The Mechanism of Galactosamine Toxicity Revisited; A Metabonomic

The Mechanism of Galactosamine Toxicity Revisited; A Metabonomic Study M. Coen,*,† Y. S. Hong,† T. A. Clayton,† C. M. Rohde,‡ J. T. Pearce,† M. D. Reily,‡ D. G. Robertson,‡ E. Holmes,† J. C. Lindon,† and J. K. Nicholson*,† Department of Biomolecular Medicine, SORA Division, Faculty of Medicine, Imperial College London, London SW7 2AZ, United Kingdom, and Metabonomics Evaluation Group, Pfizer Global R&D, Ann Arbor, Michigan 48105 Received March 22, 2007 1H

NMR spectroscopy was used to investigate the metabolic effects of the hepatotoxin galactosamine (galN) and the mechanism by which glycine protects against such toxicity. Rats were acclimatized to a 0 or 5% glycine diet for 6 days and subsequently administered vehicle, galN (500 mg/kg), glycine (5% via the diet), or both galN and glycine. Urine was collected over 12 days prior to administration of galN and for 24 hours thereafter. Serum and liver tissue were sampled on termination, 24 hours postdosing. The metabolic profiles of biofluids and tissues were determined using high-field 1H NMR spectroscopy. Orthogonal-projection to latent structures discriminant analysis (O-PLS-DA) was applied to model the spectral data and enabled the hepatic, urinary, and serum metabolites that discriminated between control and treated animals to be determined. Histopathological data and clinical chemistry measurements confirmed the protective effect of glycine. The level of N-acetylglucosamine (glcNAc) in the post-dose urine was found to correlate strongly with the degree of galN-induced liver damage, and the urinary level of glcNAc was not significantly elevated in rats treated with both galN and glycine. Treatment with glycine alone was found to significantly increase hepatic levels of uridine, UDP-glucose, and UDP-galactose, and in view of the known effects of galactosamine, this suggests that the protective role of glycine against galN toxicity might be mediated by changes in the uridine nucleotide pool rather than by preventing Kupffer cell activation. Thus, we present a novel hypothesis: that administration of glycine increases the hepatic uridine nucleotide pool which counteracts the galN-induced depletion of these pools and facilitates complete metabolism of galN. These novel data highlight the applicability of NMR-based metabonomics in elucidating multicompartmental metabolic consequences of toxicity and toxic salvage. Keywords: Metabonomics • Galactosamine Toxicity • Glycine Protection • Mechanism • NMR • O-PLS-DA

Introduction Galactosamine (galN) is an example of a classic ‘model’ hepatotoxin that has been widely used in experimental studies on liver damage.1,2 However, despite extensive investigation, the mechanism of galN-induced toxicity has not yet been fully resolved. The more established view is that the mechanism involves depletion of hepatic uridine nucleotide levels, which leads to inhibition of RNA and protein synthesis.3 Thus, it has been shown that hepatic UDP, UDP-Glc, and UTP are dramatically depleted and UDP-aminosugar levels elevated within 1 h of dosing of galN.4,5 In addition, it has been shown that the toxic effects of galN can be alleviated by the administration of UTP or its precursors.6 A more recent theory is that galN alters gut permeability and increases bacterial translocation leading to endotoxemia, and it has been postulated that the endotox* To whom correspondence should be addressed. † Imperial College London. ‡ Pfizer Global R&D. 10.1021/pr070164f CCC: $37.00

 2007 American Chemical Society

emia is responsible for the hepatic effect.7,8 GalN-induced endotoxemia occurs in rats but not in mice, which correlates to the overall lower sensitivity of the mouse to the compound. Correspondingly, mice which have been concurrently administered lipopolysacharride (LPS) are 105-fold more sensitive to the hepatotoxic effects of galN.9 Additionally, other agents such as glycine and cystamine have been shown to protect against the hepatotoxic effects of galN. The protective effects of these compounds have been attributed to various interactions in proinflammatory pathways;6,10,11 for example, it has been postulated that glycine protects against galN toxicity as it counteracts the activation of Kupffer cells and prevents cytokine release.6 Second, antibodies to inflammatory mediators such as TNFR protect against galN-induced hepatotoxicity. In earlier metabonomic studies of galN-induced hepatotoxicity in the rat, a particularly noticeable feature was the extreme inter-animal variability in the extent of the induced toxicity, and it was clearly observed that the toxic response was associated with an inability to fully metabolize the dosed Journal of Proteome Research 2007, 6, 2711-2719

