Metabolomics and Systems Biology in Human Health and Medicine

By Baljit Ubhi, Mahon Maguire, Reza Salek and 7 more

The metabolomics and systems biology approach to research can be applied to many disciplines. This book provides a solid introduction to medical metabolomics and systems biology, and demonstrates how they have been applied to studies in medicine and human health, including nutrition and pathogenic microorganisms. Following core themes of diagnosis, pathology and aetiology of disease, this book provides a reference for health care professionals interested in how to use metabolomics for medical research.

• Language: English
• Publisher: CAB International
• Release date: Oct 24, 2014
• ISBN: 9781789244212

Book preview

Contributors

Luigi Atzori, Department of Biomedical Sciences, University of Cagliari, 09124 Cagliari, Italy. Email: [e-mail]

Susan C. Connor, Department of Biochemistry and the Cambridge Systems Biology Centre, University of Cambridge, Tennis Court Road, Cambridge, CB2 1GA, UK. Email: [e-mail]

Daniel A. Dias, Metabolomics Australia (School of Botany), Building 122, Professors Walk, University of Melbourne, Parkville, Victoria 3010, Australia. Email: [e-mail]

Julian L. Griffin, Medical Research Council Human Nutrition Research, Elsie Widdowson Laboratory, 120 Fulbourn Road, Cambridge, CB1 9NL, UK, and Department of Biochemistry and the Cambridge Systems Biology Centre, University of Cambridge, Tennis Court Road, Cambridge, CB2 1GA, UK. Email: [e-mail]; [e-mail]

Robin Hesketh, Department of Biochemistry, Sanger Building, University of Cambridge, Tennis Court Road, Cambridge, CB2 1GA, UK. Email: [e-mail]

Frances Jackson, Section of Computational and Systems Medicine, Department of Surgery and Cancer, Imperial College London, South Kensington Campus, London, SW7 2AZ, UK. Email: frances. [e-mail]

Oliver A.H. Jones, School of Applied Sciences, RMIT University, GPO Box 2476, Melbourne, Victoria 3001, Australia. Email: [e-mail]

Mahon L. Maguire, BHF Magnetic Resonance Unit, Department of Cardiovascular Medicine, University of Oxford, Wellcome Trust Centre for Human Genetics, Roosevelt Drive, Oxford, OX3 7BN, UK. Email: [e-mail]

John H. Riley, Clinical Respiratory MDC, GlaxoSmithKline, Glaxo Wellcome UK Ltd, Iron Bridge Road, Stockley Park West, Uxbridge, Middlesex, UB11 1BT, UK. Email: [e-mail]

Lee D. Roberts, Harvard Medical School, Cardiovascular Research Centre, Massachusetts General Hospital, Charlestown, Boston, MA 02114, USA. Email: [e-mail]

Ute Roessner, Metabolomics Australia (School of Botany), Building 122, Professors Walk, University of Melbourne, Parkville, Victoria 3010, Australia. Email: [e-mail]

Reza M. Salek, European Molecular Biology Laboratory – European Bioinformatics Institute (EMBL-EBI), Wellcome Trust Genome Campus, Hinxton, Cambridgeshire, CB10 1SD, UK, and Department of Biochemistry and the Cambridge Systems Biology Centre, University of Cambridge, Tennis Court Road, Cambridge, CB2 1GA, UK. Email: [e-mail]; [e-mail]

Maria L. Santoru, Department of Biomedical Sciences, University of Cagliari, 09124 Cagliari, Italy. Email: [e-mail]

Jane Shearer, University of Calgary, 2500 University Drive NW, Calgary, Alberta, Canada, T2N 1N4. Email: [e-mail]

Jonathan Swann, Department of Food and Nutritional Sciences, University of Reading, Whiteknights Campus, Reading, RG6 6AP, UK. Email: [e-mail]

Baljit K. Ubhi, AB SCIEX, 1201 Radio Road, Redwood City, CA 94065, USA. Email: baljit.ubhi@ absciex.com

Aalim M. Weljie, University of Pennsylvania, 10–113 Translational Research Center, 3400 Civic Center Boulevard, Building 421, Philadelphia, PA 19104, USA. Email: [e-mail]

Anisha Wijeyesekera, Section of Computational and Systems Medicine, Department of Surgery and Cancer, Imperial College London, South Kensington Campus, London, SW7 2AZ, UK. Email: [e-mail]

