Integrating Metabolomics and Phenomics with Systems Models of Cardiac Hypoxia

businessunknownInternet και Εφαρμογές Web

12 Νοε 2013 (πριν από 4 χρόνια και 11 μήνες)

231 εμφανίσεις

Integrating Metabolomics and Phenomics with Systems
Models of Cardiac Hypoxia


Jacob D. Feala
1
, Laurence Coquin
2
,
Giovanni Paternostro
1,2
,
Andrew D.
McCulloch
1
*



1
Department of Bioengineering, University of California, San Diego, 9500 Gilman Drive,
0412, La Jolla, CA 92093
-
0412

2
Burnham Institute, 10901 North Torrey
Pines Road,
La Jolla,
CA
92037


Character count:

Running title:
Sy
stems analysis of hypoxia tolerance in
Drosophila


*
Corresponding author:

Andrew McCulloch

Mail code 04
1
2

9500 Gilman Dr.

La Jolla,
CA
92037
-
0412

Tel.
(858) 534
-
2547

Fax
(858) 332
-
1706

e
-
mail:
amcculloch@ucsd.edu



Abstract

Hypoxia is the major cause of necrotic cell death in myocardial infarc
tion. Cellular
energy supply and demand under hypoxic conditions is regulated by many interacting signaling
and transcriptional networks, which complicates studies on individual proteins and pathways. We
apply an integrated systems approach to understand

the metabolic and functional response to
hypoxia in muscle cells of the fruit fly
Drosophila melanogaster
. In addition to its utility as a
hypoxia
-
tolerant model organism,
Drosophila

also offers advantages due to its small size,
fecundity, and short life

cycle. These traits, along with a large library of single
-
gene mutations,
motivated us to develop new, computer
-
automated technology for gathering
in
-
vivo
measurements of heart function under hypoxia for a large number of mutant strains. Phenotype
data
can be integrated with
in
-
silico
cellular networks, metabolomic data, and microarrays to
form qualitative and quantitative network models for prediction and hypothesis generation. Here
we present a framework for a systems approach to hypoxia in the cardia
c myocyte, starting from
NMR metabolomics, a constraint
-
based metabolic model, and phenotypic profiles.




Keywords:
metabolomics, constraint
-
based model, hypoxia,
Drosophila melanogaster

1.
Introduction

Ischemic heart disease is the leading cause of death worldwide
(Murray and Lopez, 1997)
.
When acute coronary occlusions are survived, scarring and hypertrophic ventricular remodeling
can lead to arrhythmias or heart failure
(Frey and Olson, 2003)
. At the cellular level, the cause of
scarring and remodeling is necrosis and ap
optosis due to ischemia
-
reperfusion injury. Cellular
tolerance to ischemic injury can be enhanced


ischemic preconditioning is a strategy in which a
series of nonlethal ischemic insults provide a window of cardioprotection, so that a subsequent,
prolonge
d period of ischemia results in smaller infarct size and less tissue remodeling
(Murry et
al., 1986; Yellon a
nd Downey, 2003)
. Although many genes
(Du et al., 2006; Irie et al., 2003)

and pathways
(Kolar and Ostadal, 2004; Yellon and Downey, 2003)

are known t
o be important
for cardioprotection, the molecular basis for cellular injury versus tolerance to ischemia is not
well
understood
.

While

much effort has gone into increasing the understanding of hypoxia response
mechanisms, only recently have studies attempted to use genome
-
wide, discovery
-
based
strategies
. Global snapshots
of the transcriptome, proteome, and metabolome

have provided a
fi
rst introduction to the
wide

array of cellular changes that take place under ischemic conditions.
C
omputer models and a rich network context can help filter
large

high
-
throughput datasets to
produce new insights and hypotheses. The broad nature of the ce
llular hypoxia response,
involving multiple pathways across all three major levels of cellular information processing
(metabolic, protein signaling, and transcriptional), makes it an especially amenable and
interesting subject for applying these new techni
ques. This review describes
a new

systems
biology approach
for studying
metabolic adaptation
s

to acute myocardial hypoxia in
Drosophil
a
.


A wealth of data is publicly available for constructing system
s models. We are
combining t
he annotated
Drosophila
ge
nome
(Adams et al., 2000)

(stored in the useful online
resource FlyBa
se

(FlyBase, 2003)
), the KEGG
database of enzymes and pathways

(Ogata et al.,
1999)
, and the legacy of
Drosophila
biochemist
ry data from the li
terature
to build an
in
-
silico

reconstruction of all metabolic pathways active in
Drosophila
myocytes

(Feala et al., 2006)
.
Online d
atabases such as GRID,
DIP
, and BIND

(Bader et al., 2003; Breitkreutz et al., 2003;
Xenarios et al., 2002)

store
protein interactions, which can be overlaid onto the metabolic
network as a coarse
-
grained picture
of signaling net
works lying adjacent to metabolism.
A map
of interacting metabolic and signaling proteins serves as a

backbone onto which we

can

incorporate new data gathered in
the

lab. We integrate metabolomic and phenomic data

into the
network

model
in a way that ref
ines
it

to better represent the molecular interactions that are active
and essential for the response to hypoxia in myocytes.


The ultimate goal of this work is a genome
-
wide, data
-
driven model that includes all
genes, enzymes, metabolites, and regulatory
proteins that are involved in hypoxia defenses. This
broad scope necessarily requires a coarse
-
grained perspective
.

It is advantageous to decide
a
priori

the necessary level of detail and the corresponding modeling strategy. Our aim is to model
metaboli
sm at the level of enzyme fluxes, within the constraint
-
based framework. In the cell, a
network of signaling proteins is superimposed
onto

metabolic
pathways
, directly modulating
enzyme activity
in response

to
intra
-

and extracellular

conditions; t
his net
work will be
represented as a map of bidirectional links derived from interaction databases
. The
transcriptional cascades that regulate enzyme expression levels over longer time periods
are often

represented as a network of Boolean switches. Although it
is a daunting task to build complete
representations of these networks from the parts list, factors

that define the problem of interest,

such as
cell type (cardiac myocyte)
,
t
ime window

(acute)
, and the specific context

(
hypoxia
) help
to narrow the number
of players
. After
iterations
of experimental validation and refinement,

this
coarse
-
grained network
model may
help to find modules and targets on which to apply more
detailed modeling strategies.

O
ur first step was to develop a functional model of flight muscle metabolism.
We began
with flight

muscle for two reasons:
(1)
easy extraction of samples (a whole

fly

thorax contains
mostly flight muscle) compared with the heart tube, and
(2)
the greater
amount of biochemical
data in the literature for flight muscle.
Then
, t
his model will then be
supported

from the “bottom
-
up” with metabolomics data and
refined for heart tissue
from the “top
-
down” using genomic
cardiac
phenotypes
.

Figure 1 s
hows an overv
iew of these steps, illustrating the biological level
and tissue specificity of each data source incorporated in the model.


1.1
Hypoxia in cardiac myocytes

Hypoxia is the major insult to cell viability during ischemia, although reduced blood flow
also ca
uses cellular starvation and the
accumulation in tissue

of metabolic end products.
Myocardial cell damage during acute hypoxia in humans is thought to be caused by imbalances
such as decreased pH, altered calcium homeostasis, increased intracellular osmot
ic pressure, and
mitochondrial damage, resulting directly and indirectly from decreased ATP
(Jennings et al.,
1986; Opie, 1998)
. During hypoxia, ATP levels decrease due to restricted carbohydrate and fatty
a
cid oxidation. When pyruvate uptake into the mitochondria is decreased, pyruvate in the
cytoplasm is converted to lactate and there is an accumulation of lactate and protons in the cell
(Ingwall, 2002)
. In order to maintain the pH of the cell, energy is shifted away from myocardial
contraction towards the pumping of

protons out of the cell. Proton accumulation can also lead to
an increase in intracellular Na
+

and Ca
2+

via the Na
+
/H
+

and Na
+
/Ca
2+

exchangers
(Jennings et al.,
1986)
, possibly leading to damaging osmotic swelling and calcium overload in the m
itochondria.
Paradoxically, even more cellular damage can
occur upon reperfusion, due to

the production of
reactive oxygen species (ROS) during the influx of oxygen

(Solaini and Harris, 2005; Yellon and
Downey, 2003)
.

All cells have intrinsic defenses to maintain homeostasis during oxygen fluctuations. The
transcriptional, signaling, and metabolic pathways that carry out adaptation to hypoxia are far
-
reach
ing and complex, and yet are so highly orchestrated that they can be modeled as a single
functional unit
(Hochachka and Somero, 2002)
. The response can be separated into two
timescales, the early and l
ate response. The early response is characterized by an immediate shift
to anaerobic metabolism, as well as signaling cascades that act to decrease ATP demand
throughout the cell (e.g. by decreasing protein translation). The metabolic “master switch,”
AM
P
-
activated protein kinase (AMPK), senses decreases in the ATP:AMP ratio and triggers a
major shift in carbohydrate metabolism, limiting oxidative phosphorylation and increasing
glycolysis
(Pan and Hardie, 2002)
. Membrane ion pumps are also regulated
in an effort to
maintain pH and energy balance in the hypoxic cell. ATP
-
regulated potassium channels (K
ATP
),
both in the mitochondrial and sarcolemmal membranes, are possible cardioprotective end
effectors of ischemic preconditioning
(Yellon and Downey, 2003)
. The late response to hypoxia
involves a transcriptional cascade, triggered by hypoxia
-
inducible factor HIF
-
1α, as the cell
differentiates into a hypoxia
-
tolerant phenotype
(Hochachka and Somero, 2002; Semenza, 2001)
.


1.2
Systems biology


It is becoming clear that we have much to gain from using engineering systems analyses
to study complex networks o
f interactions in the cell. Most complex biological phenomena are
emergent properties of the system, i.e. behaviors not predictable from a detailed knowledge of the
parts. Quantitative models can help to predict and explain these emergent properties.

O
ne view of t
he systems biology approach involves
the following

steps

(adapted from
(Ideker et al., 2001)
)
:

1.

Gather high
-
throughput data to d
efine all the components
involved under the
context of interest
,

2.

Reconstruct integrated cellular networks into an
in
-
si
lico

model
,

3.

Systematically perturb and monitor the components,

4.