2711

Published on Web 06/20/2007

research articles compound.12-14 Thus, rats showing liver damage also showed much urinary galN, while urinary galN was largely absent in rats where liver damage was not evident. Furthermore, the work of Clayton et al.14,15 suggests that ATP depletion could be a key factor in galN toxicity as a result of the need to re-synthesize UDP-Glc which is depleted by the metabolism of galN. It has also been suggested14 that unusual galN-induced increases in compounds such as methionine, betaine, and urocanate might be coherently explained in terms of such ATP depletion and a consequent inability to convert methionine to S-adenosylmethionine. In the present work, we have applied a series of state-ofthe-art metabonomic16,17 approaches to lead the study of galN toxicity, and in particular, we have addressed the problem of understanding the reported protective effects of glycine.6 This study was undertaken as part of the initial efforts of the second COnsortium on MEtabonomic Toxicology (COMET 2), a followup to the initial COMET project completed in 2004,18 which developed a comprehensive metabonomic database of toxininduced metabolic changes by NMR analyses of urine, serum, and tissue. The aims of COMET 2 are to utilize metabonomics as a ‘top-down’ systems biology driver to direct research and experiments in the determination of mechanisms of toxicity and the assessment of risk factors. The present work exemplifies the power of this approach in the investigation of perturbed metabolic pathways caused by toxic insult and the ability of this approach to shed new light on the modes of action of ‘old’ toxins.

Methods Animals, Treatment, and Sample Collection. Thirty-two male 6 week old SD rats (CRL:CD (SD) IGS BR, Charles River, Wilmington, MA) were housed in a well-ventilated room at temperatures of 70-78 °F and humidity of 30-70% with a 12 h light/12 h dark cycle. During urine collection periods, rats were housed in metabolism cages. At all other times, rats were housed in individual standard stainless steel cages. Water and food was supplied ad libitum throughout the study. On day 21, rats were administered vehicle (0.9% saline, n ) 8), galN alone (n ) 8), 5% glycine (via diet, n ) 8), or a combination of galN and 5% glycine (n ) 8). While acclimatizing (Days 1-9), animals were fed powdered Lab Diet 5002 Certified Rodent Diet (Lab Diet, Purina Mills, Richmond, IN). On Day 10, the diet was changed to powdered AIN-93G (Harlan Teklad, Madison, WI) which contained 0.39% glycine (of total protein content) and was administered for 6 days. On Day 16, rats in the vehicle and galN-alone groups continued to receive powdered AIN-93G, while animals in study groups receiving 5% glycine and combined galN and 5% glycine were switched to powdered AIN-93G containing 5% glycine (w/w). Rats continued to receive these diets until study termination. GalN (galN-hydrochloride, Sigma, Lot 094K1264) was dissolved in 0.9% saline at a concentration of 50 mg/mL, and the solution was filter-sterilized using a 0.2 µm filter. A single dose was given by IP injection to each rat in a dose volume of 10 mL/kg. All animals were euthanized 24 h after galN administration on Day 22. Urine was collected for approximately 24 h beginning on the following days: Days 9, 15, 16, 17, 21, and 22. Urine samples were collected in chilled collection tubes containing sodium azide (1 mL, 1% soln. in water). Total urine volume for each 24-h period was recorded. Serum was isolated from blood samples collected at necropsy from the abdominal vena cava. 2712

Journal of Proteome Research • Vol. 6, No. 7, 2007

Coen et al.

Urine and serum samples were stored at - 40 °C pending analysis. A portion of the left lateral liver lobe was also obtained from each animal, and samples were stored at - 80 °C pending analysis. Clinical Chemistry and Histopathology. 1. Clinical Chemistry Analysis. Serum was analyzed for alanine aminotransferase, aspartate aminotransferase, and total bilirubin levels using a Vitros 950 analyzer (Ortho-Clinical Diagnostics, Rochester, NY). 2. Histological Analysis. Each sample was fixed in 10% buffered formalin, embedded in paraffin, sectioned, and stained with hematoxylin and eosin. Liver sections were then scored according to the following scale: 0 ) no hepatocellular necrosis, 1 ) minimal hepatocellular necrosis, 2 ) mild hepatocellular necrosis, 3 ) moderate hepatocellular necrosis, and 4 ) marked necrosis. NMR Spectroscopic Analyses of Urine, Serum, and Liver. 1. 1H NMR Spectroscopic Analysis of Urine. Urine samples were thawed, vortexed, and allowed to stand for 10 min prior to mixing aliquots (400 µL) with phosphate buffer (200 µL, 0.2 M containing 10% deuterium oxide (D2O), 3 mM 3-(trimethylsilyl)-[2,2,3,3-2H4]-propionic acid sodium salt (TSP), and 3 mM sodium azide) and centrifuged at 13 000 rpm for 10 min. Supernatants (550 µL) were transferred into 96 well plates (1 mL deep, Lablinks, U.K.). D2O provided a field frequency lock, and TSP provided a chemical shift reference (1H, δ 0). 1H NMR spectra were acquired on a Bruker Avance 600 spectrometer, operating at 600.13 MHz 1H frequency and a temperature of 300 K, using a Bruker flow injection probe (Bruker Biopsin, Rheinstetten, Germany) with an active volume of 120 µL and automated sample handling unit (BEST, Bruker). Samples were transferred from the 96-well plate (cooling rack at 277 K) to the NMR flow probe using a Gilson 215 automatic sample handling system (Gilson, Middleton, WI). For each sample, 500 µL of urine was injected at a rate of 3 mL/min from the well into the transfer line which was maintained at a temperature of 303 K. Urine samples were separated from subsequent samples by approximately 500 µL of push solvent (1% sodium azide in H2O), and each solution was separated by an air bubble. Blank samples were run after every tenth sample to confirm that there was no cross contamination. NMR spectra were acquired using the standard one-dimensional solvent suppression pulse sequence (relaxation delay, 90° pulse, 4 µs delay, 90° pulse, mixing time, 90° pulse, acquire FID19). For each sample, 128 transients were collected into 32 K data points using a spectral width of 12 000 Hz with a relaxation delay of 2 s, an acquisition time of 1.32 s, and a mixing time of 100 ms. The water resonance was selectively irradiated during the relaxation delay and the mixing time. A line-broadening function of 0.3 Hz was applied to all spectra prior to Fourier transformation (FT). Zero-filling was applied to one level (SI 32 K), and automated phase and baseline correction were carried out using in-house software (NMRproc v0.3). 2. 1H NMR Spectroscopic Analysis of Serum. Serum samples were thawed, vortexed, and allowed to stand for 10 min prior to mixing aliquots (200 µL) with saline containing 20% D2O (400 µL). Samples were spun at 10 000 rpm for 10 min. Samples (500 µL) were placed in NMR tubes (507-PP), and NMR spectra were acquired at a 1H observation frequency of 600.13 MHz and temperature of 300 K. Chemical shifts were referenced to that of R-glucose (1H, δ 5.23), and D2O provided a field-frequency lock. The Carr-Purcell-Meiboom-Gill (CPMG20) spin-echo pulse sequence with a fixed spin-spin relaxation delay, 2nτ,

