A metabolomics approach to understand mechanism of heat stress response in rat

1


CHAPTER 1

INTRODUCTION

1.1 General Introduction
The normal core body temperature of a healthy, resting adult human being is
stated to be at 98.6°F or 37.0°C. Though the body temperature measured on an
individual can vary, a healthy human body can maintain a fairly consistent body
temperature that is around the mark of 37.0°C. In high temperature environments, as
the body loses water, its ability to regulate temperature is greatly affected. Prolonged
dehydration can lead to heat exhaustion (paleness, dizziness, nausea, vomiting,
fainting, and a moderately increased temperature (101-102°F) or even heatstroke
(very high temeperature (106°F or higher), and delirium, unconsciousness, or
seizures). Heat stroke is a multisystem disorder that can progress to shock, circulatory
collapse and death. Heat stress can aggravate the effect of other toxins. Dehydration
and loss of minerals through sweat decreases the body's ability to detoxify chemicals.
Because the circulatory system is under strain other hazards increase. Carbon
monoxide, which reduces oxygen supply to the tissues, is of particular concern.
Heat stress is one of the leading cause for concern amongst defense personnel
in the tropics since the performance of soldiers on the battlefield is greatly influenced
by environmental factors such as ambient temperature. Fatigue, resulting from
prolonged heat exposure, causes a decline in coordination, alertness, and
performance. Understanding the physiology of heat stress, mechanisms of heat
tolerance and methods to alleviate damage due to heat stress, is the main motivation
for our studies.

2

All organisms respond to a hyperthermic stress by synthesizing a highly
conserved set of proteins known as heat-shock proteins (HSPs). An important feature
of HSPs is their role in cryoprotection and repair of cells and tissues against the
harmful effects of stress and trauma. Extensive studies have been done on HSPs and
their role in heat stress tolerance in diverse species from bacteria to humans. Study of
HSP’s as biomarkers of heat stress has been the traditional approach to investigate
heat related illnesses so far.
A biomarker is a cellular or molecular entity found in increased amounts in
blood, urine or tissues that can be used as an indicator of disease, susceptibility to
disease or exposure to any externally applied perturbation. Biomarkers are
measurements thought to be directly related to clinical outcomes. Depending on the
specific characteristic, biomarkers can be used to identify the risk of developing an
illness (antecedent biomarkers), aid in identifying disease (diagnostic biomarkers), or
predict future disease course, including response to therapy (prognostic biomarkers).
Biomarkers are nowadays routinely identified using RNA- (microarray) or protein-
(proteomics) based platforms. Both these types of markers provide possibilities that
the cell may behave in a specified manner, but they are not the endpoints of the
cellular biochemical responses. In contrast, metabolites provide several advantages.
Firstly, metabolite markers are most closely related to the cell’s final endpoint its
biochemical phenotype. Secondly, they provide more stable and longer term markers
than RNA or proteins. Thirdly, metabolome is extremely sensitive to exogenous
stimulation, hence it responds quickly and in a stable manner. Fourthly, metabolome
changes reflect the cumulative responses of cells, from signaling and transportation to
regulation; hence they show an amplification effect in the response, leading to easier
detection of changes. Since these changes also arise from convergence of multiple

3

signals to common metabolic pathways, they make the metabolite markers more

robust and representative of broader range of signalling responses. Lastly, having a
small mass, instrumentation needs for metabolite detection are more established and
less expensive than for protein-based (proteomics) methods. These reasons make
metabolite-based approaches highly desirable for monitoring purposes. In spite of
these advantages, only specific metabolite markers (those detectable by biochemical
assays) of heat stress have been extensively studied. This is mainly due to lack of
standardized or commercially available reagents or kits as compared to expression
profiling approaches.

The general aim of metabolomics is to identify, measure and interpret the
complex time-related concentration, activity and flux of endogenous metabolites in
cells, tissues, and other biosamples such as blood, urine, and saliva without any bias
for the class of molecules. Metabolic responses of cells provide the final steps of
cellular adaptation to stresses or other perturbations to the tissues or individuals.
Multiparallel techniques, allowing analyses of the levels of low molecular weight
compounds, have only just begun to be established during the past decade. This is
especially true in mammalian systems. There are a few examples of metabolic
profiling applications in medical field such as in drug metabolism in animal systems,
but reports investigating effects of heat stress with a metabolomics approach are
almost absent.

