Biomarkers of Furan Exposure by Metabolic Profiling of Rat Urine with

Feb 13, 2008 - Analysis of products of in vitro incubations of the reactive furan metabolite cis-2-butene-1,4-dial with the respective amino acid deri...
0 downloads 12 Views 307KB Size
Chem. Res. Toxicol. 2008, 21, 761–768

761

Biomarkers of Furan Exposure by Metabolic Profiling of Rat Urine with Liquid Chromatography-Tandem Mass Spectrometry and Principal Component Analysis Marco Kellert, Silvia Wagner, Ursula Lutz, and Werner K. Lutz* Department of Toxicology, UniVersity of Würzburg, Versbacher Strasse 9, 97078 Würzburg, Germany ReceiVed NoVember 30, 2007

Furan has been found in a number of heated food items and is carcinogenic in the liver of rats and mice. Estimates of human exposure on the basis of concentrations measured in food are not reliable because of the volatility of furan. A biomarker approach is therefore indicated. We searched for metabolites excreted in the urine of male Fischer 344 rats treated by oral gavage with 40 mg of furan per kg of body weight. A control group received the vehicle oil only. Urine collected over two 24-h periods both before and after treatment was analyzed by a column-switching LC-MS/MS method. Data were acquired by a full scan survey scan in combination with information dependent acquisition of fragmentation spectra by the use of a linear ion trap. Areas of 449 peaks were extracted from the chromatograms and used for principal component analysis (PCA). The first principal component fully separated the samples of treated rats from the controls in the first post-treatment sampling period. Thirteen potential biomarkers selected from the corresponding loadings plot were reanalyzed using specific transitions in the MRM mode. Seven peaks that increased significantly upon treatment were further investigated as biomarkers of exposure. MS/MS information indicated conjugation with glutathione on the basis of the characteristic neutral loss of 129 for mercapturates. Adducts with the side chain amino group of lysine were characterized by a neutral loss of 171 for N-acetyl-L-lysine. Analysis of products of in Vitro incubations of the reactive furan metabolite cis-2-butene-1,4-dial with the respective amino acid derivatives supported five structures, including a new 3-methylthio-pyrrole metabolite probably formed by β-lyase reaction on a glutathione conjugate, followed by methylation of the thiol group. Our results demonstrate the potential of comprehensive mass spectrometric analysis of urine combined with multivariate analyses for metabolic profiling in search of biomarkers of exposure. Introduction Furan is present in a number of food items, primarily as a product of heat treatment (1–3). Concentrations can come up to 100 µg/kg (4–6). High levels have often been found in baby food from small glass jars containing cooked vegetables or meat. Furan was carcinogenic in F344 rats and B6C3F1 mice (7). In rats, NTP1 reported a more than 80% incidence of cholangiocarcinoma even at the lowest tested dose of 2 mg/kg body wt per day (5 days per week for 104 weeks) as well as a doserelated induction of hepatocellular adenoma and carcinoma (8). In mice, furan induced only hepatocellular tumors. Considering the carcinogenic potency and assuming a daily intake of 100 g of a furan-containing food item by a baby of 10 kg body weight, the calculated dose of 1 µg/kg body wt per day is only by a factor of about 400 below the estimated 50% tumorigenic dose * To whom correspondence should be addressed. Professor Dr. W. K. Lutz, Department of Toxicology, University of Würzburg, Versbacher Strasse 9, 97078 Würzburg, Germany. Tel: +49(0)931/201-48402. Fax: +49(0)931/201-48446. E-mail: [email protected]. 1 TD50, tumorigenic dose 50%; NTP, National Toxicology Program; BDA, cis-2-butene-1,4-dial; LC-MS/MS, liquid chromatography tandem mass spectrometry; EMS, enhanced mass spectrometry, i.e., full scan enhanced by the use of the linear ion trap; IDA, information dependent acquisition; NL, neutral loss; EPI, enhanced product ion scan, i.e., enhanced scan by the use of the linear ion trap; MRM, multiple reaction monitoring; TIC, total ion count; RSD, relative standard deviation; PCA, principal component analysis; PC, principal component; R2X, fraction of the Sum of Squares of all the X variables explained by the current component; Q2, fraction of the total variation of the X variables predicted by a component.

(TD50) in the rat (9). This would imply an unacceptably high cancer risk if humans are as susceptible as rats and if the risk is proportional to the dose. Linear extrapolation of cancer risk to low dose is usually considered for carcinogens that can form DNA adducts per se or after metabolic activation (10). Furan is rapidly absorbed from the gastro-intestinal tract and metabolized to the reactive multifunctional intermediate cis-2butene-1,4-dial (BDA, see Scheme 1) (11, 12). All four carbon atoms of BDA exhibit electrophilic reactivity, with varying preferences in the reaction with nucleophiles. The cis-configuration of BDA is stabilized by a cyclic hydrate (12, 13). Numerous reaction products of in Vitro incubations of BDA with N-acetyl-L-lysine, N-acetyl-L-cysteine, and glutathione have been described by Peterson and co-workers and indicate complex, multistep reactions, including the formation of sulfur-substituted pyrrole and pyrrolin-2-one structures (see I, II, and III in center part of Scheme 1) (13–15). Reaction with amino acid side chains of protein could explain the cytotoxicity of furan. Liver and hepatocytes appear to be particularly susceptible, which is not surprising in view of the requirement of metabolic activation of furan to reactive intermediates (16, 17). Reaction of BDA with DNA is indicated by the in Vitro formation of nucleoside adducts at nitrogen atoms (18–22). A genotoxic contribution to the carcinogenic mechanism of action cannot therefore be excluded. In animals, metabolism has only been investigated in general terms. After oral gavage of rats with [2,5-14C]furan, one-fourth of the administered radioactivity was exhaled in the form of

10.1021/tx7004212 CCC: $40.75  2008 American Chemical Society Published on Web 02/13/2008

762

Chem. Res. Toxicol., Vol. 21, No. 3, 2008

Scheme 1. (Top) Enzymatic Activation of Furan by Cytochrome P450 Enzymes to the Electrophilic Intermediate cis-2-Butene-1,4-dial (BDA)a

Kellert et al.

structural information and characteristics concerning the time course of excretion.

