Metabonomic Study of Ochratoxin A Toxicity in Rats after Repeated

May 29, 2009 - For early detection of toxicity and improved mechanistic understanding, GC/MS-, 1H NMR-, and LC/MS-based metabonomics were applied to u...
0 downloads 10 Views 395KB Size
Download PDF
Chem. Res. Toxicol. 2009, 22, 1221–1231

1221

Articles Metabonomic Study of Ochratoxin A Toxicity in Rats after Repeated Administration: Phenotypic Anchoring Enhances the Ability for Biomarker Discovery Maximilian Sieber,† Silvia Wagner,† Eva Rached,† Alexander Amberg,‡ Angela Mally,† and Wolfgang Dekant*,† Department of Toxicology, UniVersity of Wu¨rzburg, Versbacher Strasse 9, 97078 Wu¨rzburg, Germany, and Sanofi-AVentis Deutschland GmbH, Drug Safety EValuation, Frankfurt, Germany ReceiVed December 2, 2008

For early detection of toxicity and improved mechanistic understanding, GC/MS-, 1H NMR-, and LC/ MS-based metabonomics were applied to urine samples from a rodent toxicity study on the mycotoxin and renal carcinogen ochratoxin A (OTA). OTA was administered at doses of 0, 21, 70, and 210 µg/kg body wt for up to 90 days. Urine samples were collected at 24 h intervals 14, 28, and 90 days after the start of treatment and analyzed with GC/MS, 1H NMR, and LC/MS. Principal component analysis and orthogonal projection to latent structures discriminate analysis (OPLS-DA) based on GC/MS and 1H NMR data discriminated controls from animals dosed with 210 µg/kg body wt OTA as early as 14 days and animals dosed with 70 µg/kg body wt 28 days after the start of treatment, correlating with mild histopathological changes in the kidney. Integration of histopathology scores as discriminators in OPLSDA models resulted in better multivariate model predictivity and facilitated marker identification. Decreased 2-oxoglutarate and citrate excretion and increased glucose, creatinine, pseudouridine, 5-oxoproline, and myo-inositol excretion were detected with GC/MS. Decreased 2-oxoglutarate and citrate excretion and increased amino acid excretion were found with 1H NMR. Increased urinary glucose is a well-established indicator of kidney damage, and altered excretion of TCA cycle intermediates (citrate and 2-oxoglutarate) is found as a general response to toxic insult in many metabonomics studies. Other markers are associated with cell proliferation (pseudouridine), changes in renal osmolyte handling (myo-inositol), and oxidative stress (5-oxoproline), established mechanisms of OTA toxicity. LC/MS was also able to discriminate controls and treated animals but contained more noise, and marker annotation was only speculative due to lack of reference databases. Use of multiple analytical platforms for metabonomics analysis may result in a more comprehensive metabolite coverage and may be applied to obtain mechanistic information from conventional rodent toxicity studies. Introduction Metabonomics is increasingly being used as a noninvasive method for toxicity assessment. This technology may offer a quick and comprehensive detection of biochemical perturbations in urine or plasma indicative of toxicities and may provide information on mechanisms of toxicity. Metabonomic techniques have been applied to analyze the urinary profiles of various model hepato- and nephrotoxins, such as hydrazine and mercuric chloride, and to characterize the toxicities of novel compounds in drug development (1). A variety of analytical platforms have been used for metabonomic analysis. 1H NMR was the method of choice used by the pioneers in the field (2). More recently, LC/MS has gained importance (3), especially after the introduction of ultra high * To whom correspondence should be addressed. Tel: +49(0)931/20148449. Fax: +49(0)931 /201-48865. E-mail: [email protected]. † University of Wu¨rzburg. ‡ Sanofi-Aventis Deutschland GmbH.

pressure liquid chromatography (UPLC)1 (4), resulting in superior separation, and by hydrophilic interaction chromatography (HILIC) (5), allowing efficient separation of highly polar compounds. Although the concept of metabolite profiling by GC/MS was developed in clinical analysis in the 1970s (6), more recently, GC/MS for metabonomic analysis was employed mainly in plant sciences (7, 8), whereas only a few applications of GC/MS profiling in toxicology were reported (9, 10). Besides different analytical profiling methods, there are a number of multivariate data analysis approaches used in metabonomics (11). The most commonly used is principal component analysis (PCA). It is untargeted, and the discrimination between samples is based only on the projection of variance (11). Therefore, PCA can be applied to gain rapid insight into the data structure. However, for the identification of discriminat1 Abbreviations: UPLC, ultra high-pressure liquid chromatography; PCA, principal component analysis; OPLS-DA, orthogonal projection to latent structures discriminant analysis; OTA, ochratoxin A; MSTFA, N-methylN-(trimethylsilyl)-trifluoroacetamide; TIC, total ion current; TMAO, trimethylamine-N-oxide; PC, principal component; TMS, trimethylsilyl; MO, methoxime.

10.1021/tx800459q CCC: $40.75  2009 American Chemical Society Published on Web 05/29/2009

1222

Chem. Res. Toxicol., Vol. 22, No. 7, 2009

ing markers between experimental groups, supervised approaches are superior. Orthogonal projection to latent structures discriminant analysis (OPLS-DA) models samples according to class information (e.g., treated vs control animals) attributed to the samples prior to the multivariate modeling. OPLS-DA separates discriminating information from information that does not contribute to the class separation, the so-called orthogonal information (12). Therefore, this method may be suitable to obtain information on changes in molecular composition of samples. An overview on chemometric methods used in metabonomic research is given by Trygg et al. (11). In routine toxicity testing, histopathology is still the “gold standard” for detection of chemically induced pathologic lesions. Histopathology data are acquired together with urine and plasma samples for clinical chemistry analysis. For metabonomic analysis though, only the body fluid samples are used, and the changes in urinary or plasma composition are correlated to the dose of the compound examined in the study. However, histopathology is the phenotypic anchoring of toxic lesions, and this information is often not included in the metabonomic analysis. Therefore, the aim of this study was to integrate histopathology data into metabonomics analysis to make the best use of all data routinely collected for toxicity testing. To demonstrate how additional information can be gained by metabonomics, GC/MS, 1H NMR, and LC/MS data were used to investigate metabolic changes in urine caused by ochratoxin A (OTA), a food contaminant and renal carcinogen. Full histopathology and clinical chemistry of a 90 days subchronic study on OTA have been reported (13). The suitability of different analytical platforms for metabonomics and their power to detect changes in the molecular composition of urine indicative of toxicities and its mechanisms were assessed. Chemometric analyses of urine samples separated treated animals from controls in a dose-dependent manner, and a series of molecular markers of OTA toxicity were identified.