Recent advances have made it possible to carry out an unbiased, simultaneous,
and rapid determination of metabolites in various organisms based on metabolic
profiling. Thus, metabolic profiling appears to be one of key additional tools in

4

multiparallel system analysis and plays an important role in functional genomics. The
focus of my research is to develop a metabolomics platform for understanding heat
stress response in a model animal, rat, and identify marker metabolites (intermediates

and end-points of metabolic pathways) responsible for this response. This will
ultimately help in monitoring performance and recovery of military personnel under
heat stress.

1.2 Objectives

The overall aim of this project was to understand cellular responses to
hyperthermia in multiple organs using a metabolomics approach. The specific
objectives of this study were as follows:
1) Establish a metabolomics platform for application in animal model
2) Identify metabolites from plasma and organs using a statistical approach.
3) Identify metabolic pathways affected by heat stress and their regulation.
4) Identify a comprehensive set of biomarkers from biochemical entities,
specific to heat stress.

In this thesis, the first chapter gives a brief introduction and the major objectives of
this research work. Chapter 2 is the literature review section, which provides
background information and previous as well as current research carried in the field of
heat stress effects and acclimation in mammalian systems and metabolomics. Chapter
3 provides details of the materials and methods that were used during the entire study.
Chapter 4 focuses on the extensive optimization studies of metabolomics methods,
performed using organ tissue of Rattus norvegicus (model animal, rat). It establishes a

5

metabolomics platform for ideal sample preparation and data processing techniques
using a data driven approach. This involved the use of various homogenization
buffers, ionization methods and solid phase extraction methods and their
combinations. In Chapter 5, a non-targeted approach to identify perturbational effects
is focused upon. Metabolic profiling results of the heat stressed animals after different

times of recovery, in plasma were reported and differential metabolites were
identified. Chapter 6 compares the differential expression of metabolites in various
organs at different time-points during heat stress and after times of recovery. This
chapter explores the effects of heat stress on a systems level and identifies the target
pathways. Statistical methods like t-test, ANOVA, and log base2 ratio and database
searches were used for identification of early as well as late responding markers of
heat stress and the pathways involved. Lastly, in Chapter 7, a summary of this whole
research work and scope for future research potential from the current study are
described.











6


CHAPTER 2
REVIEW OF LITERATURE
The literature reviewed here has been categorized under three parts. The first part of
the review includes heat stress metabolism, the effects of heat stress in animals, heat
acclimation, markers of heat stress and other common stressors. The second part of
this review deals with metabolomic overview, metabolomics technology platforms, its
applications and metabolomic data handling and knowledge extraction. In the third

part, metabolic pathways and networks and the applications of pathway analysis have
been highlighted.

2.1 Heat Stress Metabolism
2.1.1 Heat stress and homeostasis
Homeothermic animals must keep their body temperature within narrow limits for
optimal function. While heat is constantly generated in the body due to metabolism
and due to external factors like the air, radiant temperature, as well as humidity, the
body is equipped with adaptive mechanisms that enable a person to preserve a
constant core temperature (T
c
). The hypothalamus plays a vital role in controlling
body temperature by coordinating thermal information from all body areas and
directing the efferent signals to the appropriate heat production and heat conservation
systems. Because both thermal and several non-thermal factors will be present at all
times, it may not be appropriate to dismiss the contribution of either when discussing
the regulation of body temperature in mammals.

The sources of heat gain and heat loss to and from the body are, principally: (1)

7

Bodily heat production, or the heat of metabolism, which can vary depending on the
amount of physical activity undertaken, (2) convection and (3) radiation, either of
which may result in heat gain or heat loss depending on whether the skin temperature
is respectively below or above the ambient temperature, and (4) evaporation of sweat
from the surface of the skin, which can only result in the loss of heat from the body.
Non-thermal factors influencing heat loss and heat production responses are exercise
(Kenny et al., 1997, Thoden et al., 1994), blood glucose (Passias et al., 1993),
hydration/plasma osmolality (Ekbolm et al., 1970; Turlejska et al., 1986), sleep

(Aschoff et al., 1974), motion sickness (Mekjavic et al., 2001; Nobel et al., 2005),
fever (Bligh J., 1998), inert gas narcosis (Meklavic et al., 1992; Passias et al., 1992;
Washington et al., 1993).
Fig. 2.1 is a representation of the pathology of heat stress leading to heat stroke in
mammals (Leithead, 1978).

2.1.2 Heat stress in animals
All living creatures suffer from excessive heat and the effects of heat stress on various
species of bacteria, plants and animals have been extensively studied in the past few
decades. Metabolic adaptation of E. coli to a higher temperature via production of
heat stress proteins was reported (Weber et al., 2002). Eukaryotes like yeasts have
been known to produce heat stress proteins (Chen et al., 2003), though sphingolipids
too seem to be relevant for heat stress adaptation (Dickson et al., 1997; Jenkins et al.,
1997).