Experimental Procedures

a

The the cis-configuration is stabilized by a cyclic hydrate. (Center and bottom) Metabolites from urine of furan-treated rats. The partial structure of the four carbon atoms of BDA is highlighted in bold. The bicyclic adduct I resulted from the reaction of BDA with both the thiol and the N-terminal group of glutathione (GSH). Degradation of adduct I followed by methylation of the thiol group leads to the 3-methylthio-pyrrole derivative V. Adduct II was formed by N-acetylation of the reaction product of BDA with the side chain amino group of L-lysine. Compound III resulted from a cross-linking reaction of BDA with lysine and cysteine followed by N-acetylation. Compound IV is the sulfoxide of compound III.

[14C]CO2. Fifteen percent of the radioactivity was fixed in tissues and was found in the liver as protein adducts. Twenty percent of the radioactivity was excreted in urine as the sum of more than 10 different compounds, but structural identification was initially lacking (23). Peterson and co-workers identified one BDA-glutathione conjugate in the urine of furan-treated rats (compound I in Scheme 1) and reported on the presence of another 18 furan-derived metabolites (15). These data give hope that multivariate analysis of comprehensive analytical data could lead us to urinary excretion products and help elucidate the complex pathways of furan metabolism. This approach of metabolic profiling had been used before to detect unknown metabolites in the urine of animals treated with lipid-lowering drugs (24), citalopram (25), and the heterocyclic aromatic amine PhiP (26). Furan has a high volatility with a boiling point of 30 °C. Concentrations measured in food are therefore unlikely to provide accurate information on the ingested dose. A biomarker of exposure is needed. In view of the documented excretion of the glutathione-derived conjugate I in urine (15), the analysis of such conjugates represents a possibility for exposure biomonitoring. Here, we describe the first step toward (i) improving insight into the complex furan metabolism in ViVo and (ii) searching for specific furan-derived biomarkers of exposure. Our approach consisted of metabolic profiling of products that had been excreted in the urine of rats two 24-h periods after gavage of furan. Analysis by LC-MS/MS was combined with multivariate data analysis. Differences in chromatograms from treated and untreated rats were identified, and peaks that contributed most strongly to group separation were further analyzed for

Chemicals and Reagents. Water and acetonitrile in HPLC grade were purchased from Roth (Karlsruhe, Germany). All other chemicals were from Sigma/Fluka (Taufkirchen, Germany). Corn oil was purchased from Roth. Animals. All procedures had been approved by the animal experimentation board of the Regierung von Unterfranken. Male F344 rats of 8 weeks of age were purchased from Harlan Winkelmann (Borchen, Germany). The weight range of the animals was from 200 to 235 g. All animals were maintained in a pathogen free animal facility under a standard 12 h light/12 h dark cycle at 22 ( 2 °C with food and water ad libitum. A group of five rats was administered 40 mg/kg body wt furan in corn oil by oral gavage. A control group of five rats received 2.5 mL/kg body wt corn oil. Urine was collected for two 24-h periods, each before and after treatment by placing the rats in individual metabolism cages with food and water ad libitum. Urine was collected on an ice/water bath. The volume was measured, and 500 µL aliquots were stored at -20 °C until analysis. The specific gravity was determined by a hand-held DRC-200 refractometer (Fuzhou Link Optical Instrument Co. Ltd., Fujian, China). Creatinine, urea, sodium, potassium, and osmolarity were analyzed by the Central Laboratory of the University Hospital of Würzburg. Additional control urine was pooled from all rats before the start of the experiments for use as quality control samples. Sample Preparation. Aliquots of the 40 urine samples were thawed, vortex-mixed, and centrifuged at 4 °C and 14,000g for 20 min to remove insoluble parts prior to analysis. To equalize the concentration, the samples were diluted to a specific gravity of 1.010 g/mL with distilled water. Then to a volume of 450 µL diluted urine, 5 µL of 4 M HCl was added to achieve pH 3. Liquid Chromatography with Column Switching Unit. A detailed description of our column switching unit has previously been published (27, 28). It consisted of an autosampler with a 900 µL sample loop, two HPLC pumps (Agilent Series 1100, Waldbronn, Germany), and an electric valve. To load the sample into the system, 350 µL of the urine was transferred by the autosampler and the first HPLC pump (0.1% formic acid, 1.0 mL/min) to a trap column (ReproSil-Pur C18-AQ, 5 µm, 33 × 3 mm; Dr. Maisch, Ammerbuch, Germany). After 1.1 min, the valve was switched to backflush the analytes from the trap column onto the analytical column (ReproSil-Pur C18-AQ, 3 µm, 150 × 2 mm; Dr. Maisch). The second HPLC pump supplied the initial solvent composition of 95% formic acid (0.1%) and 5% acetonitrile at a flow rate of 0.200 mL/min. After 0.9 min, the elution was started, consisting of a linear gradient to 50% acetonitrile in 23 min, a second linear gradient to 90% acetonitrile in 2 min, followed by 2 min at 90% acetonitrile. The initial solvent composition was reached with a 2 min gradient that was followed by equilibration for 9 min. The analytes were completely eluted from the trap column to the analytical column after 10 min of total run time, and the valve was switched to disconnect the trap column. A gradient up to 90% acetonitrile was used for the regeneration of the trap column. All samples were randomized and measured in one batch. Mass Spectrometry for Metabolite Profiling. The mass spectrometer (QTRAP 2000, Applied Biosystems/MDS Sciex, Concord, Canada) was operated in the negative ion mode with an ion spray voltage of -4200 V and a source temperature of 400 °C. The declustering potential was set to -30 V, and the entrance potential to -10 V. Nitrogen was used as ion spray, drying gas, and curtain gas at a pressure of 45, 50, and 30 psi, respectively. Nitrogen was also used as collision gas. Data were recorded from 5 to 30 min. Enhanced mass spectrometry (EMS) with information dependent acquisition (IDA) of enhanced product ion scans (EPI) was used. The survey scan was in the full scan mode from m/z 100 to 500 with a scan rate of 1000 amu/s. The ions m/z 134 and 178 were excluded from the scan range since they originated from the solvents

