Susan C. Connor

Proposed minimum reporting standards for data analysis in metabolomics

The goal of this group is to define the reporting requirements associated with the statistical analysis (including univariate, multivariate, informatics, machine learning etc.) of metabolite data with respect to other measured/collected experimental data (often called meta-data). These definitions will embrace as many aspects of a complete metabolomics study as possible at this time. In chronological order this will include: Experimental Design, both in terms of sample collection/matching, and data acquisition scheduling of samples through whichever spectroscopic technology used; Deconvolution (if required); Pre-processing, for example, data cleaning, outlier detection, row/column scaling, or other transformations; Definition and parameterization of subsequent visualizations and Statistical/Machine learning Methods applied to the dataset; If required, a clear definition of the Model Validation Scheme used (including how data are split into training/validation/test sets); Formal indication on whether the data analysis has been Independently Tested (either by experimental reproduction, or blind hold out test set). Finally, data interpretation and the visual representations and hypotheses obtained from the data analyses.

High-resolution magic angle spinning 1H NMR spectroscopic studies on intact rat renal cortex and medulla

High‐resolution magic angle spinning 1H NMR (MAS‐NMR) spectroscopy was used to investigate the biochemical composition of normal renal cortex and renal papilla samples from rats, and results were compared with those from conventional 1H NMR analysis of protein‐free tissue extracts. 1H MAS NMR spectra of samples obtained from inner and outer cortex were found to be broadly similar in terms of metabolite profile, and intra‐ and inter‐animal variability was small. However, the MAS NMR spectra from renal papilla samples were qualitatively and quantitatively different from those obtained from cortex. High levels of free amino acids and several organic acids were detected in the cortex, together with choline, glucose, and trimethylamine‐N‐oxide. The dominant metabolite resonances observed in papillary tissue were from glycerophosphocholine (GPC), betaine, myo‐inositol, and sorbitol. On increasing the magic angle spinning rate from 4,200 to 12,000 Hz, the lipid MAS 1H NMR signal profile remained largely unchanged in papillary tissue, whereas “new” resonances from triglycerides appeared in the spectra of cortical tissue, this effect being reversible on returning the spinning rate to 4,200 Hz. Further investigation into the behavior of the lipid components under different spinning rates suggested that the lipids in the cortex were present in more motionally constrained environments than those in the papilla. 1H MAS NMR spectra of tissues are of value both in interrogation of the biochemical composition of whole tissue, and in obtaining information on the mobility and compartmentalization of certain metabolites. Magn Reson Med 41:1108–1118, 1999.

Effects of feeding and body weight loss on the 1H-NMR-based urine metabolic profiles of male Wistar Han Rats: Implications for biomarker discovery

For almost two decades, 1H-NMR spectroscopy has been used as an ‘open’ system to study the temporal changes in the biochemical composition of biofluids, including urine, in response to adverse toxic events. Many of these in vivo studies have reported changes in individual metabolites and patterns of metabolites that correlated with toxicological changes. However, many of the proposed novel biomarkers are common to a number of different types of toxicity. These may therefore reflect non-specific effects of toxicity, such as weight loss, rather than a specific pathology. A study was carried out to investigate the non-specific effects on urinary metabolite profiles by administering four hepatotoxic compounds, as a single dose, to rats at two dose levels: hydrazine hydrate (0.06 or 0.08 g kg−1), 1,2-dimethylhydrazine (0.1 or 0.3 g kg−1), α-napthylisothiocyanate (0.1 or 0.15 g kg−1) and carbon tetrachloride (1.58 or 3.16 g kg−1). The study included weight-matched control animals along with those that were dosed, which were then ‘pair-fed’ with the treated animals so they achieved a similar weight loss. The urinary metabolite profiles were investigated over time using 1H-NMR spectroscopy and compared with the pathology from the same animals. The temporal changes were analysed statistically using multivariate statistical data analysis including principal component analysis, partial least squares, parallel factor analysis and Fishers criteria. A number of metabolites associated with energy metabolism or which are partially dietary in origin, such as creatine, creatinine, tricarboxylic acid (TCA) cycle intermediates, phenylacetylglycine, fumarate, glucose, taurine, fatty acids and N-methylnicotinamide, showed altered levels in the urine of treated and pair-fed animals. Many of these changes correlated well with weight loss. Interestingly, there was no increase in ketone bodies (acetate and β-hydroxybutyrate), which might be expected if energy metabolism was switched from glycolysis to fatty acid β-oxidation. In some instances, the metabolites that changed were considered to be non-specific markers of toxicity, but were also identified as markers of a specific type of toxicity. For example, taurine was raised significantly in carbon tetrachloride-treated animals but reduced in the pair-fed group. However, raised urinary bile acid levels were only seen after α-napthylisothiocyanate treatment. The methodology, statistical analysis used and the data generated will help improve the identification of specific markers or patterns of urinary markers of specific toxic effects.