Experimental Procedures Chemicals and Solvents. O-Methoxyamine hydrochloride, myoinositol, dried pyridine, acetone, and chloroform were obtained from Sigma Aldrich (St. Louis, MO). Methanol was purchased from Carl Roth GmbH (Karlsruhe, Germany). N-Methyl-N-(trimethylsilyl)trifluoroacetamide (MSTFA) was purchased from AppliChem (Darmstadt, Germany). OTA (CAS 303-47-9; 99% purity) was purchased from Axxora (Gru¨neberg, Germany; batch L16528/a). Sample Acquisition. Urine samples were aquired as published previously (13). Male F344/N rats (205-225 g, n ) 5 per group, Harlan-Winkelmann, Borchen, Germany) were housed in standard Macrolon cages type 4 with wire mesh tops and softwood bedding (five animals per cage) on a 12 h light/dark cycle with a room temperature of 22 ( 2 °C and humidity of 50-60%. They had free access to pelleted standard diet (Altromin, Lage, Germany) and tap water. After a week of acclimatization, they were given OTA at doses of 0, 21, 70, or 210 µg/kg body wt in corn oil by gavage for 14, 28, or 90 days 5 days per week (13). Control animals received corn oil only. Animals were transferred to individual metabolic cages (one animal per cage) for urine collection. During urine collection, animals were fasted but had free access to drinking water. Urine samples were collected on ice for a 24 h collecting period after 14, 28, and 90 days of administration and stored at -20 °C. Once thawed, samples were prepared for metabonomic analysis within 6 h. Clinical Chemistry Analysis. Clinical chemistry analysis was carried out as published previously (13). Relevant parameters are summarized in the Supporting Information, Table 1. Histopathology Scores. Histopathology scores were obtained as published previously (13). Sections were scored by an expert

Sieber et al. pathologist, and scoring for the following end points was integrated in the multivariate data analysis: liver glycogen, liver fatty change, liver inflammatory foci, kidney hyaline inclusions, kidney tubular basophilia, kidney tubular vacuolation, kidney tubular cell apoptosis, kidney mononuclear foci, kidney tubular cast, kidney karyomegaly, and kidney interstitial fibrosis. The findings are summarized in the Supporting Information, Table 2. Sample Treatment for GC/MS Analysis. Urine samples were thawed overnight at 4 °C. After thawing, samples were briefly vortexed to remove inhomogeneities due to the freezing-thawing process. Proteins were precipitated by the addition of 100 µL of cold methanol to 50 µL of urine. After centrifugation (2.31g, 10 min), 50 µL of the supernatant was transferred to a GC autosampler vial with a microinsert. The samples were evaporated to dryness at 30 °C in a Centrivac vacuum centrifuge (Heraeus Instruments, Osterode, Germany). After 2 h, 50 µL of acetone was added to each sample to ensure complete drying of the sample. Acetone addition was repeated once. After the acetone was removed, the samples were redissolved in 50 µL of pyridine containing 20 mg/mL methoxyamine hydrochloride for methoximation of ketones and aldehydes. Samples were derivatized at 40 °C for 90 min. Then, the samples were silylated at 40 °C for 1 h with 100 µL of MSTFA and used directly for GC/MS analysis. All samples were randomized for each treatment step. Additionally, 50 µL of each sample was pooled. After every four samples, this pooled urine was worked up in the same manner as the samples and used as the quality control. The repeated workup and analysis of a pooled sample for quality control was proposed by Sangster et al. (14) to overcome the problem of monitoring sample preparation and analytical performance in untargeted analysis, where internal standards cannot easily be implemented. GC/MS Analysis. Samples were analyzed on a HP6980 gas chromatograph with a split inlet (split ratio 20:1) equipped with a J&W Scientific DB5-MS column (30 m × 0.25 mm i.d., 0.1 µm film) coupled to a HP5973 mass selective detector. Data recording and instrument control were performed by HP Chemstation version D.02.00 (Agilent GmbH, Waldbronn, Germany). Samples were introduced by a CombiPal autosampler (CTC Analytics GmbH, Zwingen, Switzerland). The inlet temperature and transfer line temperature were set to 280 °C. The oven temperature increased from 60 to 85 °C at a rate of 2.5 °C/min and from 85 to 280 °C at a rate of 9.0 °C/min and was then held at 280 °C for 4 min. The helium carrier gas flow was kept constant at 0.8 mL/min. The MS detector was switched off during the elution of the urea signal from 11.50 to 16.00 min. The detector was operated in the scan mode from m/z 60 to m/z 650 with a sampling rate of 8.69 scans/s and a threshold of 50 counts. The source temperature and quadrupole temperature were 230 and 150 °C, respectively. GC/MS Raw Data Handling and Statistical Analysis. Chromatograms were inspected visually and those that deviated strongly from the pooled quality controls due to chromatographic problems such as poor resolution, no detected peaks etc., were excluded (five samples out of 60 in total). Such samples carry no information for the following statistical analysis but may interfere with the automated peak picking and alignment process. The remaining samples were exported in the platform-independent netCDF (*.cdf) format with the ChemStation export function for further analysis. Automatic peak detection and peak alignment were performed by the XCMS software (version 1.6.1) (15) based on R-program version 2.4.0 (R-Foundation for statistical computing, www.Rproject.org). Default parameters of XCMS were used except for the following: fwhm ) 4.0 s and profmethod ) “binlinbase”. The results table containing mass spectral features as mass/retention time pairs in a tab-separated text file (*.txt) was imported into Excel work sheets (Microsoft, Unterschleissheim, Germany). Normalization to total ion current (TIC) and further data handling steps such as sorting data according to retention time were carried out in Excel prior to statistical analysis with SIMCA-P+ 11.5 (Umetrics, Umeå, Sweden). Variables were mean-centered and pareto-scaled for both PCA and OPLS-DA. The significance of the components was determined by 7-fold cross-validation, the default validation tool

Metabonomic Study of Ochratoxin A in SIMCA-P+ 11.5. Only significant components were used for further analysis. If a separation between control and dose groups was observed in the PCA scores plot, OPLS-DA was performed to highlight the differences between the groups. Potential markers for group separation were subsequently identified by analyzing the S-plot (16), which plots the covariance (p) against the correlation (pcorr). For a marker, both the contribution to the model expressed in p and the effect and reliability of this contribution expressed in p(corr) should be high; thus, the potential markers are located on the outer ends of the S-shaped point swarm. Cut-off values of p g |0.05| and p(corr) g |0.5| were used. To avoid overinterpretation of the model, markers were selected in a conservative manner so that only those markers showing a significant jack-knifed confidence interval of less than half of the variable’s covariance p were further investigated. With the mass/retention time pairs, the corresponding peak was identified in the original GC/MS chromatograms. Then, the NIST’s Automated Mass Spectral Deconvolution and Identification System (AMDIS) (17) was run for peak deconvolution, and the peak was compared to the NIST mass spectral database (National Institute of Standards and Technology, Gathersburg, MD) for identification. Confirmation of peak identity was carried out by coeluting authentic reference compounds, if available. Sample Treatment for 1H NMR Analysis. Urine samples were thawed overnight at 4 °C, and precipitated solids were removed by centrifugation (2.31g, 10 min). Urine (630 µL) was buffered with 70 µL of a 1 M phosphate buffer in D2O (pH 7.0) containing 10 mM d4-trimethylsilylpropionic acid sodium salt (TSP) as a shift lock reagent prior to transfer into a 5 mm NMR tube (Aldrich Series 30). Samples were randomized for each treatment step. 1 H NMR Analysis. Spectra were recorded on a Bruker DMX 600 spectrometer equipped with a 5 mm DCH cryoprobe using pulsed field gradients (both by Bruker Biospin GmbH, Rheinstetten, Germany). Water suppression was achieved with the noesygppr1d pulse sequence from the Bruker library. For Fourier transformation, 32 scans with an aquisition time of 2.75 s and a delay time of 2.00 s were recorded. Spectral width was 11904 Hz or 19.8 ppm. Each spectrum was manually baseline-corrected and referenced to TSP (δ ) 0.00 ppm). Instrument control, data recording, and baseline correction were carried out with the Bruker WIN-NMR Suite. 1 H NMR Raw Data Handling and Statistical Analysis. Spectra were inspected visually to exclude any outliers due to extremely diluted samples or inadequate water suppression (three samples out of 60 in total). The spectra were then exported to the Chenomx NMR Suite 4.6 (Chenomx, Edmonton, Canada) and binned into 0.04 ppm wide bins. The bins around the water resonance from 4.40 to 6.20 ppm were excluded from the analysis. Binned 1H NMR data were normalized to total integral and imported into SIMCAP+ version 11.5. Multivariate data analysis was carried out in the same manner as described for the GC/MS data. The compounds in the bins that were found to be altered due to OTA treatment were identified in the original spectra using the spectral library of the Chenomx NMR Suite 4.6. Sample Treatment for LC/MS Analysis. Urine samples were thawed, centrifuged at 14000g and 4 °C for 10 min, and diluted to an equal osmolality of 154 mosmol/kg for adjustment of salt content (final volume, 150 µL; dilution with water). An amount of 10 µL of the adjusted samples was taken to prepare a pooled quality control sample (14). This sample was analyzed four times at the beginning of the batch and then after every 10 runs of study samples. Samples were randomized for all treatment and analytical steps. LC/MS Analysis. LC/MS analysis was run on an UPLC system coupled to a Micromass LCT Premier (ESI-TOF) controlled by MassLynx software version 4.1. The analytical column was an ACQUITY BEH C18 (dimensions: 2.1 mm × 100 mm, 1.7 µm; by Waters GmbH, Eschborn, Germany) and kept at 40 °C. An injection volume of 10 µL was used with a flow rate of 0.500 mL/ min. Solvents were water with 0.1% formic acid (solvent A) and acetonitrile with 0.1% formic acid (solvent B) with the following gradient: 0 min 100% A, 4.00 min 80% A, 9.00 min 5% A, and 12.00 min 100% A. Used were the following MS parameters:

Chem. Res. Toxicol., Vol. 22, No. 7, 2009 1223 negative ionization mode; survey scan mass range, 50-1000 Da; nebulization gas (N2), 700 L/h at 450 °C; cone gas, 15-20 L/h; source temperature, 120 °C; capillary voltage, 60 V; LCT-W optics mode with 12000 resolution using dynamic range extension (DRE); data acquisition rate, 0.1 s with a 0.01 s interscan delay; lock spray, 2000 counts; lock spray method, leucine/enkephalin 50 fmol/µL; lock mass, m/z ) 556.2771; flow rate, 30 µL/min; and frequency, 5 s; data collection was performed in centroid mode averaged over 10 scans. LC/MS Raw Data Handling and Statistical Analysis. Raw data files were exported in the platform-independent netCDF (*.cdf) format with the MassLynx export function for further analysis. Automatic peak detection and peak alignment were performed by XCMS. R-program version 2.6.2 (R-Foundation for statistical computing, www.R-project.org) and XCMS version 1.11.20 (15) were used. Used were the following XCMS parameters: xcmsSet function method ) centWave, chromfiles ppm ) 50, peakwidth c(3, 15); group function bw ) 2, minfrac ) 0.5, minsamp ) 1, mzwid ) 0.05, and max ) 50. The results table containing mass spectral features as mass/retention time pairs in a tab-separated text file (*.txt) was exported to SIMCA P+ version 11.5. Multivariate data analysis was carried out in the same manner as described for the GC/MS data. Mass traces of OTA metabolites were excluded prior to any multivariate data processing steps. Identity proposals for regulated metabolites were made with the METLIN Metabolite Database (15) using the links supplied by XCMS in the results table.

Results Analytics and Statistical Analysis. Visual inspection of GC/ MS chromatograms and 1H NMR spectra showed differences between control and high-dose group animals (Figure 1a,b). However, the LC/MS TIC chromatograms of control and highdose group animals appeared almost indistinguishable as compared to the marked changes found with both GC/MS and 1 H NMR (Figure 1c). PCA models constructed with GC/MS and 1H NMR data, respectively, separated control animals from high-dose animals (210 µg/kg body wt) after 4 weeks of OTA treatment along the first and second principal component (PC) t[1] and t[2] (Figure 2a,b). GC/MS also showed a trend to separate animals of the mid-dose group (70 µg/kg body wt) from controls from the 4 week sampling point onward. However, the plot was dominated by a large variance, presumably introduced by the necessary sample workup (Figure 2a). With 1H NMR, individual animals from the high-dose group separated from controls as early as 2 weeks after the start of treatment (Figure 2b). The LC/MS model discriminated only sampling time points along t[1] and t[2] (Figure 2c), and dose-dependent information was only evident on the fourth PC (data not shown). To identify discriminating markers between control and dose groups, OPLS-DA models were constructed. In contrast to the unsupervised PCA models, where discrimination originates only from the projection of variance, the class identity in OPLS-DA is given in a Y matrix to which the spectral data are then correlated. Information discriminating between the classes is forced into the first PC, while the subsequent components contain orthogonal information, that is, information not contributing to class separation. In the OPLS-DA models, the control and low-dose animals (vehicle and 21 µg OTA/kg body wt) were combined into a new control group as 21 µg/kg body wt was established as the NOEL with regard to the classical clinical chemistry and histopathological end points (13). They were analyzed against the combined mid- and high-dose animals (given 70 and 210 µg OTA/kg body wt) as a new dose group. A Y matrix containing the values Y ) 0 for the combined control group and Y ) 1 for the combined dose group was used. This approach was chosen since model quality improves with larger

1224

Chem. Res. Toxicol., Vol. 22, No. 7, 2009

Figure 1. GC/MS chromatograms (a), 1H NMR spectra (b), and negative ion mode LC/MS chromatograms (c) of a representative control group and high-dose group animal, respectively, after 13 weeks of OTA administration showing subtle changes in urinary composition. The GC/ MS analytes are given as the derivatized compounds. MO, methoxime; and TMS, trimethylsilyl.

numbers of observations and equal group sizes (18). Mean centering and pareto scaling were applied to the data when constructing the models. This scaling procedure is a compromise between unit variance scaling, which tends to overestimate noise, and no scaling, which causes the model to be dominated by the few high-intensity signals in the chromatograms or 1H NMR spectra (18). Model characteristics of the dose-based OPLSDA models (Figure 4a-c) are summarized in Table 1. The OPLS-DA models constructed with data from all three analytical methods separate control groups from mid- and highdose groups in a dose-dependent manner along the discriminating component t[1] P. Control groups and high-dose groups are located on the outer edges of the plot, while the mid-dose group is situated in between. A time trend can be recognized in the dose groups. In the GC/MS and 1H NMR models, the highdose samples are clearly separated from the rest from the 4 week time point onward (Figure 4a,b). The LC/MS model separates high-dose samples already after 2 weeks of administration. The orthogonal component t[2] O, which models information that is not contributing to the group separation, contains information on the sampling time point. This can be observed especially in

Sieber et al.

Figure 2. PCA scores plots of GC/MS (a), 1H NMR (b), and LC/MS (c) data. Models include all time points, and variables are mean-centered and pareto-scaled. Model characteristics are as follows: (a) R2X(cum) ) 0.74, Q2(cum) ) 0.41, and seven significant PCs; (b) R2X(cum) ) 0.87, Q2(cum) ) 0.59, and nine significant PCs; and (c) R2X(cum) ) 0.68, Q2(cum) ) 0.38, and eight significant PCs.