R
econcile the experimentally observed responses with those predicted by the
model,

5.

D
esign and perform new perturbation experiments to distinguish between
competing model hypotheses
.

Although this concept has existed for decades in smaller circles of computational
biologists, the recent emergence of systems biology into the mainstream has indisputably been
driven by the appearance of large high
-
throughput datasets, and the reliance on
computer models
to integrate and decipher them. From
this

point of view, data
-
driven or ‘inductive’ reasoning is
complementary to hypothesis
-
driven or ‘deductive’ reasoning, and for large, complex systems
iterative modeling may be the best way to generate

testable hypotheses

(Kell and Oliver, 2004)
.
This general framework has seen much success in studying the single
-
cell organisms
E. coli

and
S.
cerevisae
, as well as the multicellular eukaryotes
C. elegans

and
Drosophila melanogaster
, and
shows promise as a too
l for analyzing more complex organisms.


Our systems approach to the
study of the cardiac hypoxia response in flies follows this sequence of steps.

The first concept that drives the systems approach is high
-
throughput collection of
biological data.
High
-
t
hroughput studies have examined myocardial ischemia and ischemic
preconditioning at the level of transcriptome
(Onody et al., 2003; Sehl et al., 2000; Simkhovich et
al., 2003; Stanton et al., 2000)
, metabolome
(Sabatine et al., 2005)
, and proteome
(Cuong et al.,
2006; De Celle et al., 2005; Sakai et al., 2003; White et al., 2005)
, providing a useful


albeit
noisy and incomplete


sketc
h of the cellular milieu in response to hypoxia (or hypoxia
-
reoxygenation) stimulus.
These datasets provide global snapshots of the molecular components
of the cell, but by themselves say little about function. In multicellular organisms especially,
seve
ral layers of complexity are added to the system between the cell and organ levels. Top
-
down data, quantifying the effect on the system as a whole when a component is perturbed,
provides a functional bridge between the molecular and phenotype levels. The

“phenome” is the
set of quantitative measurements of phenotypic function under a specific physiological context,
for gene perturbations covering the entire genome


i.e. a directed genetic screen. Genetic
screens have been performed for years, and they h
ave proven to be invaluab
le for discovering
disease gene homologs in
Drosophila
. The measurand in these screens has often been binary,
meaning the phenotype either displays the characteristic or not, but these methods tend to gloss
over phenotypes with on
ly slight changes.

A slight effect on phenotype could indicate a non
-
lethal but still importan
t functional role for a gene.
Q
uantitative
g
enomic phenotype information

in flies
can be integrated with molecular
-
level

data
to add an extra layer of biological context.

The second concept crucial to the systems biology paradigm is th
at

of

the dense

intracellular network, since cellular components are highly interconnected above and beyond the
traditional view of isolated pathwa
ys
(Barabasi and Oltvai, 2004)
. It has been shown that in some
intracellular systems, the topology of the network is more important for determining function than
the kinetic parameters of individual interactions
(Albert and Ot
hmer, 2003; Urbanczik and
Wagner, 2005)
. It is generally agreed that the best way to represent
genome
-
wide

datasets is by
networks of nodes and links, where nodes symbolize genes, proteins, or small molecules and
links represent interactions or reactio
ns between nodes
(Barabasi and Oltvai, 2004)
. An
advantage to the network representation is that these datasets can be easily integrated by starting
with a common set of nodes (e.g. the complete list of genes in a genome) and adding node and
link

values representing the data. Although researchers have long used cartoon networks and
systems models to describe specific cellular pathways, large
-
scale network analysis of high
-
throughput data is a recent development, spurred by the need to bypass comp
utationally intensive,
parameter
-
dependent systems of differential equati
ons for these massive datasets
(Han et al.,
2004; Ideker et al., 2002; Walhout et al., 2002)
.


The third
new

idea in this approach is the heavy reli
ance on computer models.
Researchers have built a number of kinetic models
o
f mostly smaller systems
(Le Novere et al.,
2006)
, which can be interfaced with each other in a

modular fashion using common standard
s
such as

Systems Biology Markup Language (SBML)
or Cell Markup Language (CellML)
(Hucka
et al., 2003; Lloyd et al., 2004)
. However, m
odeling the detailed dynamics of biological systems
generally requires a large set of kinetic parameters, which would be impractical to obtain for a
genome
-
wide network.
Several new tech
niques aim to infer function from large scale networks
without the reliance on detailed mechanistic models.
For metabolic
networks
in particular, the
constraint
-
based method

is a useful way to quantitatively analyze genome
-
wide
metabolic
reconstructions

without large numbers of kinetic parameters.
A metabolic reconstruction consists
of a curated set of gene
-

and protein
-
linked enzymatic reactions that completely define the
network of interest. Besides the stoichiometric restriction of this network, add
itional constraints
can be added in the form of limits on uptake rates and maximum enzyme fluxes.
Using the
assumption of steady state, this method uncovers the space of all possible enzyme flux
distributions under
the

set of physiochemical limitations im
posed on the system
(Palsson, 2000;
Papin et al., 2003)
.
A number of quantitative analyses have been developed to explore the
biological consequences

of this so
lution space
(Price et al., 2004)
.
For example, l
iving systems
have the ability to adapt to find an optimum phenotype within the solution space
, and

l
inear

programming can be used
to predict this optimum flux distribution given the constraints and an
objective for the system (for example, growth rate or ATP production)
(Ibarra et al., 2002;
Kauffman et al., 2003)
. Using these
methods
, computer simulations of microorganisms can
approximate fluxes

and growth rates seen
in
-
vivo
(Duarte et al., 2004; Edwards et al., 2001)
.
Genome
-
sc
ale constraint
-
based models have been built for a number of organisms
(Duarte et al.,
2004; Reed et al., 2003; Sheikh et al., 2005; Vo et al., 2004)
, but not yet for
Drosophila
melanogaster
.


1.3
Drosophila as a model for systems analysis of cardiac hypoxia

Fruit flies have received increasing interest as a model for cardiac research
(Bier and
Bodmer, 2004; McCulloch and Paternostro, 2005; Serluca and Fishman, 2006)
.
Drosophila

has

a
tube
-
like heart and a simple circulatory system, which, when combined with its legacy of genetic
research, makes the fly an attractive organism for studying genetic influences of cardiac function.
With its genome sequence currently in its
fourth

revisi
on
(Adams et al., 2000; FlyBase, 2003)
, the
fruit fly has one of th
e best characterized genomes of multicellular organisms. Nearly a century
of research on the fly has resulted in a large repository of knowledge on the function of fly genes
and proteins, most of which is publicly available in the online database FlyBase
(FlyBase, 2003)
.
Flies can be reared and manipulated with minimal equipment and care, and their fecundity and
ease of genetic manipulation have made them a popular subject for genetic screens. The Berkeley
Drosophil
a Genome Project has provided a public resource of
Drosophila

P
-
element insertions,
covering over 40% of the genome
(Bellen et al., 2004)
.

Genomic similarities between flies and humans suggest that
Drosophila

genes found to
influence cardiac function are likely to have a human counterpart. About half of
Drosophila
protein sequences are homologous to mammalian proteins
(Adams et al., 2
000)
, and in a survey
of 287 humans disease genes across several physiological categories,
Drosophila
was found to
have a homolog for 62%, including all 6 cardiac disease genes examined
(Fortini et al., 2000)
.
Aside from “disease genes”, flies and mammals

share many other genes underlying basic
mechanisms of heart development and function. The identification of the homeobox transcription
factor
tinman
, essential for heart vessel formation in flies
(Bodmer and Venkatesh, 1998)
,
prompted the cloning of homologues (Nkx2
-
5/Csx) which regulate cardiac development in mice
(Ikeda et al., 2002; Lints et al., 1993)
. Although flies, unlike mammals, do not use sodium
channels to create cardiac action potentials
(Gu and Singh, 1995; Johnson et al., 1998)
, the fly
pacemaker nevertheless relies on several conserved calcium and potassium channels. One
impo
rtant example is
ether
-
a
-
go
-
go
, which is similar in sequence and function to the human
HERG potassium channel which has been implicated in long QT syndrome, a potentially fatal
cardiac arrhythmia
(Curran et al., 1995; Warmke and Ganetzky, 1994)
. Mutations in the
Drosophila

version of sarco
-
endoplasmic reticulum calcium
ATPase (SERCA), a membrane
pump important for maintaining calcium homeostasis in mammalian hearts, alters heart rate and
rhythm in flies
(Sanyal et al., 2006)
.

Fruit flies have also been used as a model organism for hypoxia research. Their innate
tolerance to anoxia, being able to recover completely from up to 4 hours without oxygen
(Haddad
et al., 1997)
, has sparked recent efforts to understand the molecular basis for this remarkable
tolerance. The pres
ence of the disaccharide trehalose increases anoxia tolerance in flies, probably
by protecting against protein desiccation and aggregation
(Chen et al., 2002)
, and its protective
effects were extended to human cells by transfecting the
Drosophila
trehalose synthase enzyme
(Chen et al., 2003)
. A mutagenesis screen for anoxia sensitivity discovered the
hypnos

genes
(Hadd
ad et al., 1997)
, one of which was later identified as pre
-
mRNA adenosine deaminase
(dADAR), which edits the mRNA sequences of a number of ion channels and is expressed mainly
in neurons
(Ma et al., 2001)
.

It is likely that a core set of genes for defending cells against oxygen fluctuations also
evolved early and has been conserved in evolution from flies to humans
(O'Farrell, 2001)
. For
example, hypoxia inducible factor HIF
-
1, the metabolic “master switch” AMP
-
activated protein
kinase, and nitric

oxide signaling all coordinate hypoxia adaptation in mammalian heart
(Jones
and Bolli, 2006; Kido et
al., 2005; Pan and Hardie, 2002)
, and are all present and functional in
flies
(Lavista
-
Llanos et al., 2002; Pan and Hardie, 2002; Wingrove and O'Farrell, 1999)
.
However, it is not known whether these hypoxia tolerance mechanisms are active in fly
myocardial tissue specifically.