Jacqueline E. Wood, Department of Science and Primary Industries, Waikato Institute of Technology, Private Bag 3036, Hamilton 3240, New Zealand. Email: [e-mail]

Foreword

One of the perpetual joys of a life in science is finding oneself doing the completely unexpected, and there can be no branch in which this happens more often than biomedicine. So rapidly does the world of biology move and so perpetually astonishing are the emergent interactions that most of us would have been incredulous 10 years ago if we’d been told what we would be up to in 2014. Not merely has technical advancement dramatically gathered pace but with it entirely new subject areas have been created so that some have found themselves re-classified, at least in terms of job description.

Several years ago now I gave a talk in which I summarized some recent work that generated a computer model in an attempt to recapitulate the interconnections that occur in cell signalling – applying information theory to complex biological networks to show how they achieve a non-fuzzy response to multiple, fluctuating input signals. It was only afterwards that I realized I’d been talking about one facet of something that was beginning to crop up with increasing frequency both in the literature and in tea-rooms. Yes, it was ‘systems biology’ and I didn’t know it – and the revelation came as something of a relief because until then I’d have been a bit stuck if a student had produced, in that wide-eyed way they do, the question ‘What is systems biology?’ The short answer is of course, as Denis Noble has observed, that it’s ‘about putting together rather than taking apart’. Henceforth a more helpful response will be to get students to thrust into their hands this book, a comprehensive compilation by Oliver Jones of the views of leading figures on the impact of metabolomics and systems biology on biomedicine that opens with a clear exposition by Mahon Maguire of what the acronym-affluent field is all about and the essentials of the numerous methods now in use.

Before reviewing the subsequent chapters, two brief comments are appropriate, the first about an area not specifically covered in this book because it isn’t directly concerned with medicine, namely the emerging potential of systems biology applied to yeasts. The conservation of fundamental molecular mechanisms in essentially all eukaryotes means that this has implications for practically every field from drug development to biofuels. Records show that mankind has used yeast for baking and brewing for at least 4000 years and one of his greatest advances was to find ways of isolating pure yeast cultures in the latter part of the 19th century. This provided a rational basis by which the wondrous art of brewing could improve product quality using defined strain characteristics. From this has emerged tens of thousands of different strains, to say nothing of a complete set of mutants for the 6000 genes of Saccharomyces cerevisiae. The stage has thus been set for the development of computational models that comprehensively describe their cellular metabolism and intracellular signalling pathways and are beginning to illuminate corresponding processes in higher eukaryotes.

The second point relates to the ‘-omics’ revolution that has ensued since the completion of the Human Genome Project and which is now generating avalanches of biological data and presenting immense problems in terms of its use. In no field has this factual broadside had a more stunning effect than cancer. The capacity for deep sequencing of entire tumour genomes, coupled with expression profiling, is transforming the way in which major cancers are classified, thereby informing treatment strategies, and is leading to the creation of a comprehensive catalogue of the genomic changes involved in cancer (The Cancer Genome Atlas). It is therefore timely that Paul Workman’s group has generated an algorithm providing an objective prioritization of targets for therapeutic exploration, based on all the available biological and chemical information. Its relevance here is that, while specifically applied to the Cancer Gene Census, this model can be adapted to any human gene set emerging from large-scale ‘-omics’ studies. It is, however, important to note the authors’ caveat that their large-scale approach does not obviate the requirement for a detailed understanding of the underlying biochemistry of potential targets and pathways.