Table 1. Characteristics for OPLS-DA Models Constructed with Either Dose or Histopathology Scores (Histo) as Classifying Y Matrixa GC/MS 1

H NMR

LC/MS

Y matrix

R2Y(t[1] P)

Q2(t[1] P)

significant components

dose histo dose histo dose histo

0.42 0.51 0.49 0.53 0.52 0.54

0.32 0.43 0.39 0.45 0.15 0.26

1+0 1+1 1+0 1+0 1+2 1+0

a The R2Y(t[1] P) parameter shows how much of the variance in the discriminating component can be corellated to the Y matrix. Q2(t[1] P) is the predictivity of the discriminating component. For each model, the first, discriminating component is significant.

the 1H NMR model (Figure 4b), where the 13 week samples are located in the upper region of the plot. To further enhance the power of the multivariate statistics models, an alternative data analysis approach was developed in which information regarding the severity of histopathological changes caused by OTA treatment, that is, histopathology scores, was included. The histopathology scores of all animals at all time points and all findings examined were used as an X matrix

Metabonomic Study of Ochratoxin A

Figure 3. Scores plot (a) and the corresponding loadings plot (b) of a PCA model using histopathology scores as X matrix variables. The dose- and time-dependent separation of dose group animals from control group animals is best reflected by the finding of kidney tubular basophilia; thus, this parameter was chosen as the Y matrix for OPLSDA models. K, kidney findings; and L, liver findings.

to construct a separate PCA model (Figure 3). The resulting scores plot shows a time-dependent separation along the second PC t[2] and a dose-dependent separation of OTA-dosed animals from controls along the first two PCs t[1] and t[2] from the top right to the bottom left side of the plot. In the loadings plot, five kidney findings, that is, karyomegaly, interstitial fibrosis, tubular vacuolation, tubular basophilia, and tubular cell apoptosis, are responsible for the dose-dependent separation of samples in the scores plot. From these five findings, correlating with the administration of OTA, the tubular basophilia histopathology score was selected as the Y matrix for the construction of OPLS-DA models as it shows the strongest correlation with dose and was observed across the highest number of treated animals. Model characteristics of the OPLS-DA models constructed with histopatholgy scores as discriminators (Figure 4d-f) are compared to those of models constructed with dose in Table 1. Construction of supervised OPLS-DA models with histopathology scores as classifier (Figure 4 d-f) yielded three models with similar qualities to those constructed with dose as a classifier (Figure 4a-c). A clear dose-dependent separation along the discriminating component t[1] P can be observed with all three analytical methods (Figure 4d-f) with the control and high-dose groups on either side of the plot and the mid-dose group clustering in between. Even though the models do not differ substantially upon visual inspection, the histopathology-based OPLS models (Figure 4d-f) better represent the data structure as described by the model characteristics in Table 1. With all three analytical models, the variance modeled by the first component R2Y(t[1] P) increased when using histopathology scores instead of dose as classifiers. This parameter is indicative of how much of the information of the data that is modeled into the discriminating

Chem. Res. Toxicol., Vol. 22, No. 7, 2009 1225

component t[1] P of the model can be correlated with the discriminating Y matrix. The predictive power in 7-fold crossvalidation Q2(t[1] P) of the discriminating component also increased with all three analytical methods when using histopathology scores instead of dose as classifiers. In general, it can be stated that model quality improved when histopathology scores were included as classifiers. It needs to be recognized that the two models are not completely independent of each other; as for the “dose” model, the NOEL information was included, and in the “histopathology” model, the histopathological findings, which correlate best with dose, were used as classifiers. The point is not to classify the samples either according to dose or according to histopathology score but rather to improve the confidence in the putative metabolite markers discovered. Because the response to toxicants may be variable in the treated groups (even though doses are adjusted to body weight etc.), a correlation with the histopathology observed in the individual animal is reasonable, considering that toxicity assessment is usually based on histopathology. The S-plots of the OPLS-DA models were used for identification of potential markers of group separation as proposed by Wiklund et al. (16). In the S-plot, the range of the variables selected is highlighted with a dotted rectangle. Figure 5 illustrates the procedure with GC/MS data as an example. The variables responsible for separation are presented in Tables 2-4. For all three analytical methods, there was substantial overlap between the variables found with either dose or histopathology scores as the Y matrix. For marker identification, the original GC/MS chromatograms and 1H NMR spectra were used. For the identification of GC/ MS markers, AMDIS-deconvoluted peaks or peaks from the original HP ChemStation chromatograms were compared to the NIST mass spectral library and subsequently to chromatograms of authentic reference compounds, if available. 1H NMR peaks were identified by comparison to the spectral library of the Chenomx NMR Suite. LC/MS signals were compared to the contents of the METLIN Metabolite Database using the queries supplied by the XCMS program (15). However, because of the limited size of the METLIN Metabolite Database and the large mass tolerance used in the queries, the identities given for LC/ MS-based markers require further elucidation and should be considered only as tentative assignments. Identified Novel Markers and Biochemistry. The markers identified were found by analyzing either the dose-based (0 and 21 vs 70 and 210 µg/kg body wt) or histopathology-based OPLS-DA models for each analytical technique. There was substantial overlap for both analytical strategies. GC/MS analysis showed that excretion of the following metabolites was increased in animals given OTA and exhibiting histopathological changes in the kidney (Table 2): 5-oxoproline (2- and 5-fold increases, respectively, in high-dose samples 4 and 13 weeks after the start of treatment), D-glucose (5-fold increase in high-dose samples after 4 and 13 weeks), myo-inositol (1.5-fold increase in all treated animals after 2 weeks and 2- and 3-fold increases in high-dose animals at 4 and 13 weeks), and pseudouridine (1.3-fold increase at 4 and 13 weeks). 2-Oxoglutarate (4-fold decrease in the 4 weeks high-dose samples and a 2- and 20fold decrease in the 13 weeks mid-dose and high-dose samples, respectively) and citrate (2-fold decrease in the 4 weeks midand high-dose samples and 3-fold decrease in the 13 weeks midand high-dose samples) were decreased. The histopathologybased model also indicated an increased creatinine excretion after 13 weeks of treatment in the high-dose group.

1226

Chem. Res. Toxicol., Vol. 22, No. 7, 2009

Sieber et al.

Figure 4. Scores plots of OPLS-DA models. As discriminating Y matrix, either the dose (control and low dose animals vs mid- and high-dose animals, plots a-c) or the histopathology scores of kidney tubular basophilia (plots d-f) were used. 1

H NMR analysis showed increases in the excretion of N,Ndimethylglycine (2-, 3-, and 8-fold increases, respectively, in high-dose samples at 2, 4, and 13 weeks), lactate (4-fold increase in high-dose samples after 13 weeks), and glycine (5- and 8-fold increases, respectively, after 4 and 13 weeks in high-dose samples); excretion of 2-oxoglutarate (2- and 5-fold decreases, respectively, in mid- and high-dose samples at 4 weeks, 7-fold decrease in high-dose samples at 13 weeks) and citrate (2- and 5-fold decreases, respectively, in 4 weeks mid- and high-dose samples) was decreased. A number of bins for which no metabolite assignment could be made were also changed (Tables 3 and 4). LC/MS analysis showed a number of variables to be changed due to OTA treatment. On the basis of the spectral information, these metabolites may represent citrate and 2-oxoglutarate and aldose phosphates and steroid hormones. However, because of insufficient information obtained from the spectra, further structure elucidation was not possible (Table 5).

Discussion Analytical Chemistry and Statistical Analysis. In this study, three complementary analytical techniques were used to correlate metabonomic findings with histopathology end points from a 90 day toxicity study with OTA. In the study, histopathologic analysis gave a NOEL of 21 µg OTA/kg body wt, and minimal to mild changes in histopathology and clinical chemistry at doses

of 70 µg/kg body wt, visible as early as 4 weeks after continuous OTA administration and minimal to mild changes in histopathology as early as 2 weeks at doses of 210 µg OTA/kg body wt (13). Applying metabonomic analysis, alterations in the composition of urine samples were indicated in urine samples collected after 4 weeks in the 70 and 210 µg/kg body wt dose groups and after 2 weeks in individual animals of the 210 µg/ kg body wt dose group with both GC/MS and 1H NMR using unsupervised PCA and with all three analytical approaches using supervised OPLS-DA metabonomics. Predictivity of the OPLS-DA models increased when including histopathology scores as Y matrix, while yielding the same metabolites (decreased urinary citrate, oxoglutarate; increased urinary glucose, pseudouridine, 5-oxoproline, myo-inositol, alanine, glycine, and N,N-dimethylglycine) to be responsible for separating controls from dosed animals. This approach allows the inclusion of the information gained by histopathology, which is otherwise not used for metabonomic analysis. Moreover, it helps to overcome the problem of heterogeneous response of dosed animals to toxic insults, which is often observed in toxicity studies. In our studies, both GC/MS and 1H NMR were able to differentiate between controls and treated groups and complemented each other with regard to marker identification due to the availability of databases and regarding sensitivity and metabolite coverage. However, GC/MS analysis may be further