The fly has
been a favorite subject of systems biology studies,
due to its established
popularity as a well
-
known model organism and also

by new high
-
throughput datasets and the
FlyBase genome and literature database

(Albert and Othmer, 2003; Stuart et al., 2006)
.
RNA
interference has been used in genome
-
wide screen
s for cell viability and for more specific
phenotypes relating to

developmental signaling pathways flies
(Boutros et al., 2004; Friedman
and Perrimon, 2006)
. Large
-
scale transcriptional profiling
studies have been performed
(Arbeitman et al., 2002; Furlong et al., 2001)

and the resulting gene expression data
were

used to
construct networks of co
-
expressed genes
(Stuart et al., 2003)
. Importantly, a

genome
-
scale
Drosophila

protein interaction map

(Giot et al., 2003)
, generated using the yeast two
-
hybrid
technique,
can be

used as a scaffold with which to
reconstruct cellular networks.


2.
Cardiac phenotyping


One st
rategy for discovering the complete set of genes influencing cardiac adaptation to
hypoxia in flies is to systematically knock out each gene and measure the hypoxic heart
phenotype. Due to the recent development of libraries and tools for creating genome
-
wide
perturbations, systematic cellular phenotype screens are now possible for a number of model
organisms
(Carpenter and Sabatini, 2004)
. However, a phenotypic screen at the level of the heart
organ is only feasible by perturbing the DNA sequence
directly, since techniques such as RNA
interference or adenoviral gene vectors, for example, do not yet have the transfection efficiency to
affect all cells in a mammalian heart
(Carpenter and Sabatini, 2004)
. Of the organisms with
publicly availab
le mutant libraries of DNA
-
level perturbations, only the fruit fly
Drosophila
melanogaster

has a heart and offers the possibility of a genetic screen of
function in the
heart
organ
.


2.1 Automated cardiac phenotyping

Directed, systematic screens have seve
ral advantages over random mutagenesis: every
gene can be screened, influential genes are instantly identified without follow
-
up sequencing, and
phenotypes can be quantified for all genes rather than the top scoring ‘hits’
(Carpenter and
Sabatini, 200
4)
. However, as opposed to mutagenesis screens which can have arbitrary duration
and scope, a s
ystematic genome
-
wide screen is, by definition,
not complete until each gene
perturbation has been examined.

New technology has made this daunting tas
k conceptually
feasible. When mounted on a transparent surface with its wings spread, the fly heart is easily
visible from the dorsal side by a light microscope. Taking advantage of this property, we and
others previously developed optical imaging method
s for rapid measurement of cardiac function
in adult flies
(Choma et al., 2006; Paternostro et al., 2001; Wolf et al., 2006)
. Although these
methods are computer
-
aided

for faster throughput, none are fast enough to screen the
approximately 14,000
Drosophila
genes within a reasonable timeframe. Improvements in
computer automation would not only improve the speed of current methods, it would also serve to
eliminate human

error from repeated experiments. We
recently
developed new automation for
rapid
in
-
vivo

measurement and analysis of the cardiac hypoxia response in adult
Drosophila
melanogaster.

An overview of the system is displayed in Figure
2
.

Batches of flies can b
e crossed and raised in parallel, but with a microscope
-
based system
measurements must be taken one at a time. Automation was added to increase the throughput at
several bottlenecks, including anesthetization and mounting, locating each fly heart under th
e
microscope, environmental control and hypoxia stimulus, and gathering M
-
mode

time
-
space

representations

of each heart

tube
.
Image analysis algorithms can automatically extract
functional information such as heart rate and diameter from the raw M
-
mode im
age.
Ideally, this
system can measure up to 50 flies per hour depending on the measurement duration

(unpublished
observation)
.


2.
2

Screening cardiac hypoxia phenotypes


Before any large
-
scale screen can be initiated, it is important to first define and
characterize the endpoint of inte
rest. For
any genetic

screen, we must
clearly

define what we
mean by c
ardiac hypoxia tolerance. Previous oxygen deprivation studies using
Drosophila

have
characterized the whole
-
body response. Adult flies become motionle
ss within seconds, and
recover full mobility, viability, and fertility even after 4 hours of total anoxia
(Haddad
et al.,
1997)
, while embryos and larvae survive even longer
(Wingrove and O'Farrell, 1999)
. In their
genetic screen, Haddad et. al.
(Haddad et al., 1997)

administered total anoxia for
5 minutes, then
measured the time to recover to a prone body position as the endpoint for distinguishing hypoxia
-
sensitive mutations.

Studies of ischemic cardioprotection in mammals use five major physiological endpoints:
infarct size, myocardial stunning
, recovery of mechanical function, arrhythmias, and
electrocardiographic changes
(Yellon and Downey, 2003)
. Although infarct size and
electrocardiograms are difficult to obtain in flie
s, especially in a high
-
throughput screen, our
imaging
-
based system can measure or closely approximate the other three end points.
Myocardial stunning refers to the loss of contractile function that occurs immediately after a
sublethal ischemic insult, wh
ich does not cause infarction and will fully recover within days
(Berne and Levy, 2001)
. Recovery of mechanical function can be measured in a variety of ways,
including wall motion i
n the region of ischemia. Arrhythmias are an indirect measure of severity
of the infarct, since scarring causes conduction defects that interrupt with normal action potential
propagation.
This

technology is able to measure real
-
time heart rhythms and wal
l motion under
hypoxic stress; therefore stunning, mechanical recovery, and post
-
ischemic arrhythmias are three
measures of function in the ischemic heart can be adapted for rapid screening in
Drosophila
. For
example, flies undergo a change in heart rate
resembling myocardial stunning in mammals
, and
dilation of the fly heart tube during hypoxia mirrors
change
s

in

contractility
.

Our observations show that the response of the heart to hypoxia in Drosophila differs
from that of nerve cells, justifying a sepa
rate screen. This is consistent with the differences
between these two tissues in many of the proposed molecular mechanisms of hypoxic damage,
e.g. differences in calcium handling or in the relative fluxes of metabolic pathways. Even
different parts of th
e central nervous system seem to differ greatly in their response to oxygen
deprivation

(Donnelly et al., 1992)
.


After long periods of anoxia, signs of recovery can be seen in
the heart before any body movements are detected (unpublished observation). The time course of
hypoxia in

adult flies and mammalian hearts, shown in Figure
3
, suggests a possible range of
hypoxia durations that may expose physiological differences in
a genetic screen. Before the
actual mutant screen, automated methods can help

to systematically characterize
wild
-
type
cardiac phenotypes under hypoxic stresses of varying durations and magnitudes. Measurements
of rate, rhythm, and fractional wall shortening during stress and recovery approximate the
physiological endpoints of stunning, functional recovery, and
arrhythmias.



3.
NMR metabolomics of hypoxia

Metabolomics is the comprehensive study of endogenous metabolites with the goal of
understanding their role in systems biology. Two of the most common approaches for profiling
metabolites and their changes ar
e nuclear magnetic resonance (NMR) spectroscopy and mass
spectrometry (MS)
(Goodacre et al., 2004; Nicholson et al., 2002; Watkins and German, 2002;
Whitfield et al
., 2004)
. These techniques have been used to study isolated metabolites for decades
and, because of the greatly increased sensitivity of modern methods and instrumentation, in the
last few years have been able to provide
more comprehensive metabolic profiles.


Metabolic profiling is central to our strategy for modeling adaptation to hypoxia in flies
because metabolite fluxes are, at the molecular level, the main phenotypic endpoint for adaptation
to hypoxia by signaling ne
tworks. Fluxomics is the direct measurement of these fluxes, usually
by using

metabolomic methods in order to track molecules labeled with an isotope such as
carbon
-
13. Most current metabolomics and fluxomics efforts use one

of two technologies:
nuclear

magnetic resonance spectroscopy (NMR spectroscopy or NMR), or mass spectrometry
(MS).
Liquid chromatography combined with electrospray ionization mass spectrometry
(LC/ESI
-
MS) has several advantages over NMR, including greater sensitivity and dynamic ran
ge

(Go et al., 2003; Siuzdak et al., 1996)
. Recent advances in magnet field strength and probe
technology
(Keun et
al., 2002a)

have also expanded the capability of NMR spectroscopy as a
metabolomics tool. Some metabolites can be measured with both NMR and mass spectrometry,
but others are best measured by only one of these techniques, for examples carbohydrates do n
ot
ionize well and for them it is preferable to use NMR.

NMR results generally provide more
accurate quantitative measurements

(Keun et al., 2002b)
.

Due to the low sensitivity of NMR methods and the difficulty in harvest
ing heart tissue in
the fly, it is difficult to obtain large enough samples of
Drosophila
myocardium for NMR
spectroscopy. Flight muscle, which takes up most of the mass of the thorax,

can provide a
reasonable substitute. Most of the literature on fruit fly biochemistry focuses on flight muscle
and its metabolic adaptation for flight, therefore reconstruction of the fly metabolism is easier for
flight muscle tissue. Like mammalian car
diac muscle, flight muscle uses aerobic pathways
during exercise, although these pathways rely on oxidation of pyruvate and reducing equivalents
derived from glycolysis, rather than beta
-
oxidation of fatty acids as occurs in mammalian heart.
Drosophila
fl
ight muscle is thought to rely on carbohydrates as the major substrate, whereas heart
tissue in mammals uses
both carbohydrates and
fatty acids under normal operation
(Opie, 1998)

(it is not known what substrates are used in the fly

heart).



3.1
Metabolomic profile of hypoxic flight muscle

We used
1
H NMR spectroscopy as a starting point for understandi
ng the pathways
involved in adaptation to hypoxia in
Drosophila
myocytes. This technique provides a broad
perspective of metabolite accumulation, in this case meaning metabolites with concentrations
above 0.01 mM. We identified 21 unique compounds that h
ad at least one sample measurement
greater than 0.05 mM, of which 6 were found to change significantly over 4 hours of 0.5%
oxygen: acetate, alanine, arginine, glucose, lactate, and threonine. Three compounds (acetate,
alanine, and lactate) had significan
t changes among several time points, following a trend of
accumulation. Accumulation and saturation of lactate, alanine, and acetate was seen, and a linear
approximation of these fluxes was calculated by regression analysis

(Feala et al., 2006)
.