Urinary Biomarkers of Furan Exposure and produced a constant background. The intensity threshold was 100 counts/s, and Q0 trapping was set to on. The linear ion trap fill time was fixed to 50 ms and the collision energy to -10 V. For each data point, two EMS scans were summed up. An EPI scan of the most intense ion of each data point was acquired if the intensity exceeded 2500 counts/s. Target ions were excluded for 15 s after the first occurrence. The background ions with m/z 159, 201, 202, 203, and 212 were excluded from IDA-EPI scans. For the EPI scans, a collision energy of -30 V and a collision energy spread of 15 V were used. The scan rate was 1000 amu/s and the intensity threshold 10 counts/s. For each EPI spectrum, two scans were summed up. The mean total cycle time was 3.6 s. Data Extraction and Principal Component Analysis (PCA). EMS data were used to generate a matrix for multivariate data analyses. The 40 full scan chromatograms were extracted in one batch by MarkerView software 1.2 (Applied Biosystems/MDS Sciex). The extraction parameters were 10 counts/s for the noise threshold and 0.1 amu for the minimum spectral peak width. Retention time peak width was set to 1–20 scans. Intra- and interrun retention time shifts were tolerated within 30 s and mass shifts within 0.1 amu. To accept a peak as a variable, it had to be detected in at least 4 samples out of the 40. The maximum allowed number of extracted variables was preset to 500. To refine the peak area integration, the software was set to reintegrate the extracted variables in the raw data by a more precise algorithm (29). These parameter settings resulted in a matrix with 449 variables (mass retention time pairs). The data were exported to SIMCA-P 11.5 software (Umetrics, Umeå, Sweden), mean-centered, unit-variance scaled, and analyzed for principal components (PC). The fraction of the explained variation of a PC is given by R2X. Significance of the PCs was controlled by 7-fold cross-validation (Q2), a default function in SIMCA. On the basis of a model derived from the samples of the first post treatment period (5 treated and 5 vehicle controls), the loadings plot was explored, and the peaks (variables) were sorted according to their contribution to separate treated samples from controls. The 50 most important variables were examined in the original chromatograms and MS spectra. Variables originating from isotopes and fragments of the ionization process or signals with poor chromatographic separation were excluded from further analysis. MRM Analysis of Selected Peaks. The 13 potential biomarkers (variables) derived from multivariate data analysis were analyzed in the MRM mode of the mass spectrometer. Appropriate mass transitions were taken from the EPI scans. If possible, the ion with the highest intensity was selected for quadrupole 3. For compounds with no fragmentation, quadrupole 3 was set equal to quadrupole 1. The dwell time was set to 25 ms and the collision energy to -30 V. All other parameters were taken from the EPI method described above. Peak areas from treated and control samples differed substantially. Therefore, manual integration with the Analyst 1.4.1 software (Applied Biosystems/MDS Sciex) was necessary. Analytical reproducibility was determined on the basis of a pooled urine sample used as a quality control sample after each group of 10 analytical samples. Transitions with signals above 1 × 105 counts were integrated and the relative standard deviation (RSD) calculated. This resulted in 19% RSD for the signal of 357 f 137 at tR 13.5 min and 7% RSD for that of 217 f157 at tR 19.1 min. Statistical analysis of differences between treated samples and controls was based on the Wilcoxon test, using R 2.3.1 software (R Foundation for Statistical Computing). Preparation of cis-2-Butene-1,4-dial. cis-2-Butene-1,4-dial was prepared by hydrolysis of 2,5-diacetoxy-2,5-dihydrofuran during 24 h (21). 2,5-Diacetoxy-2,5-dihydrofuran was prepared as described elsewhere (30). Incubation of cis-2-Butene-1,4-dial with N-Acetyl-L-lysine. To 1.9 mL of potassium phosphate buffer (250 mM, pH 7.4), 50 µL of a solution of N-acetyl-L-lysine (10 mM) and 50 µL of a solution of cis-2-butene-1,4-dial (1 mM) were added. The solution was incubated for 16 h at 37 °C. The major product of this reaction had been described as R-2-(acetylamino)-6-(2,5-dihydro-2-oxo-1Hpyrrol-1-yl)-1-hexanoic acid (II) (13). Analysis of the in Vitro