8


Figure 2.1: Schematic representation of the factors leading to heat stress and
heat stroke (Redrawn based on Leithead, 1978)


Effects of heat stress on plants has been widely investigated in a variety of plant
species (Hong et al., 2001; Locato et al., 2009; Meiri et al., 2009), including
commercially important species like rice (Wang et al., 2009), lettuce (Oh et al., 2009)
and tomato (Qu et al., 2009).

Chronic exposure to environmental heat is known to improve tolerance via heat

acclimation even in lower animals like Caenorhabditis elegans (Treinin et al., 2003).
Cell lines have also been subjected to heat stress and their effects been studied
(Zimmerman et al., 1991; Gibbs et al., 2009). Effects of heat stress on mammalian

9

cell cultures (CHOK1, P19 and NIH 3T3) include changes in the cellular architecture,
and the synthesis and degradation rates, of specific proteins and during recovery from
hyperthermic shock (Roobol et al., 2009). To define better the subcellular mechanism
of heat shock induced cardioprotection, the selective expression of individual heat
stress proteins (HSPs) has been investigated (Wei et al., 2006).

Heat tolerance in higher animals is mediated by activation of the hypothalamo-
pituitary-adrenocortical (HPA) axis (Michel et al., 2007). Some studies have shown
that marked accumulation of either dopamine, serotonin or IL-1 in brain occurs in
heatstroke-induced cerebral ischemia and neuronal damage in rats. The survival of
such animals can be increased by inhibition of IL-1 receptors or monoamine system in
brain as well as by induction of heat shock proteins (Lin et al., 1997).

2.1.3 Factors affecting the outcome of heat stress
Other than environmental conditions of temperature, humidity, air movement,
insulation and clothing that may affect heat tolerance, there are several physiological
conditions that make certain individuals more vulnerable to heat stress. These
personal factors include, age, sex, obesity, sleep deprivation and diabetes. Ageing has
been shown to increase protein nitration, causing a decline in HSP induction (Oberley
et al., 2008). Studies have also shown that mitochondria in old rats are more
vulnerable to and less able to repair oxidative damage that occurs in response to
physiologically relevant heat stress (Zhang et al., 2002). Aging alters stress-induced
expression of heme oxygenase-1 in a cell-specific manner, which may contribute to
the diminished stress tolerance observed in older organisms (Bloomer et al., 2009).

Mitochondria in old animals are more vulnerable to incurring and less able to repair

10

oxidative damage that occurs in response to a physiologically relevant heat stress
(Haak et al.,2009).

The thermoregulatory capacities of men and women are mostly dependent on the
number and activity of sweat glands, sex hormones and the distribution of
subcutaneous fat. A variety of chronic pain conditions are more prevalent for females,
and psychological stress is implicated in development and maintenance of these
conditions (Kawahata, A. 1960). Understanding relationships between gender
differences in stress and pain sensitivity and sympathetic activation could shed light
on mechanisms for some varieties of chronic stress (Vierck et al., 2008).
Diabetes impairs the ability to activate the stress response partly due to, the selective
atrophy of certain muscles or muscle fiber types (Najemnikova et al., 2007). It is also
known to cause aortic stiffness and this may contribute to the increase in mortality
and morbidity associated with diabetes in rats (Ugurlucan et al., 2009). Neural
differentiation is associated with a decreased induction of the heat shock response and
an increased vulnerability to stress induced pathologies and death (Yang et al., 2008).
The activation of the hypothalamus-pituitary-adrenal axis by stress depends mainly on
the characteristics of the stressor. Moreover, the response of this axis to stress also
depends on the time of day in which the stressor is applied (Retana-Márquez et al.,
2003).
2.1.4 Heat stress response and heat acclimation
Among the variety of predisposing factors that affect thermal tolerance, only two
adaptations are directly invoked to combat heat stress: 1) the rapid heat shock
response (HSR); and 2) heat acclimation (Moseley P.L., 1997; Sawka et al., 1985).

11

Heat acclimation is a long-term developing process leading to an expanded dynamic
body temperature regulatory range due to left and right shifts in the temperature
thresholds for heat dissipation and thermal injury, respectively (Horowitz M., 2001).
In contrast, the HSR is a rapid molecular cytoprotective mechanism and involves the
production of heat shock proteins (HSP). A rise in body temperature increases
transcription of the heat shock genes, leading to rapid augmentation of their
expression. Under normothermic conditions, the resting cellular 72-kDa HSP
(HSP72) level is low. However, heat acclimation leads to a marked upregulation of
the basal level of HSP72, an inducible member of the HSP72 family that is considered
the most responsive to heat stress, and to a faster HSR (Maloyan et al., 1999).