The end products of anaer
obic metabolism in
Drosophila

were not previously known.
Lactate is the major fermentation product under hypoxic conditions in mammalian cells, but
alanine and acetate are not anaerobic end products in mammals (although some alanine is
normally secreted d
ue to amino acid breakdown in mammalian muscle). The discovery of lactate
and alanine accumulation was consistent with the fact that these are known to be byproducts of
anaerobic metabolism in other terrestrial insects
(Hoback and Stanley, 2001)
. These metabolites
do not accumulate under normoxic conditions. Even during flight, lactate is produced at a very
low rate since

most pyruvate is quickly oxidized in the mitochondria
(Gilmour, 1961)
. During
hypoxic conditions, fermentation to lactate regenerates NAD
+

for glycolysis, with the tradeoff of
decreasing pH. Certain invertebrates may have alanine and acetate fermentation pathways that
perform a similar function
(Hochachka and Somero, 2002)
, and our results suggest that
Drosophila

coordinates glycolysis with all three of these pathways of pyruvate metabolism in
order to cope with hypoxia. Modeling these pathways with flux
-
balance anal
ysis, discussed later,
provided insight as to how they might contribute to hypoxia tolerance.


3.2
Fluxomics

Flux
-
balance analysis calculates enzyme fluxes at steady state. Our first efforts used
1
H
NMR metabolomics to measure the time course of metabol
ites over 4 hours of hypoxia, for later
integration int
o the constraints
-
based model. Simulations relied on the system boundary
conditions defined by the disappearance of storage compounds (carbon influx), and the
appearance of anaerobic end products (car
bon efflux), with the network stoichiometry
constraining the possible intra
-
system flux solutions.
Although the NMR data represents
metabolite concentrations rather than fluxes, we assumed that increases in the metabolite pools
represent accumulation only
, and therefore approximated unidirectional, linear fluxes out of the
system. However, concentration measurements only provide a snapshot in time of the integral of
all fluxes into and out of a metabolite pool. Another problem with these data is that the

relatively
low sensitivity of NMR spectroscopy prevents intermediate metabolites from being quantified.
Only metabolite pools above 0.01 mM are accurately measured, which allows us to see substrates
and products but not the intermediates, which would be
t
ter identify specific pathways.


One possible technique to improve on these limitations is to use

isotopomer
-
based
fluxomics. Labeled isotopomers provide the possibility of tracing individual pathways and
accurately calculating enzyme fluxes. Carbon
-
13

is an isotope that can be measured by NMR
spectroscopy, and is naturally present at a rate of only about 1%. Enzymes do not distinguish
molecules containing
13
C from those with the natural
12
C isotope, allowing substrate molecules
such as glucose to be s
ubstituted with
13
C at various carbon positions and fed into the system., the
distribution of labeled metabolites and the positions of
13
C within those molecules can then be
measured using NMR spectroscopy. The network distribution of flux can be calculat
ed from the
data

(Sauer, 2006)
.



Some difficulties
should be

expected, and NMR spectroscopy has inherent limitations.
The method is limited by difficulties
including the necessity of water

suppression. Certain
metabolites, such as the important substrates trehalose, glycogen and fatty acids, need to be
assayed by other methods due to limitations in detection by NMR. Calculating reaction fluxes
from steady
-
s
tate isotopomer distributions is a challenge in higher organisms, however, atom
-
mapping matrices (AMMs)
(Zupke and Stephan
opoulos, 1994)

and isotopomer
-
mapping matrices
(IMMs)
(Schmidt et al., 1997)

provide a scalable mathematical framework that has been
successful in integrating these data with metabolic networks

(Vo and Palsson, 2006)
.


4.
Modeling hypoxic myocytes


Our first efforts to model metabolic adaptation to hypoxia focused on ATP
-
generatin
g
metabolism in flight muscle. Flux
-
balance analysis using the constraint
-
based model helped us to
understand how the three pyruvate pathways, hypothesized from the metabolomic data, might
stabilize pH and redox potential while maximizing ATP production u
nder low oxygen. Down
-
regulation of ATP consumption is likely to be extremely important in hypoxia tolerant organisms
as well, and anabolic metabolism must be included in the model to account for this. Hypoxia
-
induced alterations in the global flux distr
ibution (measured by
13
C NMR)
can

be used to
highlight important pathways to be carefully curated, while automated netwo
rk reconstruction
methods

can build a lower
-
confidence network scaffold around this highly curated core

(Notebaart et al., 2006)
. Then, this model of hypoxia adaptation in flight muscle can be refined
for myocardial tissue using hypoxia
-
tolerance phenotypes of enzyme mutations.


4.1
Reconstructing
Drosophila
central metabolism

We completed a draft
in
-
silico
reconstruction of ATP
-
generating metabolism in
Drosophila
flight muscle, using the biological network database and simulation softw
are
Simpheny (Genomatica,
San Diego). This model is curated, meaning that all genes and reactions
in the reconstruction were individually inspected before inclusion into the model. A legacy of
literature exists from research on
Drosophila
flight

muscle performed over the last
several

decades. These papers, plus those describing the well
-
studied and closely
-
related Diptera
Phormia terranova

(blowfly), were used to verify the existence of each reaction in
Drosophila
flight muscle. Element
-

and ch
arge
-
balanced reactions were included based on evidence from
sequence homology in the annotated genome, online enzyme databases, and the literature. Efforts
were made to only include pathways specific to muscle tissue, although generic reactions were
adde
d when necessary to close gaps. Mitochondrial
or

cytosolic
compartmentalization was

also
included when available. Reed et. al
(2006)

reviews the reconstruction process in detail.


High
-
throughput data
measured in controlled studies
can be used to suggest the existence
or

absence of enzymes when refining for a specific cell type

or condition. Vo and Palsson
(2004)

based their reconstruction of human cardiac mitochondria on a list of proteins discovered in a
previous proteomic analysis.

Transcriptomics data have been gathe
red for fly thorax

(Girardot et
al., 2006)
, and fly mitochondria were examined by proteomics methods

(Alonso et al., 20
05)
.
Although these techniques are likely to miss mRNA and proteins present in low concentrations,
many enzymes can be found in these gene and p
rotein lists, which
can

refine
the

model to better
represent the
muscle cell type.


4.2
Simulating hypoxic
flight muscle

In hypoxia, the overall reduction in energy production is thought to be less important to
cardiac dysfunction than the loss of metabolite balance within the system

(Ingwall, 2002)
. Flux
-
balance analysis (FBA) on the network reconstruction can quantitatively model these
relationships. One major advantag
e of a mathematical model is that conservation of mass is
enforced; therefore all elements and charges are balanced within the system, including electron
transport, cofactor concentration, and protons (pH). Thus, the complicated problems of balancing
redo
x potential and accounting for ATP and proton production are solved intrinsically during
simulation.
In flux
-
balance analysis, reactions and their associated biochemicals are represented
mathematically as a matrix of stoichiometric relationships that can
be manipulated with linear
algebra to reveal the solution to the equation


d
x
/dt =
S
*
v

=
0
,


where
x

is the vector of metabolites,
S

is the stoichiometric matrix and
v

is the vector of reaction
fluxes at steady state
(Schilling et al., 2000)
. The null space of S is the set of possi
ble flux
vectors that satisfy this steady state condition. See
(Papin et al., 2003)

and
(Kauffman et al.,
2003)

for

a review of the mathematics involved in flux
-
balance analysis.

FBA requires an “objective function” to maximize in order to select an optimum set of
steady
-
state fluxes from the vast solution space. In any myocyte, the objective of metabolism is
primaril
y to provide ATP to the energy
-
consuming myosin cross
-
bridges and ion pumps. In order
to isolate the effects of hypoxia, we varied O
2

uptake, leaving all other rates constant, and used
linear optimization to find the network states which maximize ATP prod
uction under c
hanges in
oxygen availability.

The result of FBA is the distribution of fluxes that maximize the objective function under
the imposed constraints. The main type of constraint commonly used in these models is restricted
metabolite influx and
efflux rates. Experimentally measured fluxes can be applied to the model
directly, reducing the degrees of freedom of the solution space and refining the model for a
specific context. In order to focus the model on hypoxic muscle metabolism, we used resu
lts
from NMR metabolomics and biochemical assays that suggested that lactate, alanine, and acetate
are the major metabolites that accumulate under hypoxic conditions. Anaerobic pathways for
generating these end products were hypothesized, and the correspo
nding reactions were linked to
existing
Drosophila
genes and built into the model. Measurements of the steady
-
state fluxes of
these compounds were integrated into the model as constraints on the flux out of the system and
simulations were conducted to inv
estigate production of ATP, H
+
, and

glucose during hypoxia.
Simulations suggested that the flexibility of pyruvate metabolism provided by these three
pathways might provide extra tolerance to hypoxia by increasing the ratio of H
+

produced per
ATP generate
d while simultaneously reducing the amount of carbohydrate fuel required per ATP

(Feala et al., 2006)
.


Flux balance analysis is not the only technique that can be use
d to study network
properties. The constraint
-
based approach includes a number of analytical tools
, suc
h as
decomposition of the network into basis vectors or
Monte Carlo sampling to estimate the shape of
the solution space, variability and correlations between fluxes throughout this space, or deletion
analysis
(Palsson, 2006; Price et al., 2004)
. Me
tabolic control analysis, which
examines the
sensitivity of certain target fluxes or flux groups to
perturbations
,
could

be used to

identify
control points and important enzymatic mediators of adaptation to hypoxia

(Stephanopoulos,
1999; Stephanopoulos et al., 1998)
.


4.3
Expanding the model


The number of published metabolic reconstructions continues to grow every year.
Althou
gh the majority of available models are for microbes, reconstructions of more complex
organisms and subsystems are becoming more common. For example, curated models of the
human cardiac mitochondria, human red blood cell, generic mouse metabolism, and mou
se
cardiomyocyte have been made publicly available
(Morel et al., 2006; Palsson et al., 1989;
Sheikh et al., 2005; Vo et al., 2004)
. Recently, computer tools have appeared that use the
complete set of ope
n reading frames (ORFs) in an annotated genome, sometimes combined with
information from existing, manually curated models, to predict reaction networks and gene
-
reaction associations
(Karp et al., 2002; Notebaart et al., 2006)
.
Automated methods may
accelerate the process of building genome
-
scale reconstruc
tions,
though the resulting network
must itself be curated.