Chem. Res. Toxicol., Vol. 21, No. 3, 2008 763 incubation with the EMS-IDA-EPI method used for the urine samples confirmed the structure of II (tR ) 14.1 min): m/z 253 [M- - H; 100%], 211 [M- - C(O)CH3; 87%], 167 [M- C(O)CH3 and - COOH; 9%], 82 [N(CH2CHdCHCdO)-; 21%]. Incubation of cis-2-Butene-1,4-dial with N-Acetyl-L-lysine and N-Acetyl-L-cysteine. To 1.85 mL of potassium phosphate buffer (250 mM, pH 7.4), 50 µL of a solution of N-acetyl-L-cysteine (10 mM), 50 µL of a solution of N-acetyl-L-lysine (10 mM), and 50 µL of a solution of cis-2-butene-1,4-dial (1 mM) were added. The solution was incubated for 16 h at 37 °C. The major product of this reaction had been described as N-acetyl-S-[1-[5-(acetylamino)-5-carboxypentyl]-1H-pyrrol-3-yl]-L-cysteine (III) (13). Analysis of the in Vitro incubation with the EMS-IDA-EPI method used for the urine samples also confirmed this structure III (tR ) 20.1 min): m/z 398 [M- - H; 100%], 269 [M- - CH2CH(NHCOCH3)COOH; 72%], 227 [M- - C4H8CH(NHCdOCH3)COOH; 65%]; 98 [pyrrole-3-thiolate; 29%]. Compound IV, also detected in the in ViVo data, is compatible with the oxidized form of III, as suggested by the fragmentation pattern. LC-MS/MS of IV (tR ) 14.4 min): m/z 414 [M- - H; 44%], 285 [M- CH2CH(NHC(O)CH3)COOH; 100%], 243 [M- - C4H8CH(NHCOCH3)COOH; 44%]; 114 [9%]. Oxidation of III by H2O2 in HPLC extracts produced compound IV. LC-MS/MS data: (tR ) 14.2 min): m/z 414 [39%], 285 [100%], 243 [48%]; 114 [10%]. Incubation of cis-2-Butene-1,4-dial with GSH. To 1.9 mL of potassium phosphate buffer (250 mM, pH 7.4), 50 µL of a solution of reduced glutathione (10 mM) and 50 µL of a solution of cis-2butene-1,4-dial (1 mM) were added. The solution was incubated for 16 h at 37 °C. The major product of this reaction had been described as N-[4-carboxy-4-(3-mercapto-1H-pyrrol-1-yl)-1-oxobutyl]-L-cysteinylglycine cyclic sulfide (I) (14). Analysis of the in Vitro incubation with the EMS-IDA-EPI method used for the urine samples also confirmed this structure. LC-MS/MS: (tR ) 15.0 min): m/z 354 [M- - H; 100%], 336 [M- - H2O; 8%]; 185 [9%]; 141 [24%]; 124 [13%]; 98 [pyrrole-3-thiolate; 12%]. Incubation of cis-2-Butene-1,4-dial with Methanethiole and Glutamic Acid. To 1.85 mL of potassium phosphate buffer (250 mM, pH 7.4), 50 µL of a solution of sodium glutamate (10 mM), 50µL of a solution of sodium methanethiolate, and 50 µL of a solution of cis-2-butene-1,4-dial (1 mM) were added. The solution was incubated for 16 h at 37 °C. The major product was S-[1-[5(acetylamino)-5-carboxypentyl]-1H-pyrrol-3-yl]-methanethiole (V) as identified by mass spectral data and 1H NMR. LC-MS/MS (tR ) 15.0 min): m/z 242 [M- - H; 100%], 112 [M- - H2O; 8%]; 97 [pyrrole-3-thiolate; 12%]. 1H NMR (400.13 MHz, acetonitriled3) δ 2.05–2.25 (m; 3H; β-CH2 and γ-CH2), 2.29 (s; 3H; SCH3), 2.30–2.45 (m; 1H; β-CH2), 4.65–4.75 (m; 1H; R-CH2), 6.13 (dd; J4,5 ) 2.91 Hz; J4,2 ) 1.74 Hz; 1H; C4-H), 6.72 (dd; J5,4 ) 2.91 Hz; J5,2 ) 2.22 Hz; 1H; C5-H), 6.75 (dd; J2,5 ) 2.22 Hz; J2,4 )1.74 Hz; 1H; C2-H).

Results Animal Treatment and Preparation of Urine. The dose of 40 mg of furan per kg of body wt selected for our single-dose study had been shown to be well tolerated by F344 rats even when given repeatedly in the 16-day NTP study (8). Urine samples were collected from each rat individually during two 24-h periods, each before and after treatment. The furan-treated rats showed a marked increase in urine volume in the first posttreatment sampling period. The increase was by a factor of 4 relative to the urine volume of the pretreatment samples. In the second post-treatment period, the urine volumes of the treated animals were about 1.6 times higher. The concentrations of creatinine, urea, sodium, and potassium as well as the osmolarity and the specific gravity of urine decreased by the same factor. Therefore, all urine samples were diluted to equal specific gravity, in order to minimize concentration-dependent matrix effects on retention times and ion suppression during LC-MS analysis.

764

Chem. Res. Toxicol., Vol. 21, No. 3, 2008

Kellert et al.

Figure 1. LC-MS/MS analysis of a urine sample collected from a furan-treated rat. (A) Chromatogram of the total ion count of the survey scan in the full scan mode. (B) Spectrum of the ions recorded at tR 20.14 with the predominant ion m/z 398 (compound III in Scheme 1). (C) Enhanced product ion spectrum of m/z 398 triggered by the full scan. NL 171, neutral loss indicating an N-acetyl-L-lysine adduct; NL 129, neutral loss indicating a mercapturic acid (N-acetyl-L-cysteine conjugate). See Scheme 2 for the fragmentation pattern.

Scheme 2. Fragmentation of the Ion with m/z 398 Eluting at tR 20.14 min (Figure 1) Identified as Compound III (Scheme 1)a

a A neutral loss (NL) of 129 is characteristic for mercapturic acids. A NL of 171 is characteristic for N-acetyl-L-lysine adducts. The fragment of m/z 98 was found to be a structural element resulting from the reaction of cis-2-butene-1,4-dial with both a thiol and an amino group.

LC-MS/MS Analysis. Figure 1 illustrates the information obtained by the chosen EMS-IDA-EPI mode of mass spectrometric analysis of urine that had been collected within 24 h after furan administration. Chart A shows the chromatogram of the total ion count summing up scans from m/z 100 to 500. Chart B shows the respective scan at retention time 20.14 min. The ion with m/z 398 triggered the acquisition of the product ion scan shown in chart C. Scheme 2 illustrates how the major fragments relate to compound III in Scheme 1. Data Extraction. Parameters were optimized to detect small peaks while keeping entries from noise to a minimum. A peak had to be present in at least four samples, that is, at least four

of the five furan-treated rats should show the potential biomarker. The used version of the MarkerView software offers reintegration of the extracted peak areas on the basis of the raw data by a second more precise algorithm (29). This process reduced the percentage of blanks (null entries) in the data matrix from 37 to 24. The resulting final matrix contained 40 observations (10 animals, 4 sampling periods) and 449 variables. Principle Component Analysis (PCA). Depending on the question addressed, subgroups were selected for the multivariate analyses. Figure 2 shows the results of the PCA for the 5 furantreated and the 5 control animals in the first post-treatment urine sampling period. The first principal component of the scores plot (Figure 2A) allowed full separation between furan-treated rats at the left-hand side (negative values of t[1]) and vehicletreated at the right-hand side. The loadings plot in Figure 2B places the 449 variables according to their contribution to separate the animals on the corresponding scores plot. As the treated rats were positioned at negative values of t[1] in the scores plots (Figure 2A), the variables that are most strongly associated with furan treatment locate at the left-hand side in Figure 2B (negative values of p[1]). The fifty variables with highest negative values in p[1] of Figure 2B were examined for chromatographic and mass spectrometric behavior. Variables originating from natural isotopes and fragments from the ionization process as well as peaks with poor chromatography were identified by the EMS-IDA-EPI information and excluded from further analysis. This reduced the number of peaks to 13, designated in Figure 2B by an open triangle. MRM Analysis of 13 Putative Biomarkers. Treatmentrelated differences in the 13 variables were verified by the more sensitive and specific mass spectrometric MRM mode. Peak areas from treated versus control animals in the first posttreatment sampling period were compared statistically. A treatment-related increase was significant at p < 0.005 for 7 of the 13 compounds. These are listed in Table 1 as potential biomarkers for furan exposure. NL 171 and/or 129 were observed with five of the seven compounds. The selected MRM transitions are shown in the last column of Table 1. Structural Information on Biomarkers. Information-dependent acquisition of EPI spectra was a major step toward structure elucidation. As an example, the structural interpretation