research articles

The Mechanism of Galactosamine Toxicity Revisited

Figure 1. Representative urine 1H NMR spectra of control, galN-treated, and galN- and glycine-treated animals 24 h post-dosing. Key: GalN, galactosamine; 2-OG, 2-oxoglutarate; DMG, dimethylglycine; TMAO, trimethylamine-N-oxide; DMA, dimethylamine; 4-PY, N-methyl-4-pyridone-5-carboxamide; NMND, N-methyl-nicotinamide; glcNAc, N-acetylglucosamine; GalNAc, N-acetylgalactosamine; Glc, glucose; dCyd, 2′-deoxycytidine; Ile/Leu/Val, isoleucine/leucine/valine; 3-HB, D-3-hydroxybutyrate; Cho/PCho, choline/phosphocholine; UDP-glcNAc, UDP-N-acetylglucosamine; UDP-GalNAc, UDP-N-acetylgalactosamine; -(CH2)n, lipoprotein methylene groups; -CH3, lipoprotein methyl groups.

of 200 ms (n ) 250, τ ) 400 µs), was applied to acquire 1H NMR spectra of all plasma samples. For each sample, 128 transients were collected into 32 K data points (SI 32K) using a spectral width of 12 000 Hz with a relaxation delay of 2 s and an acquisition time of 1.36 s. A line-broadening function of 0.3 Hz was applied to all spectra prior to FT. 3. 1H Magic Angle Spinning (MAS) NMR Spectroscopic Analysis of Liver. Liver tissue samples (median sample weight of 10 mg) were rinsed with a solution of TSP in D2O (1 mg/ mL) and placed in 4 mm zirconium oxide rotors (Bruker Biospin, Rheinstetten, Germany) and spun at 5 kHz. Spectra were acquired at a 1H observation frequency of 600.13 MHz and external sample temperature of 280 K. Chemical shifts were referenced to that of TSP (1H, δ 0.0), and D2O provided a fieldfrequency lock. The CPMG spin-echo pulse sequence20 with a fixed spin-spin relaxation delay, 2nτ, of 64 ms (n ) 128, τ ) 250 µs) was applied to acquire 1H NMR spectra of all liver samples. The CPMG experiment attenuates broad spectral resonances from high molecular weight compounds with long rotational correlation times and thus enables sharp resonances from low molecular weight compounds to be more clearly identified. For each sample, 128 transients were collected into 64 K data points using a spectral width of 12 000 Hz with a relaxation delay of 4 s and an acquisition time of 2.72 s. A line-broadening function of 1.0 Hz was applied to all spectra prior to FT. Signal assignment for representative samples was facilitated via acquisition of 2D NMR spectra (COSY, TOCSY, HSQC, and HMBC). In addition, published literature references were also utilized. Statistical Analysis of NMR Spectral Data. Full-resolution NMR data (each computer point that was generated in the spectrum was used) were imported into MATLAB (R2006a, Mathworks, Inc., 2006). The region corresponding to water/ HDO (δ 4.7-4.9) was removed from all spectra. In addition,

the TSP (δ -0.2 to 0.2) and urea (δ 5.6-6) regions were removed from urine spectra. The spectral data were then normalized to total spectral area, and pairwise, mean-centered orthogonalprojection on latent structures discriminant analysis (O-PLSDA21) models were computed. Orthogonal-Projection on Latent Structures-Discriminant Analysis (O-PLS-DA) of NMR Spectral Data. O-PLS-DA21 is a supervized pattern recognition algorithm that prefilters classification-irrelevant variation from data and improves interpretability of spectral variation between classes. O-PLS-DA extends the traditional supervised algorithm of projection on latent structures and enables maximal information to be extracted from complex spectral data. The prefiltered, structured noise in a data set is modeled separately from the class variation and can also be further interpreted via the loading matrices. The loadings coefficients are mean-centered and also ‘back-scaled’ to improve interpretability, as described by Cloarec et al.22 To prevent over-fitting of spectral data, the 7-fold cross validation method was used and the crossvalidation parameter Q2 was calculated.

