The Cooked Meat Carcinogen 2-Amino-1-methyl-6-phenylimidazo[4,5-b]pyridine Hair Dosimeter, DNA Adductomics Discovery, and Associations with Prostate Cancer Pathology Biomarkers

Abstract

Well-done cooked red meat consumption is linked to aggressive prostate cancer (PC) risk. Identifying mutation-inducing DNA adducts in the prostate genome can advance our understanding of chemicals in meat that may contribute to PC. 2-Amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP), a heterocyclic aromatic amine (HAA) formed in cooked meat, is a potential human prostate carcinogen. PhIP was measured in the hair of PC patients undergoing prostatectomy, bladder cancer patients under treatment for cystoprostatectomy, and patients treated for benign prostatic hyperplasia (BPH). PhIP hair levels were above the quantification limit in 123 of 205 subjects. When dichotomizing prostate pathology biomarkers, the geometric mean PhIP hair levels were higher in patients with intermediate and elevated-risk prostate-specific antigen values than lower-risk values < 4 ng/mL (p = 0.03). PhIP hair levels were also higher in patients with intermediate and high-risk Gleason scores ≥7 compared to lower-risk Gleason score 6 and BPH patients (p = 0.02). PC patients undergoing prostatectomy had higher PhIP hair levels than cystoprostatectomy or BPH patients (p = 0.02). PhIP-DNA adducts were detected in 9.4% of the patients assayed; however, DNA adducts of other carcinogenic HAAs, and benzo[a]pyrene formed in cooked meat, were not detected. Prostate specimens were also screened for 10 oxidative stress-associated lipid peroxidation (LPO) DNA adducts. Acrolein 1,N²–propano-2′-deoxyguanosine adducts were detected in 54.5% of the patients; other LPO adducts were infrequently detected. Acrolein adducts were not associated with prostate pathology biomarkers, although DNA adductomic profiles differed between PC patients with low and high-grade Gleason scores. Many DNA adducts are of unknown origin; however, dG adducts of formaldehyde and a series of purported 4-hydroxy-2-alkenals were detected at higher abundance in a subset of patients with elevated Gleason scores. The PhIP hair biomarker and DNA adductomics data support the paradigm of well-done cooked meat and oxidative stress in aggressive PC risk.

Graphical Abstract

Introduction

Prostate cancer (PC) is the most commonly diagnosed noncutaneous cancer and the second leading cause of cancer-related death in men in the United States.¹ The major risk factors for PC are age, family history, obesity, and African-American (AA) ethnicity.² PC incidence and mortality rates of migrant cohorts from low PC prevalence countries often increase upon their relocation to countries with high risk for PC, implying that environmental or dietary chemical exposures impact PC development.^3,4 Differences in health care and early diagnoses may also contribute to variable PC incident rates across countries. The causative factors and their mechanisms of action in PC development require further study. High-fat diets, dairy products, and alcohol are proposed risk factors for PC,⁵ but the associations between these factors and PC risk are inconsistent.^6-8 In the same vein, epidemiological studies evaluating the role of red meat consumption and PC risk also have provided equivocal results.^9-12 However, several studies reported that frequent consumption of well-done cooked red meat is linked to aggressive PC while consuming rare or medium-cooked meat is not.^11-16 Thus, the method of cooking meats may be a critical factor in PC risk. HAAs, including 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP), 2-amino-3-methylimidazo[4,5-f]quinoline (IQ), 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline (MelQx), 2-amino-3,4,8-trimethylimidazo[4,5-f]quinoxaline (4,8-DiMeIQx), 2-amino-9H-pyrido[2,3-b]indole (AαC), and polycyclic aromatic hydrocarbons (PAHs), such as benzo[a]pyrene (B[a]P), form during high-temperature frying, grilling, or barbecuing of meats.^17-19 The concentration of HAAs and PhIP, in particular, increases in meat and poultry as a function of the temperature and duration of cooking.^20-23 HAAs and PAHs are multisite carcinogens in rodents and may contribute to PC etiology.^11-13

The International Agency for Research on Cancer (IARC) classified red meat as a Group 2A colorectal carcinogen (probably carcinogenic to humans) based on a compendium of epidemiological data and mechanistic studies.²⁴ Positive associations were also seen in cohort studies and population-based case-control studies between consumption of well-done cooked red meat and cancers of the pancreas and the prostate (mainly advanced prostate cancer).²⁴ The causative agents of these cancers and their mechanisms of action require further study. IARC classified PhIP and several other HAAs as Group 2A or 2B carcinogens.²⁵ PhIP is the most abundant carcinogenic HAA formed in well-done cooked meats,¹⁸ and the only mutagen in cooked meat known to induce PC in rat models.^26-28 PhIP also induces PC in humanized CYP1A2 mice but not wild-type mice, signifying metabolism is a critical factor in PhIP carcinogenicity.²⁹ PhIP induces inflammation, glandular atrophy, high-grade prostate intraepithelial neoplasia, and oxidative stress in the prostate of F344 rat and humanized cytochrome P4501A mice.^27,29-31 These pathology features are observed in human PC.^32,33 However, the PhIP dose (200 mg/kg) used in rodents was more than one million-fold greater than the daily human intake, and the biological effects of PhIP at physiological concentrations are unknown.^34,35

Our research seeks to understand the potential role of HAAs in human cancer through mechanistic studies and to develop specific mass-spectrometry-based HAA biomarkers for implementation in epidemiology studies.^36,37 Our laboratory and several others reported that 2-hydroxyamino-1-methyl-6-phenylimidazo[4,5-b]pyridine (HONH-PhIP), the genotoxic metabolite of PhIP, undergoes efficient phase II bioactivation in prostate cells to produce reactive intermediates leading to the formation of the mutation-prone N-(2′-deoxyguanosin-8-yl)-PhIP (dG-C8-PhIP) adduct at high levels.^38-42 These data provide a biochemical mechanism for PhIP-induced mutagenesis in the human prostate. Furthermore, PhIP-induced prostate carcinogenesis is associated with oxidative stress in rodents.^27,29,30 Oxidative stress can lead to lipid peroxidation (LPO) products that can cause DNA damage.^43-45

In this current study, we employed hair as a noninvasive biospecimen to assess well-done cooked meat consumption and chronic exposure to PhIP in PC and cystoprostatectomy patients, where the cancerous bladder and the usually benign prostate are removed for medical reasons, and BPH patients. The PhIP hair dosimeter is a promising long-term biomarker to assess chronic exposure to this carcinogen.^46,47,49,50 In conjunction, we employed nanoliquid chromatography with high-resolution accurate mass spectrometry (HRAMS) to measure DNA adducts of PhIP, other carcinogenic HAAs, and B[a]P. We also screened for LPO DNA adducts as biomarkers of oxidative DNA damage (Scheme 1),^44,45,51 since PhIP induces oxidative stress in the rodent prostate.^29,32 We also applied our untargeted screening methods for DNA adductomics discovery in prostate specimens of PC and cystoprostatectomy patients.^38,52-54 The correlations between these carcinogen biomarkers and clinical biomarkers of PC, including prostate-specific antigen (PSA) values, Gleason scores, and tumor stage, were investigated.

Scheme 1.

Cooked meat carcinogen and LPO DNA adducts

Materials and Methods.

Materials.

Calf thymus DNA (CT DNA), DNase I (Type IV, bovine pancreas), Benzonase nuclease ultrapure, 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). SOLA SCX solid phase extraction (SPE) cartridges (10 mg/mL) were purchased from Thermo Scientific (Waltham, MA). 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). Soluene 350 was purchased from PerkinElmer (Waltham, MA). All Optima LC-MS grade formic acid, acetonitrile, methanol and water, silylated LC vials (PN 03-FISV), and neutral buffered formalin (10%) were purchased from Thermo Fisher Scientific (Waltham, MA). PhIP, 2-amino-1-trideuromethyl-6-phenylimidazo[4,5-b]pyridine ([²H₃C]-PhIP, >99.5% isotopic purity), MeIQx, and AαC were purchased from Toronto Research Chemicals (North York, Canada). [¹³C₁₀]-dG (99% isotopic purity), [¹⁵N₅]-dG, [¹⁵N₅]-dA, and [¹⁵N₃]-dC (>98% isotopic purities) were purchased from Cambridge Isotope Laboratory (Tewksbury, MA), and trans-4-hydroxy-2-hexenal was purchased from Cayman Chemicals (Ann Arbor, MI). N-(2′-Deoxyguanosin-8-yl)-AαC (dG-C8-AαC) and ([¹³C₁₀]-dG-C8)-AαC, N-(2′-deoxyguanosin-8-yl)-MeIQx (dG-C8-MeIQx) and dG-C8-[²H₃C]-MeIQx, N-(2′-deoxyguanosin-8-yl)-PhIP (dG-C8-PhIP), [¹³C₁₀]-dG-C8)-PhIP, and 10-(deoxyguanosin-N²-yl)-7,8,9-trihydroxy-7,8,9,10-tetrahydrobenzo[a]pyrene (dG-N²-B[a]P), and [¹³C₁₀]-dG-N²-B[a]P were synthesized as described.^55-58 LPO adduct standards heptanone-etheno-2′-deoxycytidine (HεdC), heptanone-etheno-2′-deoxyguanosine (HεdG),heptanone-etheno-2′-deoxyadenosine (HεdA), the isomeric 1,N²-propano-2′-deoxyguanosine adducts derived from acrolein (6R/S)-3-(2′-deoxyribos-1′-yl)-5,6,7,8,-tetrahydro-6-hydroxypyrimido[1,2-a]purine-10(3H)one (6-HO-PdG) and (8R/S)-3-(2′-deoxyguanosin-1′-yl)-5,6,7,8,-tetrahydro-8-hydroxypyrimido[1,2-a]purine-10(3H)one (8-HO-PdG); and the etheno adducts 3,N⁴-etheno-2′-deoxycytidine (εdC), 1,N⁶-etheno-2′-deoxyadenosine (εdA), 1,N²-etheno-2′-deoxyguanosine (εdG), and their respective [¹⁵N₅]-dG, [¹⁵N₅]-dA, [¹⁵N₃]-dC) homologs were synthesized according to reported methods.^59-65 N²-hydroxymethyl-2′-deoxyguanosine (N²-HOCH₂-dG) was synthesized by reaction of formalin with dG as described.⁶⁶ The 1,N²-dG adducts 3-(2-deoxy-β-D-erythro-pentofuranosyl)-5,6,7,8-tetrahydro-8-hydroxy-6-(1-hydroxypropyl)pyrimido[1,2-a]purin-10(3H)-one (HHE-dG) diastereomers were prepared by reaction of dG with 4-hydroxy-2-hexenal (4-HHE) as previously described.⁶⁷ The diastereomeric 1,N²-dG adducts of 4-hydroxy-nonenal (4-HNE), 3-(2-deoxy-β-D-erythro-pentofuranosyl)-5,6,7,8-tetrahydro-8-hydroxy-6-(1-hydroxyhexyl)pyrimido[1,2-a]purin-10(3H)-one (HNE-dG) and [²H₁₁]-HNE-dG were kindly provided by Dr. Carmelo Rizzo, Vanderbilt University.⁶⁸ The malondialdehyde adduct 3-(2′-deoxy-β-D-erythro-pentofuranosyl)pyrimido[1,2-a]purin-10(3H)-one (M₁dG) and [¹⁵N₅]-M₁dG were a kind gift from Dr. Lawrence Marnett, Vanderbilt University.⁶⁹ Formalin-modified CT DNA was prepared by reacting CT DNA (1 mg/mL) with formaldehyde (32 μmol) in 100 mM potassium phosphate buffer (pH 7.0). 4-HHE-modified CT was prepared by reacting CT DNA (1 mg/mL) in 100 mM sodium bicarbonate buffer (pH 9.0) with 4-HHE (2.9 μmol). Both reactions were conducted at 37 ° for 72 h. Thereafter, 0.1 vol of 5 M NaCl and two vol of ethanol were added to precipitate the DNA. The spindled DNA was washed twice with 70% ethanol/30% water.

Human Study.

The research protocol was reviewed and approved by the Institutional Review Board at the University of Minnesota. Participants were recruited during their preoperative visits. The study population consisted of men from the Upper Midwest. The inclusion criteria required that subjects were either undergoing (a) radical prostatectomy to remove the entire cancerous gland, (b) transurethral resection of the prostate (TURP) for the treatment of benign prostatic hyperplasia (BPH), or holmium laser enucleation of the prostate (HoLEP), or (c) bladder cancer patients undergoing cystoprostatectomy surgery resulting in prostate tissue removal. Patients with bladder cancer in this cohort typically have benign, cancer-free prostates and often receive 3 months of cisplatin-based chemotherapy before surgery. Patients with PC and BPH typically do not receive chemo or hormonal therapy pretreatment. The cohort's mean age (± SD) was 65.9 ± 8.6 years. The patients agreed to permit the study team to obtain a specimen of scalp or abdomen hair and their prostate tissue from the surgery for the research study, including chemical analysis of DNA. Smokers were included in the study. The smoking status was self-reported, and patients were classified as current smokers, former smokers, or never smokers. Former smokers had ceased smoking at least 6 months before surgery. Patients who dyed their hair were excluded from the study.

The nature of the study protocol was explained to all potential participants, who read and signed the informed consent using Institutional Review Board-approved procedures. All patients' identities were deidentified by assigning them a study code so that the subjects’ identities were not available to the research laboratory. Information on the subjects’ demographics and prostate pathology data were retained.

Pathology.

Normal tumor-adjacent prostate tissues from the peripheral zone (PZ) were carefully selected from the surgically excised prostatectomy and cystoprostatectomy specimens, which were snap-frozen within 30 min of removal from the patients and stored at −80 °C. Matching formalin-fixed paraffin-embedded (FFPE) samples were fixed in 10% neutral buffered formalin for 24 h at room temperature. Then, the tissues underwent serial dehydration with ethanol, followed by p-xylene cleaning, and paraffin infiltration, by a Leica ASP300S tissue processor (Leica Microsystems Inc., Buffalo Grove, IL). The FFPE sections were stained with hematoxylin and eosin (H&E) tissue stains using a Leica ST5010 Autostainer XL. All biospecimens were deidentified. The study pathologist (P. Murugan) reviewed the H&E stained FFPE sections for pathology markers associated with PC, such as atrophy, inflammation, and high-grade prostatic intraepithelial neoplasia (HGPIN) and quantified the samples as mild, moderate, or severe. Specimens selected for DNA adduct measurements were from largely tumor-free tissue. The inflammation, atrophy, and HGPIN biomarkers were evaluated later. Prognostic variables, including Gleason score and tumor stage, were recorded from the pathology database in patients diagnosed with PC, and the PSA assay was conducted by chemiluminescence using a Siemens Vista analyzer.^70-72

Assessment of Meat Consumption.