Urinary Biomarkers of Furan Exposure

Chem. Res. Toxicol., Vol. 21, No. 3, 2008 765

Figure 2. Principle component analysis of full scan data on the urine of five furan-treated (#06–10) and five vehicle-treated (#01–05) rats collected in the first 24-h post-treatment collection period. (A) Scores plot. The first principal component has an explained variation (R2X) of 0.44 and a predicted variation (Q2) of 0.27. (B) Loadings plot showing all 449 variables. Variables on the left-hand side correlated most strongly with treatment. The 13 variables designated by an open symbol were selected for refined multiple-reaction-monitoring MS.

Table 1. Urinary Biomarkers of Furan Exposure of Rats Derived from the 50 Most Important Loadings of the First Principal Componenta characteristic fragments (EPI)

peak m/z

tR [min]

m/z (loss)

354 253 398 414 242 335 371

15.0 14.1 20.1 14.4 23.3 14.6 13.7

unspecific 82 (NL 171) 269 (NL 129) 227 (NL 171) 285 (NL 129) 243 (NL 171) unspecific 206 (NL 129) 242 (NL 129)

structural components

selected MRM transition

class

#

M - H+f

lysine-adduct mercapturateb lysine-adduct mercapturateb lysine-adduct

I II III IV V

141 211 269 285 97 206 242

mercapturateb mercapturateb

a See Figure 2. Differences between furan-treated and vehicle-treated rats based on mass spectrometric MRM data were significant for all variables shown (p < 0.005). A structural class is proposed whenever a characteristic neutral loss (NL) was detected. b NL 129 is characteristic for either mercapturate or mercapturate sulfoxide.

of the EPI spectrum shown in Figure 1C for m/z 398 is illustrated in Scheme 2. The fragment with m/z 269 indicates a NL of 129, which is consistent with a mercapturic acid cleaved at the side chain sulfur atom. The fragment with m/z 227 is the result of a NL of 171, consistent with the cleavage of N-acetyl-Llysine at the side chain nitrogen. This nitrogen atom shows up as the nitrogen of a pyrrole ring that has been formed together with the four-carbon moiety of BDA. The fragment m/z 98 results from both neutral losses to leave a sulfur-substituted pyrrole ring. On the basis of this fragmentation pattern, m/z 398 was considered to represent compound III. Examination of the product ion spectra led us to postulate that five of the seven variables in Table 1 corresponded to structures I-V shown in Scheme 1. Incubations of BDA with appropriate substrates were carried out for confirmation. Bifunctional reaction of BDA with glutathione (with the thiol group of cysteine and the amino group of glutamic acid to form the pyrrole ring) resulted in compound I, a conjugate that had already been found in rat urine by Peterson and co-workers (31, 15). Reaction of BDA with the side chain amino group of N-acetyl-L-lysine led to adduct II. Reaction with both N-acetylL-cysteine and N-acetyl-L-lysine yielded III and IV. Structures II and III had been reported from in Vitro incubations (13) but not in ViVo. Compound IV probably resulted from the oxidation of III to the sulfoxide as indicated by an increase of the mass by 16 amu to m/z 414. Full characterization of IV was difficult

because it coeluted with the more abundant product II. Strong support is given by the observations that treatment of isolated III with H2O2 resulted in two products with increase in mass by 16 and 32 amu, indicative of the formation of both the sulfoxide and the sulfone. Furthermore, the fragmentation pattern was consistent with that of III. The structure of variable m/z 242 was not readily elucidated because it did not show any of the characteristic fragments. A literature search on the disposition of S-conjugates of cysteine and glutathione (32, 33) led us to postulate structure V, that is, a 3-methylthiopyrrole adduct with glutamic acid. This so far unknown metabolite of furan was then prepared by the reaction of BDA with methanethiolate and glutamic acid, and the structure was confirmed by 1H NMR spectroscopy after purification by HPLC. The 3 protons of the methyl sulfide group showed a singlet at 2.29 ppm. The chemical shifts (6.13, 6.72, and 6.75 ppm) and coupling constants (J4,5 ) 2.91 Hz, J5,2 ) 2.22 Hz, and J2,4 )1.74 Hz) of the aromatic protons were similar to those of already described 3-substituted pyrroles (13). The two diastereotopic β-methylene protons of the glutamic acid moiety gave two distinct multiplet signals at 2.05–2.25 ppm and 2.30–2.45 ppm. Both multiplets coupled with the R-methine proton (4.65–4.75 ppm) in the 1H-1H-COSY spectrum and integrated 3:1 because of the superimposed signal of the two γ-methylene protons.

766

Chem. Res. Toxicol., Vol. 21, No. 3, 2008

Kellert et al.