Metabonomic Study of Ochratoxin A

Chem. Res. Toxicol., Vol. 22, No. 7, 2009 1227

Figure 5. S-plot (a) and column plot of extracted variables with jackknifed confidence intervals (b) of an GC/MS data-based OPLS-DA model to illustrate the process of marker identification. The S-plot shows the covariance p against the correlation p(corr) of the variables of the discriminating component of the OPLS-DA model. Cut-off values for the covariance of p g |0.05| and for the correlation of p(corr) g |0.5| were used; the variables thus selected are highlighted in the S-plots with dotted rectangles (a). In order not to overinterpret the model, the markers were selected in a conservative manner by investigating only those variables showing a jack-knifed confidence interval less than half of the variable’s value (b).

improved with the use of sampling robots to minimize variance introduced by the derivatization procedure (combined variance of the sample workup and analysis procedure for this study ranged around 15% depending on the intensity of the speak as determined from quality controls) and the introduction of twodimensional GC coupled to time-of-flight detectors (GC × GC/ TOF-MS) (19, 20). LC/MS analysis contained more noise, and identification of putative markers requires substantial effort due to lack of characteristic mass spectral features and spectral databases. Thus, LC/MS appears to be more suited for targeted analysis of specific compound classes. Advantages and disadvantages of all three analytical methods for metabonomics have been reviewed recently (21), and a detailed discussion goes beyond the scope of this study. Interestingly, multivariate models of both 1H NMR and GC/ MS suggested an increase in urinary creatinine excretion (Tables

2 and 3) in OTA-dosed animals; urinary creatinine was found to be increased with GC/MS in the 13 weeks high-dose samples upon normalization to total integral, and the bin containing the creatinine signal (3.96-4.00 ppm) in 1H NMR spectra normalized to total integral was increased in 4 weeks high-dose samples and 13 weeks mid- and high-dose samples. Urinary volume, however, was not significantly changed between groups (Supporting Information, Table 1), and urinary clinical chemistry analysis did also not show any significant changes in 20 h creatinine excretion rates (13). How these findings can be explained remains unclear, and in light of these findings, we report the excretion levels of the urinary metabolites in Table 4 as 20 h excretion rates. Metabolic Changes and Relation to OTA Toxicity. The combined GC/MS- and 1H NMR-based metabonomic analysis of urine samples from rats administered OTA showed a decrease in the urinary excretion of the Krebs cycle intermediates citrate and oxoglutarate and an increased excretion of glucose, creatinine, pseudouridine, 5-oxoproline, myo-inositol, alanine, glycine, and N,N-dimethylglycine (Tables 2-4). These changes may be rationalized based on the welldescribed OTA toxicity and general considerations on kidney toxicity. In histopathology, OTA toxicity is characterized by damage to the S3 segment of the proximal tubule of the kidney with accompanying impairment of renal function, resulting in alteration of amino acid and osmolyte excretion as an early toxic effect (13, 22, 23). These changes are reflected by a significant increase of the excretion of the amino acids alanine and glycine (2- and 8-fold, respectively, after 13 weeks of treatment, Table 4) as well as in the excretion of the most abundant renal osmolyte myo-inositol (24) (1.5-fold, Table 2) in the 210 µg/ kg body wt group as early as 2 weeks after the start of treatment. Whether the altered osmolyte excretion and aminoaciduria is due to impaired excretory function or simply due to damaged, leaky cells remains unclear; glucosuria and aminoaciduria have been observed for a number of tubular toxins (25, 26). After application of OTA at higher doses and a shorter period of observation, Mally et al. observed increased excretion of the renal osmolyte trimethylamine-N-oxide (TMAO) as the dominating molecular marker by 1H NMR metabonomics (27). GC/ MS analysis of these samples also yielded a 9-fold increase of myo-inositol excretion in dosed animals (unpublished results). Because there was no alteration of TMAO excretion observed in this study, myo-inositol seems to be a more sensitive indicator of disturbed renal osmolyte handling in OTA toxicity than TMAO.

Table 2. Main Fragments and Chromatographic Retention Time (RT) of Metabolic Changes Induced by OTA Administration, Identified Using GC/MSa main fragments (m/z)

RT (s)

vVb

fcc

identification

156, 157, 258 205, 217 100, 115, 143, 329 292 277, 278 103, 117, 133, 205 191, 204 305 217, 218, 357 112, 156, 229, 304 273, 347, 363, 465

1046 1049 1075 1082 1207 1354 1412 1472 1619 1113 1303

v v v v v v v v v V V

5.3** 1.2* 1.5 1.3* 1.7* 4.2*** 2.0 3.0*** 1.3* 16.7*** 2.6**

5-oxoproline 2TMS C4-polyol 4TMS (erythritol 4TMS)d creatinine enol 3TMSe 2,3,4-trihydroxybutyric acid 4TMS (5-hydroxy-1-H-indole 2TMS)d,f D-glucose MO5TMS D-glucose 5TMS myo-inositol 6TMS pseudouridine 5TMS 2-oxoglutarate MO2TMS citrate 4TMS

a Metabolites are given as the actual analyte, that is, the MO and TMS derivate. Metabolite identification was carried out by comparison to the NIST mass spectral database. b v indicates increased excretion in exposed groups as compared to the control group, and V indicates decreased excretion in exposed groups. c Fold change in the high-dose group after 13 weeks. Statistical significance levels were determined by ANOVA and Dunnett’s posthoc: *p < 0.05, **p < 0.01, and ***p < 0.001. d Metabolites in parentheses are only putatively assigned. e Observed in the histopathology-based model only. f Observed in the dose-based model only.

1228

Chem. Res. Toxicol., Vol. 22, No. 7, 2009

Sieber et al.

Table 3. Bins of Dose-Based and Histophathplogy-Based (Histo) Models and Corresponding Chemical Shifts Altered upon OTA Administration, Identified Using 1H NMRa bin dose 20 23 26 39 43 50, 51 53 56, 57 62 64 74 76 78 79 81-88 89

bin histo 20 22 23 39 43 50, 51 53 56, 57 62 64 69 73 74 76 78 79 81-88

∂ (ppm)

vVb

1.20-1.24 1.32-1.36 1.36-1.40 1.44-1.48 1.96-2.00 2.12-2.16 2.40-2.48 2.52-2.56 2.64-2.72 2.88-2.92 2.96-3.00 3.16-3.20 3.32-3.36 3.36-3.40 3.44-3.48 3.52-3.56 3.56-3.60 3.64-3.94 3.96-4.00

v v V v V V V V V v V v v v v v v v v

identification (lactate)c (alanine)c acetate 2-oxoglutarate citrate citrate N,N-dimethylglycine 2-oxoglutarate (glucose)c (glucose)c (glucose)c taurine (glucose)c creatinine

a Metabolite identification was carried out by comparison to the Chenomx NMR Suite 4.6 spectral database. b v indicates increased excretion in exposed groups as compared to the control group, and V indicates decreased excretion in exposed groups. c Signals were not strong enough for unambiguous identification and are only putatively asigned.