An abbreviated self-administered meat food frequency questionnaire (FFQ) was employed to assess meat consumption and adapted from the module of Anderson and co-workers.⁷³ The patients completed a questionnaire on red meat (beef), processed meat (sausage and bacon), poultry, and seafood consumed during the past week. HAA exposure was assessed employing the PhIP hair biomarker as a proxy for well-done cooked meat consumption.^46,47,50

Hair collection of hair and LC/MS measurement of PhIP.

Newly grown hair (50 – 100 mg) was collected by study staff from the back of the subjects' heads, near the hairline (nape of the neck, to collect freshly grown hair) just before surgery. Many patients had closely cropped hair and lacked sufficient scalp hair for analyses, so hair was also taken from the lower abdomen as an alternative specimen just before surgery. Hair was stored in Ziploc bags in the dark at room temperature before analyses. A previously validated analytical method was employed to isolate PhIP from hair except that a SOLA SCX SPE cartridge replaced the Oasis MCX cartridge (Waters).⁴⁸ Hair samples were finely minced with professional hair clippers. Duplicate hair samples (25 mg) were placed in Eppendorf tubes (1.5 mL) rinsed with aqueous and organic solvents, dried in a ventilated hood, and hydrolyzed in 1N NaOH at 80 °C for 1 h. The internal standard [²H₃C]-PhIP was spiked at a 25 pg/25 mg hair level before base hydrolysis. PhIP was isolated from the hydrolysate using solvent extraction followed by mixed-mode SPE. The extract was vacuum centrifuged to dryness and resuspended in 5 mM NH₄HCO₃, pH 9.0 (30 μL)

LC/MS analysis was performed on a TSQ Quantiva mass spectrometer interfaced by a HESI II source to a Dionex RSLCnanoUltiMate 3000 UHPLC (Thermo Scientific, San Jose, CA). The HPLC separation was done with a BEH 130 Shield RP18 column (0.3 x 100 mm, 1.7 μm, Waters Corporation, New Milford, MA). The aqueous mobile phase (solvent A) contained 5 mM NH₄HCO₃ (pH 9.0), and the organic mobile phase (solvent B) was CH₃CN. The flow rate was 5 μL/min, and the injection volume was 5 μL. A linear gradient was employed, commencing at 10% B and reaching 99% B at 10 min. The spray voltage was 3.3 kV, the ion transfer tube temperature was 400 °C, and the vaporizer temperature was set at 70 °C. The nitrogen sheath and auxiliary gas were at 8 and 1 arbitrary units. The Q1 and Q3 resolutions were set at 0.7 full width of the peak at its half-maximum (fwhm), and the dwell time was 100 ms. The collision gas was argon and set at 1.5 mTorr. PhIP and [²H₃C-]-PhIP were measured by selective reaction monitoring (SRM) of m/z 225.1 → 210.1 and m/z 228.1 → 210.1, respectively, at a collision energy (CE) of 29 eV. A qualifier ion at m/z 140.1 was used to confirm the identity and purity of PhIP and [²H₃C]-PhIP and monitored at 48 eV CE. A relative abundance ratio of the qualifier ion to the target ion was set at 0.25 ± 0.05 for PhIP and [²H₃C]-PhIP to confirm that the analytes were devoid of isobaric interferences. The level of PhIP in hair was expressed as pg/g hair or as ng/g melanin to adjust the PhIP hair levels for pigmentation.^47,48 The calibration curve was constructed with 25 mg of hair from a vegetarian with no detectable PhIP, using 0, 10, 20, 50, 100, 200, 500 pg/g hair employing [²H₃C]-PhIP at 1000 pg/g hair. The calibration curve showed good linearity (slope = 0.9969 ± 0.0053, r²=0.99). The limit of quantification (LOQ) was 25 pg of PhIP/g of hair with between-day and within-day coefficient of variations of <10%.⁴⁸

Spectrophotometric Characterization of the Melanin Content in Hair.

Hair (25 mg) was digested in Soluene 350:H₂O (9:1 v/v, 1 mL) by heating at 95 °C, as described.⁴⁸ The visible spectrum was recorded in the 400–800 nm range. The absorbance at 500 nm was used to quantify the total combined amount of melanin (eumelanin and pheomelanin). The melanin estimate was based on the absorbance of 0.99 at 500 nm by a 100 μg/mL melanin (eumelanin) solution.⁷⁴

PhIP Accrual in the Hair of Healthy Adults Participating in a 4-Week Semicontrolled Feeding Study.

This study was conducted at the University of Hawaii Cancer Center, Honolulu, HI.⁴⁶ Forty-one nonsmokers (32 males and nine females) aged 18–63 years (median: 25 years) completed a semicontrolled diet ingesting ground beef cooked to different doneness levels, reflecting a low or high concentration of PhIP in the range found in the American diet.^34,35 The participants ate cooked ground beef preparations at a concentration of PhIP at 0, 0.38, 0.46, 1.3, 1.5, 2.3, 3.5, or 5.5 μg per serving (150 or 200 g cooked ground beef) for 5 days a week for 1 month.⁴⁶ The "0" value represents the unknown basal level of PhIP ingested before the feeding study commenced. Recently grown hair was clipped from the back of the subjects' heads, near the hairline (nape of the neck). This newly grown hair captures the previous 2-month exposure of PhIP.^46,47,49 The PhIP concentrations in the cooked meat diet were determined by LC/MS², employing [²H₃C]-PhIP as the internal standard.²³ Full details of the feeding study were reported.⁴⁶

Isolation and Enzymatic Digestion of Prostate DNA.

Normal tumor-adjacent prostate tissues of the peripheral zone (PZ) were flash-frozen in liquid nitrogen and stored at −80 °C until assayed. The prostate tissues (−100 mg) 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 (the equivalent of 40 mg wet tissue weight) 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 addition of 0.1 vol of 10% SDS, and 0.4 mg proteinase K for 2 h. Protein was removed by the addition of chilled Puregene protein precipitation solution (250 μL) containing 10 mM βME (Qiagen, Germantown, MD), and residual lipids were removed by extraction with chloroform/isoamyl alcohol (24:1). DNA was precipitated by adding 0.1 vol. of 5 M NaCl followed by 1 volume of cold isopropanol containing 10 mM βME and stored at −20 °C overnight. The DNA was retrieved by centrifugation, and the pellets were dissolved in 5 mM HEPES buffer pH 8.0 containing 2.5 mM βME. The concentration was estimated by UV spectra, assuming 1 unit of A₂₆₀ was equivalent to 50 μg of ds-DNA.

For hydrophobic, bulky-type aromatic DNA adducts, DNA (20 μg) was spiked with isotopically labeled internal standards at 1 adduct per 10⁸ nts levels before the digestion: [¹³C₁₀]-dG-C8-4-ABP, [¹³C₁₀]-dG-C8-AαC, dG-C8-[²H₃C]-MeIQx, [¹³C₁₀]-dG-C8-IQ, [¹³C₁₀]-dG-C8-PhIP, [¹³C₁₀]-dG-N²-BPDE, and [²H₁₁]-HNE-dG. DNA samples were digested using a thermomixer at 37 °C, 900 rpm in 5 mM HEPES buffer pH 8.0 containing 2.5 mM βME, 15 mM MgCl₂, and 10 mM CaCl₂ with the following nucleases: benzonase (300 U) and nuclease P1 (0.2 U) for 3.5 h, followed by phosphodiesterase 1 (24 mU) and alkaline phosphatase (40 mU) for 15 h. The digest was diluted 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 (30 μL), sonicated, centrifuged at 21,000 g for 5 min, and transferred to silylated borosilicate glass inserts for LC-MS analysis employing on-line trapping to removed non-modified 2′-deoxynucleosides.^52,54 Calibration curves employed calf thymus DNA as the matrix.^52,75 The LOQ values for dG-C8-PhIP, dG-C8-MeIQx, dG-C8-AαC, and dG-N²-B[a]P were <3 adducts per 10⁹ nts.⁵²

For polar LPO adducts, DNA (20 μg) was spiked with isotopically labeled internal standards at 1.3 adducts per 10⁷ nts before digestion: [¹⁵N₃]-εdC, [¹⁵N₅]-εdA, [¹⁵N₅]-εdG, [¹⁵N₃]-HεdC, [¹⁵N₅]-HεdA, [¹⁵N₅]-HεdG, [¹⁵N₅]-6-HO-PdG, [¹⁵N₅]-8-HO-PdG, [¹⁵N₅]-M₁dG, and [²H₁₁]-HNE-dG. LPO DNA was digested using our published method, followed by treatment with adenosine deaminase (12 mU) at 25 °C for 1.5 h to convert dA to 2′-deoxyinosine.⁵³ Bond Elut C18 SPE cartridges were conditioned with CH₃CN (1 mL) and equilibrated with LC-MS grade water (1 mL) before DNA digests were applied to the SPE cartridge. The SPE cartridges were washed with LC-MS grade water (4 mL), followed by 1% CH₃CN in water (1 mL), and LPO and other adducts were eluted with 90% CH₃CN:10% H₂O (1 mL). After drying by vacuum centrifugation, the SPE eluants were reconstituted in 90% CH₃CN:10% H₂O and transferred to the LC-MS vials and dried a second time. LC-MS grade water (12 μL) was used to reconstitute the residue for LC-MS analysis. Calibration curves employed rat liver DNA from an untreated rat as the matrix adding internal standards at 1.3 adducts per 10⁷ nts before digestion.⁵³ The LOQ values for 6-HO-PdG, 8-HO-PdG, and other LPO adducts are ≤1 adduct per 10⁸ nts.⁵³

Targeted Mass Spectrometric Analysis of DNA Adducts.

The targeted DNA adduct analyses were performed on an Orbitrap Lumos™ Tribrid™ MS interfaced to an UltiMate 3000 RSLCnano UHPLC system (Thermo Fisher Scientific, San Jose, CA). For hydrophobic and bulky aromatic DNA adducts, a PicoChip nanoESI source and a PicoChip column (Reprosil-Pur C18-AQ, 0.075 × 105 mm, 3 μm particle size, 120 Å, New Objective, Woburn, MA) were used for the chromatographic separation. An Acclaim PepMap trap cartridge RP C18 (0.3 × 5 mm, 5 μm particle size, 100 Å, Thermo Fisher Scientific, San Jose, CA) was employed for online DNA adduct enrichment. The LC solvents were (A) 0.05% HCO₂H in H₂O and (B) 0.05% HCO₂H in 95% CH₃CN. The DNA digests (7 μL) were injected onto the trap column and washed with solvent A for 4 min at a flow rate of 12 μL/min by the loading pump to remove the polar, non-modified nucleosides. After trapping, the adducts were flushed onto the analytical column at a flow rate of 0.6 μL/min using a linear gradient from 1% B to 99% B over 35 min, followed by 3 min washing at 99% B and equilibration at 1% B for 5 min. Data was acquired by Xcalibur™ version 4.4 (Thermo Fisher Scientific, San Jose, CA). MS parameters were as follows: spray voltage, 2200 V; ion transfer tube temperature, 275 °C; quadrupole isolation, 1.6 m/z; Orbitrap resolution, 60,000 (fwhm) at m/z 200); activation type, higher-energy C-trap dissociation (HCD); collision energy, 25%; maximum injection time, 118 ms; AGC target, 4 × 10⁵. The targeted MS² ion transitions used to construct the extracted ion chromatograms (EICs) for adduct quantifications were: dG-C8-AαC (m/z 449.2 → 333.1207), [¹³C₁₀]-dG-C8-AαC (m/z 459.2 → 338.1375); dG-C8-MeIQx (m/z 479.2 → 363.1425), [²H₃C]-dG-C8-MeIQx (m/z 482.2 → 366.1613); dG-C8-PhIP (m/z 490.2 → 374.1470), [¹³C₁₀]-dG-C8-PhIP (m/z 500.2 → 379.1640); 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); HNE-dG (m/z 424.2 → 308.1718), and [²H₁₁]-HNE-dG (m/z 435.2 → 319.2408). All EICs were constructed with a 5 ppm mass tolerance window.

A Nanospray Flex ion source (Thermo Fisher Scientific) and a homemade nanocolumn using a New Objective PicoFrit emitter (75 μm x 200 mm, 10 ± 1 μm orifice, New Objective, Woburn, MA) packed with Hydro RP phase (4 μm particle size, 80 Å, Phenomenex Corp. Torrance, CA) were used for the separation of polar LPO DNA adducts. The LC solvent A was 2 mM NH₄CH₃CO₂ and B was CH₃CN. The DNA digests (2 μL) were injected on the nanocolumn at a flow rate of 0.6 μL/min. The LC gradient was 0-7 min, 0.6 μL/min at 1% B; 7-22 min, 0.3 μL/min 1-25% B; 22-36 min, 0.3 μL/min, 25-60% B; column washing (2 min at 95% B) and equilibration (6 min at 1% B) were both at 0.6 μL/min. MS parameters were as follows: spray voltage, 2200 V; ion transfer tube temperature, 275 °C; quadrupole isolation, 1.6 m/z; Orbitrap resolution, 30,000; injection time, 54 ms; AGC target, 4 x 10⁵. The targeted MS2 ion transitions and CEs used to construct the EICs for adduct quantifications were: εdC (m/z 252.1 → 136.0506, CE10%), [¹⁵N₃]-εdC (m/z 255.1 → 139.0417, 10%); εdA (m/z 276.1 → 160.0618, 15%), [¹⁵N₅]-εdA (m/z 281.1 → 165.0470, 15%); εdG (m/z 292.1 → 176.0567, 10%), [¹⁵N₅]-εdG (m/z 297.1 → 181.0419, 10%); HεdC (m/z 364.2 → 248.1394, 10%), [¹⁵N₃]-HεdC (m/z 367.2 → 251.1305, 10%); HεdA (m/z 388.2 → 272.1506, 20%), [¹⁵N₅]-HεdA (m/z 393.2 → 277.1358, 20%); HεdG (m/z 404.2 → 288.1455, 10%), [¹⁵N₅]-HεdG (m/z 409.2 → 293.1307, 10%); and HNE-dG (m/z 424.2 → 308.1718, 18%), and [²H₁₁]-HNE-dG (m/z 435.2 → 319.2408, 18%). The isomeric acrolein 1,N²–propano-dG adducts were measured at the MS³ scan stage because some samples contained isobaric interferences of the aglycone ion: 6-HO-PdG (m/z 324.2 → 208.1 → 152.0567 and 190.0723, CE10%, 30%), [¹⁵N₅]-6-HO-PdG (m/z 329.2 → 213.1 → 157.0419 and 195.0575, CE10%, 30%); 8-HO-PdG (m/z 324.2 → 208.1 → 164.0567, CE10%, 30%), [¹⁵N₅]-8-HO-PdG (m/z 329.2 → 213.1 →169.0419, CE10%, 30%).⁵³ The M₁dG glycoside linkage was labile under the MS source conditions and its aglycone was measured at the MS² scan stage (m/z 188.1 → 97.0396, 106.0400, 133.0509, 160.0618, 60%), and [¹⁵N₅]-M₁dG (aglycone m/z 193.1 → 99.0337, 109.0311, 137.0390, 165.0470, 60%).