It is wholly appropriate that our first specific section considers the brain, under the guidance of Reza Salek, because it highlights the astonishing feats of ingenuity that, through next-generation sequencing, have given us an almost complete picture of the genomic difference between humans and our relatives, both near and more distant. From this emerges a remarkable parallel with cancer. Humans differ from other species in vast numbers of genetic variants, as do tumour cells from their normal counterparts, but the majority of these differences almost certainly have no effect on phenotypic evolution, either of species or neoplasm. Accordingly, identifying the relatively few specific molecular ‘drivers’ of human cognition remains a challenge only beginning to be met by faltering footsteps focusing, for example, on more highly mutated regions from which have been teased genes associated with brain size and speech. Many genes associated with cognitive dysfunction show similar differential expression among primates. The field is therefore open for metabolomics to advance our understanding from preliminary studies showing distinct metabolic profiles in humans, chimpanzees and macaques to predictive signatures for neurodegenerative diseases. To this may be added a corresponding requirement for brain tumours. Malignant gliomas are the most common cancers of the central nervous system and medulloblastoma is the most prevalent malignant, childhood brain tumour. Whole genome sequencing, expression profiling and copy number analyses have revealed the molecular heterogeneity of these tumours. As in other cancers, anomalously active signal pathways are identifiable that permit sub-type classification with hitherto undreamed of precision. However, as yet there are no adequate biomarkers for brain cancers, few chemotherapeutic options and the prognosis for these conditions, which cause about 175,000 deaths per year worldwide, remains dismal.

The spotlight then focuses on Jacqueline Wood who reviews the cardiovascular diseases and in particular the pursuit of biomarkers for the multiple defects that this broad category embraces. Arrhythmias are a particularly appropriate example in the present context because for some forms, for example long QT syndrome, causative mutations in the ion-channel genes SCN5A and KCNH2 have been identified. This has led to the generation of both transgenic mice and stem cell-based models for studying the pathway perturbations involved, through which electro-physiological measurements have been added to the portfolio of ‘systems’. The causes of other heterogeneous conditions such as sudden cardiac death and atrial fibrillation are only slowly emerging but several variants (single nucleotide polymorphisms) have been implicated in the former and there is evidence that the transcription factor PITX2c can play a role in the latter. In mice PITX2c is required for the development of the pulmonary myocardial sleeve that surrounds pulmonary veins and is the site at which abnormal electrical activity occurs when atrial fibrillation is stimulated. For the present, the central point is that a range of inputs, broad even by the standards of systems biology, is being integrated into well-established computer models of heart function that take account of the activities of individual ion channels and transporters to transform treatment strategies for cardiac arrhythmias.

Lung development and its repair after injury is increasingly the subject of investigation by genomic, proteomic, metabolomic and epigenomic methods and with Baljit Ubhi et al. we come next to pulmonary and respiratory disorders. This chapter focuses on chronic obstructive pulmonary disease although we might note the concern in this field over the rise in the incidence of tuberculosis together with the emergence of drug-resistant and multidrug-resistant variants. To this must be added, of course, lung cancer, the most common cause of cancer-related mortality in the world. For the two main types, small cell lung carcinoma (SCLC) and non-SCLC (NSCLC), the principal driver mutations have been identified by genomic sequencing. Notable is the fact that 10% of NSCLCs in western populations have somatic mutations in the EGFR kinase domain and the small-molecule inhibitor erlotinib has increased survival by up to 18 months in 65% of patients with such mutations. Nevertheless, drug resistance invariably develops, emphasizing the requirement for early detection. While several sets of prognostic gene expression signatures have been obtained for NSCLC, none has thus far proved reliable, although useful markers, for example CADM1, may be beginning to emerge. An alternative approach of mass spectrometry-based proteomics offers the prospect that protein sets with diagnostic and prognostic value for lung cancers will eventually be resolved.

The major role played by the liver in metabolism and disease is the subject of Luigi Atzori and Maria Laura Santoru’s review. Fatty liver disease, a widespread, reversible condition denoted by triacylglycerol accumulation in hepatocytes, is commonly associated with alcohol use or metabolic syndrome. The general pattern of hepatic lipid metabolism can now be discerned by MRI, opening the possibility of identifying lipids linked to insulin resistance. The inflammation accompanying liver disease can give rise to hepatic steatosis, a condition that may progress to cirrhosis, the largest risk factor for hepatocellular carcinoma (HCC). This is the third most common cause of cancer-related mortality worldwide but, despite the massive burden of 700,000 deaths per year, it has received far less research attention than other major cancers and is consequently lagging on both the genetic and biomarker fronts. Even so, mutational patterns reflecting aetiology (e.g. exposure to aflatoxins or hepatitis B virus infection) are emerging and a four-gene signature, derived from microRNA expression profiling, has been defined for the rare childhood tumour hepatoblastoma. The currently used biomarkers for HCC are manifestly inadequate but analysis of serum or urine metabolites by metabonomics and proteomics is beginning to produce sets that distinguish between HCC patients, healthy controls and cirrhotic individuals. Nevertheless, the patient groups thus far analysed have been small in number and disparate in nature and HCC screening is still in its infancy. An alternative for cancer detection and therapy monitoring is the use of silicon chip technology for antibody-directed selection of circulating tumour cells that have detached from a primary tumour and entered the bloodstream. This remarkable technology may offer both the most promising way to early tumour detection and of determining responses to chemotherapy. It also provides a bridge between proteomic and genomic technologies because DNA, extracted from the captured cells, can be used for whole genome sequencing. If this system evolves to be able to acquire cells from most major types of tumour it will provide a rapid route from early detection through genomic analysis to tailored chemotherapy without the requirement for tumour biopsies. For HCC, however, as things stand there are no reliable biomarkers nor have chip methods been successfully applied.