Details such as substrate specificity, exact reaction stoichiometry, cofactor
usage, and compartmentalization are extremely important for flux
-
balance analysis
(Reed et al.,
2006)

but are unlikely to be produced correctly by any c
omputer algorithm.


4.4
Refining for cardiac myocytes

Fluxes measured by isotopomer labeling
can

be integrated into the metabolic model as
upper and lower bounds on reaction rates, following the example of Vo and Palsson
(2006)
,
resulting in a model experimentally refined to represent
the hypoxic myocyte.
To adapt this
model to the cardiac myocyte, r
eactions from the existing models of mouse cardiac myocyte and
human cardiac mitochondria

can be used
as a guide for molding the reaction network to resemble
these eukaryotic heart models.


Cardiac phenotypes of enzyme mutants can also provide information about reactions
specific for myocardial cells.


Systematic perturbation analysis


measurements
of cardiac
phenotypes for a
ll available enzyme mutations


can
be compared with the results
of an
in
-
silico

deletion analysis. The results of simulated deletions using flux
-
balance analysis will be compared
to cardiac function in an all
-
or
-
nothing fashion: if the mutation produces a significant decrease in
hypoxia tolerance, the simulated enzym
e deletion should produce a change in ATP production.
Heterozygous mutants present a problem since the level of gene expression can be variable, but
sensitivity analysis can be performed for heterozygous mutants to help understand how ATP
output relates t
o small changes in enzyme activity. Discrepancies between experiment and
simulation can lead to changes to the pathways or flux constraints in the model.


5.
Integrating
signaling

networks


During acute hypoxia, there are many changes in metabolic energy production and
utilization pathways as the system adapts to the anaerobic state. These changes happen quickly,
on the order of minutes in
Drosophila
flight muscle, according to our prelimina
ry research. It is
not possible for new genes to be expressed and translated into proteins on these short timescales;
therefore, the networks involved in metabolic regulation are strictly composed of enzymes,
signaling proteins and allosteric interactions

with metabolites. Although no high
-
throughput
methods have yet been developed to measure global protein
-
metabolite interactions, the yeast
-
two
-
hybrid method has been successful at cataloguing immense numbers of protein
-
protein
interactions for the constr
uction of global networks. These

networks
have drawbacks su
ch as
high false
-
positive rate,

low coverage, and no cellular compartmentalization

(Ito et al., 2002)
, but
they can nevertheless be useful in creating a scaffold for interpreting
new

data.

Current methods for quantitatively modeling protein signaling are gen
erally restricted to
small numbers of variables, requiring many kinetic parameters.
An alternative is to f
ocus on the
discovery of new regulatory modules rather than detailed modeling of

well
-
known players.
Regulatory control

of metabolism
must

be modele
d within a separate framework from the
constraint
-
based model, but the relationship between reaction fluxes and enzyme proteins
can

be
used to integrate the two.


Though it is currently impossible to quantify all of the kinetics of specific protein
interac
tion in genome
-
wide network models, other numerical data can be integrated with
qualitative interaction maps in order to build statistical models of function. For example,
bidirectional protein interaction networks have been incorporated with microarray d
ata
(Goodacre
et al., 2004; Ideker et al., 2002)
, quantitative phenotype measu
rements
(Begley et al., 2002; Han
et al., 2004)
, an
d genetic interactions
(Kelley et al., 2004)

to infer function. In these studies,
statistical algorithms are used to extract hypothetical functional modules from these integrated

networks.


The open
-
source
Cytoscape

software
(Shannon et al., 2003)

(with the useful online
database plug
-
in cPath)
(Cerami et al., 2006)
,
or other biological network software can be used to
combine, store, and analyze interacti
on datasets for
Drosophil
a
from public
databases
. Known
hypoxia response pathways from the literature can be manually entered into this network as well.
The software keeps track of the data source and nature (e.g. metabolite
-
protein or protein
-
protein)
o
f each node and link, allowing follow
-
up of features of interest. Under the (
somewhat
limited)
assu
mption that these networks have good
coverage of the
true

‘interactome,’ most proteins
responsible for mediating the metabolic adaptation to hypoxia would a
ppear as the first or second
neighbors of the regulated enzymes. Therefore, all genes coding for proteins one or two links
from enzymes in the metabolic network
could

be added to
a

list of candidate genes to be screened
for cardiac hypoxia phenotype.


T
he
Drosophila
protein
interaction network

can be combined

with cardiac phenotype
data and
differential fluxes from
simulation
s of normoxia and hypoxia
. The general overview of
this network integrat
ion is illustrated in Figure 3.
The network can

then

be a
nalyzed to extract
and vi
sualize high
-
scoring modules. A

publicly available Cytoscape plug
-
in, 'ActiveModules

(Ideker et al., 2002)
,

implements a greedy annealing algorithm to simulate a “shake out” of high
-
scoring modules from the network.


Modules discovered by this method represent new hypotheses for proteins and p
athways
that regulate metabolism during hypoxia. Nodes with significant phenotypes may be either
enzymes or signaling proteins, while closely connected nodes with high flux scores represent the
end point of regulatory signaling. More importantly, this ap
proach can be used to predict
functions of additional proteins. Although individual inspection of the datasets exposes all

significant

“hits” discovered experimentally,
analyzing

the results in the context of the network
can reveal important, closely conn
ected nodes that were either not measured or not significant.
As another example, it is possible for an extracted module to include no high
-
scoring nodes but
many closely connected proteins with mid
-
range scores. These are insights that would be missed
b
y inspecting each dataset individually.


Iteration and detailed follow
-
up

Systems analysis of the cardiac hypoxia response generate
s

hypotheses for essential
pathways and proteins that can be easily tested using mutants or biochemical methods. For
example
, we used metabo
lomics and modeling to predict

pathway
s

for the anaerobic production
of acetate
and alanine
under hypoxic conditions. This is being tested with iterative application of
our high
-
throughput techniques, gathering metabolomic profiles and car
diac phenot
ypes of
mutants in these pathways.


Hypoxia sensitive mutants can be further verified by more detailed experimental methods,
for example by producing transgenic flies for temporal control of gene expression, using cardiac
-
specific promoters can
be used for tissue specificity. In this way, molecular biology has a
complementary role in coarse
-
grained systems analyses, because any results from the model are
only hypotheses until they are validated by standard techniques. Valid results can be enter
ed
back into the model, improving its accuracy for future simulations.


Discussion

Our systems analysis of the hypoxia response in
ATP
-
producing pathways in
Drosophila
flight muscle provided several insights and hypotheses for molecular mechani
sms contributing to
their remarkable hypoxia tolerance. A metabolomic assay of high
-
concentration small molecules
revealed that acetate and alanine accompany lactate as an anaerobic end product.
Modeling this
activity within a
metabolic network reconstru
ction


built from the annotated genome, the KEGG
gene and
pathway database, and
Drosophila

biochemistry literature



produced these end
products under oxygen deprivation and showed why
the corresponding

pathways might aid
survival.

T
he other half of the
anaerobic shift
,
down
-
regulation of ATP
-
consuming pathways,
was not
modeled

but plays a crucial role in survival during hypoxia.
Therefore our model

is

only
a first step toward developing a global model

that represents both ATP supply and demand, with
a g
reater relevance to the heart.

In order to extend this approach to cardiac myocytes, both the model and its underlying
data
can be expanded
to
represent
global
regulation of metabolism in

the heart.
One possible
step
is to refine the reactions in the mode
l using
more myocyte
-
specific data


starting with
existing
microarray and proteomic

data for fly thorax (flight muscle),
adding fluxomic data from a
13
C
isotopomer study,
then by perturbation analysis using cardiac phenotypes of
enzymatic mutants.
This e
xperimentally
-
based, global model of hypoxic flight muscle metabolism can

then

be
studied further using a variety of constraint
-
based techniques.
Next, to expand the model to
encompass possible signaling networks acting on enzymes,
it is possible to
integ
rate a protein
interaction network (from online interaction databases) and measure cardiac phenotypes for all
close neighbors to enzymes in the metabolic model.
A
lthough the

dense interaction network
cannot be functionally simulated in the same way as the

metabolic model, statistical methods
can

be used to discover highly connected modules enriched

with large phenotype changes. E
ach of
the steps in this approach
ha
s had previous success under some other context
. This review
discussed

prior

work
to
justif
y the value and feasibility of each step, and attempted to show how
they might be combined into a cohesive methodology for studying hypoxic metabolism in cardiac
myocytes.

Once a genome
-
scale systems model of the hypoxia response is constructed, it can be
used to test other factors influencing this defense system. An important example is aging. We
previously showed that aging genes are more highly connected within biologic
al networks, and
showed that restoration of highly
-
connected ‘hubs’ restored function in a generic network model
of generalized degradation
(Ferrarini et al., 2005)
.
T
his method
can be repeated
using the model
described here, in order to try to understand the dysregulation of hypoxia response that would be
expected to occur with aging. Metabolomic and phenomic data from young and old f
lies
could

be
gathered and interpreted within the network for comparison.


E
fforts to bring the data and models together under the context of acute cardiac hypoxia
could be replicated for other systems of interest, possibly using some of the data or networ
ks
presented here. Conversely, it would be possible for future work on

Drosophila

cardiac hypoxia
to use this integrated model as a framework with which to incorporate and interpret further
models or experiments. A major goal for systems biology is to pr
ovide scaffolds for storing and
integrating the overwhelming data both in the existing literature and in new high
-
throughput
experiments. For example, our network incorporates the
Drosophila
annotated genome with
decades of literature on fly metabolism as

well as our newly gathered metabolomic and
phenotype data.

We have taken a

discover
y
-
based approach to find sys
tem
-
wide properties of the cardiac
hypoxia response, integrating both top
-
down and bottom
-
up data. Network m
odels are a way of
bridging the

g
ap across multiple functional scales. The incorporation of genome
-
wide data
provides a coarse
-
grained view of the system, complementary to a more detailed biochemical and
molecular biology approach. By ensuring that all important components are accounted

for, the
global approach is an unbiased way to focus
future,

more detailed
efforts

on the most important
parts of the system.