DNA Adductomics Data Acquisition: Untargeted Adduct Screening Using Wide Selected Ion Monitoring) Tandem Mass Spectrometry and Data-Dependent Acquisition.

Untargeted DNA adduct screening and identification employed the same LC-MS system and chromatographic conditions used for bulky aromatic and LPO adducts. A wide selected ion monitoring/tandem mass spectrometry (wide-SIM/MS²) DNA adductomics method was used.⁵⁴ This method screens for the coelution of the modified nucleoside precursor ions ([M+H]⁺) in the wSIM scan mode and their corresponding aglycone ions ([BH₂]⁺) at the MS² scan stage produced by the loss of dR under collision-induced fragmentation. For bulky aromatic adducts, the wide-SIM/MS² method was comprised of 20 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 30 m/z, and the sequential MS² scan event fragmented all ions in the same mass range with a product ion detection ranging from m/z 100 to 650, resulting in a mass range of m/z 330–630 screened for both the wide-SIM and MS² data acquisition. For LPO adducts, the m/z range of 240-480 was scanned with 16 alternating SIM and MS² events of 30 m/z isolation width. The MS parameters were as follows: isolation mode, quadrupole; isolation width, 34 m/z (including a 2 m/z overlap on both ends of the isolation ranges); RF-lens, 40% (bulky adducts) or 35% (LPO adducts); resolution, 120,000; maximum injection time, 246 ms; AGC, 5 × 10⁵ for wide-SIM and 4 × 10⁶ for MS²; activation type, HCD; collision energy, 25% (bulky adducts) or 15% (LPO adducts); data type, profile.

DNA Adductomics Data Analysis: Searching for DNA Adducts Using wSIM-City.

Thermo MS.raw data files were converted to the mzML centroided data format using the Proteowizard v.3.0.19098 (2019-4-8) msconvert.exe raw file converter with the following command: msconvert –filter “peakPicking true 1-“ –simAsSpectra –mzML <raw file name>. These mzML files were used as input into the wSIM-City search algorithm.⁷⁶ Each mzML file was separated into 10 separate wide-SIM and MS² R data objects, one for each m/z acquisition range, using the R (v4.0.0) package mzR (v.2.16.2). Each MS level (wide-SIM or MS²) m/z range was processed separately for feature finding and alignment, producing 20 separate data analysis experiments (vide infra). The raw MS data used for spectral library searching were converted using msconvert.exe to .MGF format. The algorithm wSIM-City automates the peak picking process and the statistical analysis.

DNA Adducts and Postsearch Alignments, and DNA Adduct Scoring Criteria.

wSIM-City completes the searches for both bulky and LPO fractions using the following parameters: delta_search_mass = −116.0473, ppm_tol = 20, rt_tol = 12, mzmin = 135, mzmax = 634, mzwid = 0.1 and type = ‘SIM’.⁷⁷ The following parameters were used to identify putative DNA adducts and scoring criteria.⁷⁶ The MS¹ and MS² m/z values were selected from the maximum intensity signal for each DNA adduct and its aglycone ([BH₂]⁺). MS² signals generally had superior signal-to-noise ratios than the precursor ions in wide-SIM scan mode and were used in the remaining data alignment steps and statistical analysis.⁷⁶ The algorithm required identical retention times of the EIC of the MS¹ (precursor ion) and MS² (aglycone) peaks at maximum intensity. However, a retention time difference of the signals’ apexes by one or two duty cycles was found to exclude some high-quality putative DNA adducts. Thus, putative adducts were retained for downstream analyses when the neutral loss of dR was observed in a minimum of five sequential scans across the peaks.⁷⁶

wSIM-City scores putative DNA adducts by comparing the observed versus the calculated difference in m/z values between the precursor ions and the aglycones ([BH₂]⁺) acquired in wide-SIM/MS² using the neutral loss of deoxyribose (dR) with the additional loss of hydrogen (116.0743 Da, C₅H₈O₃). Possible molecular formulas (MF) for each m/z of the precursor and the aglycone peaks were calculated within 5 ppm mass error. Those putative adducts with a formula change of -C₅H₈O₃ between the precursor and aglycone peaks were retained for further analysis. The assigned MF with the lowest ppm mass error was selected in cases where more than one MF pair was calculated. The data were subsequently filtered by using the score_mz (Table S1) and the Pearson correlation of their peak intensities showing similar peak shapes. A permissive minimum score_mz of 0.7 and a minimum correlation of peak shapes > 0.5 were used to capture some additional putative adducts. DNA adducts with a score_mz of 0.7 are equal to ~15 ppm error of the detected dR loss (ppm_tol = 20), and a score_mz of 0.9 corresponds to ~10 ppm m/z error of the dR loss. These remaining adducts were then filtered by the observed ratio of the maximum MS² over MS¹ peak intensities (|MS²/MS¹| < 10.⁷⁷ The peaks detected across the samples were aligned with the create_ref_table and join_align functions, using a retention time tolerance (rt_tol) = 1.5 and parts per million tolerance (ppm_tol) = 10. The blank sample peaks were comprised of buffer and nucleases without DNA in the matrices and removed from the data.

Data Filtering and Imputation of Missing Values.

Quality control (QC) samples (n = 4) comprised of pooled aliquots of each sample were included in the adductomics experiment. The sample sequence order was randomized to mitigate technical bias. The putative DNA signals were required to be present in a minimum of 50% of the QC samples of the combined high and low Gleason score patients and used for imputation and statistical analysis. Since partial least squares discriminant analysis (PLS-DA) requires no missing data, adducts with 0 intensity were imputed using the random number generator for a Gaussian distribution in R with the following command: rnorm(n, mean = Pk_min, sd = 0.2 * Pk_min), where n was the total number of intensity values to impute, sd is the standard deviation in the function, 0.2 is the assumed coefficient of variation for low intensity peaks, and Pk_min is the minimum peak intensity observed in the data matrices.

Untargeted DNA Adductomics Statistics and Plotting.

All statistics were completed using the R v 4.0.0 programming language.⁷⁸ All tests of significance used Welch’s two-tailed t-test for unequal variances and sample size. Due to the variability in signal intensities and the limited number of samples assayed by wSIM/MS², the significance level was selected at p < 0.10. Partial least squares discriminant analyses PLS-DA were completed using the ropls package, to discern contributing factors that distinguish differences between low and high Gleason score patients.^79,80 The quality of the PLS-DA models was calculated by the internal repeated (7 repetitions) cross-validation check with 20 permutations and 5 predictive components.⁷⁹ R2X, R2Y, and Q2Y represented the fraction of variance of the X matrix, Y matrix, and predictive accuracy, respectively. Root-mean-squared error of estimation (RMSEE) between the actual and predicted responses during PLS-DA model evaluation was automatically calculated by the ropls PLS-DA function. Variables of importance in projection (VIP) scores were obtained using the ropls::getVipVn function, and VIP scores > 1.5 were considered sufficient for group discrimination. DNA adduct maps were plotted using the bubble plotting function using the ggplot2 (v3.3.3) and plotly (v4.9.2.1) R packages. EICs were produced using wSIMCity.⁷⁶ The wSIM-City web browser provides plots and data tables for inspection of putative DNA adducts.

Putative DNA Adduct Characterization.

Putative adducts identified by wide-SIM/MS² and wSIM-City analysis in low and high Gleason score patients were further characterized by data-dependent acquisition-constant neutral loss MS³ scanning (DDA-CNL/MS³). Targeted mass inclusion lists for the polar LPO and the hydrophobic bulky aromatic DNA adduct fractions were used and consisted of high-confident score adducts identified in ≥50% of the QC samples with log2-fold change differences >1 or <−1 and p-values <0.10. The DDA-CNL/MS³ method contained an MS scan from m/z 330 to 660 (hydrophobic and bulky aromatic adducts) or m/z 240 to 480 (polar LPO adducts), followed by a data-dependent MS² (ddMS²) scan for ions that had an intensity above 1 × 10⁴ and charge state of 1 or undetermined. The “perform dependent scan on most intense ion if no targets are found” feature was used to potentially identify putative DNA adducts not included in the targeted mass inclusion list. A dynamic exclusion of 6 s was employed to eliminate the continuous fragmentation of the same adduct precursor (with a mass tolerance of 10 ppm) after each MS² scan. The ddMS² scan was followed by an MS³ scan when the neutral loss of deoxyribose (dR, 116.0473 Da) was detected in the MS² spectra within a mass tolerance of 20 ppm. The MS parameters were as follows: RF-lens, 40% (bulky adducts) or 35% (LPO adducts); resolution, 120,000 for MS, 30,000 for ddMS² and ddMS³; isolation mode, quadrupole; isolation width, 1.6 m/z for ddMS² and 2.0 for ddMS³; maximum injection time, 50 ms for MS, 54 ms for ddMS², and 200 ms for ddMS³; AGC, 4 × 10⁵ for MS, 5 × 10⁴ for ddMS², and 1.6 × 10⁵ for ddMS³; activation type, HCD; CE, 25% (ddMS²) and 50% (ddMS³) for bulky adducts or 15% (ddMS²) and 40% (ddMS³) for LPO adducts; data type, profile.

Statistical Analysis of PhIP, HO-PdG and DNA Adductomics Biomarkers, and Prostate Pathology.

The purpose of the analysis was to correlate biomarkers of cooked red meat consumption, including PhIP levels in hair, PhIP and LPO DNA adducts, and untargeted DNA adductomics, to the demographics and clinical pathologies of PC. All analyses were performed with the average values of PhIP obtained from the scalp and abdomen when both hair specimens were available or solely from scalp or abdomen hair. If PhIP hair levels were below the LOQ value, PhIP levels were assigned with ½ the LOQ value (12.5 pg PhIP/g hair) and employed to adjust PhIP hair levels to the melanin content (ng PhIP/g melanin). HO-PdG adducts below the LOQ of 1 adduct per 10⁸ nts were imputed to 0.5 adducts per 10⁹ nts. Biomarker levels were characterized using basic summary statistics (i.e., mean, median, geometric mean, etc.). Biomarkers were log-transformed for analysis, and log-transformed biomarker values were correlated with demographics and clinical characteristics using linear regression with the log-transformed biomarkers as outcomes and summarized by the ratio of the geometric means between groups. The one exception was melanin content, which was analyzed on the original scale and compared between groups using Welch’s two-sample t-test assuming unequal variances. Summary statistics were calculated using GraphPad Prism version 8.4.0 for Windows, GraphPad Software (San Diego, CA), and correlative analyses were completed using the R statistical programming language.⁷⁸ Spearman rank correlation coefficients (r_s) were used to assess the associations between PhIP hair levels (pg/g hair or pg/g melanin) and weekly frequency of total meat, poultry, and fish intake, and the associations between PhIP hair levels and dG-C8-PhIP adduct levels in prostate tissue. P-values < 0.05 were assumed to be statistically significant unless otherwise noted.

Results

Description of the Cohort.

The study population (N = 325) was mainly men from Minnesota, with some patients from Wisconsin, South Dakota, North Dakota, and Iowa. The cohort included patients diagnosed with PC who underwent radical prostatectomy, bladder cancer patients who underwent cystoprostatectomy where the usually benign prostate was removed for medical reasons, and patients with BPH who underwent treatment by TURP or HoLEP. The patient population distribution was predominantly Caucasian (89%), followed by (AA (9%), and mirrored Minnesota's racial composition. The demographics and clinical characteristics of the cohort are reported in Table 1.

Table 1:

Demographic and Clinical Characteristics of the Prostate Patients^a,b,c

Variable	Patients, n	Mean, SD	Pathology	n	Mean, SD
Age years)	322	65.9 ± 8.6	PSA (ng/mL)	273	9.1 ± 11.1
BMI (kg/m2)	322	28.4 ± 4.7	Low (<4)	75
<25 kg/m2 (healthy weight)	88		Moderate (4 - 10)	125
25 - 30 kg/m2 (overweight)	125		High (>10)	73
>30 kg/m2 (obese)	109
Ethnicity		Mean age, SD	Gleason score
White/Non-Hispanic Latino	287	65.9 ± 8.4	Benign, Gleason 6	127
African-American	29	65.4 ± 9.0	Medium risk Gleason 7 (3 + 4)	90
Others	6	71.3 ± 10.3	Medium risk Gleason (4 + 3)	48
			High risk Gleason 8 (4+4)	11
Smoking status			Very high risk, Gleason 9/10	34
Never	161
Former	106		Cancer stage
Current	37		Benign	95
			Local (T1N0, T2N0)	104
Surgical procedure			Regional (T3aN0, T3bN0)	89
Prostatectomy	192		Distant (T2N1, T3aN1, T3bN1)	21
Cystoprostatectomy	40
TURP	61
HoLEP	22
No tissue	7

^aThe cohort was comprised of 325 patients; age and BMI were reported for 322 patients.

^bThe n is the number of subjects per group.

^cThe cancer stage was collapsed to three stages, which reflects the extent of the cancer spread.

T1 is nonpalpable tumor confined to prostate, T2 is palpable tumor confined to prostate, T3a is tumor spreading to extraprostatic tissue, T3b is tumor spreading to seminal vesicle. N0 is absence of spread to lymph nodes, N1 is tumor metastatic to lymph nodes

PhIP Levels in Scalp Hair of Healthy Adults on a Semicontrolled 4 Week Meat Feeding Study: The PhIP Hair Dosimeter.

These data recapitulate our key published findings.^46,47 We investigated the relationship between PhIP intake and PhIP hair levels and the modulating effects of melanin pigmentation and cytochrome P4501A2 (P4501A2) metabolism with healthy adult volunteers (N = 41) on a 4-week semicontrolled meat diet ingesting known quantities of PhIP.^46,47 The incorporation of some chemicals in hair is influenced by pigmentation, which is a potential confounding factor for exposure assessment using the hair dosimeter.⁸¹ For example, the accrual of PhIP in rodent and canine fur and human hair depends on the pigmentation content.^47,82,83 Thus, PhIP levels are reported per hair weight and adjusted per gram melanin. MeIQx and AαC, two other prevalent HAAs formed in well-done cooked meat, accrue in rodent fur and human scalp hair at considerably lower levels than PhIP and are infrequently detected in human hair.^47,48,84 The PhIP hair dosimeter data from our published feeding study are shown in Figure 1A,B.⁴⁶ The PhIP scalp hair levels are plotted versus PhIP intake (μg/day). Thirty-six of the 41 volunteers had PhIP scalp hair levels above the LOQ (25 pg/g hair), and 40 of the volunteers had PhIP at levels >0.5*LOQ at the study's onset (92 ± 79 pg/g hair, mean ± SD) when they were on a free-choice diet.⁴⁶ The Spearman rank correlation coefficient (r_s) showed a strong increase in PhIP hair levels as a function of PhIP intake (r_s = .76, p < 0.0001). The two volunteers with the highest PhIP intake harbored lower levels of PhIP in scalp hair compared to other subjects with lesser PhIP intake levels. However, after adjusting PhIP levels for the melanin content, these two subjects' PhIP hair values were more closely aligned to the PhIP intake (r_s = .82, p < 0.0001). The PhIP hair levels measured in this semicontrolled feeding study with a known intake of PhIP were used as the dosimeter to estimate the recent dietary intake of PhIP in the prostate cohort under treatment at the University of Minnesota.