Results Representative 600 MHz 1H NMR spectra of urine, serum, and liver from control, glycine-treated, galN-treated, and combined galN- and glycine-treated animals (for each class, the urine, serum, and liver spectra represent samples from the same animal) are shown in Figures 1, 2 and 3, respectively. A wide range of metabolites can be assigned in each spectral matrix that provide complementary information on a global systems response to both galN-toxicity and glycine protection. To provide a definitive interpretation of the metabolic signature changes reflective of galN toxicity and the protective effects of glycine, a series of pattern recognition methods were employed. Principal Components Analysis (PCA23) was initially Journal of Proteome Research • Vol. 6, No. 7, 2007 2713

research articles

Coen et al.

Figure 2. Representative 1H NMR serum spectra of control, galN-treated, and galN- and glycine-treated animals 24 h post-dosing. See key for Figure 1.

Figure 3. Representative 1H MAS NMR liver tissue spectra of control, galN-treated, and galN- and glycine-treated animals 24 h postdosing. See key for Figure 1.

applied to the spectral data to visualize inherent clustering between control and treated classes. All treated samples separated clearly from the control samples (data not shown). In this paper, we wish to present the results following application of a supervised pattern recognition method, O-PLSDA, to the spectral data. O-PLS-DA loadings matrices from fullresolution NMR data enable simple interpretation of significant metabolic changes between classes due to the conserved spectral format and the color-coding of resonances with respect to discriminatory weight (coefficient of determination, r2). Pairwise models were constructed from urine collected 24 h post-galN dosing from all classes, that is, control (n ) 8), glycine-treated (n ) 8), galN-treated (n ) 8), and galN- and 2714


glycine-treated (n ) 8), and from serum and liver collected on termination (also 24 h post-dosing). O-PLS-DA models were constructed for each treated class versus the control class. Urinary Response to GalN Toxicity and Glycine Protective Effects. In Figure 4a, the statistics for a model differentiating the urine spectra of galN-treated samples from controls reveal high R2 and Q2 values (>0.9), and a cross-validated O-PLS-DA scores plot (Figure 4b) shows clear differentiation between classes. The O-PLS-DA loadings plot shown in Figure 4c reveals the metabolites responsible for discrimination between the urine spectra of control and galN-treated animals 24 h postdosing. The color map corresponds to the strength of the coefficient of determination (r2) with those metabolites colored


research articles

Figure 4. (a) Model statistics, (b) cross-validated scores, and (c) O-PLS-DA coefficient loadings of full-resolution urinary NMR data revealing metabolites responsible for discrimination between controls and galN-treated samples. The color scale represents correlation (r2) to the discriminant variable. The upper section of the loadings plots represents metabolites increased in the treated class, whereas the lower part represents metabolites decreased in intensity.

in red being highly significant in discriminating controls from treated animals. In the loadings plot, the upper half of the plot represents metabolites increased in the treated class, whereas the lower half represents metabolites decreased in intensity. The intensity of these discriminatory resonances corresponds to the degree of change. On full analysis of the O-PLS-DA loadings plot, the discriminatory metabolites responsible for differentiation of the galN-treated class from controls were revealed to be elevated levels of galN, glcNAc, and urocanic acid together with decreased levels of 2-oxoglutarate, N-methylnicotinamide, 2′-deoxycytidine, and phenylacetylglycine. The O-PLS-DA loadings plot (not shown) for a model differentiating the urine spectra of the galN- plus glycinetreated group from controls indicates the metabolites respon-

sible for differentiation were elevated levels of galN, formate, and isovalerylglycine together with decreased levels of Nmethylnicotinamide and 2′-deoxycytidine. It was interesting to note that levels of glcNAc were not significant in discriminating controls from the co-treated galN and glycine samples. In addition, the model revealed that urocanic acid was not elevated, and 2-oxoglutarate, D-3-hydroxybutyrate, and phenylacetylglycine were not depleted. The profile representing cotreatment with galN and glycine is much closer to the control profile than that representing galN treatment, which is also apparent on inspection of the representative spectra in Figure 1. The NMR visible level of glcNAc correlated strongly with the degree of liver damage ascertained from the liver histopatholgy score, which suggests it plays a significant role in the heaptotoxicity of galN. Journal of Proteome Research • Vol. 6, No. 7, 2007 2715

research articles

Coen et al.