ET-1A receptor antagonism can alleviate symptoms of heatstroke, like hyperthermia,
arterial hypotension, decreased cardiac output, increment of tumor necrosis factor-γ,
and increment of cerebral ischemia (e.g., glutamate and lactate/pyruvate ratio) and
injury (e.g., glycerol) markers in rat (Chang et al., 2004). When rats were exposed to
high environmental temperature (e.g., 42 or 43°C), hyperthermia, hypotension, and
cerebral ischemia and damage occurred during heat stroke were associated with
increased production of free radicals (specifically hydroxyl radicals and superoxide
anions), higher lipid peroxidation, lower enzymatic antioxidant defenses, and higher
enzymatic pro-oxidants in the brain of heat stroke-affected rats (Chang et al., 2007).
The breaching of the blood-cerebrospinal fluid barrier in hyperthermia significantly
contributes to cell and tissue injuries in the CNS (Sharma et al., 2007) and induction
of heat shock protein, antagonism of interleukin-1 or N-methyl-D-aspartate receptors
or depletion of brain monoamines protects against the heatstroke-induced arterial
hypotension and cerebral ischemic injury (Lin et al., 1999).

12

Heat stress has been shown to influence normal bodily functions like digestion,

vision, smell and sleep depending on the intensity, duration and the mode of exposure
to heat (Sinha et al., 2006; Maloiy et al., 2008). Some studies even suggest that sleep
may be necessary for effective thermoregulation (Rechtschaffen et al.,2002).

2.1.5 Markers of Heat Stress
Although referred to as heat shock or stress proteins, we now know that most of these
proteins are expressed constitutively in normal unstressed cell and participate in
several important biological pathways. A select few of these stress proteins however,
are expressed only in times of trouble, and hence their appearance is often diagnostic
of some stress or trauma undergone by the cell. Over the years, heat shock proteins,
especially the HSP 70 family, have been the biomarkers of stress in diverse species.
In fact, HSP 72 over expression protects against hyperthermia, circulatory shock, and
cerebral ischemia during heatstroke (Yang et al., 1998; Maloyan et al., 2002; Lee et
al., 2006). It has also been shown that after the onset of heatstroke, the hypotension
and altered protein profiles displayed by animals can be reversed by whole body
cooling (Cheng et al., 2008).
There were significant differences in the concentrations of glucose, blood urea
nitrogen, sodium, potassium, calcium, inorganic phosphorus, triiodothyronine (T3)
and thyroxine (T4) and the activities of aspartate aminotransferase (AST), alanine
aminotransferase (ALT), alkaline phosphatase (ALP), creatine kinase (CK) and
lactate dehydrogenase (LD) in heat stressed camels (Gheisari et al., 1999), quails
(Yenisey et al., 2004), rats (Lee et al., 2008)


13

Presently, more than 100 genes (including HSPs) have been found to be affected by
heat stress. Expression profiling has contributed to this effort by identifying many
elements not previously known to be involved in the cellular response to thermal
stress. Changes in gene expression represent only a part of the overall response to

thermal stress. A full understanding of the cellular physiology of stress requires an
integrative approach that includes understanding the function and interactions of all of
the involved elements, proteins and others.

2.1.6 Heat stress management
Simultaneous measurement of heart rate, blood pressure and temperature, has not
previously been reported in unrestrained heat-acclimated rats. Measurement of these
variables without the confounding effects of restraint or handling has increased the
validity of the rat as a model for human heat acclimation. Better understanding of heat
stress mechanism and metabolism in rat can eventually help in better management of
heat related disorders.

Studies have already shown that pretreatment with anti-inflammatory dose of aspirin
can provide protection against heat stroke in rats, which may be associated with the
inhibition of elevation of plasma IL-1beta levels by aspirin (Song et al., 2004). Also,
dietary supplementation of chromium as chromium nanoparticles significantly
decreased serum concentrations of insulin and cortisol, increased sera levels of
insulin-like growth factor I and immunoglobulin G, and enhanced the
lymphoproliferative response and phagocytic activity of peritoneal macrophages in
heat-stressed rats (Zha et al., 2008). In a similar study, it was suggested that hyper-
HAES (hydroxyethyl starch) is superior to 7.2% NaCl or HAES alone in resuscitation

14

of heatstroke. The benefit of hyper-HAES during heatstroke is related to restoration of
normal multi-organ function (Liu et al., 2009). Discovery of early markers of heat
stress will lead to early diagnosis and treatment of heat stress and can even aid in
monitoring physiological status of target groups like defense personnel, miners,
foundry workers etc.