Figure 1.

Correlation of post-feeding PhIP scalp hair levels in healthy volunteers with dietary intake of PhIP following a four-week semicontrolled feeding study. (A) PhIP (pg/g hair) and (B) PhIP (ng/g melanin) versus PhIP intake (μg/day). The “0” intake value represents the onset of the study, where volunteers were on a free, free-choice diet, and PhIP intake was not known. Adapted with permission from reference 46.

PhIP Levels in Scalp and Abdomen Hair of Prostate Patients.

Hair was obtained from 205 of the 325 patients. However, many patients had insufficient scalp hair. Therefore, hair was collected from the lower abdomen as an alternative biospecimen. Scalp hair was obtained from 11 patients, matching scalp and abdomen hair was obtained from 35 patients, and abdomen hair from 159 patients. PhIP was above the LOQ in 123 patients (60%). Twenty-four subjects with matching scalp and abdomen hair had PhIP hair levels > 0.5*LOQ (12.5 pg/g hair). The PhIP hair levels in these matching scalp and abdomen hair specimens were plotted using a least-squares regression curve with equal weightings (Figure 2). The correlations of PhIP levels in scalp and abdomen hair were highly significant (PhIP pg/g hair: y = 1.26x – 9.89; r² = 0.92, p < 0.0001, and PhIP ng/g melanin: y = 0.99x – 0.75, r² = 0.94, p < 0.0001). One patient harbored high PhIP hair levels in matching scalp and abdomen hair. The p-values and the goodness of fit values (r²) were lower when removing this single high-PhIP hair subject; nonetheless, the correlations between scalp and abdomen PhIP hair levels remained significant, and stronger when PhIP levels were adjusted for melanin content (PhIP pg/g hair: y = 0.77x −10.9, r² = 0.32, p = 0.005, and PhIP ng/g melanin: y = 0.77x – 0.42, r² = 0.70, p < 0.0001). We conclude that PhIP levels in scalp and abdomen hair are highly correlated. Representative mass chromatograms of PhIP present in abdomen hair are shown in Figure S1.

Figure 2.

Correlation of PhIP hair levels reported as PhIP (pg/g hair) or adjusted for melanin content PhIP (ng/g melanin) in matching scalp and abdomen hair of prostate patients (n = 24). The data sets include PhIP (ng/g hair) ≥ 0.5*LOQ. The high PhIP-hair subject was removed from the regression analysis of the inserted curves. The dashed lines represent the 95% CI of the slope. The goodness of fit measures r² and p-values are reported.

Histograms and scatter plots are shown for PhIP hair levels (pg/g hair) (Figure 3A) and adjusted for melanin content (ng/g melanin) in the prostate cohort (Figure 3B). The median level of PhIP is 33.2 pg/g hair and 64.7 and 132 pg/g hair, respectively, at the 75^th and 90^th percentile in the cohort subgroup assayed (n = 205). The median level of PhIP adjusted for melanin content was 3.70 ng/g melanin, and 7.10 and 12.5 ng/g melanin, respectively, at the 75^th and 90^th percentiles (n = 205) (Figure 3B). The PhIP hair levels are not normally distributed, with a positively skewed distribution shifted to low PhIP-hair levels. The PhIP hair levels in the prostate cohort (57.8 ± 77.4 pg/g hair, mean ± SD) are significantly lower than the hair levels of young, healthy adults (n = 41) during the non-restricted, free-choice diet phase of the semi-controlled meat feeding study conducted at the University of Hawaii (92.0 ± 79.2 pg/g hair, p =.01) (Figure 3).⁴⁶ The melanin content reportedly influences PhIP accrual in hair.^47,81-84 The prostate cohort is comprised of older-aged men (mean ± SD, 65.9 ± 8.6 years of age) whose hair follicles produce less color than hair follicles of the University of Hawaii primarily comprised of younger adults (18–63 years, median: 25 years). Thus, the scalp hair melanin content is lower in the prostate cohort (7.62 ± 5.71, mg/g hair, mean ± SD) than the volunteers on the semi-controlled feeding study (17.7 ± 9.08), p < 0.0001 (Welch's two-tailed t-test). The melanin content of abdomen hair in the prostate cohort is also significantly lower (10.8 ± 6.06) than the melanin content in scalp hair of the healthy younger volunteers p < 0.0001. The majority of the PC patients (168 of 205 subjects, 82%) had PhIP hair levels below the mean PhIP hair level of the University of Hawaii cohort during the free-choice diet phase. More than 95% of the prostate cohort had PhIP hair levels below 235 pg/g of hair, a level which corresponds to a daily PhIP intake of 0.38 – 0.45 μg for volunteers on the semi-controlled feeding study (Figures 2 and 3). When adjusting PhIP hair levels per melanin content, 71% of the prostate cohort had levels below the PhIP mean of the University of Hawaii cohort (6.13 ng PhIP/g melanin) on the free-choice diet, and 87% of the prostate cohort had PhIP hair levels below the PhIP mean (10.9 ng PhIP/g melanin) of the University of Hawaii cohort with a daily PhIP intake of 0.38 – 0.45 μg. We conclude that current PhIP consumption by the large majority of prostate cohort patients, based on the PhIP hair dosimeter, is below the average daily intake of PhIP reported for young adult males (>30 years of age) in the United States, which ranges from 0.49 – 0.98 μg/70 kg adult.^34,35

Figure 3.

Histograms and scatter plots of PhIP hair levels in (A) PhIP (pg/g hair), and (B) PhIP hair levels adjusted for melanin (ng/g melanin) in the University of Minnesota prostate cohort. PhIP hair values below the LOQ (25 pg/g hair) are reported as 0.5*LOQ (12.5 pg/g hair) and used for reporting the PhIP adjusted for melanin content (ng/g melanin). The mean level of PhIP hair levels (pg/g hair and ng/g melanin) of the University of Hawaii cohort at the basal level (free-choice diet) and following consumption between 0.38 - 0.46 μg PhIP five days per week for four weeks are depicted by red arrows (see Fig. 1 and reference ⁴⁶)

Association of PhIP Hair Levels with Meat, Poultry and Fish Intake.

The meat, poultry, and fish consumption per week is reported for the subjects with PhIP hair measurements (Table 2). The median red meat and poultry servings were three times per week, two weekly servings for processed meat, and one for fish. Almost half the patients consumed cooked red meat and processed meats five or more times per week. The four patients who were strict vegetarians had PhIP hair levels below the LOQ, confirming our earlier PhIP hair biomarker data with vegetarians.⁴⁸ The correlation between PhIP hair levels with weekly servings of total meat, poultry, and fish intake is shown in Figure 4. Although some patients have substantially high levels of PhIP in their hair, the correlation between the frequency or meat, poultry, and fish consumption and PhIP hair levels is weak (Spearman correlation coefficient (r_s) for PhIP hair levels (ng/g) and melanin content (ng/g melanin) were, respectively r_s =.09 (p = 0.22) and r_s =.05 (p = 0.54). The correlation was not strengthened when excluding subjects with PhIP hair levels below the LOQ value. PhIP forms across many types of well-done cooked meats, poultry, and grilled seafood.^{17,18,20,85-88} The portion sizes of these meat dishes and the degree of meat well-doneness are uncertain. The HAA concentrations were not measured in our cohort's diet, contributing to the poor correlation between the self-reported intake of these cooked foods and PhIP hair levels.

Table 2:

Meat, Poultry and Fish Servings per Week^a,b

Consumption per week	Red meat, processed meat, poultry and fish combined	Red meat	Processed meat	Poultry	Fish
Minimum	0.0 (4)	0.0 (12)	0.0 (39)	0.0 (18)	0.0 (49)
25% Percentile	7.6 (39)	1.0 (34)	1.0 (36)	2.0 (75)	0.0 (49)
50% Percentile	10 (54)	3.0 (56)	2.0 (28)	3.0 (40)	1.0 (74)
75% Percentile	13.0 (32)	5.0 (40)	4.0 (32)	4.0 (23)	2.0 (27)
Maximum	26.5 (43)	14.0 (30)	8.5 (37)	15.0 (34)	12.0 (22)

^an = 172 patients with PhIP hair measurements who completed the FFQ

^bthe numbers in the parentheses report the subject number in the quartile

Figure 4.

Correlation of (A) PhIP hair levels (pg/g hair) and (B) PhIP hair levels adjusted for melanin (ng/g melanin) content with weekly servings of red and processed meat, poultry, and fish. Spearman correlation coefficients r_s with p-values are reported. The PhIP hair values of the four strict vegetarians are depicted as red circles.

PhIP Hair Levels in Caucasian and African American Men.

We examined the PhIP hair levels in AA and Caucasians of the prostate cohort. Scalp hair was only available for two AA patients, but abdomen hair was obtained from 16 AA and 177 Caucasian patients (Table 3, Figure 5). Fifteen of the 16 AA subjects (93.8%) had PhIP hair levels above the LOQ, while 97 of the Caucasian subjects (55%) had PhIP levels above the LOQ. The PhIP geometric mean level reported as pg/g hair was significantly higher in AA than Caucasians (104 versus 29.9, p = 0.009). However, there was no statistically significant difference between PhIP geometric mean hair levels in AA and Caucasians when values were adjusted for melanin content (PhIP ng/g melanin) (4.8 versus 3.4, p = 0.27). The conclusions on PhIP hair levels and exposure differences between AA and Caucasians should be interpreted with caution due to the low number of AA subjects. Furthermore, hair pigmentation and density differ among ethnic groups, which influence PhIP accrual in hair.^47,89

Table 3:

PhIP Hair Levels and Prostate Pathology Biomarkers^a,b,c,d

	PhIP (pg/g hair)			PhIP (ng/g melanin)			PhIP (ng/g melanin)d
Variable	n	Geometric mean (IQR)	p	n	Geometric mean (IQR)	p	n	Geometric mean (IQR)	p
Ethnicity
White	185	30.3 (38.5)	<0.001	185	3.7 (5.1)	0.29	178	3.5 (4.6)	0.22
Black	16	105 (56.7)		16	4.8 (3.4)		16	4.8 (3.4)
Other	3	24.3 (24.1)		3	2.2 (3.4)		3	2.2 (3.4)
Smoking status
Never	104	38.1 (63.5)	0.09	104	4.0 (5.3)	0.30	102	3.9 (5.1)	0.31
Former	64	31.6 (42.5)		64	3.8 (3.6)		60	3.4 (3.6)
Current	28	24.6 (36.3)		28	3.0 (5.9)		27	2.9 (6.0)
Surgical procedure
Cystoprostatectomy	38	25.1 (35.0)	0.02	38	3.0 (3.4)	0.04	37	2.9 (3.5)	0.09
BPH (HoLEP/TURP)	24	24.0 (29.8)		24	2.9 (2.6)		23	2.7 (2.9)
Prostatectomy	141	38.4 (58.2)		141	4.2 (6.3)		135	3.9 (5.7)
PSA ng/mL (binary)
PSA < 4	44	26.6 (36.5)	0.03	44	3.3 (3.7)	0.14	40	2.9 (3.6)	0.07
PSA ≥ 4	133	38.6 (60.9)		133	4.1 (5.7)		131	3.9 (5.2)
Gleason score (binary)
Benign and 6	71	27.0 (38.4)	0.02	71	3.1 (3.6)	0.03	68	2.9 (3.8)	0.02
7 and above	131	37.7 (59.5)		131	4.1 (5.9)		127	3.9 (5.6)
Cancer status
- Benign	52	26.2 (33.5)	0.03	52	3.1 (3.6)	0.09	49	3.0 (3.8)	0.09
- Malignant	150	36.5 (59.9)		150	4.0 (5.7)		146	3.8 (5.3)
Cancer stage
Benign	52	26.2 (33.5)	0.20	52	3.1 (3.6)	0.41	49	3.0 (3.8)	0.32
Local	69	37.1 (56.3)		69	4 (5.0)		69	3.9 (5.2)
Regional	66	36.2 (66.6)		66	3.9 (5.2)		62	3.5 (4.2)
Distant	14	37.9 (46.7)		14	4.3 (6.0)		14	4.3 (6.0)

^aThe associations between PhIP (pg/g hair) and PhIP (ng/g melanin) levels used the average values of PhIP obtained from the scalp and abdomen when both hair specimens were available or from scalp or abdomen hair. PhIP values below the LOQ value (25 pg/g hair) were assigned ½ the LOQ (12.5 pg/g hair).

^bn = number of subjects assayed per group for the PhIP hair biomarker

^cThe geometric mean of PhIP is reported with the interquartile range (IQR)

^dThe geometric mean of PhIP is reported with the interquartile range (IQR) reported melanin content values > 3.5 mg melanin/g hair

Figure 5.

PhIP levels in the abdomen hair of AA and Caucasians (A) PhIP pg/g hair and (B) PhIP ng/g melanin. The geometric means and 95% CI are depicted of PhIP (pg/g hair) and PhIP (ng/g melanin). **p = 0.009 Welch’s t-test AA vs Whites (PhIP pg/g hair).

Targeted Analysis of Cooked Meat Carcinogen DNA Adducts Formed in the Prostate Peripheral Zone by Nanoliquid chromatography–high resolution tandem mass spectrometry (nanoLC/MS²).

DNA adducts were measured in the peripheral zone (PZ), the most common site of PC development.³² We previously characterized PhIP-DNA adducts in matching freshly frozen and formalin-fixed paraffin-embedded prostate tissue.^38,52,90 This study extended our previous measurements of the DNA adducts of PhIP and other cooked meat carcinogens, including MeIQx, AαC, and B[a]P (Scheme 1) to 149 PC patients. A representative EIC and MS³ product ion spectrum of dG-C8-PhIP identified in PC patients are shown in Figure S2. Thirteen out of the 149 PC patients harbored dG-C8-PhIP at values above the LOQ, and one patient had dG-C8-PhIP above the LOD of 1 adduct per 10⁹ nts for a total 9.4% positivity. The data are depicted as a scatter plot (Figure 6A).^38,52 Three of the 18 AA patients were positive for dG-C8-PhIP (16.7%), while 11 out of 131 Caucasian patients (8.4%) were positive. Anecdotally, the two subjects with the highest levels of dG-C8-PhIP were AA.

Figure 6.