In reviewing molecular nutrition research Anisha Wijeyesekera, Frances Jackson and Jonathan Swann introduce yet another member of the club – ‘nutrigenomics’ – a field directed to optimizing diet, the implementation of which would impact on type 2 diabetes, obesity, cardiovascular diseases and cancers. The requirement for a holistic approach to this subject is evident not only from the limitless complexity of the human diet but also from the fact that some dietary factors are known to exert epigenetic effects. Furthermore, genome sequencing now permits the populations of microorganisms in food to be quantified, opening the possibility of controlling the balance between ‘good’ and ‘bad’ microbial types in food. To these multiple facets may be added the finding that ingested plant microRNAs can enter the circulation and hence modulate, for example, low-density lipoprotein metabolism – an astonishing example of cross-kingdom regulation.

Infectious diseases are a major cause of death and disability worldwide, including being responsible for at least 20% of cancers, and in Chapter 7 Daniel Dias and Ute Roessner turn their attention to the systematic studies of the underlying interactions between pathogenic microorganisms and hosts. The continuing accumulation of genome sequences for pathogens and the extension to transcriptomics is facilitating the delineation of pathogen–host interactions during infection. The problem of antibiotic-resistant strains of tuberculosis has already been mentioned but equally challenging is the increased tolerance to antibiotics that occurs in biofilms. These arise, for example, in staphylococcal infections and in cystic fibrosis patients with chronic Pseudomonas aeruginosa lung infection. As these are largely untreatable at present there is an immense need for a systemic approach to how the expression of virulence factors is controlled as a step towards novel therapies.

Lee Roberts discusses the combination of disorders referred to as metabolic syndrome that raises the risk of heart disease and diabetes. For the latter in 90% of cases that means type 2 diabetes, a multifactorial disease that in both its overt and precursor forms shows great heterogeneity in pathology, progression and treatment response. This complexity implies limitations in specificity and sensitivity to the traditional diagnostic and prediction methods based on glucose assays. Some understanding has come from a number of rodent models for obesity and insulin resistance, both genetic and dietary, and the latter in particular reflect disease progression in humans. It is comforting, therefore, that the metabolic changes they have revealed are broadly consistent with those emerging as metabolomics is applied to the human disease. Several such studies have identified sets of half a dozen or so circulating proteins or amino acids from which can be derived a diabetes risk score that gives greatly enhanced specificity for the identification of individuals at high risk of progression to overt disease. Notably, elevated levels of amino acids have been detected 12 years before the onset of overt disease – implying a confluence of at least some of their metabolic pathways with insulin signalling and hence glucose metabolism. If this magnitude of metabolic mayhem was not enough, systems biology is also unveiling the role of the human gut microbiome, which may not only change in response to the progression of diabetes but also contribute to its onset, for example through its effect on adipocyte metabolism. To this pot-pourri is added the continuing hunt for associated genetic variants: some 40 have now been tracked down but these account for only about 30% of the heritability of type 2 diabetes.

The penultimate chapter by Jane Shearer and Aalim Weljie considers systems approaches to the study of muscle. For skeletal muscle in particular, playing an important part in the overall regulation of metabolism, the molecular mechanisms controlling insulin sensitivity and the formation of new mitochondria are gradually being pieced together. This extends to a discussion of the group of over 70 autoimmune diseases that arise from the interplay of genetic and environmental factors and are of gradually increasing incidence in the western world. The multifaceted nature of these conditions requires systems-based strategies for the much-needed development of reliable biomarkers and effective therapeutic combinations.