2.1.7 Other common stressors
The biochemical manifestations of stress in mammals is often similar, irrespective of
the type of stress. For example, in response to short-term treadmill running, rats show
signs of systemic stress including increase in serum corticosterone and HSP72.
(Brown et al.,2007). Immobilization is a severe stressor that elicits extremely large
elevations of plasma epinephrine (EPI), norepinephrine (NE), and corticosterone
levels (Mravec et al., 2008). Combined biochemical, proteomic and histological
evidence suggests that the effects of spaceflight on the liver may be similar to mild
cold stress or fasting (Baba et al., 2008; Luo et al., 2008).
General anesthesia is a major stressor and it causes suppression of thermoregulatory
mechanisms. Some research indicates that isoflurane anesthesia significantly
increases the esophageal temperature triggering thermoregulatory sweating, but that
the sensitivity and maximum sweating rate are maintained at normal levels relatively
well (Washington et al., 1993).
Ischemia followed by reperfusion presents a stress in mammalian tissue, that is very
similar to heat stress in its metabolic outcome and like in heat stressed animals, the
ischemia-reperfusion tolerance can be improved by mild exercise Wang et al., 2001)
or presence of reactive oxygen species scavengers (Lee et al., 1999).


15

2.2 Metabolomics
2.2.1 Metabolism and metabolomics
Since the systematic genome sequencing of the first microbe, we have seen the advent
of the ‘omics’ technologies, in which investigators seek to understand complex
biological systems on a large scale. The macromolecular omics (especially the
transcriptomics and proteomics) were the first to gain widespread attention. However,
metabolomics, is one of the more recently introduced ‘omics’ technologies. The
general aim of metabolomics is to identify, measure and interpret the complex time-

related concentration, activity and flux of endogenous metabolites in cells, tissues,
and other biosamples such as blood, urine, and saliva. In the last few years, there has
been an increased focus on the application of metabolomics for functional genomics.
The application of metabolomics for functional genomics was first discussed (Oliver
et al., 1998). In the same year, the term metabolome analysis was mentioned in the
context of analysis of metabolites for phenotypic profiling of Escherichia coli cells
(Tweeddale et al., 1998). A clear definition for metabolome analysis and terms for
other approaches to measure cellular metabolites was later introduced (Fiehn et al.,
2001).
The metabolome is the complete set of metabolites in a cell or tissue (Fiehn et al.,
2001; Goodacre et al., 2004), consists of low-molecular weight chemical
intermediates (Oliver et al., 1998), which can be considered to be the end products of
gene expression. It is a well-established fact that while change in gene (protein)
expression levels will have only small effects on metabolic fluxes, they must have
large effects on metabolite concentrations. Moreover, metabolic responses of cells
provide the final steps of cellular adaptation to stresses or other perturbations to the
tissues or individuals. Metabolomics thus represents an ideal level at which to

16

analyse change in biological system sensitively (Goodacre et al, 2004), under
conditions in which there may be negligible effects on the gross phenotype (Cornish-
Bowden and Ca´rdenas, 2001; Raamsdonk et al., 2001). Qualitative and quantitative
metabolome analyses also provide a view of the biochemical status of an organism
under specific conditions. For this reason, in the context of functional genomics,
metabolomics is now regarded as a viable counterpart to proteomics and
transcriptomics.

2.2.2 Metabolic profiling and metabolomics technology platforms
As the definition of the metabolome above suggests, in a metabolomics experiment

one would like to quantify all the metabolites in a cellular system, which could be
cells, tissues or biofluids in a given state, at a particular time point. For the analysis of
mRNA and proteins one ‘only’ needs to know the genome sequence of the organism
and exploit this information using nucleic acid hybridization or protein separation
followed by mass spectrometry (although post-translational modifications are
problematic). However, the analysis of metabolites is not as straightforward. Whilst
triple quadrupole MS instruments can be calibrated for accurate quantification of
specific metabolites of known structure, in general, for unknown analytes there is a
lack of simple automated analytical techniques that can measure hundreds to
thousands of metabolites quantitatively in a reproducible and robust way. In contrast
with transcriptome analysis (but in common with protein analysis) methods are not
available for amplification of metabolites and therefore sensitivity is a major issue.
Metabolites are generally labile species, by their nature are chemically very diverse,
and often present in a wide dynamic range. All of these challenges need to be
adequately addressed by the analysis strategy employed. This is currently a very

17

active area within metabolomics and in particular is presenting opportunities for novel
analytical instrument manufacture. Finally, in contrast to transcripts or protein
identification, metabolites are not organism specific (that is to say, sequence
dependent), thus when one has learnt how to measure the metabolite once, the
analytical protocol is equally applicable to prokaryotes, fungi, plants and animals
(Goodacre et al.,2007).