(A) Scatter plots for dG-C8-PhIP in the prostate, (B) Spearman correlation for dG-C8-PhIP levels versus PhIP hair levels (pg/g hair) and (C) Spearman correlation for dG-C8-PhIP levels versus PhIP hair levels adjusted for melanin.

Ten PC patients with measured PhIP hair levels were positive for dG-C8-PhIP in prostate DNA. The PhIP hair values were not normally distributed (Shapiro-Wilk test). The Spearman correlations for dG-C8-PhIP adduct and PhIP hair levels are shown in Figure 6B,C.. PhIP hair levels represent the unmetabolized dose, whereas dG-C8-PhIP is a biomarker of the biologically reactive intermediate. Its levels are influenced by the collective activities of P450 and phase II enzymes involved in the bioactivation and detoxication of PhIP.^36,49 Also, dG-C8-PhIP is a substrate for the nucleotide excision repair (NER) DNA repair pathway, and the time interval since PhIP consumption is another variable influencing DNA adduct levels.⁹¹ There was a positive trend for higher dG-C8-PhIP adducts in this subset of patients with increasing PhIP pg/g hair levels (r_s =.60, p = 0.073) and adjusted for melanin (ng/g melanin) (r_s =.43, p = 0.22); however, more DNA adduct-positive subjects are required to investigate the potential significance of the association with PhIP hair levels. The dG-C8-MeIQx, dG-C8-AαC, and dG-N²-B[a]PDE adducts were below the LOD in the PZ zone of all PC patients and below the LOD in the transition zone of 15 patients treated for BPH.^38,52,90

Targeted Analysis of LPO Adducts in the Prostate PZ by Nanoliquid Chromatography–High Resolution Tandem Mass Spectrometry.

Inflammation and oxidative stress are hallmark features of PC and PhIP-induced prostate carcinogenesis in rodents.^29,32 We measured oxidative stress-induced LPO DNA adducts in the PZ of 88 subjects who underwent prostatectomy (n = 52) or cystoprostatectomy (n = 36). 1,N²-dG adducts of acrolein were frequently detected, as previously reported in our pilot study.⁵³ 6-HO-PdG was above the LOQ value in 25 subjects (28.4%), and 8-HO-PdG was present above the LOQ in 48 (54.5%) of the 88 subjects assayed. The mean adduct level and 95% CI per 10⁸ nts for 6-HO-PdG was 6.99 (1.64 – 12.3); 8-HO-PdG was 5.94 (2.65 – 9.24), and 6-HO-PdG + 8-HO-PdG combined was 12.9 (4.44 – 21.42). The EIC of HO-PdG adducts and their product ion spectra at the MS³ scan stage are shown in Figure 7. The 8-HO-PdG mass spectrum has a major product ion at m/z 164.0564 and is absent in the spectrum of 6-HO-PdG. The scatter plots of the HO-PdG adduct levels in the prostate of PC patients who underwent prostatectomy and benign prostates of cystoprostatectomy patients are summarized in Figure 8. The HεdC, HεdG, HεdA, εdA, εdG, εdC, HNE-dG, and M₁dG LPO adducts were detected sporadically, generally occurring at levels near or below the LOQ values (≤ 1 adduct per 10⁸ nts) (Figure 9). However, an isomer of HNE-dG was detected at t_R 26.2 min in the prostate DNA of many patients. Further characterization of this adduct is reported (vide infra).

Figure 7.

EIC (5 ppm m/z tolerance) of 6-HO-PdG and 8-HO-PdG in (A) DNA of a prostate cancer patient positive for HO-PdG adducts and (B) DNA of a prostate cancer patient with adducts below the LOQ (1 adduct per 10⁸ nts). The 6-HO-PdG eluted earlier as two interchanging peaks, whereas 8-HO-PdG eluted as a single peak. (C) The MS³ product ion spectra of HO-PdG adducts and their [¹⁵N₅]-HO-PdG internal standards are shown in the right panel.

Figure 8.

6-HO-PdG and 8-HO-PdG, and total (t)-HO-PdG levels (mean and SD) in the prostate specimens of prostatectomy and cystoprostatectomy patients.

Figure 9.

Extracted ion chromatograms of LPO adducts and their internal standards at the MS² scan stage [M+H]⁺ →[M+H −116.0473]⁺ except for M₁dG and [¹⁵N₅]-M₁dG which underwent in-source fragmentation to the aglycone followed by MS². Internal standards were added at a level of 1.3 adducts per 10⁷ nts before enzymatic digestion of DNA. The * peak is an isomer of HNE-dG.

Associations of PhIP Hair and HO-PdG DNAAdduct Levels with Prostate Pathology Biomarkers and Surgical Treatments.

The relationships between PhIP hair levels with PSA blood values, Gleason scores, prostate malignancy, BPH, surgical procedures, and tumor stage are reported in Table 3. The associations of HO-PdG DNA adduct levels with these pathology biomarkers are summarized in Table 4. The PSA blood test screens for PC, but the test does not provide precise diagnostic information about prostate pathology because cancerous and noncancerous prostate tissue produce the PSA protein. Approximately 27% of men with PC had a PSA test with normal protein values (< 4 ng/ml) (Table 1), consistent with previous reports.⁹² Intermediate and high PSA levels may indicate the presence of PC. However, other conditions, such as an enlarged or inflamed prostate, can also increase PSA levels. PSA values also increase as a function of age, which can confound the reliability of this protein as a diagnostic biomarker of PC.^93,94 The Gleason grading system is used to evaluate the men's prognosis of PC and is based on the prostate biopsy's microscopic appearance. Gleason scores range from 6 to 10. Gleason scores 6 (Grade Group 1) are low or very low-risk groups, Gleason scores of 7 ((3+4) or (4+3)) (Grade Groups 2 and 3) are intermediate-risk, and scores of 8 – 10 (Groups 4 and 5) are considered high to very high risk.^70,95 The PC patients with a higher Gleason score have more aggressive cancers with poorer prognoses.

Table 4:

HO-PdG Adducts and Prostate Pathology Biomarkers^a,b,c,d,e

	6-HO-PdG x 10−8 nts			8-HO-PdG x 10−8 nts			(6-HO-PdG + 8-HO-PdG) x 10 −8 nts
Variable	n	Geometric mean (IQR)	p	n	Geometric mean (IQR)	p	n	Geometric mean (IQR)	p
Ethnicity
White	78	1.1 (1.3)	0.38	78	1.3 (3.3)	0.18	78	2.5 (4.6)	0.23
Black	10	1.7 (5.1)		10	2.6 (11.5)		10	4.6 (26.3)
Smoking status
Never	44	1.5 (4.5)	0.27	44	1.8 (5.8)	0.22	44	3.5 (9.8)	0.24
Former	31	1.0 (0)		31	1.3 (2.6)		31	2.4 (3.6)
Current	11	0.7 (0)		11	0.8 (0)		11	1.6 (0.0)
Surgical procedure
Cystoprostatectomy	36	1.3 (2.5)	0.46	36	1.5 (2.7)	0.75	36	3.0 (6.1)	0.65
BPH (HoLEP/TURP)		NA			NA			NA
Prostatectomy	52	1.1 (1.1)		52	1.4 (3.6)		52	2.6 (5.1)
PSA ng/mL (binary)
PSA < 4	19	1.2 (1.2)	0.93	19	1.5 (1.7)	0.91	19	2.9 (3.8)	0.90
PSA ≥ 4	49	1.1 (2.8)		49	1.5 (4.3)		49	2.7 (8.5)
Gleason score (binary) b
Benign and 6	42	1.1 (1.2)	0.61	42	1.2 (1.7)	0.38	42	2.4 (3.3)	0.42
7 and above	46	1.3 (4.2)		46	1.6 (4.2)		46	3.1 (8.4)
Cancer status
Benign (cystoprostatectomy patients)	27	1.1 (0.8)	0.90	27	1.3 (1.7)	0.67	27	2.5 (3.5)	0.72
Malignant	61	1.2 (2.8)		61	1.5 (3.9)		61	2.8 (5.7)
Cancer stage
Benign	27	1.1 (0.8)	0.63	27	1.3 (1.7)	0.52	27	2.5 (3.5)	0.58
Local	31	0.9 (0.0)		31	1.3 (2.2)		31	2.3 (3.1)
Regional	26	1.6 (5.1)		26	2.0 (10.1)		26	3.8 (16.2)
Distant	4	1.0 (1.9)		4	0.8 (0.8)		4	1.9 (2.7)

^an = number of subjects assayed for HO-PdG in prostate

^bBenign prostate specimens are from bladder cancer patients who underwent cysotprostatectomy

^cThe geometric mean is reported with the interquartile range (IQR)

^dThere were 7 current smokers in prostatectomy patient group (6 subjects had combined HO-PdG levels <LOQ, one subject harbored combined HO-PdG = 11.8 adducts per 10⁸ nts). There were 4 current smokers in the cysoprostatectomy group (3 subjects had combined HO-PdG levels < LOQ, one subject harbored combined HO-PdG = 11.8 adducts per 10⁸ nts)

^eNA = not analyzed

The PhIP hair levels expressed as pg/g hair or ng/g melanin were not associated with the PSA values or Gleason scores when the prostate pathology biomarkers were analyzed as continuous variables (Table 3). Therefore, pathology biomarkers were dichotomized into clinically meaningful threshold prostate pathology scores for PC risk. Patients were grouped with PSA values < 4 ng/mL versus patients with elevated and high-risk PSA levels ≥ 4 ng/mL.⁷² For Gleason scores, the low-risk patients with an indolent Gleason score of 6 and patients with BPH were grouped together, and the patients with intermediate and high-risk Gleason scores 7 – 10, who undergo treatment with radiation or surgery, served as the second group.^70,96,97 PhIP levels were higher in patients with PSA values ≥ 4 ng/mL (geometric mean, 38.6 versus 26.6 pg/g hair, p = 0.03). PhIP hair levels were also higher in patients with Gleason scores ≥ 7 (geometric mean 37.7 versus 27.0 pg/g, p = 0.02) and when PhIP hair levels were adjusted for melanin (geometric mean 4.1 versus 3.1 ng/g melanin, p = 0.03). PC patients undergoing prostatectomy had significantly higher PhIP hair levels (38.4 pg/g hair) than those patients under treatment for BPH (24.0 pg/g hair) and bladder cancer patients (25.1 pg/g hair, p = 0.02). PhIP hair levels were associated with cancer status and were higher in patients with PC than the patients with benign disease (p = 0.03). PhIP hair levels were not associated with PC stage groupings.

Elevated HO-PdG adduct levels were not associated with elevated PSA levels, aggressive Gleason scores, cancer status, surgical treatments, or PC stage groupings (Table 4). Two bladder cancer patients who underwent cystoprostatectomy had combined HO-PdG adduct levels in the prostate exceeding 240 adducts per 10⁸ nts or about ~10-fold higher than the mean combined HO-PdG levels detected in PC patients. However, excluding these two patients from the analysis did not alter the association (Figure 8). The 6-HO-PdG (p = 0.88) and 8-HO-PdG (p = 0.64) or combined HO-PdG levels (p = 0.69) were not significantly different between PC and cystoprostatectomy patients. The combined HO-PdG adduct levels did not differ among AA and Caucasians (p = 0.23).

DNA adduct discovery in the prostate using untargeted DNA adductomics.

We examined global DNA damage in the prostate patients’ genome, employing our wideSIM/MS² DNA adductomics screening method and wSIM-City for data analysis.^54,76,98,99 A subset of 12 prostate DNA samples were selected from PC patients with a high Gleason score ≥ 7 (n = 6) and low Gleason score 6 (n = 6) for analysis of the DNA adduct profiles. This technology screens for putative DNA adducts by extracting the coeluting signals of modified nucleoside precursor ions ([M+H]⁺) in the wSIM scan mode and their aglycone ions [BH₂]⁺ ([M+H₂-dR]⁺) at the MS² scan stage. ^54,76 The DNA adductomics experiments successfully detected several hundred potential adducts using wSIM-MS² scanning technology and the wSIM-City algorithm for data analysis. The DNA adduct list increased as a function of the number of subjects. Some DNA adducts were unique to each patient, and the levels of many DNA adducts were highly variable. However, some adducts were not included in the statistical analyses because of the stringent scoring criteria used by wSIM-City for DNA adduct identification. Adducts not observed in at least 50% of the pooled QC samples of low and high Gleason score patients were excluded from the statistical analysis during the filtering step, a criterion established for DNA adductomics analyses. This exclusion does not imply that some of the omitted DNA adducts are not biologically relevant. For example, mutation-prone HO-PdG adducts were only detected in three of the twelve patients assayed. The adduct levels were substantially diluted in the pooled QC samples and not detected. Thus, HO-PdG adducts were not included in the final data set for statistical analysis by wSIM-City. Other adducts, such as the putative 4-hydroxy-2-alkenal 1,N²-dG adducts (discussed below) in the polar LPO DNA adduct fraction, were detected by wSIM-City (vide infra). However, several did not pass quality checks for inclusion in the final DNA adductomics data set because of an insufficient number of neutral loss scans detected (minimum of 5 scans across the peak) or slight retention time differences between the apexes of the precursor and the aglycone peaks. These criteria are employed by wSIM-City to minimize the false-positive detection rate of putative DNA adducts. Further development of the wSIM-City algorithm will improve its robustness for automated DNA adduct detection. A larger subject number may be required for DNA adductomics and associating specific DNA adducts with disease risk.

A list of potential DNA adducts was generated for each subject, and the m/z precursor ion values were mined using a database comprised of 334 DNA adducts formed with exogenous genotoxicants and endogenous electrophiles.^76,100 However, most putative DNA adducts identified by wSIM-City are not reported in the literature and are of unknown origin.^76,100,101 After filtering the polar LPO-DNA adduct fraction list using the DNA adduct criteria score, 1152 putative DNA adducts remained as candidate DNA adducts. The adduct list narrowed to 68 putative adducts based on significant differences in ion abundance between high and low Gleason score patients (p < 0.1), and 39 putative adducts remained when filtered for p < 0.1 and VIP > 1.5. After filtering, the bulky aromatic DNA adduct fraction list contained 1270 putative adducts remained; 78 putative adducts showed a significant difference between high and low Gleason score (p <0.1), and 33 adducts fit the criteria for p < 0.1 and VIP > 1.5. The list of the putative DNA adducts identified by wSIM-City is reported in Table S1.

The ion abundances for many presumed adducts were low and insufficient to successfully conduct DDA-CNL/MS³ or targeted MS³ to characterize the aglycone structures. Moreover, the product ion spectra at the MS³ scan stage were often aggregate spectra due to the isobaric interferences bordering the aglycones, making spectral interpretation of the adducts difficult (unpublished observations, R. Turesky). Longer gradients are required to separate DNA adducts from isobaric interferences with an increased number of scans across the peaks to improve the quality of the MS³ spectra of the aglycones. Bioinformatic approaches will be required to deconvolute many mass spectra to aid in spectral interpretation.

