNA adducts were estimated at 6 adducts per 10⁶ nts down to 3 adducts per 10⁹ nts employing 4 μg DNA digest assayed on the column. Our technology employs an online trap column to remove the non-modified 2′-deoxynucleosides present at ~million-fold excess to the adducted nucleoside. The large quantities of non-modified 2′-deoxynucleosides can contaminate the source, resulting in a rapid decline in the sensitivity and performance of the Orbitrap. Thus, our approach does not capture all DNA adducts. For example, oxidative DNA damage and polar lipid DNA adducts are not retained on the online trap and are not measured in this assay. An offline SPE approach is required to screen for these classes of DNA adducts.^51,88

We previously detected many unknown putative DNA adducts employing wide-SIM/MS² and characterized several novel lipid peroxidation DNA adducts derived from 4-hyroxy-2-alkenals by DDA-CNL/MS³ in the prostate genome of prostate cancer patients.⁸⁸ Totsuka and co-workers conducted a DNA adductome study on the human esophagus used DIA with the Quadrupole time-of-flight (QTOF) MS with multiscale entropy (MS^E) and detected a DNA adduct of N-nitrosopiperidine, a rodent esophageal carcinogen, among hundreds of other putative DNA adducts of unknown structures.⁹⁹ The Motwani laboratory employed a wide-SIM/MS² approach and reported more than 100 putative, unknown DNA modifications in amphipods, which serve as a sentinel species for monitoring chemical pollution in marine sediments in various regions of the Baltic Sea.¹⁰³ Some putative DNA adducts detected in these studies are likely false positives. However, the findings reveal that the landscape of DNA damage in rodent, marine, and human genomes is complex, with many putative unknown DNA adducts. In our study, we detected many putative DNA adducts in the liver and kidney of rats exposed to 13-carcinogens study by wide-SIM² with wSIM-City identification (Figure 11); some may be previously unreported DNA adducts of this 13-carcinogen mixture through novel biotransformation pathways,¹⁰⁴ or from endogenous chemicals or the result of oxidative stress; others may be false-positives. High-resolution accurate mass MSⁿ scanning is required to characterize the modified nucleobases and to deduce or identify chemical structures. However, identifying and annotating unknown DNA adducts is complex and requires in-silico methods for structure elucidation, which is beyond the scope of our study. Metabolomics and exposomics also have similar challenges, where most chemicals detected are unknown.⁸³

Our analyses show that AA and NBA containing substituted nitrophenanthrene moieties form the highest levels of DNA adducts in the kidneys of treated rats 24 h after an intraperitoneal injection of the 13-carcinogen mixture. In contrast, the levels of dG-C8–2-AF (m/z 447.1775) formed by nitro reduction of 2-NF to the reactive N-hydroxy-2-AF metabolite were relatively low in the kidneys and liver. Some aromatic amines, including 2-AF, undergo N-acetylation; the N²-acetylated dG-C8–2-AF adduct (m/z 489.1881) was not detected.⁶⁸ The DNA adduct levels formed by aromatic amines, HAAs, and B[a]P were also relatively low in the kidneys compared to AA-I and NBA DNA adduct levels. AA-I is a potent renal carcinogen in humans and is classified as a Group 1 carcinogen by IARC.¹⁰⁵ Several nitroarenes, including NBA, are found in diesel and gasoline engine exhausts and classified as Group 2A carcinogens (probably carcinogenic to humans) or Group 2B carcinogens (possibly carcinogenic to humans) by IARC. Gasoline and diesel engine exhaust exposures are associated with increased lung and possibly kidney cancer risks.^70,106 Quantitative LC-MS² measurements showed that the highest levels of NBA DNA adducts were formed in rat lungs, kidneys, and pancreas 48 h after intratracheal installation.⁴⁹ Similar findings were reported by ³²P-postlabeling.¹⁰⁷ Under our dose regimen, NBA was the most potent DNA-damaging agent of the 13-carcinogen mixture in the rat kidney and liver.

Based on previous data from our laboratory and others, we expected to detect dA-N⁶-AL-I and dG-N²-AL-I in the liver and kidneys of the 13-carcinogen-dosed rats.^47,82,108 Untargeted wide-SIM/MS² scanning with the wSIM-City algorithm also discovered the putative dC-N⁴-AL-I adduct and several major novel DNA adducts formed with the AA-III isomer. The role of dC-N⁴-AL-I adduct in AA-I mutagenesis requires further study.²⁵ The toxicity data reported on AA-III is limited. Balachandran investigated the structure-activity relationships of 18 AA analogs isolated from A. fangchi and A. contorta in cultured renal LLC-PK1 epithelial cells from the pig.¹⁰⁹ The nephrotoxic potential of AA analogs was assessed using the neutral red dye exclusion assay and the induction of caspase 3/7 activity. AA-I was the most toxic analog, followed by AA-II; however, AA-III displayed no toxicity in either assay. The authors concluded that shifting the methoxy group of AA from the 8-position (AA-I) to the 10-position (AA-III) of the phenanthro-(3,4-d)-1,3-dioxolo-5-carboxylic acid makes the compound non-toxic. Our results in the rat model do not support that conclusion. AA-III undergoes bioactivation and is a potent DNA-damaging agent in rat kidneys. The amount of AA-III is ~10-fold lower than AA-I in the herbal extract assayed. Nevertheless, DNA adduct formation with AA-III in rat kidneys occurs at ~25 to 40-fold greater levels than those formed with AA-I when normalized per mg of AA isomer dosed/kg body weight at 24 h post-dose treatment.⁵⁷ The AA-III isomer preferentially forms dA and dG DNA adducts in the kidney than in the liver. In contrast, the levels of dG-N²-AL-I and dA-N⁶-AL-I are comparable in the kidney and liver, although the novel putative dG-AL-I isomer was only detected in the kidney (Figure 10). Thus, AA-III may be a more potent DNA-damaging agent than AA-I in rat kidneys. The kinetics of AA-III DNA adduct formation and persistence requires study compared to AA-I to understand the impact of the O-methyl group position on the AA analog genotoxicity in rodents and human kidney cells.⁸²

Only a few reports exist in the literature on AA-III concentrations in Arisolochia plants. AA-III was measured in several Aristolochia species in South America: A. bridgesil, A. argentina, and A. chilensis: the amounts of AA-III reported in roots of A. Chilensis was 196 mg/kg, while AA-I was 32 mg/kg.^63–65 Thus, the variation in AA-III levels found in different parts of the plant (root, stem, and fruit) used for medicinal treatments can be significant. Traditional herbal medicines have been used world-wide for treating a variety of ailments.¹⁰⁵ Historically, A. chilensis was used as a medicinal plant for women suffering from certain menstrual disorders. Further studies on exposure assessment, genotoxicity, and potential cancer risk of AA-III are warranted.

Conclusion

Our untargeted wide-SIM/MS² and DDA-CNL/MS³ scanning technologies have identified many DNA adducts in rat kidneys and liver formed with 13 potential human carcinogens. The detection sensitivity for many DNA adducts is less than 1 adduct per 10⁸ nts, assaying 4 μg DNA. Our results show that several substituted nitrophenanthrenes are far more potent DNA-damaging agents than some prototypical aromatic amines, HAAs, or B[a]P in the rat model. Our goal is to continue optimizing and refining wide-SIM/MS² and DDA-CNL/MS³ scanning technologies, including GPF ion trap fragmentation, to screen DNA damage in human kidneys and other organs and advance our knowledge of potential etiological agents of cancer causation.