Table 2. Concordance of Chromatographic and Mass Spectral Information between Urinary Metabolites of Furan-Treated Rats and Reaction Products Generated in Vitro by Incubation of cis-2-Butene-1,4-dial (BDA) with Glutathione, N-Acetyl-L-lysine, N-Acetyl-L-cysteine, and Methanethiole variables

retention time and fragments tr; m/z (relative abundance %)

compound (Scheme 1)

m/z

tR [min]

354

15.0

I

253 398 414 242

14.1 20.1 14.4 23.3

II III IV V

in Vitro

rat urine 15.0 min; 354 124 (17); 98 14.1 min; 253 20.1 min; 398 14.4 min; 414 23.3 min; 242

(100); 336 (14); 185 (10); 141 (25); (13) (100); 211 (92); 167 (9); 82 (19) (100); 269 (78); 227 (78); 98 (40) (45); 285 (100); 243 (41); 114 (11) (40); 198 (15); 112 (34); 97 (100)

We cannot propose a structure for the mercapturate with m/z 335. For the variable with m/z 371, a hypothesis based on the R-keto acid form of lysine will be discussed. Mass spectral data of all in Vitro reaction products were compared with the spectra obtained from the in ViVo samples. Table 2 shows the high concordance for both retention time and fragmentation of the EMS-IDA-EPI analyses. Chromatography of the mixtures of urine and in Vitro incubations resulted in the coelution of all 5 compounds. Time Course of Excretion. Considering the rapid uptake, distribution, and metabolism of furan, one would have expected that the urinary concentrations of the furan metabolites were much lower in the second post-treatment collection period. However, a significant decrease was seen only for the bicyclic glutathione conjugate I, its secondary product V, and the unidentified mercapturic acid with m/z 335 and tR 14.6 min. The other variables listed in Table 1 did not markedly decrease between 24 and 48 h. Note that the latter are all lysine adducts.

Discussion Metabolic Profiling and PCA. Since acidic conjugation products predominate in the urinary excretion of xenobiotics, our chromatography and mass spectrometry focused on the analysis of the respective classes of metabolites, such as mercapturic acids (34, 35), glucuronides (36), or sulfates (37). The column-switching unit allowed for large urine volumes to be analyzed, which resulted in higher concentrations of analytes on the analytical column without affecting retention time or enlarging peak shape. A second positive effect was the reduction of ion suppression in the source by removing urinary salts and small charged molecules. This resulted in a gain of sensitivity for the polar acidic compounds that were retained on the trap column under the acidic conditions of the solvent system. Scaling of the data is an important step before multivariate analysis. Variables were mean-centered and unit variance-scaled, which gives all variables the same weight, irrespective of the magnitude of the signal. PCA performed without scaling or with Pareto scaling failed to detect compound IV among the first 50 treatment-related variables. With unit variance-scaling, low-level metabolites with a small peak area have a chance to be detected. Background noise, however, could influence the model inappropriately. This problem was confined in the data matrix by setting an intensity threshold during acquisition and by the requirement that variables had to be detected in at least 4 samples. Glutathione (Cysteine) Conjugates. All but one of the excretion products shown in Scheme 1 could result from an addition reaction of the thiol group of cysteine to the 2,3-double bond of BDA. While compound I contains a complete glutathione moiety, compound III is a mercapturate, that is, it contains only N-acetyl-L-cysteine; compound IV is its sulfoxide derivative. The formation of compound V requires additional steps. A β-lyase-mediated cleavage of the -CH2-S bond has

15.4 min; 354 124 (13); 98 14.4 min; 253 20.5 min; 398 14.8 min; 414 22.5 min; 242

(100); 336 (8); 185 (9); 141 (24); (12) (100); 211 (87); 167 (9); 82 (21) (100); 269 (72); 227 (65); 98 (29) (44); 285 (100); 243 (44); 114 (9) (27); 198 (11); 112 (32); 97 (100)

been described for a number of cysteine conjugates (32, 33), and methylation of thiol groups by S-adenosylmethionine has been reported a long time ago (38, 39). Compound V is very likely formed from glutathione conjugate I as indicated by the glutamic acid/pyrrole moiety and the similar time course of excretion for both compounds. Whether the cysteine conjugate β-lyase pathway contributes to the cell type-specific profile of furan-mediated toxicity and carcinogenicity in the liver remains to be investigated. Interaction of the free thiol with disulfide bridges in protein, formation of thiyl radicals, or interference with DNA methylation may be thought of as specific mechanisms of toxic action. The central question will be whether celltype specific processes favor the formation and accumulation of this conjugate. Lysine Adducts as Biomarkers. All five metabolites shown in Scheme 1 contain a five-membered heterocyclic ring structure. The nitrogen atom is derived from the side chain amino group of lysine in three cases (II, III, and IV) and from the amino terminal of glutamic acid in glutathione (I and V). Lysine adducts and their N-acetylated derivatives had already been discussed as biomarkers in the context of malondialdehyde (40, 41) and of the dialdehyde formed in the metabolism of aflatoxin B1 (42, 43). The lysine-containing adducts showed similar concentrations in urine of the first and the second 24-h collection periods. This protracted time course of excretion indicates that adducts were not formed with the free amino acid but with lysine incorporated in protein. This supports the idea that lysine adducts may play an important role in reactions of electrophiles with cellular macromolecules, which may ultimately be responsible for cytotoxicity. The NL of 171 amu observed for all identified N-acetyl-Llysine adducts could be highly characteristic for this class of metabolites. Together with the NL of 129 for mercapturic acids, this information could help identify metabolites in urine that indicate the formation of electrophilic toxicants. With long-lived protein as targets, such as globin, measurement of the excretion of the respective products in urine could even serve as markers of past exposure. It might be interesting to note here that the compound with m/z 371 tR 13.7 min in Table 1 could have a structure similar to that of compound IV (m/z 414). The mass difference of 43 is consistent with the difference between N-acetyl-L-lysine and the R-keto acid of lysine, that is, a product of L-amino acid oxidase action. This could also explain the fact that this compound showed two neutral losses of 129 amu, one to be attributed to N-acetyl-L-cysteine, the second to a cleavage of the R-keto acid at the pyrrole nitrogen (minus C6H9O3). Biomarkers in Controls. With the improved sensitivity of MRM analysis, compounds II, III, and m/z 371 were above the limit of detection also in control urine. If this was due to small amounts of furan present in the rat chow, one would expect that all three variables showed a similar ratio for the

Urinary Biomarkers of Furan Exposure

difference between treated samples and controls. This was not seen; ratios spanned from 30 for compound II to 1500 for compound III. We cannot offer any convincing explanation for the background but favor a formation unrelated to furan exposure. 5′-Oxidation of deoxyribose in DNA has for instance been shown to produce trans-2-butene-1,4-dial (44). It should be noted in this context that the cis–trans configuration no longer plays a role when the double bond is saturated by the addition of a thiol. Whatever the source, it did not interfere with our search for biomarkers of furan exposure.