One polar adduct in the LPO-DNA adduct fraction occurred at a 1.79-fold-higher level in the high Gleason score group (p = 0.05) and was detected in all 12 patients. The adduct's MS² and MS³ product ion spectra are shown in Figure 10. The spectra are an excellent match to synthetic N²-hydroxymethyl-2′-deoxyguanosine (N²-HOCH₂-dG), an adduct of dG formed with formaldehyde.⁶⁶ The precursor ion is detected at m/z 298.1140 (C₁₁H₁₅N₅O₅⁺) (2.0 ppm relative error to the theoretical m/z). The MS² spectrum contains the aglycone at m/z 182.06720 (C₆H₈O₂⁺) (−0.5 ppm relative error to the theoretical m/z) and an ion at m/z 164.0567, attributable to the loss of H₂O from the aglycone (C₆H₆N₅O)⁺, and assignable as the iminium cation (0 ppm relative error to the theoretical m/z). A product ion at m/z 152.0566 was also observed and assignable to protonated guanine (C₅H₆N₅O⁺) (−0.7 ppm relative to the theoretical m/z). The MS³ spectrum acquired at elevated HCD shows fragmentation of the aglycone with product ions observed at m/z 164.0564, 152.0566, and 135.0301 (C₅H₃N₄O⁺) (0.0 ppm relative to the theoretical m/z) and m/z 110.0346 (C₄H₄N₃O⁺) (0.0 ppm relative to the theoretical m/z), attributed, respectively, to the losses of NH₃ and H₂NCN from guanine.¹⁰² The ion at m/z 153.0403 is a water cluster of m/z 135.0298 (theoretical m/z 153.0407)

Figure 10.

DNA adduct discovery in the prostate genome using wSIM/MS² and wSIM-city data analysis. (A) EICs for precursor ion m/z 298.1146 (MS¹ blue line) and aglycone m/z 182.0673 (MS² red line) of the putative N²-HOCH₂-dG adduct. (B) DDA CNL/MS² and (C) DDA CNL/MS³ spectra for putative N²-HOCH₂-dG in prostate DNA, and (D, E) MS² and MS³ spectra of synthetic N²-HOCH₂-dG under the same collision energies. The ion at m/z 153.0403 is a water cluster of m/z 135.0298 (exact mass: 153.0407). The HCD collision energies were 25 for MS² and 40 for MS³.

N²-HOCH₂-dG is unstable and can undergo deformylation during DNA isolation.⁶⁶ Thus, we did not expect to detect this adduct in the prostate genome. N²-HOCH₂-dG and the structurally related adduct N²-hydroxyethyl-2′-deoxyguanosine formed with acetaldehyde are in equilibrium with their Schiff bases and often reduced with NaBH₃CN to form the stable N²-methyl-dG and N²-ethyl-dG adducts before enzymatic digestion of DNA and mass spectral analysis.^103-105 Although, N²-HOCH₂-dG and the structurally related N⁶-hydroxymethyl-2′-deoxyadenosine adduct have been detected in cellular DNA in their nonreduced forms.^66,106 We analyzed CT-DNA modified with formaldehyde, and some portion of the N²-HOCH₂-dG was recovered following the nuclease digestion of DNA. The EICs and MS³ spectra of synthetic N²-HOCH₂-dG and the adduct recovered from CT DNA are shown in Figure S3. Formaldehyde is a metabolite of many carcinogens, alkaloids, including nicotine, caffeine, and drugs.^107,108 Formaldehyde is also produced endogenously from the metabolism of several amino acids, and the endogenous concentration of formaldehyde in human blood is estimated at 100 μM.¹⁰⁹ Studies employing DNA pretreated with NaBH₃CN are required to study further labile aldehyde-derived DNA adducts in the prostate genome.^104,105

Characterization of a Series of Putative 4-Hydroxy-2-Alkenal-like DNA Adducts.

Wide-SIM/MS² detected five DNA adducts with precursor ions at m/z 382.1734, 396.1894, 410.2045, 424.2191, and 438.2368 that overlapped in the polar LPO and bulky hydrophobic aromatic DNA adduct fractions (Figure 11). The mass differences increased by 14.0157 ± 0.0011 Da (CH₂), suggesting a series of related DNA adducts differing in the carbon chain length. All the aglycones underwent fragmentation at the MS³ scan stage during DDA CNL/MS³ analysis to lose one and two molecules of water [BH₂-H₂O]⁺ and [BH₂-2H₂O]⁺ (Figure 12). Other product ions of the aglycones include fragments with the neutral loss of H₂O and CO or HCO₂H [BH₂-CH₂O₂]⁺, and ions at m/z 152.0564 (C₅H₆N₅O⁺) attributable to protonated Gua (−2.6 ppm relative error) and m/z 135.0299 (C₅H₃N₄O⁺) attributable to [Gua-NH₃]⁺ (−2.2 ppm relative error).¹⁰² The ion at m/z 164.0564 (C₆H₆N₅O)⁺ (0 ppm relative error) is the proposed iminium cation and detected at low abundance for all adducts. The accurate m/z of the precursor and aglycone ions and the product ions in the DDA CNL/MS³ spectra suggest cyclic 1,N²-propano-dG adducts had formed with a series of 4-hydroxy-2-alkenal homologues. The adducts with precursor ions at m/z 382.1734 and 424.2191 are within 3.1 ppm mass accuracy for HHE-dG and HNE-dG, respectively, and the mass accuracies of the aglycones are within 4.4 ppm mass accuracy (Table 5).^67,110 HHE-dG and HNE-dG are formed through Michael adduction of the N² amino group of dG and the C-3 position of the 4-HHE and 4-HNE, followed by ring closure between N1 of dG and the C1 atom of the 4-hydroxy-2-alkenals to form two pairs of diastereomeric adducts (Scheme 2).^67,110,111 Michael addition in the opposite pathway, with initial bond formation between N1 atom of dG and C3 atom of these 4-HO-alkenals followed by ring closure at the N² amino group of dG were not detected.⁶⁷ In contrast, acrolein reacts with dG by both pathways to form 6-HO-PdG and 8-HO-PdG (Scheme 1).¹¹² Neither 4-HHE nor 4-HNE form the cyclic N²,3 propano-dG adducts in contrast to vinyl chloride, which reacts with dG to form the N²,3 etheno adduct in high yield.¹¹³ We hypothesize that these 4-hydroxy-2-alkenal-like dG adducts may have undergone rearrangement to their cyclic hemiacetal adducts.¹¹⁴

Figure 11.

DNA adduct discovery using wSIM/MS₂ scanning and wSIM-city data analysis. EICs for MS¹ (blue line) and MS² (red line) for putative 4-hydroxy-2-alkenal 1,N²-dG adducts detected at m/z 382.1734 (t_R: 13.94 min), 396.1891 (t_R: 16.21 min), 410.2045 (t_R: 18.25 min), 424.2203 (t_R: 20.29 min), and 438.2368 (t_R: 22.16 min). The precursor ions (blue sticks) and aglycones (red sticks) are shown with smoothed EICs. The monoisotopic MS precursor and its first isotopologue peak (blue sticks), and the aglycone's protonated ion, and its first isotopologue peak (red sticks) are shown in the wSIM and MS² spectra.

Figure 12.

DDA CNL/MS³ of putative 4-hydroxy-2-alkenal-like dG adducts. Left panel: mass spectra for (A) m/z 382.1734, (B) m/z 396.1891, (C) m/z 410.2045, (D) m/z 424.2203 and (E) m/z 438.2368. The right panel depicts the MS³ spectra of synthetic (F) HHE-dG and (G) HNE-dG acquired under the same CE conditions. The mass spectra of the synthetic adducts are composites of all isomers.

Table 5:

Mass Accuracy Measurements of Proposed 4-Hydroxy-2-enal 1,N² propano-dG adducts Detected in Human Prostate DNA^a,b

	MS1			MS2
Adduct	[M+H]+	theoretical m/z	error (ppm)	[BH2]+	theoretical m/z	error (ppm)
4-HO-2-Hexenal	382.1733	382.1721	3.1	266.1258	266.1248	3.8
4-HO-2-Heptenal	396.1895	396.1878	4.3	280.1422	280.1404	6.4
4-HO-2-Octenal	410.2045	410.2034	2.7	294.1573	294.1561	4.1
4-HO-2-Nonenal	424.2204	424.2191	3.1	308.1730	308.1717	4.2
4-HO-2-Decenal	438.2368	438.2347	4.8	322.1889	322.1874	4.7
	MS3
Adduct	[BH2-HCO2H-NH3]+	[BH2-HCO2H]+	[BH2-H2O]+	[BH2-2H2O]+	[Gua]+	[Gua-NH3]+
4-HO-2-Hexenal	203.0924 (−2.0)	220.1189 (−1.8)	248.1137 (−2.0)	230.1033 (−1.3)	152.0564 (−2.0)	135.0300 (−0.7)
4-HO-2-Heptenal	217.1080 (1.8)	234.1346 (−1.3)	262.1297 (−0.4)	244.1189 (−1.2)	152.0564 (−2.0)	135.0300 (−0.7)
4-HO-2-Octenal	231.1236 (−2.2)	248.1501 (−2.0)	276.1450 (−1.8)	258.1342 (−2.7)	152.0563 (−2.6)	135.0298 (−2.2)
4-HO-2-Nonenal	245.1392 (−2.0)	262.1657 (−1.9)	290.1602 (−3.1)	272.1501 (−1.5)	152.0564 (−2.0)	135.0299 (−1.5)
4-HO-2-Decenal	259.1548 (−2.3)	276.1815 (−2.0)	304.1762 (−2.0)	286.1658 (−1.4)	152.0563 (−2.6)	135.0299 (−1.5)

^aerror from theoretical m/z values for MS and MS² reported in ppm

^berror from theoretical m/z values for MS³ is reported in ppm in parentheses

Scheme 2.

Cyclic 4-hydroxy-2-nonenal (HNE) Michael addition adducts formed at the N1 and N² positions of dG and ring-opening chemistry to form the hemiacetal adducts. There are four diastereoisomers for the HNE-1,N²-dG, and HNE hemiacetal adducts (stereoisomer bonds not shown). The numbering scheme of the HNE-dG adduct is adapted from Huang et al.¹¹⁴ The HNE N²,3-dG adducts have not been reported.

The MS³ spectra of synthetic HHE-dG and HNE-dG acquired at the same CE are shown alongside the 4-hydroxy-2-alkenal-like DNA adducts of HHE and HNE in the prostate (Figure 12). The aglycones of HHE-dG and HNE-dG undergo HCD to produce Gua at m/z 152.0564 (C₅H₆N₅O⁺) and [Gua-NH₃]⁺ at m/z 135.0299 (C₅H₃N₄O⁺). Both aglycones also lose one and two molecules of H₂O but at a much higher relative abundance than the isomeric adducts detected in prostate DNA. The synthetic HHE-dG and HNE-dG aglycones undergo fragmentation to produce ions with the neutral loss of 46.0055 Da ([BH₂-CH₂O₂]⁺), detected at m/z 220.1189 and 262.1657, respectively, at very low abundance compared to the proposed HHE-dG and HNE-dG adducts in prostate DNA. In contrast, ions observed at m/z 230.1023 and 246.1336 in the MS³ spectra for synthetic HHE-dG and HNE-dG, respectively, were very low in the spectra of the prostate DNA adducts (Figure 12). The MS³ spectra of synthetic HNE-dG and [²H₁₁]-HNE-dG are shown in Figure S4. Several proposed pathways of HCD fragmentation of these adducts and the putative HNE-dG cyclic hemiacetals isolated from prostate DNA are reported in Figure S5.

8-HO-PdG and HNE-dG undergo ring-opening to form the aldehydes under alkaline pH and transform to stable alcohols following reduction with NaBH₄.^112,115 This chemistry has been employed to screen for these adducts in rodent and human tissues. High-resolution NMR spectroscopy also revealed 8-HO-PdG and diastereomeric HNE-dG adducts incorporated into 11- or 12-mer oligodeoxynucleotide duplex undergo ring-opening to their corresponding aldehydic forms under neutral pH conditions when matched with dC in the complementary DNA strand.^114,116 The ring-opened aldehydic forms of HNE-dG underwent further rearrangement to form their stereoisomeric cyclic hemiacetals, which became the predominant species present at equilibrium (Scheme 2).¹¹⁴ To our knowledge, the occurrence and chemical stability of the HNE-dG cyclic hemiacetals at the nucleoside level following enzymatic digestion of 4-hydroxy-2-enal-modified oligomers or polymeric DNA have not been reported. However, glutathione reacts with 4-HNE to form stable diastereomeric hemiacetal conjugates as primary products.¹¹⁷ We propose these 4-hydroxy-2-alkenal-like DNA adducts of HHE and HNE, and 4-hydroxy-2-heptenal (MH⁺ at m/z 396.1878), 4-hydroxy-2-octenal (MH⁺ at m/z 410.2034), and 4-hydroxy-2-decenal (MH⁺ at m/z 438.2347) in prostate DNA may be cyclic 1,N²-propano-dG adducts that underwent rearrangement to form cyclic hemiacetals.

We investigated the chemical stabilities of HHE- and HNE-1,N²-propano-dG adducts. dG was reacted with 4-HHE in 20 mM NaHCO₃ (pH 8.5) to form cyclic 1,N²-propano-dG-diastereomers.⁶⁷ After reaction for 72 h, the mixture was diluted 100-fold in water and analyzed by LC/MS employing the bulky aromatic analysis conditions (0.05% HCO₂H) with the Reprosil-Pur C18-AQ picochip column. The EICs of the adduct at the MS² and MS³ scan stages are shown in Figure 13. There was no evidence for the HHE-dG-like adducts detected in prostate DNA. However, four minor isomeric adducts appeared following reanalysis of the synthetic HHE-dG adduct 48 h after the diluted reaction mixture had been stored in the autosampler at 4 °C. We employed a shallower gradient to analyze synthetic HHE-dG adducts, the adducts in nuclease-digested 4-HHE-modified CT DNA, and freshly isolated pooled human prostate DNA. The cyclic HHE 1,N²-propano-dG diastereomeric adducts were at the limit of detection in the CT DNA modified with 4-HHE. However, four later eluting isomeric peaks were identified in CT DNA and the pooled human prostate DNA at the same retention time and displayed the same MS³ spectra (with characteristic fragment ion at m/z 220.1189) as the rearranged synthetic HHE-dG adducts (Figure 13).

Figure 13.

