NMR Metabonomics of Urine from Wistar Rats Dosed Separately with Kidney and Liver Toxicants
R. D. Beger1 , L. K. Schnackenberg1 , Y. P. Dragan1 , M. D. Reily2 , D. G. Robertson2 , 1NCTR, FDA, Jefferson, AR, 2Pfizer Global Research & Development, Ann Arbor Laboratories, Ann Arbor, MI
A collaboration between Pfizer and the Center of Metabolomics Research at NCTR was initiated to study the ability of NMR in combination with pattern recognition techniques to rapidly detect toxicity through analysis of urine samples. Single low and high doses of 4-aminophenol (PAP; 15 mg/kg, 150 mg/kg), galactosamine (Gal; 60 mg/kg, 600 mg/kg), and thiacetamide (THI; 20 mg/kg, 200mg/kg) were administered to male Wistar rats in groups of 4 for each toxin and dose level. Further, saline was administered to four rats as the drug vehicle control, for three toxins. Predose urine samples (Day 0) and samples from post-dosing (Day 2 and 3) were collected for each rat in metabolism cages at Pfizer and monitored by 1D 1H NMR at NCTR. Chenomx Eclipse software (Chenomx, Edmonton, Canada) was used to spectrally identify 198 metabolites in the 1D NMR spectra of urine. Partial least squares-discriminant function (PLS-DF) models of liver and kidney toxicity were developed from 8 metabolites. The metabolites used to build both the high and low dose PLS-DF models of toxicity were 3-hydroxyphenylacetate, homocysteine, 2-phosphoglycerate, 5-hydroxylysine, creatine, indole-3-acetate, sarcosine, and myo-inositol. The average of four training models of liver and kidney toxicity built from the high dose data was 98.2%. The average of four training models of liver and kidney toxicity built using the low dose data was 86.2 %.