The strategies of sampling and sample preparation are diverse. Both invasive (blood,
intra-cellular metabolites in plants and microbes) and non-invasive (urine, volatile
components) sampling can be performed. Extra-cellular metabolites from urine
(metabolic footprint), depict a picture over a period of metabolic activity and are
normally stored at low temperatures to inhibit metabolic reactions. The extraction of

intra-cellular metabolites provides a snapshot of the metabolome, but can be time
consuming, and is subject to certain difficulties when compared to other sampling
strategies.

Metabolic processes are rapid (reaction times less than 1 second), hence rapid
inhibition of enzymatic processes is required and subsequent storage at -80°C. For
unicellular organisms or biofluids this is usually achieved by spraying the biomass
into very cold (<-40°C) 60% buffered methanol (Tweeddale et al., 1998). Whilst for
animal and plant tissues, liquid nitrogen is used to snap freeze the sample with
subsequent storage at -80°C (Whittmann et al., 2004). The storage of samples is
important, as the continued freeze/thawing of samples can be detrimental to stability
and composition (Wilson et al., 1997).


18

For the extraction of the metabolites from the matrix, there are many different
methods. The most common ones are: acid extraction using perchloric acid or nitric
acid, followed by freeze thawing, then neutralization with potassium hydroxide; alkali
extraction typically using sodium hydroxide, followed by heating (80°C); and
ethanolic extraction by boiling the sampling in ethanol at 80°C. (Buchholz et al.,
2001; Nielsen et al., 2005). However, these approaches can result in a severe
reduction in the number of metabolites detected and degradation compounds not
stable at extreme pH. Polar/non-polar extractions are the most frequently applied
method and are performed by physical/chemical disruption of the cells, removal of the
cell pellet by centrifugation and distribution of metabolites to polar (methanol/water)
and non-polar (chloroform) solvents. Recently, in microbial metabolomics
metabolites naturally excreted from intra-cellular volumes to the extra-cellular
supernatant are analysed (Tava et al., 2000). Sampling and collection of volatile
compounds from plants has also been performed (Weckwerth et al., 2003), and a

procedure for extraction and separation of metabolome, proteome and transcriptome
has also been reported (Rossi et al., 2002). Depending on nature of sample and
downstream metabolomic application, further sample preparation maybe necessary.
This may involve protein precipitation with organic solvents (Xu et al., 2008) and
further isolation from the sample matrix by solid phase extraction (SPE) or liquid–
liquid extraction (LLE) (Repetto et al., 2001). Indeed, with complex matrices like
mammalian tissue, sample preparation becomes a critical step before metabolomic
analysis actually begins. Tissue homogenization in appropriate buffers, protein
precipitation, centrifugation and sample filtration are steps that ultimately determine
the level of chemical information obtained from the sample.


19

The extract once ready for analysis, there are many different methods and approaches
that one could use. The choice of analytical technology is based on the level of
chemical information required from the metabolites, keeping in mind the speed and
resolution of the analysis. Realistically, no single technology is ‘all encompassing’
and there are currently no ‘set’ protocols to study metabolomics. In metabolomics
(also termed metabonomics for NMR- based clinical applications), NMR
spectroscopy provides a rapid, unbiased, reproducible, non-destructive, high-
throughput method that requires minimal sample preparation (Lindon et al., 2004;
Reo et al., 2002). The technique, especially
1
H NMR is used extensively in clinical
and pharmaceutical applications since it provides detailed structural information of
small organic molecules and, as such, has enabled a large number of biofluid
constituents to be identified and catalogued or listed (Lentner C., 1981; Lindon et al.,
2004).

Although not as sensitive as other techniques, such as mass spectrometry, useful data
can still be generated by NMR, from small sample volumes. Typically, biofluids such
as urine, bile, and blood plasma have been investigated, but also tissue extracts (Coen
et al., 2003). Studies as varied as toxicology (Keun et al., 2002), disease progression
(Makinen et al.,2008), drug efficacy monitoring (Griffin et al., 2004) and biomarker
discovery (Williams et al., 2005; Kim et al., 2003) have been successfully carried out
using NMR. However, this approach has limited sensitivity, resolution, and dynamic
range, resulting in only the most abundant components being observed.