High-resolution ^1H-NMR and magic angle spinning NMR spectroscopic investigation of the biochemical effects of 2-bromoethanamine in intact renal and hepatic tissue

The metabolic consequences of xenobiotic‐induced toxicity were investigated using high‐resolution magic angle spinning (MAS) NMR spectroscopy of intact tissue. Renal papillary necrosis (RPN) was induced in Sprague‐Dawley rats (n = 12) via a single i.p. dose of 250 mg/kg 2‐bromoethanamine (BEA) hydrobromide. At 2, 4, 6, and 24 h after treatment with BEA, three animals were killed and tissue samples were obtained from liver, renal cortex, and renal medulla. Tissue samples were also removed at 2 and 24 h from matched controls (n = 6). 1H MAS NMR spectroscopic techniques were used to analyze samples of intact tissue (∼10 mg). Decreased levels of nonperturbing renal osmolytes (glycerophosphocholine, betaine, and myo‐inositol) were observed in the renal papilla of BEA‐treated animals at 6 and 24 h postdose (p.d.), concomitant with a relative increase in the tissue concentration of creatine. Increased levels of glutaric acid were found in all tissues studied in BEA‐treated animals at 4 and 6 h p.d., indicating the inhibition of mitochondrial fatty acyl CoA dehydrogenases and mitochondrial dysfunction. Increased levels of trimethylamine‐N‐oxide occurred in the renal cortex at 6 h p.d. Changes in the metabolite profile of liver included an increase in the relative concentrations of triglycerides, lysine, and leucine. The novel application of 1H MAS NMR to the biochemical analysis of intact tissues following a toxic insult highlights the potential of this technique as a toxicological probe in providing a direct link between urinary biomarkers of toxicity and histopathological evaluation of toxicological lesions. Magn Reson Med 45:781–790, 2001.

Peak alignment of urine NMR spectra using fuzzy warping.

Proton nuclear magnetic resonance (1H NMR) spectroscopic analysis of mixtures has been used extensively for a variety of applications ranging from the analysis of plant extracts, wine, and food to the evaluation of toxicity in animals. For example, NMR analysis of urine samples has been used extensively for biomarker discovery and, more simply, for the construction of classification models of toxicity, disease, and biochemical phenotype. However, NMR spectra of complex mixtures typically show unwanted local peak shifts caused by matrix and instrument variability, which must be compensated for prior to statistical analysis and interpretation of the data. One approach is to align the spectral peaks across the data set. An efficient and fast warping algorithm is required as the signals typically contain ca. 32,000-64,000 data points and there can be several thousand spectra in a data set. As demonstrated in our study, the iterative fuzzy warping algorithm fulfills these requirements and can be used on-line for an alignment of the NMR spectra. Correlation coefficients between the aligned and target spectra are used as the evaluation function for the algorithm, and its performance is compared with those of other published warping methods.