We come at last to the chapter contributed by Julian Griffin and devoted specifically to cancer, but such is the dominance of these diseases in contemporary biomedical science that several examples of their intersection with systems biology have already been mentioned. In particular these have noted the delineation of aberrantly acting signal pathways in a number of major cancers. In addition, the power of deep sequencing now permits the construction of evolutionary trees for individual tumours and the resolution of mutational signatures that differentiate not only primary tumours from their metastases but also localized regions within primaries. From expression profiling have come prognostic signatures based on a few tens of genes that are informing the design of treatment strategies for breast and colon cancer, acute myeloid leukaemia, diffuse large-B-cell lymphoma, Burkitt’s lymphoma and other cancers. The genetic mayhem thus laid bare is indeed staggering, and unveiling the vast panoply of potentially ‘druggable’ targets therein represents a great advance. Its significance lies in the fact that the major challenge confronting cancer science is dealing with metastasis, for it is disseminated tumours that are responsible for 90% of the death toll and these can only be treated by chemotherapy. The feasibility of this approach was demonstrated in the immediate wake of the human genome project when, in a kinome screen, BRAF was identified as the most frequently mutated gene in malignant melanoma, a discovery followed within a few years by the production of vemurafenib, a small-molecule inhibitor specific for the mutant form of BRAF that was effective against metastatic disease. Seductive though this story is, it is a rare example – and the astonishing repertoire of biochemical tricks that has subsequently been revealed by which tumour cells can short-circuit the action of this and other specific inhibitors serves as a stark warning that all that glistens in the drug cabinet may not be solid gold. Moreover, this wonderful science has also served to emphasize that the cabinet is in fact pretty bare – a situation unlikely to show dramatic improvement on any useful timescale. It therefore seems reasonable to suggest that a major effort should be focused on earlier detection with the aim of pre-empting metastasis.

Julian Griffin’s review of cancer biomarkers has as its starting point the fact that none currently in use even approaches an ideal specification, as exemplified by the well-known shortcomings of prostate-specific antigen and cancer antigen 125 that is elevated in expression in about 90% of advanced cases of ovarian cancer. The field is therefore open for the metabolomic pursuit of entities that are sensitive, reliable indicators of early disease. For prostate cancer in particular one focus of attention has been the glycome: members of the galectin family of glycan-binding proteins show expression patterns that correlate with disease severity, particularly in advanced stages. Other methods include ¹H nuclear magnetic resonance spectroscopy, used in the eTUMOUR study to provide automated diagnosis of brain tumours, and hyperpolarized ¹³C magnetic resonance imaging to quantify the differential metabolism that characterizes cancer cells. Perhaps the most promising of these embryonic approaches is the capture of circulating tumour cells, referred to earlier, or of tumour-derived nucleic acids from a range of solid cancers, from which DNA can subsequently be sequenced.

Nobody opening this book is likely to be unaware that systems biology is in its infancy nor that the full gamut of ‘-omics’ methods is now being applied to many of the major problems confronting mankind including, in addition to the biomedical fields reviewed in this book, crop provision and biofuels. Whole genome sequencing is already being used to catalogue the causative mutations in hereditary monogenic diseases and as an essentially non-invasive neonatal screen wherein a baby’s DNA, acquired from a small sample of maternal blood, is subjected to repeated sequencing so that it can be distinguished from the parental genomes. The vistas that science is opening are truly stunning but it behoves all its practitioners to consider the social implications of what we do and how the information obtained may be presented to the public. Widespread non-invasive prenatal screening for genetic disorders and the sequencing of cancer genomes will present serious problems for clinicians when, for the most part, there is little on offer in terms of therapy; but they will pale before the dilemmas arising when we begin to unearth, for example, genetic variants associated with specific skills – yet alone with intelligence.

Robin Hesketh*

Department of Biochemistry

University of Cambridge

Cambridge, UK

1 An Introduction to Metabolomics and Systems Biology

Mahon L. Maguire*

University of Oxford, Oxford, UK

1.1 Introduction

Personalized medicine is anticipated to become a crucial paradigm in the future of healthcare. Key to the ability to tailor medicine to the individual patient is an understanding of the intricacies of an individual’s metabolism and its potential interaction(s) with potential treatments. The rise of functional genomics and systems biology has opened the door to understanding the interactions between genome, transcriptome, proteome and metabolome not only on a cellular level but also on a holistic level (i.e. the study of the individual as a complete system) and, despite being relatively new fields of research, both metabolomics and systems biology have a lot to offer to medical science. This chapter will outline the background of the fields of metabolomics and systems biology and explore their potential to generate new insights into human health and medicine.