References


Adams, M. D., Celniker, S. E., Holt, R. A., Evans, C. A., Gocayne, J. D., Amanatides, P.
G., Scher
er, S. E., Li, P. W., Hoskins, R. A., Galle, R. F., George, R. A., Lewis, S.
E., Richards, S., Ashburner, M., Henderson, S. N., Sutton, G. G., Wortman, J. R.,
Yandell, M. D., Zhang, Q., Chen, L. X., Brandon, R. C., Rogers, Y. H., Blazej, R.
G., Champe, M.,

Pfeiffer, B. D., Wan, K. H., Doyle, C., Baxter, E. G., Helt, G.,
Nelson, C. R., Gabor, G. L., Abril, J. F., Agbayani, A., An, H. J., Andrews
-
Pfannkoch, C., Baldwin, D., Ballew, R. M., Basu, A., Baxendale, J.,
Bayraktaroglu, L., Beasley, E. M., Beeson, K.
Y., Benos, P. V., Berman, B. P.,
Bhandari, D., Bolshakov, S., Borkova, D., Botchan, M. R., Bouck, J., Brokstein,
P., Brottier, P., Burtis, K. C., Busam, D. A., Butler, H., Cadieu, E., Center, A.,
Chandra, I., Cherry, J. M., Cawley, S., Dahlke, C., Davenpor
t, L. B., Davies, P.,
de Pablos, B., Delcher, A., Deng, Z., Mays, A. D., Dew, I., Dietz, S. M., Dodson,
K., Doup, L. E., Downes, M., Dugan
-
Rocha, S., Dunkov, B. C., Dunn, P., Durbin,
K. J., Evangelista, C. C., Ferraz, C., Ferriera, S., Fleischmann, W., Fos
ler, C.,
Gabrielian, A. E., Garg, N. S., Gelbart, W. M., Glasser, K., Glodek, A., Gong, F.,
Gorrell, J. H., Gu, Z., Guan, P., Harris, M., Harris, N. L., Harvey, D., Heiman, T.
J., Hernandez, J. R., Houck, J., Hostin, D., Houston, K. A., Howland, T. J., Wei
,
M. H., Ibegwam, C., et al. (2000) The genome sequence of Drosophila
melanogaster.
Science

287,

2185
-
95.

Albert, R. and Othmer, H. G. (2003) The topology of the regulatory interactions predicts
the expression pattern of the segment polarity genes in Droso
phila melanogaster.
J Theor Biol

223,

1
-
18.

Alonso, J., Rodriguez, J. M., Baena
-
Lopez, L. A. and Santaren, J. F. (2005)
Characterization of the Drosophila melanogaster mitochondrial proteome.
J
Proteome Res

4,

1636
-
45.

Arbeitman, M. N., Furlong, E. E., Ima
m, F., Johnson, E., Null, B. H., Baker, B. S.,
Krasnow, M. A., Scott, M. P., Davis, R. W. and White, K. P. (2002) Gene
expression during the life cycle of Drosophila melanogaster.
Science

297,

2270
-
5.

Bader, G. D., Betel, D. and Hogue, C. W. (2003) BIND: t
he Biomolecular Interaction
Network Database.
Nucleic Acids Res

31,

248
-
50.

Barabasi, A. L. and Oltvai, Z. N. (2004) Network biology: understanding the cell's
functional organization.
Nat Rev Genet

5,

101
-
13.

B
egley, T. J., Rosenbach, A. S., Ideker, T. and Samson, L. D. (2002) Damage recovery
pathways in Saccharomyces cerevisiae revealed by genomic phenotyping and
interactome mapping.
Mol Cancer Res

1,

103
-
12.

Bellen, H. J., Levis, R. W., Liao, G., He, Y., Carls
on, J. W., Tsang, G., Evans
-
Holm, M.,
Hiesinger, P. R., Schulze, K. L., Rubin, G. M., Hoskins, R. A. and Spradling, A.
C. (2004) The BDGP gene disruption project: single transposon insertions
associated with 40% of Drosophila genes.
Genetics

167,

761
-
81.

B
erne, R. and Levy, M. (2001) Cardiovascular Physiology. In
Mosby Physiology
Monograph Series
, Mosby, St. Louis, Missouri.

Bier, E. and Bodmer, R. (2004) Drosophila, an emerging model for cardiac disease.
Gene

342,

1
-
11.

Bodmer, R. and Venkatesh, T. V. (199
8) Heart development in Drosophila and
vertebrates: conservation of molecular mechanisms.
Dev Genet

22,

181
-
6.

Boutros, M., Kiger, A. A., Armknecht, S., Kerr, K., Hild, M., Koch, B., Haas, S. A., Paro,
R. and Perrimon, N. (2004) Genome
-
wide RNAi analysis o
f growth and viability
in Drosophila cells.
Science

303,

832
-
5.

Breitkreutz, B. J., Stark, C. and Tyers, M. (2003) The GRID: the General Repository for
Interaction Datasets.
Genome Biol

4,

R23.

Carpenter, A. E. and Sabatini, D. M. (2004) Systematic genome
-
wide screens of gene
function.
Nat Rev Genet

5,

11
-
22.

Cerami, E. G., Bader, G. D., Gross, B. E. and Sander, C. (2006) cPath: open source
software for collecting, storing, and querying biological pathways.
BMC
Bioinformatics

7,

497.

Chen, Q., Behar, K. L.,

Xu, T., Fan, C. and Haddad, G. G. (2003) Expression of
Drosophila trehalose
-
phosphate synthase in HEK
-
293 cells increases hypoxia
tolerance.
J Biol Chem

278,

49113
-
8.

Chen, Q., Ma, E., Behar, K. L., Xu, T. and Haddad, G. G. (2002) Role of trehalose
phosph
ate synthase in anoxia tolerance and development in Drosophila
melanogaster.
J Biol Chem

277,

3274
-
9.

Choma, M. A., Izatt, S. D., Wessells, R. J., Bodmer, R. and Izatt, J. A. (2006) Images in
cardiovascular medicine: in vivo imaging of the adult Drosophila

melanogaster
heart with real
-
time optical coherence tomography.
Circulation

114,

e35
-
6.

Cuong, D. V., Kim, N., Youm, J. B., Joo, H., Warda, M., Lee, J. W., Park, W. S., Kim, T.,
Kang, S., Kim, H. and Han, J. (2006) Nitric oxide
-
cGMP
-
protein kinase G
signa
ling pathway induces anoxic preconditioning through activation of ATP
-
sensitive K+ channels in rat hearts.
Am J Physiol Heart Circ Physiol

290,

H1808
-
17.

Curran, M. E., Splawski, I., Timothy, K. W., Vincent, G. M., Green, E. D. and Keating,
M. T. (1995) A
molecular basis for cardiac arrhythmia: HERG mutations cause
long QT syndrome.
Cell

80,

795
-
803.

De Celle, T., Vanrobaeys, F., Lijnen, P., Blankesteijn, W. M., Heeneman, S., Van
Beeumen, J., Devreese, B., Smits, J. F. and Janssen, B. J. (2005) Alterations
in
mouse cardiac proteome after in vivo myocardial infarction: permanent ischaemia
versus ischaemia
-
reperfusion.
Exp Physiol

90,

593
-
606.

Donnelly, D. F., Jiang, C. and Haddad, G. G. (1992) Comparative responses of brain
stem and hippocampal neurons to O2
deprivation: in vitro intracellular studies.
Am J Physiol

262,

L549
-
54.

Du, Q., Jovanovic, S., Clelland, A., Sukhodub, A., Budas, G., Phelan, K., Murray
-
Tait,
V., Malone, L. and Jovanovic, A. (2006) Overexpression of SUR2A generates a
cardiac phenotype res
istant to ischemia.
Faseb J

20,

1131
-
41.

Duarte, N. C., Herrgard, M. J. and Palsson, B. O. (2004) Reconstruction and validation of
Saccharomyces cerevisiae iND750, a fully compartmentalized genome
-
scale
metabolic model.
Genome Res

14,

1298
-
309.

Edwards, J.

S., Ibarra, R. U. and Palsson, B. O. (2001) In silico predictions of Escherichia
coli metabolic capabilities are consistent with experimental data.
Nat Biotechnol

19,

125
-
30.

Feala, J., Coquin, L., McCulloch, A. and Paternostro, G. (2006) Flexibility in e
nergy
metabolism supports hypoxia tolerance in Drosophila flight muscle: metabolomic
and computational systems analysis.
Mol Syst Biol
-

submitted
.

Ferrarini, L., Bertelli, L., Feala, J., McCulloch, A. D. and Paternostro, G. (2005) A more
efficient search

strategy for aging genes based on connectivity.
Bioinformatics

21,

338
-
48.

FlyBase. (2003) The FlyBase database of the Drosophila genome projects and
community literature.
Nucleic Acids Res

31,

172
-
5.

Fortini, M. E., Skupski, M. P., Boguski, M. S. and Har
iharan, I. K. (2000) A survey of
human disease gene counterparts in the Drosophila genome.
J Cell Biol

150,

F23
-
30.

Frey, N. and Olson, E. N. (2003) Cardiac hypertrophy: the good, the bad, and the ugly.
Annu Rev Physiol

65,

45
-
79.

Friedman, A. and Perrimon
, N. (2006) A functional RNAi screen for regulators of
receptor tyrosine kinase and ERK signalling.
Nature

444,

230
-
4.

Furlong, E. E., Andersen, E. C., Null, B., White, K. P. and Scott, M. P. (2001) Patterns of
gene expression during Drosophila mesoderm de
velopment.
Science

293,

1629
-
33.

Gilmour, D. (1961) The biochemistry of insects, pp. 343 p., Academic Press, New York,.

Giot, L., Bader, J. S., Brouwer, C., Chaudhuri, A., Kuang, B., Li, Y., Hao, Y. L., Ooi, C.
E., Godwin, B., Vitols, E., Vijayadamodar, G.
, Pochart, P., Machineni, H., Welsh,
M., Kong, Y., Zerhusen, B., Malcolm, R., Varrone, Z., Collis, A., Minto, M.,
Burgess, S., McDaniel, L., Stimpson, E., Spriggs, F., Williams, J., Neurath, K.,
Ioime, N., Agee, M., Voss, E., Furtak, K., Renzulli, R., Aane
nsen, N., Carrolla, S.,
Bickelhaupt, E., Lazovatsky, Y., DaSilva, A., Zhong, J., Stanyon, C. A., Finley, R.
L., Jr., White, K. P., Braverman, M., Jarvie, T., Gold, S., Leach, M., Knight, J.,
Shimkets, R. A., McKenna, M. P., Chant, J. and Rothberg, J. M. (2
003) A protein
interaction map of Drosophila melanogaster.
Science

302,

1727
-
36.