Figure 5. O-PLS-DA loadings plot of full-resolution MAS NMR data from liver revealing the metabolites that differentiate between control and (a) glycine-treated in the aliphatic and sugar spectral region, (b) glycine-treated in the aromatic spectral region, and (c) glycine-treated model statistics; (d) galN-treated in the aliphatic and sugar spectral region, (e) galN-treated in the aromatic spectral region, and (f) galN-treated model statistics; (g) galN- and glycine-treated in the aliphatic and sugar spectral region, (h) galN- and glycine-treated in the aromatic spectral region, and (i) galN- and glycine-treated model statistics. The upper half of each plot represents metabolites increased in the treated class, whereas the lower half represents metabolites decreased in the treated class. Note: The resonances that have been assigned to ‘Lipid Triglycerides’ may also have contributions from non-esterified fatty acids.

To conduct a comprehensive study of the protective effect of glycine on galN toxicity, the determination of the metabolic effects of treatment following administration of glycine alone was also included in this study. O-PLS-DA analyses (loadings plot not shown) of urine from animals treated with glycine alone revealed elevated levels of glycine, isovalerylglycine, isobutyrylglycine, 2-methylbutyrylglycine, 2-oxoglutarate, orotate, and fumarate together with reduced levels of 2′-deoxycytidine, N-methylnicotinamide, and its catabolite N-methyl-4pyridone-3-carboxamide (4-PY). Serum Response to GalN and Glycine. The corresponding serum data were also analyzed using O-PLS-DA (plots not shown), and the metabolites responsible for differentiation between control and galN-treated animals were determined to be elevated levels of tyrosine, phenylalanine, lactate, betaine, D-3-hydroxybutyrate, and dCyd together with reduced levels of lipids. Conversely, the profile for treatment with galN and glycine remained much closer to that of a typical control profile, in that the changes identified were simply elevated levels of glycine and D-3-hydroxybutyrate. Elevated levels of glycine were apparent in the serum following administration of glycine (loadings plots not shown), and no other changes were observed. The increase in tyrosine in the serum of galN-treated 2716


animals was dramatic (see Figure 2) and has previously been identified following administration of galN and attributed to reduced hepatic tyrosine aminotransferase activity.15 Hepatic Response to GalN and Glycine. The analysis of liver tissue by magic angle spinning (MAS) NMR added a further key dimension to this study as it provided information on liverspecific metabolic changes resulting from the treatments given. Representative MAS NMR spectral profiles of liver for each sample class are given in Figure 3, and marked changes in a host of metabolites are evident relative to control levels following treatment with galN. Interestingly, uridine was very clearly increased following treatment with glycine, and the UDP-amino sugars (UDP-glcNAc and UDP-galNAc) were dramatically elevated following co-treatment with glycine and galN. O-PLS-DA analysis of the MAS NMR data enabled the metabolic changes that occurred in the liver following treatment with glycine, galN, or both galN and glycine to be more simply visualized. The O-PLS-DA loadings representing the discrimination of MAS NMR spectra of glycine-treated and control animals are given in Figure 5a,b and reveal increased levels of uridine in the treated class as evidenced from the aromatic region of the spectrum (Figure 5b). In addition,


research articles

increased levels of UDP-galactose and UDP-glucose (Figure 5b) and depleted levels of unassigned resonances at δ 2.66 and δ 4.19 were evident (Figure 5a, annotated as ‘U’). The hepatic metabolic consequences of galN dosing are represented in Figure 5c,d, which shows increased levels of lipid triglycerides and UDP-glcNAc/UDP-galNAc together with reduced levels of glucose and choline/phosphocholine (Figure 5c,d). In addition, depletion of resonances assigned to adenosine-type moieties (Figure 5d) was also apparent. The O-PLS-DA loadings that differentiate control liver samples from those of animals treated with both galN and glycine are given in Figure 5e,f). These loadings reveal a markedly different profile from the loadings for galN treatment alone (Figure 5c,d)) and reflect significant increases in hepatic UDP-glcNAc/UDP-galNAc in the galN and glycine group. In addition, decreased levels of glucose and increased levels of lipids were seen, although the scale of these changes was far less marked than that seen for treatment with galN alone. The scale of the increases in uridine nucleotides observed in the MAS NMR spectra following treatment with glycine alone and following co-treatment with glycine and galN were further investigated. On inspection of the normalized median spectra for each class, it was apparent that the increase in the level of hepatic uridine relative to controls was greatest for the glycine only group (approximately 4× greater than control levels) with the glycine plus galN dose group also showing increased uridine at approximately 3× greater than the control levels. The increases observed in hepatic UDP-glcNAc and UDP-galNAc at 24 h post-dosing were greatest following coadministration of galN plus glycine (approximately 9× greater than the control level) with a smaller increase being observed in response to galN alone (approximately 3× greater than control levels). It is noteworthy that depleted levels of adenosine-type moieties following administration of galN alone were also observed. In addition, there was a slight increase in UDPgalactose and UDP-glucose following administration of glycine alone. Clinical Chemistry and Histopathology. The clinical chemistry data revealed that glycine co-treatment protected against galN-induced increases in ALT, AST, and bilirubin (Figure 6 a-c). The histopathological data also confirmed the protective effect of glycine (Figure 6d). Animals that were treated with galN alone displayed moderate (n ) 5/8) and marked (n ) 3/8) hepatocellular necrosis. In comparison, animals that had been treated with both galN and glycine displayed predominantly mild (n ) 5/8) but also moderate (n ) 3/8) hepatocellular necrosis.