LC/MS analysis of isomeric 4-hydroxy-2-hexenal-like dG adducts (m/z 382.1734). The EIC of m/z 266.1248 at the MS² scan stage is the aglycone [M+H –116.0473]⁺ ([BH₂]⁺) of synthetic HHE -dG and the proposed rearranged isomers. The EIC at m/z 220.1119, is attributed to [BH₂-HCO₂H]⁺, and the EIC at m/z at 152.0567 is protonated guanine at the MS³ scan stage. (A) EICs of synthetic HHE-dG obtained by reaction of dG and 4-HHE in 20 mM NaHCO₃ buffer (pH 8.5) for 72 h at 37 °C. (B) Reanalysis of the reaction product after 48 h storage in the autosampler at 4 °C. (C) Reanalysis of the reaction product after 49 h storage in the autosampler using a longer gradient. (D) Analysis of 4-HHE adducts formed by reaction of CT DNA with 4-HHE in 100 mM NaHCO₃ buffer (pH 9.0) for 72 h at 37 °C. (E) Analysis of pooled human prostate DNA (25 μg of DNA digest on column). (A) and (B) were performed with a linear gradient reaching 95% CH₃CN in 15 min following online trapping, and (C, D, and E) employed a linear gradient reaching 95% CH₃CN in 34 min following online trapping, using the Reprosil-Pur C18-AQ picochip column for hydrophobic and bulky aromatic DNA adducts.

HNE-dG was not detected in human prostate DNA; however, a later eluting adduct was detected at t_R 26.1 min when employing the LPO analysis conditions (2 mM NH₄CH₃CO₂) with the Hydro RP phase column (Figure 9). We reanalyzed HNE-dG employing the bulky aromatic analysis conditions (0.05% HCO₂H) with the Reprosil-Pur C18-AQ picochip column (Figure 14). Synthetic HNE-dG added to the diluted reaction solution of 4-HHE-dG did not transform to the rearranged isomers as observed for HHE-dG after 48 h of storage in the autosampler. The synthetic HNE-dG diastereomers spiked post-digestion in prostate DNA digest elute earlier than the isomeric prostate HNE-dG-like adducts in the prostate. The [²H₁₁]-HNE-dG internal standard was spiked before enzymatic digestion of prostate DNA and did not rearrange to other isomeric forms (Figure 14). 4-HNE reacted with CT DNA to form the diastereomeric HNE 1,N²-propano dG adducts;⁵³ however, the 4-hydroxy-2-nonenal-like dG adducts were present in untreated CT DNA but did not noticeably increase immediately after treatment with 4-HNE (J. Guo and R. Turesky, unpublished observations). The rearrangement of HNE-dG adducts to the proposed cyclic hemiacetal forms (t_R 19.2 min) may take longer to occur than for HHE-dG (Fig. 13). We conclude that 1,N²-propano-dG adducts of several 4-hydroxy-2-alkenals appear to form in human DNA, which may have rearranged to cyclic hemiacetal adducts in vivo or possibly during storage or processing of DNA. Further studies on the chemistry and stability of these proposed 4-hydroxy-2-alkenal-dG adducts are required to elucidate the adducts' identities in the prostate genome.

Figure 14.

LC/MS analysis of synthetic HNE-dG (m/z 424.2191) and [²H₁₁]-HNE-dG (m/z 435.2881). (A) The EIC of m/z 308.1717 at the MS² scan stage is the aglycone [M+H –116.0473]⁺, attributed to the loss of dR from HNE-dG (dG-[²H₁₁]-HNE is not present in this sample). The EIC at m/z 262.1622, is attributed to [BH₂-H₂O-CO]⁺, and the EIC at m/z at 152.0567 is protonated guanine at the MS³ scan stage. The EICs of HNE-dG were acquired following storage in the autosampler for 48 h at 4 °C with the 100-fold diluted reaction of HHE-dG (described in Fig. 13). (B) Analysis of pooled human prostate DNA (25 μg of DNA digest on column). [¹¹H₂]-HNE-dG was spiked before and HNE-dG was added post-nuclease digestion at 1.3 adducts per 10⁷ nts and 3.0 adducts per 10⁷ nts, respectively. The presumed rearranged hemiacetals of HNE-dG elute at t_R 19.2 min. Chromatography was performed as described in Materials and Methods except a linear gradient reaching 95% CH₃CN in 34 min after the online trapping, using the Reprosil-Pur C18-AQ picochip column for hydrophobic and bulky aromatic DNA adducts.

Volcano and PLS-DA Plots and VIP Scores of the DNA Adductomics Data in the Peripheral Zone of PC patients.

The volcano plot and the log2-fold-change of putative DNA adducts (for low/high Gleason score patients) versus the −1 log(p-values) are shown in Figure 15A. The volcano plot displays p values versus log 2-fold changes. P values are plotted as −1*ln (p-value) so that the level of significance is plotted higher on the y axis. Each dot represents an individual adduct. There were 49 putative DNA adducts in the LPO fraction with p < 0.10 and log2-fold change> 1.0 (25 adducts were higher in the high Gleason (H) group, and 24 presumed adducts were increased in the low Gleason (L) group (plotted as negative log2 FC). There were 36 significantly altered DNA adducts in the bulky aromatic group with p < 0.10 and log2 FC > 1.0 (27 adducts were increased in the high Gleason (H) group, and 9 adducts were higher in the low Gleason (L) group. PLS-DA was performed to determine which putative DNA adducts may explain differences between Gleason H and L score groups and the quality control (QC) group. For the PLS-DA plots, the qualitative metrics for the model were determined.^79,80 The response variances (R²X, R²Y) and predictive accuracy (Q²Y) were 0.55, 0.999, and 0.879 for the LPO fraction, and 0.56, 0.999, and 0.813 for the bulky fraction. The RMSEE between the actual and predictive models were 0.022 (LPO) and 0.019 (bulky). Given the small subject number per group (n = 6), the clustering results are satisfactory. The H and L Gleason score groups and QC groups were well-separated clusters (Figure 15B). The high VIP scores describe DNA adducts that distinguish the high and low Gleason score groups and agree with their significant p-values (Figure 15C). The N²-HOCH₂-dG adduct (shown as a green circle) is detected in the LPO DNA adduct fraction (p =0.05 and VIP = 1.8). Overall, informatics analysis of the wide-SIM/MS² data by wSIM-City shows the technology’s potential to discover novel DNA adducts.

Figure 15.

DNA adductomics of putative polar LPO and hydrophobic bulky aromatic DNA adduct fractions present in the PZ of PC patients with low (L) and high (H) Gleason scores. (A) Volcano plot of log₂ (fold change H/L, Gleason score), vertical blue dashed lines show the cutoff of log2 FC = 1, the horizontal dashed line shows the significance cutoff of p = 0.10 as −1*ln(0.1). The red dots are significantly altered DNA adducts (p < 0.10) with fold-change (FC) > 2. (B) PLS-DA showing the separation of Gleason H and L score groups and the pooled QC samples. Red, blue, and green circles show the 95% confidence ellipsoids for the H, L, and QC groups, respectively. The black circle is the 95% confidence ellipsoid of all samples. T1 and T2 are the first two principal components plotted for the PLS-DA analysis. (C) The comparison of significance testing DNA adduct p values as −1*ln(p value) versus variables of importance in projection (VIP) scores: blue dashed lines show the cutoff of p = 0.10 as −1*ln(0.1) and VIP = 1.5. The red line is a linear fit between the −1*ln(p) values and the VIP scores, was positively correlated, and the Pearson correlation coefficient was r = 0.68 for both HAA and LPO fractions. The N²-HOCH₂-dG adduct derived from formaldehyde in the LPO DNA adduct fraction is circled in green and enlarged for visualization.

Discussion

PhIP is the only cooked meat mutagen known to induce PC in rodent models.^26,28,29 However, the doses of PhIP employed were more than a million-fold higher than the daily human intake ofPhIP, and the potential deleterious biological effects of PhIP in the rodent or human prostate at chronic dietary exposure levels are unknown. The human prostate cell line (LNCaP) and normal epithelial transition cells of BPH patients efficiently bioactivate HONH-PhIP to form DNA adducts.^38,39 Our data also show that mutation-inducing PhIP DNA adducts are present in the prostate genome of some PC patients.^38,52,90 In addition to its genotoxic effects, PhIP binds and activates the androgen receptor (AR) in LNCaP, a human prostate AR-positive cell line.¹¹⁸ In PC-3, an AR negative human prostate cell line, PhIP induces cell proliferation and migration by activating the mitogen-activated protein kinase extracellular signal-related kinase pathway.¹¹⁹ Moreover, PhIP-induced prostate carcinogenesis is associated with oxidative stress in rodents.^27,29,30 Thus, PhIP may act as a PC initiator and a tumor promoter. Collectively, these observations support PhIP’s potential role in human PC etiology.

Epidemiological studies have reported inconsistent associations between cooked red meat consumption, HAA exposure, and PC risk.^10-13,24 A significant limitation in these studies is their reliance on an FFQ to estimate HAA exposure accurately. PhIP is the most abundant carcinogenic HAA formed in well-done cooked meat and can occur over a 100-fold concentration range in cooked beef, poultry, and fish, depending on the temperature and cooking duration.^18,20-23,120 Estimating the HAA content in cooked meat at the ppb level through questionnaires is difficult. Previous studies estimated HAA intake from 23 meat/poultry and fish items employing multiple food diaries with photographs of meat dishes cooked to varying degrees of well-doneness and found that the PhIP intake was severely underestimated employing the FFQ module.^121,122 Weak correlations were also observed between FFQ meat modules and short-term HAA urinary biomarkers.^123,124 Thus, using an FFQ module in the absence of quantitative measurements of PhIP in the cohort’s diet can misclassify subjects' exposures. There is a critical need to establish long-lived biomarkers of PhIP and other HAAs to assess these dietary carcinogens' exposure and health risk.

The PhIP level in hair is the sole long-term biomarker available to assess exposure to this cooked meat carcinogen and may serve as a surrogate for exposure to other HAAs.^46,47,49,50 One study from Japan with detailed FFQ data on caloric intake, with known portion sizes of meat/poultry and fish and measured HAA levels, reported a positive correlation between meat intake and PhIP scalp hair levels of healthy volunteers.^50,125 However, the associations between the meat intake and PhIP hair levels above the LOQ value were only significant when the diet was adjusted for total energy intake.⁵⁰ The patients in our prostate cohort completed a questionnaire on their weekly intake of red and processed meats/poultry and fish, but the portion sizes and total caloric intake were not included in the survey, nor were PhIP measurements conducted on meat/poultry and fish dishes commonly eaten by the cohort. These variables likely explain the weak correlation between the meat diet and PhIP hair levels (Figure 4). The hair dosimeter can be a superior estimate of PhIP intake than an FFQ, particularly when the quantity of meat and cooking methods are in doubt and HAA measurements in the cohort’s diet are lacking.

PhIP levels in hair are influenced by pigmentation and may be impacted by metabolism.^46,47 Hepatic P4501A2 is the major enzyme involved in PhIP metabolism in humans, accounting for more than 70% of the PhIP biotransformation.¹²⁶ We hypothesized that subjects with rapid P4501A2 phenotype would have lower levels of PhIP accrued in hair than slow P4501A2 phenotype individuals because there would be less unmetabolized PhIP in the bloodstream after the first-pass metabolism for binding to hair. However, the P4501A2 activity did not impact PhIP hair levels in the University of Hawaii cohort, based on the P4501A2 phenotyping using the urinary caffeine metabolic ratio.^46,47 The intraindividual metabolic phenotype varied widely over the time-course of the feeding study and precluded stratification of subjects' P4501A2 phenotype activity.^46,47 We did not phenotype individuals for P4501A2 activity in the prostate cohort.

Newly grown scalp hair from the nape of the neck captures the PhIP accrued during the previous ~10 weeks of a meat diet.⁴⁶ Based on the PhIP hair dosimeter, the daily consumption of PhIP by our prostate cohort at the time of surgery was below the daily intake of PhIP for healthy volunteers during the free-choice diet phase of the semicontrolled meat feeding study conducted at the University of Hawaii (Figures 1 and 2) and below the average daily intake of PhIP reported for younger adults in the United States, estimated between 0.49 – 0.98 μg per day.^34,35 Gray hair with a low melanin content binds PhIP less efficiently than dark pigmented hair, and hair pigmentation may confound the estimated PhIP intake in our aged cohort even though the PhIP hair levels were adjusted for melanin content.^47,82,83 Dietary patterns also can change with increasing age, with less consumption of energy-dense foods, including red meat.¹²⁷ Therefore, the dietary intake of PhIP and other meat-related genotoxicants may have been higher earlier in life. Longitudinal studies measuring PhIP hair levels to assess changes in an individual’s consumption patterns of well-done cooked meats at different stages of life may improve our understanding of the potential role of a well-done cooked meat diet on the development of PC that occurs later in life.

Nevertheless, elevated PhIP hair levels are positively associated with intermediate and high-risk PSA values and Gleason pathology scores. The PhIP hair levels are also higher in patients undergoing prostatectomy surgery for PC than bladder cancer patients undergoing cystoprostatectomy or BPH patients undergoing TURP or HoLEP treatment. The associations between PhIP hair levels adjusted for melanin content with pathology biomarkers were less robust than PhIP pg/g hair reported. However, there was a significant association between elevated PhIP hair levels adjusted for melanin and patients with intermediate and high-risk Gleason scores, and patients undergoing surgery for PC. Excluding the six patients with the lowest melanin content (≤ 3.5 mg/g hair), a value for blond/gray-haired pigmentation,⁷⁴ marginally changed the associations between PhIP hair levels (PhIP ng/g melanin) and PC pathology scores. The PhIP hair biomarker is a well-established validated biomarker in healthy young and middle-aged adults with brown and black pigmented scalp hair, and PhIP hair levels are relatively constant over a 6 month time interval.^46-48,50 However, semicontrolled feeding studies with known PhIP intake in a cohort with a wide array of hair colors, including individuals with gray hair, are required to determine the precise relationship between PhIP accrual in light and dark pigmented hair and the most accurate method for adjusting chemical levels for variable hair pigmentation

AA men are at a two-time higher risk for PC than Caucasians.² PhIP intake by AA men is reportedly several-fold greater than for Caucasian men because of a preference for eating well-done cooked meats and poultry.^34,35 The data on HAA biomarkers in different racial groups are limited. The geometric mean urinary level of PhIP was 2.8-fold higher in AA than in Caucasians in one study.¹²³ In another study, well-done meat intake was associated with elevated PSA levels in a prospective clinic-based study among AA men, providing support for a role of a well-done cooked meat diet in aggressive PC risk.¹²⁸ Our PhIP hair biomarker data show AA men have higher PhIP hair levels than Caucasians, suggesting higher PhIP intake. PhIP hair levels remained higher in AA men when adjusted for melanin content, but the difference was not statistically significant (Table 2 and Figure 5). A larger cohort is required to determine the extent of PhIP exposure and if increased PhIP hair levels occur in AA men with aggressive PC than Caucasians.