Tryptophan–NAD+ pathway metabolites as putative biomarkers and predictors of peroxisome proliferation

The present study was designed to provide further information about the relevance of raised urinary levels of N-methylnicotinamide (NMN), and/or its metabolites N-methyl-4-pyridone-3-carboxamide (4PY) and N-methyl-2-pyridone-3-carboxamide (2PY), to peroxisome proliferation by dosing rats with known peroxisome proliferator-activated receptor α (PPARα) ligands [fenofibrate, diethylhexylphthalate (DEHP) and long-chain fatty acids (LCFA)] and other compounds believed to modulate lipid metabolism via PPARα-independent mechanisms (simvastatin, hydrazine and chlorpromazine). Urinary NMN was correlated with standard markers of peroxisome proliferation and serum lipid parameters with the aim of establishing whether urinary NMN could be used as a biomarker for peroxisome proliferation in the rat. Data from this study were also used to validate a previously constructed multivariate statistical model of peroxisome proliferation (PP) in the rat. The predictive model, based on 1H nuclear magnetic resonance (NMR) spectroscopy of urine, uses spectral patterns of NMN, 4PY and other endogenous metabolites to predict hepatocellular peroxisome count. Each treatment induced pharmacological (serum lipid) effects characteristic of their class, but only fenofibrate, DEHP and simvastatin increased peroxisome number and raised urinary NMN, 2PY and 4PY, with simvastatin having only a transient effect on the latter. These compounds also reduced mRNA expression for aminocarboxymuconate-semialdehyde decarboxylase (ACMSDase, EC 4.1.1.45), the enzyme believed to be involved in modulating the flux of tryptophan through this pathway, with decreasing order of potency, fenofibrate (−10.39-fold) >DEHP (−3.09-fold) >simvastatin (−1.84-fold). Of the other treatments, only LCFA influenced mRNA expression of ACMSDase (−3.62-fold reduction) and quinolinate phosphoribosyltransferase (QAPRTase, EC 2.4.2.19) (−2.42-fold) without any change in urinary NMN excretion. Although there were no correlations between urinary NMN concentration and serum lipid parameters, NMN did correlate with peroxisome count (r2=0.63) and acyl-CoA oxidase activity (r2=0.61). These correlations were biased by the large response to fenofibrate compared to the other treatments; nevertheless the data do indicate a relationship between the tryptophan–NAD+ pathway and PPARα-dependent pathways, making this metabolite a potentially useful biomarker to detect PP. In order to strengthen the observed link between the metabolites associated with the tryptophan–NAD+ pathway and more accurately predict PP, other urinary metabolites were included in a predictive statistical model. This statistical model was found to predict the observed PP in 26/27 instances using a pre-determined threshold of 2-fold mean control peroxisome count. The model also predicted a time-dependent increase in peroxisome count for the fenofibrate group, which is important when considering the use of such modelling to predict the onset and progression of PP prior to its observation in samples taken at autopsy.

Optimisation of collection, storage and preparation of rat plasma for 1H NMR spectroscopic analysis in toxicology studies to determine inherent variation in biochemical profiles

Biofluid 1H NMR spectroscopy has been assessed as a tool for toxicological investigations for almost two decades, with most studies focussing on urinary changes. This study has examined variations in the 1H NMR spectroscopy spectra of plasma collected from control rats at different times of the day. The collection, preparation and storage of samples were optimised and potential sources of variation in samples taken for toxicology studies identified. Plasma samples were collected into heparinised containers and analysed following a standard dilution with D(2)O. The value of deproteinising plasma with acetonitrile to look at low molecular weight metabolites has also been assessed. Variations in lactate and citrate levels in whole blood plasma were found and are consistent with the observation that lactate is one of the most variable metabolites in human plasma. Lipids levels also varied, in particular higher levels of lipids were found in spectra from male rats compared to female rats, and in samples collected in the morning following the feeding period. No significant changes were identified in samples which were snap-frozen and stored for up to 9 months at -80 degrees C. More changes were observed after storage at 4 degrees C or room temperature, including an increase in glycerol and choline levels, which may have resulted from lipid hydrolysis.