Discussion To date, the alleviation of galN-toxicity by glycine has been attributed to it preventing activation of Kupffer cells and the ensuing release of TNF-R.11,24,25 Glycine has also been shown to have anti-inflammatory and immunomodulatory properties, in addition to being cytoprotective against a wide range of compounds.26 However, the increased urinary level of orotate and increased hepatic levels of uridine, UDP-glucose, and UDPgalactose following administration of glycine via the diet has led us to propose a novel hypothesis for the role of glycine in protecting against the hepatotoxic effects of galN. We propose that glycine pretreatment increases the hepatic nucleotide pool and hence counteracts, to some extent, the depletion of uridine nucleotides by galN. Uridine-monophosphate (UMP) is synthesized de novo from glycine/ammonia-donating substrates

Figure 6. GalN increases levels of serum liver injury markers and liver tissue necrosis, but these effects are ameliorated by glycine administration. Serum and liver tissue analyses were performed on samples collected 24 h after galN treatment. All gray bar values are expressed as mean ( SD, n ) 8 per group. Black circles represent individual animal measurements. (a) Serum alanine aminotransaminase levels (ALT); (b) serum aspartate aminotransferase levels (AST); (c) serum total bilirubin levels; (d) liver histopathology scores. Livers were scored as described in the Methods section.

via the orotate pathway27,28 with the liver being the major site of synthesis. The first and most important regulatory step in de novo synthesis of pyrimidines is production of carbamoyl phosphate (CP) from ammonia/glutamine, ATP, and carbonate, Journal of Proteome Research • Vol. 6, No. 7, 2007 2717

research articles catalyzed by carbamoyl phosphate synthetase (CPS29). Most eukaryotes have two types of CPS; CPS I is a mitochondrial enzyme which is ammonia-specific and is involved in arginine biosynthesis and the urea cycle. CPS II is located in the cytosol and contributes to the synthesis of pyrimidines with glutamine as the preferential substrate, although it also utilizes ammonia at high concentrations. The second step in pyrimidine biosynthesis is the formation of N-carbamoylaspartate from CP and aspartate. N-Carbamoylaspartate is subsequently dehydrated to form dihydroorotate which is oxidized to the pyrimidine base orotate, and then the phosphoribosyl group of 5-phospho-Dribosyl-1-pyrophosphate is added to orotate, forming orotidine 5′-monophosphate which is decarboxylated to form UMP. We propose that glycine increases the rate of CP synthesis which leads to increased UMP production as evidenced from the increased urinary levels of orotate and the increased hepatic levels of uridine, UDP-glc, and UDP-gal. Orotate has previously been shown to prevent galN-induced hepatotoxicity.5,30 Previous studies in Saccharomyces cerevisiae have shown that glycine had a marked stimulatory effect on the rate of ammoniadependent CP synthesis.31,32 Furthermore, the administration of ammonia or glycine has been shown to increase both hepatic synthesis and urinary excretion of orotate.33 Glycine-induced uridine nucleotide accumulation has also been reported in Staphylococcus aureus.34 It has been reported35 that, 30 min post-dosing with galN (400 mg/kg GalN-HCL), maximal depletion of uridine and UDPconjugates (UMP, UDP, UTP, UDP-glc, and UDP-gal) occurs in the liver as a result of rapid production of UDP-aminosugars (UDP-glcNAc and UDP-galNAc). In this study, the liver was sampled 24 h after galN dosing and inspection of the relevant spectra combined with O-PLS analysis indicated depleted levels of adenosine, which is consistent with the metabolism of galN to UDP-conjugates leading to depleted ATP levels. The galNderived UDP-amino sugars would be expected to rise rapidly within 1 h of dosing,3,4 but once the ready supply of UDPglucose becomes exhausted, the rate of formation of the UDPamino sugars would be expected to cease or to decrease dramatically. Thus, it was reasonable that we observed only a slight increase in hepatic UDP-aminosugars, relative to controls, 24 h after galN dosing. Metabonomic markers of liver damage were also evident in that we observed increased lipid triglycerides and a dramatic reduction in glucose and glycogen levels 24 h after GalN dosing. Such findings are consistent with ATP depletion, which would be expected to impair lipoprotein synthesis. The MAS NMR spectra of liver samples taken at 24 h after co-treatment with glycine and galN were significantly different to those obtained after treatment with galN alone, as much higher levels of uridine and UDP-amino sugars (UDP-glcNAc/ UDP-galNAc) were present. Furthermore, the liver samples representing treatment with glycine alone revealed marked increases in uridine and UDP-sugars (UDP-glc/UDP-gal). The increased uridine is attributed to the effect of glycine on increasing flux through the carbamoyl phosphate/orotate pathway, as was previously discussed. In comparison to the galN-treated group, the elevated levels of the UDP-amino sugars seen in the livers of the galN- and glycine-treated group indicate that the uridine nucleotide pool has not been so readily depleted and that the metabolism of galN via conjugation to UDP-sugars has continued for longer. An additional and very significant finding was the novel observation of a marked increase in urinary glcNAc levels 2718


Coen et al.