Increased excretion of glucose is associated with an impairment of renal function. Although no significant changes in urinary glucose excretion were found by clinical chemistry analysis of the study, an impairment of kidney function is indicated by an increase in serum creatinine after 13 weeks was observed in the 210 µg/kg body wt group (13). 1H NMR as well as GC/MS analysis both indicated increased glucose as the discriminating feature between controls and high-dose animals in OPLS-DA models, and a statistically significant 5-fold increase in urinary glucose levels is observed with GC/

MS analysis already 4 weeks after the start of treatment. Such changes in the excretion levels of a number of metabolites contribute to group separation in multivariate data analysis, even if these alterations are not statistically significant when considered isolated for each single metabolite. Thus, multivariate data analysis may be more sensitive than classical clinical chemistry regarding small and variable changes of effect markers, if a number of these effect markers are changed simultaneously. On a cellular level, impairment of mitochondrial function, oxidative stress, and inhibition of protein synthesis have been described as toxic effects of OTA (28-31). These effects may be linked to altered levels of Krebs cycle intermediates, increased 5-oxoproline excretion, and altered amino acid excretion as described below. A decrease in citrate and 2-oxoglutarate excretion together with an increased lactate excretion as observed with 1H NMR (Tables 3 and 4) is widely attributed to alterations in cellular energy production, which may be linked to impairment of mitochondrial function. An impaired energy metabolism results in a depletion of ATP in the cells, and ATP depletion as a consequence of mitochondrial toxicity is a prominent feature of OTA-induced renal toxicity in rodents (28). Klawitter et al. also discussed transporters of Krebs cycle intermediates, lactate and glucose, located in the tubuli to be involved in the altered metabolite excretion (32). However, impairment of the Krebs cycle is described in many metabonomic studies (1) and may rather be an effect of stress response and general toxicity than of specific kidney toxicity or OTA effects, and the mechanism remains unclear. 5-Oxoproline is involved in the γ-glutamyl-cycle, which is responsible for glutathione synthesis (33), and 5-oxoprolinuria in rats is induced by several agents, which induce glutathione depletion (34, 35). The observation of increased secretion of 5-oxoproline is consistent with a reduction of detoxifying capacity via inhibition of Nrf2-dependent gene expression and reduced glutathione concentrations observed after OTA treat-

Table 4. Twenty Hour Excretion Rates and Relative Changes of Significantly Altered Urinary Metabolites as Determined with the Chenomx NMR Databasea dose µg OTA/kg bw

week 2

week 4

week 13

week 2

week 4

week 13

a

citrate V

0 21 70 210 0 21 70 210 0 21 70 210

107.7 ( 26.7 103.8 ( 69.7 95.1 ( 51.3 90.0 ( 38.1 101.8 ( 19.1 87.6 ( 36.4 54.7 ( 16.1* 33.8 ( 17.9** 77.6 ( 36.8 101.2 ( 25.2 66.0 ( 39.9 29.1 ( 8.8

0 21 70 210 0 21 70 210 0 21 70 210

1.00 ( 0.25 0.96 ( 0.65 0.88 ( 0.48 0.84 ( 0.35 1.00 ( 0.19 0.86 ( 0.31 0.54 ( 0.15* 0.33 ( 0.18** 1.00 ( 0.47 1.30 ( 0.32 0.85 ( 0.51 0.38 ( 0.11

2-oxo-glutarate V

N,N-dimethylglycine v

alanine v

glycine v

20 h excretion rates (µmol/20 h) 71.7 ( 24.1 4.70 ( 1.81 48.3 ( 30.4 4.62 ( 1.09 53.1 ( 29.8 4.08 ( 1.43 43.2 ( 17.5 5.48 ( 1.20 75.4 ( 10.2 4.09 ( 1.13 63.2 ( 25.9 3.94 ( 2.44 41.5 ( 11.6* 4.18 ( 1.26 13.6 ( 6.5*** 5.44 ( 2.24 56.0 ( 26.7 2.81 ( 0.95 62.5 ( 8.65 4.08 ( 2.23 31.1 ( 25.0 4.35 ( 1.53 8.02 ( 2.08* 10.54 ( 1.36***

lactate v

1.46 ( 0.78 0.97 ( 0.42 1.51 ( 0.68 3.22 ( 1.59* 1.24 ( 0.31 1.20 ( 0.83 1.79 ( 0.59 3.65 ( 1.64** 0.54 ( 0.20 0.63 ( 0.24 1.32 ( 0.35 4.21 ( 0.89***

4.17 ( 1.42 4.11 ( 1.17 3.86 ( 1.18 5.16 ( 1.33 3.89 ( 0.90 3.48 ( 1.88 3.80 ( 1.32 4.12 ( 1.46 2.47 ( 0.65 4.06 ( 1.48 3.86 ( 1.78 4.87 ( 1.95

9.40 ( 3.03 8.72 ( 2.31 8.84 ( 2.19 13.55 ( 3.67 8.02 ( 2.14 9.32 ( 3.84 9.43 ( 1.68 36.9 ( 15.2*** 6.39 ( 2.25 7.42 ( 2.46 11.60 ( 3.41 49.02 ( 6.56***

fold 1.00 ( 0.34 0.67 ( 0.43 0.74 ( 0.42 0.60 ( 0.24 1.00 ( 0.13 0.84 ( 0.31 0.55 ( 0.14* 0.18 ( 0.09*** 1.00 ( 0.48 1.12 ( 0.15 0.56 ( 0.45 0.14 ( 0.04*

1.00 ( 0.53 0.67 ( 0.29 1.04 ( 1.08 2.20 ( 1.08* 1.00 ( 0.25 0.97 ( 0.64 1.45 ( 0.66 2.95 ( 1.33** 1.00 ( 0.38 1.17 ( 0.44 2.43 ( 0.64 7.78 ( 1.64***

1.00 ( 0.34 0.99 ( 0.28 0.93 ( 0.28 1.24 ( 0.32* 1.00 ( 0.23 0.89 ( 0.41 0.98 ( 0.42 1.06 ( 0.37 1.00 ( 0.26 1.64 ( 0.60 1.56 ( 0.72 1.97 ( 0.79

1.00 ( 0.32 0.93 ( 0.25 0.94 ( 0.25 1.44 ( 0.39 1.00 ( 0.27 1.16 ( 0.31 1.18 ( 0.37 4.59 ( 1.89*** 1.00 ( 0.35 1.16 ( 0.38 1.82 ( 0.53 7.67 ( 1.03***

changes 1.00 ( 0.38 0.98 ( 0.23 0.87 ( 0.30 1.17 ( 0.25 1.00 ( 0.28 0.96 ( 0.50 1.02 ( 0.45 1.33 ( 0.55 1.00 ( 0.34 1.45 ( 0.80 1.55 ( 0.55 3.76 ( 0.48***

v indicates increased excretion in exposed groups as compared to the control group, and V indicates decreased excretion in exposed groups. Statistical significance levels are determined by ANOVA and Dunnett’s posthoc: *p < 0.05, **p < 0.01, and ***p < 0.001.