Mass spectrometry is the most widely applied technology in metabolomics, as it
provides a blend of rapid, sensitive and selective qualitative and quantitative analyses

20

with the ability to identify metabolites. This approach, coupled with gas
chromatography (GC/MS) is the method of choice for plant metabolomics (Hall et al.,
2002). And owing to its success in plant metabolomics, mass spectrometry has been
adapted to mammalian (Welthagen et al., 2005) and yeast (Moheler et al., 2007)
metabolomics. Although GC/MS is biased against non-volatile, high-MW
metabolites, all metabolites can be analyzed after chemical derivatisation at elevated
temperatures prior to analysis. Due to hard ionization energies, results are highly
consistent between laboratories and datasets from different laboratories can be shared,
leading to construction of standard databases. However, the limitations as to the size
and types of metabolites that can be analyzed and the extensive preparation and
derivatization required is a big concern.

Liquid chromatography mass spectrometry (LC/MS) is ideal for metabolite profiling
as biofluids such as urine can be directly injected whereas samples such as plasma
need minimal pretreatment (protein precipitation). LC/MS is also capable of moderate
to high throughput, has a reasonable dynamic range combined with good potential for

biomarker identification (based on the spectral data generated), is not specific to
particular classes of compounds and can be extremely sensitive. The electrospray
ionization (ESI) technique has made polar molecules accessible to direct analysis by
MS. Quantification of multiple compounds in crude extracts can, in principle, be
achieved in the same way as described for GC/MS, although automation of the
procedure presents greater practical difficulties. LC/MS/MS provides additional
structural information that can be a very useful aid in the identification of new or
unusual metabolites, or in the characterization of known metabolites in cases where
ambiguity exists.

21


Fourier transform ion cyclotron resonance (FT-ICR) MS analysis has become
popular in the recent years because it enables the rapid, non-destructive, reagentless
and high-throughput analysis of a diverse range of sample types. It is a sensitive
technique and, with its high mass resolution (>10
6
) coupled to software that can
exploit the information in isotope patterns, can produce the empirical formulae for
metabolites directly (Aharoni et al., 2002). Due to its holistic nature, FT-IR
spectroscopy is a valuable metabolic fingerprinting/ footprinting tool owing to its
ability to analyse carbohydrates, amino acids, lipids and fatty acids as well as proteins
and polysaccharides simultaneously (Harrigan G. and Goodacre R., 2003). Relative
quantitation could also be achieved by comparing the absolute intensities of each
mass using internal calibration. Whilst selectivity is not as high as the other methods,
the rapidity of spectral collection and the fact that FT-IR readily lends itself to high-
throughput analyses is highly advantageous.

Some metabolomics work has also been carried out using Raman spectroscopy. This

is an emerging technology with significant potential for monitoring metabolites
(Mahadevan-Jansen et al., 19984. The metabolic fingerprinting potential of near IR
(NIR) spectroscopy should also be recognised, and studies undertaken using NIR
include measurement of lactate in human blood (Lafrance et al., 2003) as well as the
investigation of metabolites in faeces (Nakamura et al., 1998).

The application of liquid chromatography-mass spectrometry in the field of
metabolomics in recent years has been increased further by the development of higher
pressure systems such as Ultra Performance Liquid Chromatography-Mass

22

Spectrometry (UPLC-MS) (Plumb et al., 2005; Wilson et al., 2005; Kind et al., 2007).
UPLC-Tof MS has also been extensively used for toxicity, drug metabolism and
biomarkers studies (Crockford et al., 2006; Plumb et al., 2004; Yin et al., 2006).

Detection and quantitation of high molecular weight biomolecules from biofluids and
intact tissue (plant and animal) by matrix-assisted laser desorption ionization time-of-
flight mass spectrometry (MALDI-ToFMS) shows promise as a metabolomics
technology for several applications (Bucknall et al., 2002; Fraser et al., 2007). In
studies where ESI-MS would present complex spectra due to multiple charging
effects, quantitative MALDI-ToFMS can prove useful. The predominance of the
singly charged species in a complex mixture enables easier interpretation of spectra,
and the ability of MALDI-ToFMS to analyze complex biological samples aids in
eliminating the need for chromatographic steps.

Successful de novo identifications of biomarker metabolites have already been
demonstrated in animals exposed to various perturbations and stresses (Soga et
al.,2006; Loftus et al., 2008) coupled to database queries. Most recently, high-
throughput profiling of metabolic snapshots has been demonstrated in various rat

tissues (Ding et al., 2004; Chu et al., 2004). The future challenge will be to integrate
the observed metabolic alterations into hypotheses about changes in biochemical
pathways and gene expression levels.


2.2.3 Data handling and knowledge extraction
As with all functional analyses, a typical metabolomics experiment can generate
mountains of data (samples times variables times metabolites) and critical steps must

23

be taken to turn these data (information) into knowledge. In particular, we need well-
curated databases, very good data to populate them, and even better algorithms to turn
these metabolite data into knowledge.