Figure 7. The metabolism of galN; 1 represents the production of UDP-aminosugars, 2 represents the inhibition of 1 and the ensuing increase in glcNAc when UDP-glc levels are depleted, 3 represents the proposed mode of glycine protection where pretreatment with glycine leads to increased levels of uridine nucleotides. The colored arrows indicate changes in the levels of metabolites as determined from the metabonomic data.

following treatment with galN which correlated strongly with the liver histopathology score, suggesting it is both toxicologically and mechanistically important, and we postulate that the urinary excretion of glcNAc is a measure of the extent to which GalN fails to be metabolized by the normal preferred pathway. Thus, the production of UDP-amino sugars from galN would continue to deplete UDP-glucose until the supply of UDPglucose was effectively exhausted. Then, unable to metabolize further galN by that route, the remaining galN would be epimerized and N-acetylated prior to excretion as glcNAc. Alternatively, the galN may first be N-acetylated and then epimerized to glcNAc. However, following pretreatment with glycine, where the levels of UDP-glucose and its precursors were enhanced, the administered galN was primarily metabolized by the normal route, and there was little production of glcNAc. Hence, these urinary observations are fully consistent with our interpretation of those seen in the liver. The current pathway of galactosamine metabolism is summarized in Figure 7 with the addition of the novel mechanistic observations elucidated in this study (red and green). We show that metabolism of galN leads to an increase in UDP-glcNAc and UDP-galNAc (1, blue); however, the resultant depletion of UDP-glc inhibits this route and metabolism of galN proceeds via epimerization and N-acetylation to glcNAc (2, red). The protective effects of glycine are shown in the observed increase in uridine, UDP-glc, and UDP-gal (3, green) which facilitates metabolism of galN via conjugation with UDP-glc.

Conclusion The application of the metabonomic approach to the ‘muchstudied’ hepatotoxin galactosamine has resulted in some unexpected findings, which strongly suggest that the mechanism by which glycine protects against galN-induced liver damage is related to the uridine nucleotide pool. This study demonstrates the investigative power of NMR-based metabonomics, which does not need to be led by prior hypothesis.

research articles


Abbreviations: O-PLS-DA, orthogonal-projection on latent structures discriminant analysis; glcNAc, N-acetylglucosamine; galN, galactosamine; UDP, uridine 5′-diphosphate; UDPglcNAc, UDP-N-acetylglucosamine; UDP-galNAc, UDP-Nacetylgalactosamine; UTP, Uridine 5′-triphosphate; UMP, Uridine 5′-monophosphate; LPS, lipopolysacharride; TNFR, tumor necrosis factor-alpha; MAS, magic angle spinning; NMR, nuclear magnetic resonance; 2-OG, 2-oxoglutarate; DMG, dimethylglycine; TMAO, trimethylamine-N-oxide; DMA, dimethylamine; 4-PY, N-methyl-4-pyridone-5-carboxamide; NMND, N-methylnicotinamide; Glc, glucose; dCyd, 2′-deoxycytidine; Ile/Leu/Val, isoleucine/leucine/valine; 3-HB, D-3-hydroxybutyrate; Cho/ PCho, choline/phosphocholine; UDP-glc, UDP-glucose; UDPgal, UDP-galactose; Glc-1-P, Glucose-1-phosphate; GalN-1-P, Galactosamine-1-phosphate; GlcN, glucosamine; galNAc, Nacetylgalactosamine; ADP, Adenosine 5′-diphosphate; ATP, Adenosine 5′-triphosphate; PPi, Pyrophosphate; CP, carbamoyl phosphate; CPS, carbamoyl phosphate synthetase.

Acknowledgment. The authors thank Dr. J. Trygg for use of the O-PLS-DA algorithm and Dr. O. Cloarec for providing the in-house software for application of O-PLS-DA. This work received financial support from Pfizer, Bristol-Myers-Squibb, Sanofi-Aventis, Servier and Waters as part of COMET 2. The authors acknowledge the Korean Government (MOEHRD) for Korea Research Foundation Grant (KRF-2006-214-F00024) for Y. S. Hong. Drs. T. Ebbels and H .Keun are acknowledged for use of NMRproc spectral processing software. Also, thanks to Dr. A. Maher for assistance with setup of the automated Bruker flow-injection system. Dale F. Wells is acknowledged for conducting the animal work at Pfizer. References (1) Keppler, D.; Lesch, R.; Reutter, W.; Decker, K. Exp. Mol. Pathol. 1968, 9, 279-90. (2) McMillan, J. M.; McMillan, D. C. Toxicology 2006, 222, 175-84. (3) Keppler, D. O.; Pausch, J.; Decker, K. J. Biol. Chem. 1974, 249, 211-16. (4) Keppler, D.; Frohlich, J.; Reutter, W.; Wieland, O.; Decker, K. FEBS Lett. 1969, 4, 278-80. (5) Keppler, D.; Rudigier, J.; Reutter, W.; Lesch, R.; Decker, K. HoppeSeyler’s. Z. Physiol. Chem. 1970, 351, 102-4. (6) Stachlewitz, R. F.; Seabra, V.; Bradford, B.; Bradham, C. A.; Rusyn, I.; Germolec, D.; Thurman, R. G. Hepatology 1999, 29, 737-45.