4,4-Dimethyl-4-silapentane-1-ammonium trifluoroacetate (DSA), a promising universal internal standard for NMR-based metabolic profiling studies of biofluids, including blood plasma and serum

Nuclear magnetic resonance (NMR)-based metabolic profiling of biofluids and tissues are of key interest to enhance biomarker discovery for disease, drug efficacy and toxicity studies. Urine and blood plasma/serum are the biofluids of most interest as they are the most accessible in both clinical and preclinical studies. However, proteinaceous fluids, such as blood serum or plasma, represent the greatest technical challenge since the chemical shift (δ) and line-width (ν1/2) of internal standards currently used for aqueous NMR samples are greatly affected by protein binding. We have therefore investigated the suitability of 4,4-dimethyl-4-silapentane-1-ammonium trifluoroacetate (DSA) as a universal internal standard for biofluids. Proton (1H) NMR spectroscopy was used to determine the effect of serum pH (3, 7.4 and 10) and DSA concentration on the overall lineshape and position of the trimethylsilyl resonance of DSA. The results were compared to that of 3-(trimethylsilyl)propionic acid sodium salt (TSP). Both the chemical shift and line-width of the DSA peak were not significantly affected by pH or DSA concentration, whereas these parameters for TSP showed large variations due to protein binding. Furthermore, the peak area of DSA correlated linearly with its concentration under all pH conditions, whilst no linear correlation was observed with TSP. Overall, in contrast to TSP, these results support the use of DSA as an accurate universal internal chemical shift reference and concentration/normalisation standard for biofluids. In the case of proteinaceous biofluids such as serum, where no current standard is available, this offers a considerable saving in both operator and spectrometer time.

Metabolomic study of the LDL receptor null mouse fed a high-fat diet reveals profound perturbations in choline metabolism that are shared with ApoE null mice

Failure to express or expression of dysfunctional low-density lipoprotein receptors (LDLR) causes familial hypercholesterolemia in humans, a disease characterized by elevated blood cholesterol concentrations, xanthomas, and coronary heart disease, providing compelling evidence that high blood cholesterol concentrations cause atherosclerosis. In this study, we used (1)H nuclear magnetic resonance spectroscopy to examine the metabolic profiles of plasma and urine from the LDLR knockout mice. Consistent with previous studies, these mice developed hypercholesterolemia and atherosclerosis when fed a high-fat/cholesterol/cholate-containing diet. In addition, multivariate statistical analysis of the metabolomic data highlighted significant differences in tricarboxylic acid cycle and fatty acid metabolism, as a result of high-fat/cholesterol diet feeding. Our metabolomic study also demonstrates that the effect of high-fat/cholesterol/cholate diet, LDLR gene deficiency, and the diet-genotype interaction caused a significant perturbation in choline metabolism, notably the choline oxidation pathway. Specifically, the loss in the LDLR caused a marked reduction in the urinary excretion of betaine and dimethylglycine, especially when the mice are fed a high-fat/cholesterol/cholate diet. Furthermore, as we demonstrate that these metabolic changes are comparable with those detected in ApoE knockout mice fed the same high-fat/cholesterol/cholate diet they may be useful for monitoring the onset of atherosclerosis across animal models.

Development of a multivariate statistical model to predict peroxisome proliferation in the rat, based on urinary 1H-NMR spectral patterns

A previous report of this work (Ringeissen et al. 2003) described the use of nuclear magnetic resonance (NMR) spectroscopy coupled with multivariate statistical data analysis (MVDA) to identify novel biomarkers of peroxisome proliferation (PP) in Wistar Han rats. Two potential biomarkers of peroxisome proliferation in the rat were described, N-methylnicotinamide (NMN) and N-methyl-4-pyridone-3-carboxamide (4PY). The inference from these results was that the tryptophan-nicotinamide adenine dinucleotide (NAD+) pathway was altered in correlation with peroxisome proliferation, a hypothesis subsequently confirmed by TaqMan® analysis of the relevant genes encoding two key enzymes in the pathway, aminocarboxymuconate-semialdehyde decarboxylase (EC 4.1.1.45) and quinolinate phosphoribosyltransferase (EC 2.4.2.19). The objective of the present study was to investigate these data further and identify other metabolites in the NMR spectrum correlating equally with PP. MVDA Partial Least Squares (PLS) models were constructed that provided a better prediction of PP in Wistar Han rats than levels of 4PY and NMN alone. The resulting Wistar Han rat predictive models were then used to predict PP in a test group of Sprague Dawley rats following administration of fenofibrate. The models predicted the presence or absence of PP (above on arbitrary threshold of >2-fold mean control) in all Sprague Dawley rats in the test group.