Metabonomic Study of Ochratoxin A

Chem. Res. Toxicol., Vol. 22, No. 7, 2009 1229

Table 5. Mass Retention Time Pairs of Dose-Based and Histopathology-Based (Histo) Models Altered upon OTA Administration Identified Using LC/MSa variable ID dose [M (m/z) T (RT)]

variable ID histo [M (m/z) T (RT)]

vVb

fcc

M145T40 M145T46

M145T40 M145T64 M181T153 M191T51 M191T62 M192T51 M201T303 M206T260 M217T281 M231T141 M231T79

V V v V V V V v V v V v v V v v v V V V V v V V V v v V v V V V V V V V V v V v v v v v v v v v

20*** 14*** 14*** 1.7* 2.0** 1.7* 1.4 38*** 1.4 2.33*** 2.1*** 1.1 4.0* 8.3*** 56*** 10*** 23*** 7.1*** 17*** 14* 10* 100*** 20*** 1.3** 1.3* 3.9*** 2.9*** 1.1 11** 10*** 4.8*** 17*** 1.5 17*** 6.7*** 43*** 1.1 19*** 3.6** 66*** 758*** 201*** 2.14 8.8*** 2.4*** 11* 10* 1.6**

M191T51 M192T51 M206T260 M231T141 M231T79 M232T143 M237T339 M244T117 M251T32 M259T306 M259T325 M260T325 M261T113

M287T367 M301T309 M302T131 M303T248 M303T260 M317T253 M318T253 M350T135 M372T111 M411T194

M244T117 M249T234 M251T32 M252T260 M259T102 M259T325 M260T325 M260T90 M261T113 M261T138 M262T138 M269T344 M273T283 M275T174

M303T260 M317T253 M318T253 M323T284 M337T313 M350T135 M409T372 M411T194 M412T194

M417T299 M445T248 M461T167 M473T318

M473T318 M474T318 M523T138

proposed structured 2-oxoglutarate, adipate, lysine 2-oxoglutarate, adipate, lysine citrate citrate (isotope peak) acetylphenylalanine,phenylpropinylglycine melatonine aldose phosphates aldose phosphates aldose phosphates aldose phosphates aldose phosphates (isotope peak)

steroid hormone steroid hormone steroid hormone

a Mass traces of OTA and metabolites were excluded in the analysis. Metabolites are only putatively assigned. b v indicates increased excretion in exposed groups as compared to the control group, and V indicates decreased excretion in exposed groups. c Fold change in the high-dose group after 13 weeks. Statistical significance levels were determined by ANOVA and Dunnett’s posthoc: *p < 0.05, **p < 0.01, and ***p < 0.001. d Identity proposals are made by the METLIN Metabolite Database using the linking function in XCMS.

ment (36). The increase of N,N-dimethylglycine observed with 1 H NMR (3-fold in the 210 µg/kg body wt group after 4 weeks of treatment, Table 4) may also be related to oxidative stress and increased glutathione synthesis. There is evidence that the glycine needed for GSH synthesis may also be supplied by dietary choline, which is transferred to glycine via betaine, dimethylglycine, and sarcosine (37). Induction of this pathway may explain the observed increase in N,N-dimethylglycine excretion. Increased urinary pseudouridine excretion (1.2-fold starting in the 210 µg/kg body wt dose group at week 2) as identified by the GC/MS analysis may be related to stimulation of cell proliferation in the kidney induced by OTA as observed by BrdU staining (13). Pseudouridine is a modified nucleoside in sRNA and tRNA. When RNA is degraded to nucleosides and bases, pseudouridine cannot be reused for de novo nucleotide synthesis and is excreted into the urine; pseudouridine excretion with urine

is indicative of RNA turnover and has been postulated as a cancer marker (38). An increase in RNA turnover is consistent with increased DNA synthesis as a result of the stimulation of renal cell proliferation observed in the study (13). Although LC/ MS allows the measurement of metabolite classes not accessible by GC/MS or 1H NMR, these alterations are not discussed further due to the putative metabolite assignments for LC/MS data. Although the biochemical perturbations observed in this study reflect rather general alterations and none of the metabolites observed in this study is predictive on its own, they correlate well with observations made in other studies on OTA toxicity. With metabonomics, the onset of toxic lesions could be observed at the same time points as with histopathology. The advantage of the omics approach is the wealth of information such as the indications of disturbed amino acid and osmolyte excretion, increased GSH production, and cell proliferation obtained by

1230

Chem. Res. Toxicol., Vol. 22, No. 7, 2009

one global analysis, which otherwise had to be drawn from many different specific measurements. Although the biochemical reasons for the observed perturbances described here are speculative and may also have other causes and therefore require confirmation by targeted experiments, such findings could be used effectively for building hypotheses on mode-of-action and expected toxicities, especially for compounds where a broad knowledge base such as for OTA does not exist.

Conclusion The integrated multiplatform metabonomics approach with GC/MS, 1H NMR, and LC/MS shows promise for the investigation of changes in urinary metabolite pattern, since good correlation exists between metabonomic group discrimination and histopathological findings, especially since the histopathological changes were rather subtle (13). The metabolites identified as markers are more a pattern of “usual suspects” (1) than specific markers but correlate with the reported observations made on OTA toxicity. The data presented here show that an inclusion of histopathology scores in a metabonomic analysis improves model quality and allows greater confidence in discriminating markers found by metabonomics. The data analysis strategy presented here can be used for the analysis of existing toxicology studies to draw more information and to gain mechanistic insight of observed toxicities. Acknowledgment. We thank Dr. Matthias Gru¨ne for the recording of the NMR spectra, Nataly Bittner for LC/MS maintenance, and Ursula Tatsch for sample preparation. Parts of the author’s work were supported by RCC Ltd. (Itingen, Switzerland). All analytical work was performed using equipment obtained through grants from the Deutsche Forschungsgemeinschaft and the State of Bavaria. Supporting Information Available: Body and kidney weight and clinical chemistry parameters after repeated administration of OTA to male F344 rats (Table 1) and summary of histopathological observations in kidneys of male F344/N rats treated with OTA (Table 2). This material is available free of charge via the Internet at http://pubs.acs.org.

Sieber et al.

(10) (11) (12) (13)

(14)

(15)

(16)

(17) (18) (19)

(20) (21) (22)

(23) (24) (25)

References (1) Robertson, D. G. (2005) Metabonomics in toxicology: A review. Toxicol. Sci. 85, 809–822. (2) Coen, M., Holmes, E., Lindon, J. C., and Nicholson, J. K. (2008) NMRbased metabolic profiling and metabonomic approaches to problems in molecular toxicology. Chem. Res. Toxicol. 21, 9–27. (3) Lu, X., Zhao, X., Bai, C., Zhao, C., Lu, G., and Xu, G. (2008) LCMS-based metabonomics analysis. J. Chromatogr. B 866, 64–76. (4) Wilson, I. D., Nicholson, J. K., Castro-Perez, J., Granger, J. H., Johnson, K. A., Smith, B. W., and Plumb, R. S. (2005) High resolution “ultra performance” liquid chromatography coupled to oa-TOF mass spectrometry as a tool for differential metabolic pathway profiling in functional genomic studies. J. Proteome Res. 4, 591–598. (5) Idborg, H., Zamani, L., Edlund, P.-O., Schuppe-Koistinen, I., and Jacobsson, S. P. (2005) Metabolic fingerprinting of rat urine by LC/ MS. Part 1. Analysis by hydrophilic interaction liquid chromatographyelectrospray ionization mass spectrometry. J. Chromatogr. B 828, 9– 13. (6) Gates, S. C., and Sweeley, C. C. (1978) Quantitative metabolic profiling based on gas chromatography. Clin. Chem. 24, 1663–1673. (7) Kopka, J. (2006) Current challenges and developments in GC-MS based metabolite profiling technology. J. Biotechnol. 124, 312–322. (8) Halket, J. M., Waterman, D., Przyborowska, A. M., Patel, R. K. P., Fraser, P. D., and Bramley, P. M. (2005) Chemical derivatization and mass spectral libraries in metabolic profiling by GC/MS and LC/MS/ MS. J. Exp. Bot. 56, 219–243. (9) Ni, Y., Su, M., Qiu, Y., Chen, M., Liu, Y., Zhao, A., and Jia, W. (2007) Metabolic profiling using combined GC-MS and LC-MS

(26)

(27)

(28)

(29)

(30)