Metabolomics attempts to quantify all of the small-molecule metabolites in a tissue, cell, biofluid or indeed whole organism. The term metabolome was first suggested by both Oliver et al. and Tweeddale et al. in the late 1990s (Oliver et al., 1998; Tweeddale et al., 1998); Oliver defined metabolomics as ‘measuring the concentrations of as many metabolites as possible to produce a metabolic snapshot’ and Tweeddale as measuring ‘the total complement of metabolites in a cell’. Nicholson et al. (at Imperial College London) defined the related concept of metabonomics at around the same time (Nicholson et al., 1999). While metabolomics and metabonomics have very similar definitions – being generally used to describe the use of analytical chemistry techniques, coupled with statistical analysis, to study the changes in a metabolome caused by a disease, perturbation or time course – the term metabolomics is more commonly employed in the literature and will be used here. More recently, the related concept of lipidomics has also arisen (Han and Gross, 2003), which specifically attempts to identify and quantify all of the lipids present in a biological sample. In 2007, Professor David Wishart’s team at the University of Alberta in Canada finished its first draft of the human metabolome database (http://www.hmdb.ca/). The latest draft (version 3.0, released in 2013) of this chemical counterpart of the human genome contains details of more than 40,000 metabolites, with 3100 compounds in urine alone (http://www.urinemetabolome.ca/). A diagram showing the ‘typical’ workflow of a metabolomics experiment is shown in Fig. 1.1.

The cellular metabolite pool serves as the ultimate expression of biological phenotype as it is influenced by the genome and proteome as well as external environmental, nutritional and xenobiotic factors (Nicholson et al., 2005; Dumas, 2012; Jones et al., 2012). Metabolomics can thus give a snapshot of the underlying biochemistry of a sample and, with repeated measurements taken over time, of cellular metabolism. Changes in metabolite levels can be observed as the result of a disease before clinical signs present (Chen and Snyder, 2012). Similarly, early changes in metabolism can be seen as a result of drug toxicity, presenting the opportunity to withdraw or change medication before the onset of pathological symptoms (Kaddurah-Daouk et al., 2007; Nicholson et al., 2012). Xenobiotic drug metabolism is also measurable (Clayton et al., 2009) and as metabolites are generally conserved across species, methods developed in academic or preclinical contexts can be readily translated to a human context.

Fig. 1.1. An outline of a ‘typical metabolomics study’. (NMR, nuclear magnetic resonance spectroscopy; GC-MS, gas chromatography–mass spectrometry.)

1.2 Sample Preparation for Metabolomic Analysis

Once a tissue sample has been taken, metabolic processes must be stopped. If metabolism is permitted to continue after sampling, the metabolite concentrations in the sample will alter and may no longer reflect the processes in vivo. Typically, samples are flash frozen in liquid nitrogen and then stored below –20°C, or ideally below –70°C. This prevents enzymatic activity and slows metabolite breakdown.

If a tissue sample is then to be subjected to metabolite extraction, the sample is required to be homogenized. A pestle and mortar cooled by liquid nitrogen can be used, or for larger samples, an electric tissue homogenizer can provide a convenient alternative as it homogenizes tissue directly into the extraction solvent. The sample must not be allowed to thaw during the homogenization process, so addition of liquid nitrogen or dry ice to the pestle and mortar may help. Proteins/enzymes in the sample are then precipitated and metabolites extracted using either acid or cold organic solvents. Due to the wide variety of metabolites, each treatment has its merits and limitations. Perchloric acid, methanol–water, acetonitrile–water and methanol–chloroform–water are all in common use (Pears et al., 2005; Stentiford et al., 2005). Perchloric acid extraction has proven popular in the past and has been shown to achieve reasonable hydrophilic metabolite extraction, although it has been observed to give poor reproducibility between replicates (Lin et al., 2007). Organic solvent extractions are relatively straightforward and produce reasonable results. Acetonitrile extractions have good reproducibility but poor metabolite fractionation. Extractions based on methanol–chloroform–water demonstrate good reproducibility and yield for both hydrophilic and hydrophobic metabolites (Lin et al., 2007; Beltran et al., 2012); the organic fraction will contain lipids and lipid-soluble metabolites, while the aqueous fraction will contain some lipids and aromatic compounds, amino acids, choline metabolites, sugars, nucleic acids and creatine. The same solvent-based metabolite extraction methods used for homogenized tissue samples are also applicable to biofluids such as urine, plasma and cerebrospinal fluid. Solvent-based metabolite extraction can be used to produce samples that are suited to analysis by both nuclear magnetic resonance spectroscopy (NMR) and mass spectrometry (MS) (Beltran et al., 2012).