Conclusions Our results demonstrate the usefulness of comprehensive mass spectrometric analysis of urine combined with multivariate analyses in the search for biomarkers of exposure to chemicals that form an electrophilic intermediary metabolite, even under the conditions of a highly complex mixture of excretion products. Our investigation also allowed for the elucidation of so far unknown furan metabolites. The approach should be particularly helpful for furan because of the numerous potentially unknown sources and routes of exposure and the problem of volatility that renders measurements in food items unreliable. Acknowledgment. This work was supported by the Swiss Federal Office of Public Health (grant no. 05.002738 and 06.004619; monitoring scientist: Dr. Josef Schlatter). We thank Dr. Wolfgang Völkel for helpful discussions and support. Marion Friedewald is acknowledged for excellent technical assistance. MarkerView software was kindly provided by Applied Biosystems/MDS Sciex (Concord, Canada).

References (1) Maga, J. A. (1979) Furans in foods. CRC Crit. ReV. Food Sci. Nutr. 11, 355–400. (2) Perez Locas, C., and Yaylayan, V. A. (2004) Origin and mechanistic pathways of formation of the parent furan: a food toxicant. J. Agric. Food Chem. 52, 6830–6836. (3) Yaylayan, V. A. (2006) Precursors, formation and determination of furan in food. Journal fuer Verbraucherschutz und Lebensmittelsicherheit 1, 5–9. (4) European Food Safety Authority (2004) Report of the Scientific Panel on Contaminants in the Food Chain on Furan in Food. The EFSA Journal Vol. 137, EFSA, Parma, Italy. (5) US Food and Drug Administration (2004) Furan in food, Thermal Treatment. www.cfsan.fda.gov/∼lrd/fr040510.html. (6) Reinhard, H., Sager, F., Zimmermann, H., and Zoller, O. (2004) Furan in foods on the Swiss market - method and results. Mitt. Lebensmittelunters. Hyg. 95, 532–535. (7) International Agency for Research on Cancer (1995) Dry Cleaning, Some Chlorinated Solvents and Other Industrial Chemicals. IARC Monographs, Vol. 63, IARC World Health Organisation, Lyon, France. (8) National Toxicology Program (1993) Toxicology and Carcinogenesis Studies of Furan in F344 Rats and B6C3F1 Mice Natl. Toxicol. Program Tech. Rep. Ser. Vol. 402 , U.S. Department of Health and Human Services, Public Health Service, National Institutes of Health, Research Triangle Park, NC. (9) Gold, L. S. and Zeiger, E. (1997) Handbook of Carcinogenic Potency and Genotoxicity Databases, CRC Press, New York. (10) Lutz, W. K. (1990) Dose-response relationship and low dose extrapolation in chemical carcinogenesis. Carcinogenesis 11, 1243–1247. (11) Kedderis, G. L., Carfagna, M. A., Held, S. D., Batra, R., Murphy, J. E., and Gargas, M. L. (1993) Kinetic analysis of furan biotransformation by F-344 rats in vivo and in vitro. Toxicol. Appl. Pharmacol. 123, 274–282. (12) Chen, L. J., Hecht, S. S., and Peterson, L. A. (1995) Identification of cis-2-butene-1,4-dial as a microsomal metabolite of furan. Chem. Res. Toxicol. 8, 903–906. (13) Chen, L. J., Hecht, S. S., and Peterson, L. A. (1997) Characterization of amino acid and glutathione adducts of cis-2-butene-1,4-dial, a reactive metabolite of furan. Chem. Res. Toxicol. 10, 866–874.