(31) (32)

provides a systems understanding of aristolochic acid-induced nephrotoxicity in rat. FEBS Lett. 581, 707–711. Zhang, Q., Wang, G., Du, Y., Zhu, L., and A, J. (2007) GC/MS analysis of the rat urine for metabonomic research. J. Chromatogr. B 854, 20–25. Trygg, J., Holmes, E., and Lundstedt, T. (2007) Chemometrics in metabonomics. J. Proteome Res. 6, 469–479. Trygg, J., and Wold, S. (2002) Orthogonal projections to latent structures (O-PLS). J. Chemom. 16, 119–128. Rached, E., Hard Gordon, C., Blumbach, K., Weber, K., Draheim, R., Lutz Werner, K., Ozden, S., Steger, U., Dekant, W., and Mally, A. (2007) Ochratoxin A: 13-week oral toxicity and cell proliferation in male f344/n rats. Toxicol. Sci. 97, 288–298. Sangster, T., Major, H., Plumb, R., Wilson, A. J., and Wilson, I. D. (2006) A pragmatic and readily implemented quality control strategy for HPLC-MS and GC-MS-based metabonomic analysis. Analyst 131, 1075–1078. Smith, C. A., Want, E. J., O’Maille, G., Abagyan, R., and Siuzdak, G. (2006) XCMS: Processing mass spectrometry data for metabolite profiling using nonlinear peak alignment, matching, and identification. Anal. Chem. 78, 779–787. Wiklund, S., Johansson, E., Sjoestroem, L., Mellerowicz, E. J., Edlund, U., Shockcor, J. P., Gottfries, J., Moritz, T., and Trygg, J. (2007) Visualization of GC/TOF-MS-based metabolomics data for identification of biochemically interesting compounds using OPLS class models. Anal. Chem. 80, 116–123. Stein, S. E. (1999) An integrated method for spectrum extraction and compound identification from gas chromatography/mass spectrometry data. J. Am. Soc. Mass Spectrom. 10, 770–781. Eriksson, L., Johannson, E., Kettaneh-Wold, N., Trygg, J., Wilkstrom, C., and Wold, S. (2006) MultiVariate and MegaVariate Data Analysis: Basic Principles and Applications (Part I), Umetrics. Pasikanti, K. K., Ho, P. C., and Chan, E. C. Y. (2008) Development and validation of a gas chromatography/mass spectrometry metabonomic platform for the global profiling of urinary metabolites. Rapid Commun. Mass Spectrom. 22, 2984–2992. Pierce, K. M., Hoggard, J. C., Mohler, R. E., and Synovec, R. E. (2008) Recent advancements in comprehensive two-dimensional separations with chemometrics. J. Chromatogr. A 1184, 341–352. Lenz, E. M., and Wilson, I. D. (2007) Analytical strategies in metabonomics. J. Proteome Res. 6, 443–458. Mally, A., Keim-Heusler, H., Amberg, A., Kurz, M., Zepnik, H., Mantle, P., Voelkel, W., Hard, G. C., and Dekant, W. (2005) Biotransformation and nephrotoxicity of ochratoxin B in rats. Toxicol. Appl. Pharmacol. 206, 43–53. Boorman, G. A., McDonald, M. R., Imoto, S., and Persing, R. (1992) Renal lesions induced by ochratoxin A exposure in the F344 rat. Toxicol. Pathol. 20, 236–245. Nakanishi, T., Balaban, R. S., and Burg, M. B. (1988) Survey of osmolytes in renal cell lines. Am. J. Physiol. 255, C181-191. Xu, E. Y., Perlina, A., Vu, H., Troth, S. P., Brennan, R. J., Aslamkhan, A. G., and Xu, Q. (2008) Integrated pathway analysis of rat urine metabolic profiles and kidney transcriptomic profiles to elucidate the systems toxicology of model nephrotoxicants. Chem. Res. Toxicol. 21, 1548–1561. Lenz, E. M., Bright, J., Knight, R., Wilson, I. D., and Major, H. (2004) A metabonomic investigation of the biochemical effects of mercuric chloride in the rat using 1H NMR and HPLC-TOF/MS: Time dependant changes in the urinary profile of endogenous metabolites as a result of nephrotoxicity. Analyst 129, 535–541. Mally, A., Voelkel, W., Amberg, A., Kurz, M., Wanek, P., Eder, E., Hard, G., and Dekant, W. (2005) Functional, biochemical, and pathological effects of repeated oral administration of ochratoxin A to rats. Chem. Res. Toxicol. 18, 1242–1252. Aleo, M. D., Wyatt, R. D., and Schnellmann, R. G. (1991) Mitochondrial dysfunction is an early event in ochratoxin A but not oosporein toxicity to rat renal proximal tubules. Toxicol. Appl. Pharmacol. 107, 73–80. Gautier, J. C., Holzhaeuser, D., Markovic, J., Gremaud, E., Schilter, B., and Turesky, R. J. (2001) Oxidative damage and stress response from ochratoxin A exposure in rats. Free Radical Biol. Med. 30, 1089– 1098. Hoehler, D., Marquardt, R. R., McIntosh, A. R., and Hatch, G. M. (1997) Induction of free radicals in hepatocytes, mitochondria and microsomes of rats by ochratoxin A and its analogs. Biochim. Biophys. Acta 1357, 225–233. Rached, E., Pfeiffer, E., Dekant, W., and Mally, A. (2006) Ochratoxin A: Apoptosis and aberrant exit from mitosis due to perturbation of microtubule dynamics? Toxicol. Sci. 92, 78–86. Klawitter, J., Bendrick-Peart, J., Rudolph, B., Beckey, V., Klawitter, J., Haschke, M., Rivard, C., Chan, L., Leibfritz, D., Christians, U., and Schmitz, V. (2009) Urine metabolites reflect time-dependent

Metabonomic Study of Ochratoxin A

(33)

(34)

(35)

(36)

effects of cyclosporine and sirolimus on rat kidney function. Chem. Res. Toxicol. 22, 118–128. Skordi, E., Yap, I. K. S., Claus, S. P., Martin, F.-P. J., Cloarec, O., Lindberg, J., Schuppe-Koistinen, I., Holmes, E., and Nicholson, J. K. (2007) Analysis of time-related metabolic fluctuations induced by ethionine in the rat. J. Proteome Res. 6, 4572–4581. Ghauri, F. Y. K., McLean, A. E. M., Beales, D., Wilson, I. D., and Nicholson, J. K. (1993) Induction of 5-oxoprolinuria in the rat following chronic feeding with N-acetyl 4-aminophenol (paracetamol). Biochem. Pharmacol. 46, 953–957. Waters, N. J., Waterfield, C. J., Farrant, R. D., Holmes, E., and Nicholson, J. K. (2006) Integrated metabonomic analysis of bromobenzene-induced hepatotoxicity: Novel induction of 5-oxoprolinosis. J. Proteome Res. 5, 1448–1459. Cavin, C., Delatour, T., Marin-Kuan, M., Holzhaeuser, D., Higgins, L., Bezencon, C., Guignard, G., Junod, S., Richoz-Payot, J., Gremaud,

Chem. Res. Toxicol., Vol. 22, No. 7, 2009 1231 E., Hayes, J. D., Nestler, S., Mantle, P., and Schilter, B. (2007) Reduction in antioxidant defenses may contribute to ochratoxin A toxicity and carcinogenicity. Toxicol. Sci. 96, 30–39. (37) Friesen, R. W., Novak, E. M., Hasman, D., and Innis, S. M. (2007) Relationship of dimethylglycine, choline, and betaine with oxoproline in plasma of pregnant women and their newborn infants. J. Nutr. 137, 2641–2646. (38) Tebib, J. G., Reynaud, C., Cedoz, J. P., Letroublon, M. C., and Niveleau, A. (1997) Relationship between urinary excretion of modified nucleosides and rheumatoid arthritis process. Br. J. Rheumatol. 36, 990–995.

TX800459Q