Girardot, F., Lasbleiz, C., Monnier, V. and Tricoire, H. (2006) Specific age
-
related
signatures in Drosophila body parts transcriptome.
BMC Genomics

7,

69.

Go, E. P., Prenni,
J. E., Wei, J., Jones, A., Hall, S. C., Witkowska, H. E., Shen, Z. and
Siuzdak, G. (2003) Desorption/ionization on silicon time
-
of
-
flight/time
-
of
-
flight
mass spectrometry.
Anal Chem

75,

2504
-
6.

Goodacre, R., Vaidyanathan, S., Dunn, W. B., Harrigan, G. G. a
nd Kell, D. B. (2004)
Metabolomics by numbers: acquiring and understanding global metabolite data.
Trends Biotechnol

22,

245
-
52.

Gu, G. G. and Singh, S. (1995) Pharmacological analysis of heartbeat in Drosophila.
J
Neurobiol

28,

269
-
80.

Haddad, G. G., Wyma
n, R. J., Mohsenin, A., Sun, Y. and Krishnan, S. N. (1997)
Behavioral and Electrophysiologic Responses of Drosophila melanogaster to
Prolonged Periods of Anoxia.
J Insect Physiol

43,

203
-
210.

Han, J. D., Bertin, N., Hao, T., Goldberg, D. S., Berriz, G. F.,

Zhang, L. V., Dupuy, D.,
Walhout, A. J., Cusick, M. E., Roth, F. P. and Vidal, M. (2004) Evidence for
dynamically organized modularity in the yeast protein
-
protein interaction network.
Nature

430,

88
-
93.

Hoback, W. W. and Stanley, D. W. (2001) Insects in
hypoxia.
J Insect Physiol

47,

533
-
542.

Hochachka, P. W. and Somero, G. N. (2002) Biochemical adaptation : mechanism and
process in physiological evolution, pp. xi, 466 p., Oxford University Press, New
York.

Hucka, M., Finney, A., Sauro, H. M., Bolouri, H.,

Doyle, J. C., Kitano, H., Arkin, A. P.,
Bornstein, B. J., Bray, D., Cornish
-
Bowden, A., Cuellar, A. A., Dronov, S., Gilles,
E. D., Ginkel, M., Gor, V., Goryanin, II, Hedley, W. J., Hodgman, T. C., Hofmeyr,
J. H., Hunter, P. J., Juty, N. S., Kasberger, J.
L., Kremling, A., Kummer, U., Le
Novere, N., Loew, L. M., Lucio, D., Mendes, P., Minch, E., Mjolsness, E. D.,
Nakayama, Y., Nelson, M. R., Nielsen, P. F., Sakurada, T., Schaff, J. C., Shapiro,
B. E., Shimizu, T. S., Spence, H. D., Stelling, J., Takahashi,
K., Tomita, M.,
Wagner, J. and Wang, J. (2003) The systems biology markup language (SBML): a
medium for representation and exchange of biochemical network models.
Bioinformatics

19,

524
-
31.

Ibarra, R. U., Edwards, J. S. and Palsson, B. O. (2002) Escherichi
a coli K
-
12 undergoes
adaptive evolution to achieve in silico predicted optimal growth.
Nature

420,

186
-
9.

Ideker, T., Galitski, T. and Hood, L. (2001) A new approach to decoding life: systems
biology.
Annu Rev Genomics Hum Genet

2,

343
-
72.

Ideker, T., Ozi
er, O., Schwikowski, B. and Siegel, A. F. (2002) Discovering regulatory
and signalling circuits in molecular interaction networks.
Bioinformatics

18 Suppl
1,

S233
-
40.

Ikeda, Y., Hiroi, Y., Hosoda, T., Utsunomiya, T., Matsuo, S., Ito, T., Inoue, J., Sumiyos
hi,
T., Takano, H., Nagai, R. and Komuro, I. (2002) Novel point mutation in the
cardiac transcription factor CSX/NKX2.5 associated with congenital heart disease.
Circ J

66,

561
-
3.

Ingwall, J. (2002) ATP and the Heart, Kluwer Academic Publishers, Boston, Lo
ndon.

Irie, H., Krukenkamp, I. B., Brinkmann, J. F., Gaudette, G. R., Saltman, A. E., Jou, W.,
Glatz, J. F., Abumrad, N. A. and Ibrahimi, A. (2003) Myocardial recovery from
ischemia is impaired in CD36
-
null mice and restored by myocyte CD36
expression or m
edium
-
chain fatty acids.
Proc Natl Acad Sci U S A

100,

6819
-
24.

Ito, T., Ota, K., Kubota, H., Yamaguchi, Y., Chiba, T., Sakuraba, K. and Yoshida, M.
(2002) Roles for the two
-
hybrid system in exploration of the yeast protein
interactome.
Mol Cell Proteomics

1,

561
-
6.

Jennings, R. B., Reimer, K. A. and Steenbergen, C. (1986) Myocardial ischemia revisited.
The osmolar load, membrane damage, and reperfusion.
J Mol Cell Cardiol

18,

769
-
80.

Johnson, E., Ringo, J., Bray, N. and Dowse, H. (1998) Genetic and pharmac
ological
identification of ion channels central to the Drosophila cardiac pacemaker.
J
Neurogenet

12,

1
-
24.

Jones, S. P. and Bolli, R. (2006) The ubiquitous role of nitric oxide in cardioprotection.
J
Mol Cell Cardiol

40,

16
-
23.

Karp, P. D., Paley, S. and
Romero, P. (2002) The Pathway Tools software.
Bioinformatics

18 Suppl 1,

S225
-
32.

Kauffman, K. J., Prakash, P. and Edwards, J. S. (2003) Advances in flux balance analysis.
Curr Opin Biotechnol

14,

491
-
6.

Kell, D. B. and Oliver, S. G. (2004) Here is the evi
dence, now what is the hypothesis?
The complementary roles of inductive and hypothesis
-
driven science in the post
-
genomic era.
Bioessays

26,

99
-
105.

Kelley, B. P., Yuan, B., Lewitter, F., Sharan, R., Stockwell, B. R. and Ideker, T. (2004)
PathBLAST: a tool

for alignment of protein interaction networks.
Nucleic Acids
Res

32,

W83
-
8.

Keun, H. C., Beckonert, O., Griffin, J. L., Richter, C., Moskau, D., Lindon, J. C. and
Nicholson, J. K. (2002a) Cryogenic probe 13C NMR spectroscopy of urine for
metabonomic studi
es.
Anal Chem

74,

4588
-
93.

Keun, H. C., Ebbels, T. M., Antti, H., Bollard, M. E., Beckonert, O., Schlotterbeck, G.,
Senn, H., Niederhauser, U., Holmes, E., Lindon, J. C. and Nicholson, J. K.
(2002b) Analytical reproducibility in (1)H NMR
-
based metabonomic
urinalysis.
Chem Res Toxicol

15,

1380
-
6.

Kido, M., Du, L., Sullivan, C. C., Li, X., Deutsch, R., Jamieson, S. W. and Thistlethwaite,
P. A. (2005) Hypoxia
-
inducible factor 1
-
alpha reduces infarction and attenuates
progression of cardiac dysfunction after my
ocardial infarction in the mouse.
J Am
Coll Cardiol

46,

2116
-
24.

Kolar, F. and Ostadal, B. (2004) Molecular mechanisms of cardiac protection by
adaptation to chronic hypoxia.
Physiol Res

53 Suppl 1,

S3
-
13.

Lavista
-
Llanos, S., Centanin, L., Irisarri, M., Ru
sso, D. M., Gleadle, J. M., Bocca, S. N.,
Muzzopappa, M., Ratcliffe, P. J. and Wappner, P. (2002) Control of the hypoxic
response in Drosophila melanogaster by the basic helix
-
loop
-
helix PAS protein
similar.
Mol Cell Biol

22,

6842
-
53.

Le Novere, N., Bornst
ein, B., Broicher, A., Courtot, M., Donizelli, M., Dharuri, H., Li, L.,
Sauro, H., Schilstra, M., Shapiro, B., Snoep, J. L. and Hucka, M. (2006)
BioModels Database: a free, centralized database of curated, published,
quantitative kinetic models of biochemi
cal and cellular systems.
Nucleic Acids
Res

34,

D689
-
91.

Lints, T. J., Parsons, L. M., Hartley, L., Lyons, I. and Harvey, R. P. (1993) Nkx
-
2.5: a
novel murine homeobox gene expressed in early heart progenitor cells and their
myogenic descendants.
Developme
nt

119,

419
-
31.

Lloyd, C. M., Halstead, M. D. and Nielsen, P. F. (2004) CellML: its future, present and
past.
Prog Biophys Mol Biol

85,

433
-
50.

Ma, E., Gu, X. Q., Wu, X., Xu, T. and Haddad, G. G. (2001) Mutation in pre
-
mRNA
adenosine deaminase markedly att
enuates neuronal tolerance to O2 deprivation in
Drosophila melanogaster.
J Clin Invest

107,

685
-
93.

McCulloch, A. D. and Paternostro, G. (2005) Cardiac systems biology.
Ann N Y Acad Sci

1047,

283
-
95.

Morel, S., Milano, G., Ludunge, K. M., Corno, A. F., Sam
aja, M., Fleury, S., Bonny, C.,
Kappenberger, L., von Segesser, L. K. and Vassalli, G. (2006) Brief
reoxygenation episodes during chronic hypoxia enhance posthypoxic recovery of
LV function : Role of mitogen
-
activated protein kinase signaling pathways.
Bas
ic
Res Cardiol

101,

336
-
45.

Murray, C. J. and Lopez, A. D. (1997) Alternative projections of mortality and disability
by cause 1990
-
2020: Global Burden of Disease Study.
Lancet

349,

1498
-
504.

Murry, C. E., Jennings, R. B. and Reimer, K. A. (1986) Precondit
ioning with ischemia: a
delay of lethal cell injury in ischemic myocardium.
Circulation

74,

1124
-
36.