Chem. Res. Toxicol., Vol. 21, No. 3, 2008 767 (14) Peterson, L. A., Cummings, M. E., Vu, C. C., and Matter, B. A. (2005) Glutathione trapping to measure microsomal oxidation of furan to cis2-butene-1,4-dial. Drug Metab. Dispos. 33, 1453–1458. (15) Peterson, L. A., Cummings, M. E., Chan, J. Y., Vu, C. C., and Matter, B. A. (2006) Identification of a cis-2-Butene-1,4-dial-derived glutathione conjugate in the urine of furan-treated rats. Chem. Res. Toxicol. 19, 1138–1141. (16) Wilson, D. M., Goldsworthy, T. L., Popp, J. A., and Butterworth, B. E. (1992) Evaluation of genotoxicity, pathological lesions, and cell proliferation in livers of rats and mice treated with furan. EnViron. Mol. Mutagen. 19, 209–222. (17) Carfagna, M. A., Held, S. D., and Kedderis, G. L. (1993) Furan-induced cytolethality in isolated rat hepatocytes: correspondence with in vivo dosimetry. Toxicol. Appl. Pharmacol. 123, 265–273. (18) Gingipalli, L., and Dedon, P. C. (2001) Reaction of cis- and trans-2butene-1,4-dial with 2′-deoxycytidine to form stable oxadiazabicyclooctaimine adducts. J. Am. Chem. Soc. 123, 2664–2665. (19) Byrns, M. C., Predecki, D. P., and Peterson, L. A. (2002) Characterization of nucleoside adducts of cis-2-butene-1,4-dial, a reactive metabolite of furan. Chem. Res. Toxicol. 15, 373–379. (20) Bohnert, T., Gingipalli, L., and Dedon, P. C. (2004) Reaction of 2′deoxyribonucleosides with cis- and trans-1,4-dioxo-2-butene. Biochem. Biophys. Res. Commun. 323, 838–844. (21) Byrns, M. C., Vu, C. C., and Peterson, L. A. (2004) The formation of substituted 1,N6-etheno-2′-deoxyadenosine and 1,N2-etheno-2′-deoxyguanosine adducts by cis-2-butene-1,4-dial, a reactive metabolite of furan. Chem. Res. Toxicol. 17, 1607–1613. (22) Byrns, M. C., Vu, C. C., Neidigh, J. W., Abad, J. L., Jones, R. A., and Peterson, L. A. (2006) Detection of DNA adducts derived from the reactive metabolite of furan, cis-2-butene-1,4-dial. Chem. Res. Toxicol. 19, 414–420. (23) Burka, L. T., Washburn, K. D., and Irwin, R. D. (1991) Disposition of [14C]furan in the male F344 rat. J. Toxicol. EnViron. Health 34, 245–257. (24) Plumb, R. S., Stumpf, C. L., Gorenstein, M. V., Castro-Perez, J. M., Dear, G. J., Anthony, M., Sweatman, B. C., Connor, S. C., and Haselden, J. N. (2002) Metabonomics: the use of electrospray mass spectrometry coupled to reversed-phase liquid chromatography shows potential for the screening of rat urine in drug development. Rapid Commun. Mass Spectrom. 16 1991–1996. (25) Idborg, H., Edlund, P. O., and Jacobsson, S. P. (2004) Multivariate approaches for efficient detection of potential metabolites from liquid chromatography/mass spectrometry data. Rapid Commun. Mass Spectrom. 18, 944–954. (26) Chen, C., Ma, X., Malfatti, M. A., Krausz, K. W., Kimura, S., Felton, J. S., Idle, J. R., and Gonzalez, F. J. (2007) A comprehensive investigation of 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP) metabolism in the mouse using a multivariate data analysis approach. Chem. Res. Toxicol. 20, 531–542. (27) Kellert, M., Scholz, K., Wagner, S., Dekant, W., and Volkel, W. (2006) Quantitation of mercapturic acids from acrylamide and glycidamide in human urine using a column switching tool with two trap columns and electrospray tandem mass spectrometry. J. Chromatogr., A 1131, 58–66. (28) Wagner, S., Scholz, K., Sieber, M., Kellert, M., and Voelkel, W. (2007) Tools in metabonomics: an integrated validation approach for LCMS metabolic profiling of mercapturic acids in human urine. Anal. Chem. 79, 2918–2926. (29) Sangster, T. P., Wingate, J. E., Burton, L., Teichert, F., and Wilson, I. D. (2007) Investigation of analytical variation in metabonomic analysis using liquid chromatography/mass spectrometry. Rapid Commun. Mass Spectrom. 21, 2965–2970. (30) Holzapfel, C. W., and Williams, D. B. G. (1995) A facile route to 3a,8a-dihydrofuro[2,3-b]benzofuran. Tetrahedron 51, 8555–8564. (31) Peterson, L. A. (2006) Electrophilic intermediates produced by bioactivation of furan. Drug Metab. ReV. 38, 615–626. (32) Tateishi, M., Suzuki, S., and Shimizu, H. (1978) Cysteine conjugate beta-lyase in rat liver. A novel enzyme catalyzing formation of thiolcontaining metabolites of drugs. J. Biol. Chem. 253, 8854–8859. (33) Commandeur, J. N., Stijntjes, G. J., and Vermeulen, N. P. (1995) Enzymes and transport systems involved in the formation and disposition of glutathione S-conjugates. Role in bioactivation and detoxication mechanisms of xenobiotics. Pharmacol. ReV. 47, 271– 330. (34) Scholz, K., Dekant, W., Voelkel, W., and Paehler, A. (2005) Rapid detection and identification of N-acetyl-L-cysteine thioethers using constant neutral loss and theoretical multiple reaction monitoring combined with enhanced product-ion scans on a linear ion trap mass spectrometer. J. Am. Soc. Mass Spectrom. 16, 1976–1984. (35) Wagner, S., Scholz, K., Donegan, M., Burton, L., Wingate, J., and Voelkel, W. (2006) Metabonomics and biomarker discovery: LC-MS metabolic profiling and constant neutral loss scanning combined with

768

(36)

(37)

(38) (39) (40)

Chem. Res. Toxicol., Vol. 21, No. 3, 2008 multivariate data analysis for mercapturic acid analysis. Anal. Chem. 78, 1296–1305. Lutz, U., Lutz, R. W., and Lutz, W. K. (2006) Metabolic profiling of glucuronides in human urine by LC-MS/MS and partial least-squares discriminant analysis for classification and prediction of gender. Anal. Chem. 78, 4564–4571. Lafaye, A., Junot, C., Ramounet-Le Gall, B., Fritsch, P.,., Ezan, E., and Tabet, J.-C. (2004) Profiling of sulfoconjugates in urine by using precursor ion and neutral loss scans in tandem mass spectrometry. J. Mass Spectrom. 39, 655–664. Bremer, J., and Greenberg, D. M. (1961) Enzymic methylation of foreign sulfhydryl compounds. Biochim. Biophys. Acta 46, 217–224. Weisiger, R. A., and Jakoby, W. B. (1979) Thiol S-methyltransferase from rat liver. Arch. Biochem. Biophys. 196, 631–637. McGirr, L. G., Hadley, M., and Draper, H. H. (1985) Identification of Na-acetyl-e-(2-propenal)lysine as a urinary metabolite of malondialdehyde. J. Biol. Chem. 260, 15427–15431.

Kellert et al. (41) Piche, L. A., Cole, P. D., Hadley, M., Van den Bergh, R., and Draper, H. H. (1988) Identification of N-e-(2-propenal)lysine as the main form of malondialdehyde in food digesta. Carcinogenesis 9, 473–477. (42) Sabbioni, G., Ambs, S., Wogan, G. N., and Groopman, J. D. (1990) The aflatoxin-lysine adduct quantified by high-performance liquid chromatography from human serum albumin samples. Carcinogenesis 11, 2063–2066. (43) Guengerich, F. P., Williams, K. M., Sutter, T. R., Hayes, J. D., Johnson, W. W., Arneson, K. O., Voehler, M., Deng, Z., and Harris, T. M. (2004) Competing reactions of aflatoxin B1 dialdehyde: Enzymatic reduction versus adduction with lysine. ACS Symp. Ser. 865, 171–182. (44) Chen, B., Bohnert, T., Zhou, X., and Dedon, P. C. (2004) 5′-(2Phosphoryl-1,4-dioxobutane) as a product of 5′-oxidation of deoxyribose in DNA: elimination as trans-1,4-dioxo-2-butene and approaches to analysis. Chem. Res. Toxicol. 17, 1406–1413.

TX7004212