(7) Kasravi, F. B.; Wang, L.; Wang, X. D.; Molin, G.; Bengmark, S.; Jeppsson, B. Hepatology 1996, 23, 97-103. (8) Tiegs, G. Acta Gastroenterol. Belg. 1997, 60, 176-79. (9) Galanos, C.; Freudenberg, M. A.; Reutter, W. Proc. Natl. Acad. Sci. U.S.A. 1979, 76, 5939-43. (10) Wang, B.; Ishihara, M.; Egashira, Y.; Ohta, T.; Sanada, H. Biosci. Biotechnol. Biochem. 1999, 63, 319-22. (11) MacDonald, J. R.; Thayer, K. J.; White, C. Toxicol. Appl. Pharmacol. 1987, 89, 269-77. (12) So, P. W. Ph.D. Thesis, University of London, 1996. (13) Beckwith-Hall, B. Ph.D. Thesis, University of London, 1998. (14) Clayton, T. A. Ph.D. Thesis, University of London, 2001. (15) Clayton, T. A.; Lindon, J. C.; Everett, J. R.; Charuel, C.; Hanton, G.; Net, J. L.; Provost, J. P.; Nicholson, J. K. Arch. Toxicol. 2006, 201-10. (16) Nicholson, J. K.; Lindon, J. C.; Holmes, E. Xenobiotica 1999, 29, 1181-89. (17) Nicholson, J. K.; Connelly, J.; Lindon, J. C.; Holmes, E. Nat. Rev. Drug Discovery 2002, 1, 153-61. (18) Lindon, J. C.; Keun, H. C.; Ebbels, T. M.; Pearce, J. M.; Holmes, E.; Nicholson, J. K. Pharmacogenomics 2005, 6, 691-99. (19) Jeener, J. Chem. Phys. 1979, 71, 4546-53. (20) Meiboom, S.; Gill, D. Rev. Sci. Instrum. 1958, 29, 688-91. (21) Trygg, J.; Holmes, E.; Lundstedt, T. J. Proteome. Res. 2007, 6, 46979. (22) Cloarec, O.; Dumas, M. E.; Trygg, J.; Craig, A.; Barton, R. H.; Lindon, J. C.; Nicholson, J. K.; Holmes, E. Anal. Chem. 2005, 77, 517-26. (23) Jackson, J. E. A. A Users Guide to Principal Components; Wiley: New York, 1991. (24) Bruck, R.; Wardi, J.; Aeed, H.; Avni, Y.; Shirin, H.; Avinoach, I.; Shahmurov, M.; Hershkoviz, R. Liver Int. 2003, 23, 276-82. (25) Ikejima, K.; Iimuro, Y.; Forman, D. T.; Thurman, R. G. Am. J. Physiol 1996, 271, G97-103. (26) Zhong, Z.; Wheeler, M. D.; Li, X.; Froh, M.; Schemmer, P.; Yin, M.; Bunzendaul, H.; Bradford, B.; Lemasters, J. J. Curr. Opin. Clin. Nutr. Metab. Care 2003, 6, 229-40. (27) Connolly, G. P.; Duley, J. A. Trends Pharmacol. Sci. 1999, 20, 21825. (28) Scriver, C. R. Metabolic Basis of Inherited Disease, 2000. (29) Makoff, A. J.; Radford, A. Microbiol. Rev. 1978, 42, 307-28. (30) Scharnbeck, H.; Schaffner, F.; Keppler, D.; Decker, K. Exp. Mol. Pathol. 1972, 16, 33-46. (31) Pierard, A.; Schroter, B. J. Bacteriol. 1978, 134, 167-76. (32) Lim, A. L.; Powers-Lee, S. G. J. Biol. Chem. 1996, 271, 11400-9. (33) Vasudevan, S.; Laconi, E.; Rao, P. M.; Rajalakshmi, S.; Sarma, D. S.; La, P. G.; Fransvea, E.; Marzulli, D.; Lofrumento, N. E. Eur. J. Biochem. 1998, 251, 597-604. (34) Strominger, J. L.; Birch, C. L. J. Bacteriol. 1965, 89, 1124-27. (35) Keppler, D.; Decker, K. Eur. J. Biochem. 1969, 10, 219-25.

PR070164F

Journal of Proteome Research • Vol. 6, No. 7, 2007 2719

The Mechanism of Galactosamine Toxicity Revisited; A Metabonomic

Recommend Documents