The dG-C8-PhIP adduct was detected in 9.4% of the prostate cohort assayed using a highly specific HRAMS method. The low detection frequency may be attributed to the low PhIP intake, where the PhIP DNA adduct levels were below the LOQ; alternatively, DNA adducts may have been repaired by NER.^91,129 In contrast, studies employing immunohistochemistry (IHC) with a polyclonal antibody raised against dG-C8-PhIP-modified calf thymus DNA frequently detected elevated levels of putative PhIP DNA adducts in the prostate of PC patients.^130-133 The mean intensity staining for the putative dG-C8-PhIP adduct was higher in nontumor specimens of AA with high-risk Gleason scores than lower-risk Gleason scores. This association was not seen for Caucasians, and PhIP-DNA adduct levels in prostate cells did not vary significantly by race. ¹³³ Putative PAH-DNA adducts were also detected at high frequency in the prostate of PC patients by IHC methods using different antibodies raised against dG-N²-B[a]P from two different laboratories.^134,135 We did not detect the dG-N²-B[a]P adduct in our cohort. The results of our study are fundamentally at odds with published results obtained by IHC screening, even though the LOQ values for HRAMS are 10-fold or lower than the LOQ of the IHC methods.^52,135,136 The levels and frequency of detection of cooked meat carcinogen DNA adducts may differ between cohorts and depend on the cohort's dietary preferences for consuming well-done meats and the analytical methods employed to measure DNA adducts. IHC is less specific than HRAMS and may be detecting various lesions in addition to dG-C8-PhIP and dG-N²-B[a]P. There is a critical need to cross-validate immunodetection and HRAMS methods for confirming DNA adduct identities in the same tissue samples. We expect such direct comparisons to support further the need to employ specific MS methods to measure DNA adducts in PC cohorts. HRAMS DNA adduct biomarker data can then be examined for associations with cancer pathology biomarkers and correlations to genetic polymorphisms in genes encoding enzymes of carcinogen metabolism and DNA repair that impact PC risk.

dG-C8-PhIP, the principal adduct formed with PhIP,^55,137,138 induces C:G>A:T and C:G>T:A transversions, and the C:G>G:C transition in rodents, transgenic target genes, mouse embryo fibroblasts, human-induced pluripotent stem cells, and site-specific mutagenesis studies.^42,139-143 The PhIP mutational signature shows similarity to COSMIC SBS4, SBS8, SBS18, and SBS29 mutational signatures found in several types of human tumors, including the prostate.¹⁴⁴ However, these signatures are not unique to PhIP, and DNA adducts of other environmental, dietary, and endogenous genotoxicants are likely to contribute to these mutations.^142,143 Other components in red meat, such as heme iron, can damage organs through free radical formation, and LPO products formed in cooked meat also may form DNA adducts or serve as tumor promoters.¹⁴⁵

Inflammation and oxidative stress are features in PC development.^32,33 Our and other laboratories have shown that antioxidant scavengers, such as βME or glutathione, are required to block the artifactual formation of LPO adducts during the isolation and enzymatic digestion of DNA.^53,146,147 We screened for 10 prototypical LPO DNA adducts in prostate specimens.^44,45,51,148 DNA adducts of acrolein (HO-PdG) occurred at a relatively high frequency in the prostate; however, DNA adducts of other known LPO products were detected sporadically and often below the LOQ (Scheme 1). 6-HO-PdG and 8-HO-PdG were detected in over 50% of the patients assayed. The combined mean and SD were 12.9 ± 40.1 HO-PdG adducts per 10⁸ nts in PC and cystoprostatectomy patients. However, the PZ specimens selected for analyses were largely tumor-free with no or mild inflammation, which may explain the absence of some LPO adducts. Thus, the frequent occurrence of HO-PdG adducts was unforeseen. We did not observe a statistically significant difference (p = 0.65) between the total HO-PdG adduct levels in the bladder cancer patients (n = 36) undergoing cystoprostatectomy and PC patients (n = 52) undergoing prostatectomy (Table 4). Nor were HO-PdG adduct levels associated with prostate surgical treatment, high-risk PSA, Gleason scores, or tumor stage. These data should be interpreted with caution due to the small sample size and wide range of HO-PdG adduct levels, particularly for cystoprostatectomy patients, where cisplatin-based chemotherapy-induced reactive oxygen species may have impacted DNA adduct levels.

The Hecht laboratory recently measured 6-HO-PdG and 8-HO-PdG by HRAMS in lung DNA of current smokers and reported at 8.6 ± 2.7 and 19.9 ± 14.3 adducts per 10⁹ nts, respectively.¹⁴⁷ The combined HO-PdG values were 28.5 ± 14.9 adducts per 10⁹ nts. Similar adduct levels were present in nonsmokers, signifying that these basal DNA adduct levels occur from endogenous sources of acrolein and not cigarette smoke. The Tang laboratory reported putative HO-PdG adducts by ³²P-postlabeling in human bladder epithelium at 63 ± 25 adducts per 10⁷ nts or ~10 to 50-fold higher than the levels we detected in the prostate genome.¹⁴⁹ Human exposure to acrolein occurs through cigarette smoke (18–98 μg/cigarette), the environment, and also occurs endogenously from oxidative stress.^150-152 IARC classified acrolein as probably carcinogenic to humans (Group 2A) based on "sufficient" evidence of carcinogenicity in experimental animals and "strong" mechanistic evidence.¹⁵⁰ Acrolein forms DNA and protein adducts, and at elevated concentrations, induces oxidative stress, mitochondrial disruption, membrane damage, endoplasmic reticulum stress, immune dysfunction,^152-154 and inhibits NER and base excision repair in human urothelial cells.¹⁵⁵ Acrolein-treated pSP189 plasmids containing HO-PdG adducts replicated in human normal lung fibroblasts induced C:G>A:T transversion mutations followed by C:G>T:A transitions in the supF gene.¹⁵⁶ Further investigations on the biological effects of acrolein and its possible role in PC are warranted.

Our untargeted DNA adductomics data reveals that the prostate genome contains a complex mixture of DNA adducts derived from many xenobiotics and endogenous electrophiles. The identities of most putative DNA adducts are unknown. The mass spectral data show that formaldehyde and multiple 4-hydroxy-2-alkenals form adducts with dG. Wide-SIM/MS² detected N²-HOCH₂-dG and five 4-hydroxy-2-alkenal-like dG adducts in the subset of low and high Gleason score patients assayed. The N²-HOCH₂-dG was elevated in the high Gleason score group (p = 0.05). Several of the 4-hydroxy-2-alkenal-like dG adducts were also elevated in the high Gleason score patients based on the manually integrated EIC signals of the aglycones’ m/z ions acquired by wide-SIM/MS² (p < 0.05, Welch’s two-tailed t-test, Figure S6). Quantitative adduct measurements with internal standards in a larger number of patients are required to confirm this association. Further studies, including chemical syntheses and mechanisms of 4-hydroxy-2-alkenal-like dG adduct rearrangement to the presumed cyclic hemiacetals s, are necessary for unambiguous identification of these lesions and their possible role in PC etiology. Advances in structural elucidation of the many other DNA adducts will require the application of different computational approaches, such as spectral similarity searching of MSⁿ spectra, which is a promising approach to annotate unknown metabolic features in MS-based untargeted metabolomics.^157,158 The complex landscape of the DNA damage in the prostate genome is consistent with previous adductomics studies reported for the human lung and esophagus, where hundreds of putative DNA adducts of unknown structures were discovered.^101,159,160 The samples' heterogeneity and variability in DNA adduct detection necessitate larger subject numbers for robust associations of specific DNA adducts with disease risk. Nevertheless, some putative DNA adduct biomarkers discovered by wSIM-city in this study are associated with elevated Gleason scores. Elucidating the identities of some of these previously unreported DNA adducts can advance the discovery of potential cancer-causing agents and their mechanisms of DNA damage in the genome.¹⁶¹

Biomonitoring of DNA adducts is most relevant at the time when tumor formation commences rather than many years later when cancer has been diagnosed. However, environmental pollution, tobacco smoking, and diet often represent long-term exposures. Current levels of DNA adducts from some of these exposures are likely to correlate with adduct levels that existed during the time of tumor initiation. At present, aflatoxin B₁ and aristolochic acid are among the few environmental carcinogens where their DNA adducts are firmly linked to well-characterized mutational signatures in highly-exposed human cohorts in Asia.^143,162-166 However, cohorts in the United States and other economically developed countries are exposed to lower levels of an array of potentially hazardous chemicals. The role of many of these chemicals in cancer etiology remains uncertain.¹⁶⁷ Characterizing DNA adduct signatures through DNA adductomics is a first step associating specific chemical exposures to cancer risk. DNA adductomics is a developing technology.^{54,105,160,161,168,169} Once it is a fully mature technique, some adducts identified in the DNA adductome could be linked to mutational signatures in cancer-driver genes and strengthen associations between chemical exposures and cancer etiology.^{143,144,170-172}

Conclusion

PC is a multifactorial disease; genetics and multiple environmental and dietary risk factors, and oxidative stress contribute to this cancer. Some epidemiological studies have linked frequent consumption of well-done cooked red meat containing HAAs with aggressive PC. Our study is the first to demonstrate associations between elevated PhIP hair levels, a proxy for well-done meat consumption, with intermediate and high-risk PSA values, and Gleason pathology scores. These findings support the paradigm for well-done cooked red meat consumption in aggressive PC risk. A larger-sized cohort with a documented history of well-done cooked meat consumption and a higher proportion of AA and other minority groups is required to investigate the impact of PhIP and other cooked meat-related genotoxicants in PC risk and racial disparities of this disease.

Several oxidative stress-associated LPO DNA adducts are elevated in a subset of PC patients with intermediate and high Gleason scores. Are these increases in oxidative DNA adduct levels attributed to endogenous ROS processes of the disease, or do lifestyle factors, such as the meat diet, contribute to ROS and the evolution of PC? Notably, PhIP-induced prostate carcinogenesis in rodents is associated with inflammation and oxidative stress.

The biological effects of HAAs, including PhIP and other chemicals in cooked meat, can be modulated by polymorphic phase I and II xenobiotic metabolism enzymes. The DNA adducts formed from metabolically reactive intermediates are substrates for DNA repair enzymes. Mechanistic studies are warranted to assess the impact of genetic polymorphisms on enzymes involved in DNA adduct formation and DNA repair on PC risk. Such data are critical to advancing our knowledgehow chemicals in cooked meat may contribute to prostate mutagenesis and cancer risk.

Biographies

Jingshu Guo obtained a Ph.D. in Chemistry from the University of Toledo. She was a Research Associate and Research Assistant Professor in the Turesky laboratory, Department of Medicinal Chemistry, University of Minnesota. Her research interests focus on biomonitoring human carcinogens by liquid chromatography-mass spectrometry (LC-MS) approaches. She has developed highly sensitive and selective LC-MS-based methods to detect carcinogen biomarkers in human biospecimens as molecular evidence to understand cancer etiology. She is currently a Product Applications Specialist, Clinical Research and Toxicology, Chromatography and Mass Spectrometry Division, Thermo Fisher Scientific.

Joe Koopmeiners received a Ph.D. in Biostatistics from the University of Washington. He is currently a Mayo Professor of Public Health and Head of the Division of Biostatistics in the University of Minnesota School of Public Health. His research interests include Bayesian adaptive methods for clinical trials, causal inference, and statistical methods for biomedical imaging. He has extensive experience as a collaborative biostatistician, having been a Faculty Statistician in the Biostatistics Shared Resource of the University of Minnesota Masonic Cancer Center from 2009 – 2018. He collaborates with basic, population, and clinical scientists in the Masonic Cancer Center.

Scott Walmsley obtained his Ph.D. in Cell and Molecular Biology from Colorado State University. He is a Computational Scientist within the Institute of Health Informatics and the Masonic Cancer Center at the University of Minnesota, aiding research efforts toward DNA adduct discovery. He developed the wide selected ion monitoring/tandem mass spectrometry data software ‘wSIM-City’ for DNA adductomics used in this study. His research interests are geared towards developing informatics software solutions for proteomics, metabolomics, and DNA adductomics to analyze liquid chromatography – high-resolution mass spectrometry data.

Peter Villalta received a Ph.D. in Chemistry from the University of Minnesota. He is currently the Director of the Masonic Cancer Center’s Analytical Biochemistry Shared Resource at the University of Minnesota and a Research Professor in the Department of Medicinal Chemistry. He is supported by an R50 NCI Research Specialist award to provide Masonic Cancer Center researchers with expertise in the use of advanced mass spectrometric tools to advance their research, including the study of how cancer is initiated and the development of strategies for cancer prevention and/or treatment of cancer.

Lihua Yao obtained an M.D. from the Peking University Health Science Center in Beijing, China. Her research in China broadly encompassed immunotherapy for breast cancer treatment and prevention. She is currently a research scientist in the Turesky laboratory at the Department of Medicinal Chemistry, University of Minnesota. Her work primarily involves conducting sample preparation of biospecimens and bioanalytical support for projects biomonitoring carcinogens in humans. She also provides administrative support and supervises animal toxicology studies conducted in the Turesky laboratory.

Paari Murugan is an M.D. and an Associate Professor at the University of Minnesota Medical Center. He is an American Board of Pathology certified anatomic pathologist whose expertise is in the surgical pathology of genitourinary tumors. He has contributed over 70 peer-reviewed publications, numerous scientific meeting presentations, and book chapters. He has lectured nationally and internationally. He is involved in multiple NIH research grants and is associate editor of Diagnostic Pathology. He also serves as an elected member of the College of American Pathologists' Cancer Committee and as Medical Director of the University of Minnesota's Biorepository and Laboratory Services.

Resha Tejpaul is a research professional at the University of Minnesota, Minneapolis. She has a Bachelor of Arts in Physiology and Biochemistry, and her research interests are in digital medicine and urologic oncology. Her role in this study was clinical research coordinator. She screened and consented patients, conducted biospecimen procurement logistics, and entered the data collected into the database.

Christopher Weight is the Center Director for Urologic Oncology and directs the SUO Urologic Oncology Fellowship at the Cleveland Clinic. He received an M.D. from the University of Utah School of Medicine and completed his residency at Cleveland Clinic and fellowship at the Mayo Clinic. He has published over 125 articles and is an NIH-funded researcher on four R01 grants. His research areas have focused on environmental exposures that may lead to genitourinary cancers and utilizing artificial intelligence to help personalize the treatment of patients. His role in this research was clinical urologists and a co-I for this project.

Rob Turesky received a Ph.D. in Nutrition and Food Science from M.I.T. He is a Professor in the Department of Medicinal Chemistry and holds the inaugural Masonic Chair in Cancer Causation at the University of Minnesota. His research seeks to understand the role of environmental and dietary genotoxicants in human cancer through mechanistic studies and human biomonitoring. He has developed specific mass spectrometry-based biomarkers of many hazardous chemicals for implementation in epidemiology studies that seek to understand the role of chemical exposures in cancer etiology.