1.3 Analytical Methodologies

Due to the diversity of metabolites present in most biological samples and their wide range of physical properties (e.g. lipids, nucleic acids, amino acids, sugars), it is difficult to envisage a single analytical technique capable of quantifying the entire metabolome. As a result a number of different technologies have been employed with the most common being NMR and MS. Although these techniques are the most commonly used, in practice any method capable of quantifying a large number of metabolites can be employed, including Fourier transform ion cyclotron resonance spectroscopy (FT-ICR), high-pressure liquid chromatography (HPLC), thin layer chromatography (TLC) and capillary electrophoresis (CE). An overview of the more common techniques is given in Table 1.1 and below.

Table 1.1. A summary of common analytical techniques used in metabolomics.

1.3.1 Nuclear magnetic resonance spectroscopy (NMR)

Solution-state NMR

NMR has long been a mainstay of metabolomic analysis and can detect metabolites down to a concentration of 5–10 μmol/l (Salek et al., 2007; Lindon and Nicholson, 2008; Heather et al., 2013). It provides a relatively simple method for measuring the ensemble of metabolites present in a solution with minimal sample preparation. Every nucleus has a quantum mechanical property called spin, which can take a value in increments of ½. NMR makes use of those nuclei with a noninteger spin, such as ¹H, ¹³C and ³¹P. It is helpful to imagine the nuclear spin as a vector that aligns with the direction of the applied magnetic field. This vector precesses about the applied magnetic field vector with a characteristic frequency (Larmor frequency). If a nucleus is excited with a radio-frequency oscillating magnetic field, we can then record its exponential decay back to equilibrium – the free induction decay (FID). A Fourier transform of the FID converts the time-domain FID into a frequency-domain spectrum with a peak corresponding to the precessional frequency of the nucleus; this frequency, or chemical shift, is commonly reported in magnetic field-independent units of parts per million (ppm). The amplitude of the peak is directly proportional to the number of nuclei of that type present in the solution and can therefore be used to quantify the concentration of a compound in solution.

The above description is a gross simplification of NMR but serves as a useful model for understanding the method; a more complete description of NMR including the quantum mechanics involved can be found elsewhere (Keeler, 2005). NMR is relatively insensitive, especially when compared with MS; however, it is typically reliable and, because the sample is introduced into the spectrometer in a tube, the spectrometer does not need cleaning, performance is maintained and the sample is not destroyed. Samples with a high salt content, such as urine, are also readily analysed except in very extreme cases.

The chemical shift of a nucleus is dependent on the exact magnetic field experienced, which is governed by the applied magnetic field and by the effects of the electrons surrounding the nucleus. As the electronic environment of the nucleus is determined by the chemical structure of the molecule, each molecule gives rise to a characteristic pattern of chemical shifts in its NMR spectrum. This pattern of chemical shifts can therefore be used to identify the compound in the spectrum. As a metabolomic sample typically contains a complex mixture of compounds, the resulting NMR spectrum will contain a complex pattern of peaks. The chemical shift range for ¹H spectra is small; these peaks may overlap, complicating the identification of the compounds. The solvent used to dissolve the sample may also have a peak in the spectrum. As the solvent, typically water, has a very high concentration compared with that of the metabolites in solution (110 mol/l versus 1 mmol/l), the solvent peak can dominate the spectrum. This distorts the shape of the metabolite peaks and compresses them into the noise baseline of the spectrometer’s analogue-to-digital converter (ADC). Solvent suppression methods exist to mitigate the solvent peak allowing the amplification of the metabolite signal without overflowing