Irrespective of the analytical technique used, the analysis of the data is essentially
performed in three stages. Initially the raw data need to be preprocessed to convert
them to a suitable form. Secondly it may be useful to subject these modified data to
data reduction so that only the most relevant input variables are used in the
subsequent data analysis. The objective of the third stage of the data analysis is to find
patterns within the data which give useful biological information that can be used to
generate hypotheses that can be further tested and refined.

For the metabolome, because the biological differences between samples sometimes
arises from comparatively small differences in many metabolite concentrations,
recognizing the pattern and interpreting it is not straightforward. The methods
available for metabolome analysis can be placed in four main (and partly overlapping)
categories – univariate and multivariate statistical, unsupervised learning (which
looks at the overall pattern or structure of the data), supervised learning (which uses
known information to help guide the classification of the data (Yeang et al., 2001;

Hastie et al., 2001), and system-based analyses which use theories such as MCA
(Thomas et al., 1997) to help interpret the data in terms of the biological networks that
generated them (Kell et al., 2004). Many unsupervised learning methods are
equivalent to clustering methods and are often statistically based, while supervised
methods come in many varieties (Weiss and Kulikowski 1991; Mitchell, 1997),
including statistical, neural, rule-based, evolutionary and so on.

24


The concept of multivariate biomarker profiles has become reality (van der Greef J.,
2004) and hence, more powerful supervised learning multivariate analysis methods
are needed (Hastie et al., 2001). In supervised learning an algorithm (vide infra) is
used to transform the multivariate data from metabolite profiles into something of
biological interest, usually of much lower dimensionality, which as discussed above
can be categorical (diseased vs. healthy) or quantitative (severity of disease).
Discriminant analysis (DA) is a particularly popular algorithm, which is a cluster
analysis-based method and involves projection of test data into cluster space (Manly
B.F.J., 1994). This is a categorical method and loadings matrices can give an
indication of important inputs (metabolites). Partial least squares (PLS) is a very
popular linear regression-based method (Martens et al., 1989). The algorithm can be
programmed in a quantitative way (PLS1) or categorical (PLS2 or PLS-DA), and as
for DA, loadings matrices can give an indication of important metabolites. Artificial
neural networks (ANN) are very popular based machine learning methods, which in
contrast to DA and PLS can learn non-linear as well as linear mappings (Bishop C.M.,
1995).

2.2.4 Application of metabolomics in animal studies
Metabolic profiling has been in use from the early 1970s (Farreet et al., 2001). It has
been extensively used in medical applications such as the screening of blood and

urine samples. The use of NMR spectroscopy was the first step in the development of
‘metabolic fingerprinting’ technologies. NMR spectroscopic analysis of biofluids,
cells or tissues enables generation of spectral profiles of a wide range of low
molecular weights (MW) metabolites that reflect the metabolic status of an organism.

25


Nowadays a variety of metabolomics technologies are employed to several
applications for studying mammalian systems. By far the most extensive application
of metabolomics is in the medical/clinical field. ‘Clinical metabolomics’ aims at
evaluating and predicting health and disease risk in an individual by investigating
metabolic signatures in body fluids or tissues, which are influenced by genetics,
epigenetics, environmental exposures, diet, and behaviour (Ceglarek et al., 2008).
Metabolic signatures of cancer (Spratlin et al.,2009), Celiac disease (Bertini et al.,
2009), gut microbiota (Jacobs et al., 2009) have been investigated using this platform.
Most metabolomics studies in rats involve toxicology studies (Stokvis et al., 2004)
and drug metabolism (Tong et al., 2006). One study even attempted to determine
hepatopathy signatures in dogs (Whitfield et al., 2005). Zuker rats have been used for
obesity related research using metabolomics consistently in the past few years
(Welthagen et al.,2005; Loftus et al., 2008). Of late, attempts are being made to
integrate metabolomic and transcriptomic data for a systems level perspective (Li et
al., 2009; Weckwerth et al., 2008).

2.2.5 Metabolite biomarkers
A biomarker is a molecule or a set of molecules that indicate an alteration of the
physiological state of an individual in relation to health or disease state, drug
treatment, toxins, and other challenges of the environment. Biomarkers are routinely
identified using RNA (microarray) or protein (proteomics) based platforms. Although,
both these types of markers provide possibilities that the cell may behave in a

specified manner, they are not the endpoints of the cellular biochemical responses. In
contrast, metabolites provide several advantages. Firstly, metabolite markers are most