Nicholson, J. K., Connelly, J., Lindon, J. C. and Holmes, E. (2002) Metabonomics: a
platform for studying drug toxicity and gene function.
Nat Rev Drug Dis
cov

1,

153
-
61.

Notebaart, R. A., van Enckevort, F. H., Francke, C., Siezen, R. J. and Teusink, B. (2006)
Accelerating the reconstruction of genome
-
scale metabolic networks.
BMC
Bioinformatics

7,

296.

O'Farrell, P. H. (2001) Conserved responses to oxygen de
privation.
J Clin Invest

107,

671
-
4.

Ogata, H., Goto, S., Sato, K., Fujibuchi, W., Bono, H. and Kanehisa, M. (1999) KEGG:
Kyoto Encyclopedia of Genes and Genomes.
Nucleic Acids Res

27,

29
-
34.

Onody, A., Zvara, A., Hackler, L., Jr., Vigh, L., Ferdinandy, P.

and Puskas, L. G. (2003)
Effect of classic preconditioning on the gene expression pattern of rat hearts: a
DNA microarray study.
FEBS Lett

536,

35
-
40.

Opie, L. (1998) The Heart, Raven Press, New York.

Palsson, B. (2000) The challenges of in silico biology
.
Nat Biotechnol

18,

1147
-
50.

Palsson, B. (2006) Systems biology : properties of reconstructed networks, pp. xii, 322 p.,
Cambridge University Press, Cambridge ; New York.

Palsson, B. O., Narang, A. and Joshi, A. (1989) Computer model of human erythrocyte
metabolism.
Prog Clin Biol Res

319,

133
-
50; discussion 151
-
4.

Pan, D. A. and Hardie, D. G. (2002) A homologue of AMP
-
activated protein kinase in
Drosophila melanogaster is sensitive to AMP and is activated by ATP depletion.
Biochem J

367,

179
-
86.

Papin, J.

A., Price, N. D., Wiback, S. J., Fell, D. A. and Palsson, B. O. (2003) Metabolic
pathways in the post
-
genome era.
Trends Biochem Sci

28,

250
-
8.

Paternostro, G., Vignola, C., Bartsch, D. U., Omens, J. H., McCulloch, A. D. and Reed, J.
C. (2001) Age
-
associa
ted cardiac dysfunction in Drosophila melanogaster.
Circ
Res

88,

1053
-
8.

Price, N. D., Reed, J. L. and Palsson, B. O. (2004) Genome
-
scale models of microbial
cells: evaluating the consequences of constraints.
Nat Rev Microbiol

2,

886
-
97.

Reed, J. L., Famil
i, I., Thiele, I. and Palsson, B. O. (2006) Towards multidimensional
genome annotation.
Nat Rev Genet

7,

130
-
41.

Reed, J. L., Vo, T. D., Schilling, C. H. and Palsson, B. O. (2003) An expanded genome
-
scale model of Escherichia coli K
-
12 (iJR904 GSM/GPR).
Ge
nome Biol

4,

R54.

Sabatine, M. S., Liu, E., Morrow, D. A., Heller, E., McCarroll, R., Wiegand, R., Berriz,
G. F., Roth, F. P. and Gerszten, R. E. (2005) Metabolomic identification of novel
biomarkers of myocardial ischemia.
Circulation

112,

3868
-
75.

Sakai,

J., Ishikawa, H., Kojima, S., Satoh, H., Yamamoto, S. and Kanaoka, M. (2003)
Proteomic analysis of rat heart in ischemia and ischemia
-
reperfusion using
fluorescence two
-
dimensional difference gel electrophoresis.
Proteomics

3,

1318
-
24.

Sanyal, S., Jenning
s, T., Dowse, H. and Ramaswami, M. (2006) Conditional mutations in
SERCA, the Sarco
-
endoplasmic reticulum Ca(2+)
-
ATPase, alter heart rate and
rhythmicity in Drosophila.
J Comp Physiol [B]

176,

253
-
63.

Sauer, U. (2006) Metabolic networks in motion: 13C
-
base
d flux analysis.
Mol Syst Biol

2,

62.

Schilling, C. H., Letscher, D. and Palsson, B. O. (2000) Theory for the systemic
definition of metabolic pathways and their use in interpreting metabolic function
from a pathway
-
oriented perspective.
J Theor Biol

203,

229
-
48.

Schmidt, K., Carlsen, M., Nielsen, J. H. and Villadsen, J. (1997) Modeling isotopomer
distributions in biochemical networks using isotopomer mapping matrices.
Biotechnol Bioeng

55,

831
-
840.

Sehl, P. D., Tai, J. T., Hillan, K. J., Brown, L. A., Godd
ard, A., Yang, R., Jin, H. and
Lowe, D. G. (2000) Application of cDNA microarrays in determining molecular
phenotype in cardiac growth, development, and response to injury.
Circulation

101,

1990
-
9.

Semenza, G. L. (2001) Hypoxia
-
inducible factor 1: oxygen h
omeostasis and disease
pathophysiology.
Trends Mol Med

7,

345
-
50.

Serluca, F. C. and Fishman, M. C. (2006) Big, bad hearts: from flies to man.
Proc Natl
Acad Sci U S A

103,

3947
-
8.

Shannon, P., Markiel, A., Ozier, O., Baliga, N. S., Wang, J. T., Ramage, D.
, Amin, N.,
Schwikowski, B. and Ideker, T. (2003) Cytoscape: a software environment for
integrated models of biomolecular interaction networks.
Genome Res

13,

2498
-
504.

Sheikh, K., Forster, J. and Nielsen, L. K. (2005) Modeling hybridoma cell metabolism
us
ing a generic genome
-
scale metabolic model of Mus musculus.
Biotechnol Prog

21,

112
-
21.

Simkhovich, B. Z., Marjoram, P., Poizat, C., Kedes, L. and Kloner, R. A. (2003) Brief
episode of ischemia activates protective genetic program in rat heart: a gene chip

study.
Cardiovasc Res

59,

450
-
9.

Siuzdak, G., Bothner, B., Yeager, M., Brugidou, C., Fauquet, C. M., Hoey, K. and Chang,
C. M. (1996) Mass spectrometry and viral analysis.
Chem Biol

3,

45
-
8.

Solaini, G. and Harris, D. A. (2005) Biochemical dysfunction in
heart mitochondria
exposed to ischaemia and reperfusion.
Biochem J

390,

377
-
94.

Stanton, L. W., Garrard, L. J., Damm, D., Garrick, B. L., Lam, A., Kapoun, A. M., Zheng,
Q., Protter, A. A., Schreiner, G. F. and White, R. T. (2000) Altered patterns of
gene e
xpression in response to myocardial infarction.
Circ Res

86,

939
-
45.

Stephanopoulos, G. (1999) Metabolic fluxes and metabolic engineering.
Metab Eng

1,

1
-
11.

Stephanopoulos, G., Aristidou, A. A. and Nielsen, J. H. (1998) Metabolic engineering :
principles
and methodologies, pp. xxi, 725 p., Academic Press, San Diego.

Stuart, J. M., Segal, E., Koller, D. and Kim, S. K. (2003) A gene
-
coexpression network
for global discovery of conserved genetic modules.
Science

302,

249
-
55.

Stuart, L. M., Boulais, J., Charri
ere, G. M., Hennessy, E. J., Brunet, S., Jutras, I., Goyette,
G., Rondeau, C., Letarte, S., Huang, H., Ye, P., Morales, F., Kocks, C., Bader, J.
S., Desjardins, M. and Ezekowitz, R. A. (2006) A systems biology analysis of the
Drosophila phagosome.
Nature
.

Urbanczik, R. and Wagner, C. (2005) Functional stoichiometric analysis of metabolic
networks.
Bioinformatics

21,

4176
-
80.

Vo, T. D., Greenberg, H. J. and Palsson, B. O. (2004) Reconstruction and functional
characterization of the human mitochondrial metabo
lic network based on
proteomic and biochemical data.
J Biol Chem

279,

39532
-
40.

Vo, T. D. and Palsson, B. O. (2006) Isotopomer analysis of myocardial substrate
metabolism: a systems biology approach.
Biotechnol Bioeng

95,

972
-
83.

Walhout, A. J., Reboul, J.
, Shtanko, O., Bertin, N., Vaglio, P., Ge, H., Lee, H., Doucette
-
Stamm, L., Gunsalus, K. C., Schetter, A. J., Morton, D. G., Kemphues, K. J.,
Reinke, V., Kim, S. K., Piano, F. and Vidal, M. (2002) Integrating interactome,
phenome, and transcriptome mapping

data for the C. elegans germline.
Curr Biol

12,

1952
-
8.

Warmke, J. W. and Ganetzky, B. (1994) A family of potassium channel genes related to
eag in Drosophila and mammals.
Proc Natl Acad Sci U S A

91,

3438
-
42.

Watkins, S. M. and German, J. B. (2002) Metab
olomics and biochemical profiling in
drug discovery and development.
Curr Opin Mol Ther

4,

224
-
8.

White, M. Y., Cordwell, S. J., McCarron, H. C., Prasan, A. M., Craft, G., Hambly, B. D.
and Jeremy, R. W. (2005) Proteomics of ischemia/reperfusion injury in
rabbit
myocardium reveals alterations to proteins of essential functional systems.
Proteomics

5,

1395
-
410.

Whitfield, P. D., German, A. J. and Noble, P. J. (2004) Metabolomics: an emerging post
-
genomic tool for nutrition.
Br J Nutr

92,

549
-
55.

Wingrove, J.

A. and O'Farrell, P. H. (1999) Nitric oxide contributes to behavioral,
cellular, and developmental responses to low oxygen in Drosophila.
Cell

98,

105
-
14.

Wolf, M. J., Amrein, H., Izatt, J. A., Choma, M. A., Reedy, M. C. and Rockman, H. A.
(2006) Drosophi
la as a model for the identification of genes causing adult human
heart disease.
Proc Natl Acad Sci U S A

103,

1394
-
9.

Xenarios, I., Salwinski, L., Duan, X. J., Higney, P., Kim, S. M. and Eisenberg, D. (2002)
DIP, the Database of Interacting Proteins: a re
search tool for studying cellular
networks of protein interactions.
Nucleic Acids Res

30,

303
-
5.

Yellon, D. M. and Downey, J. M. (2003) Preconditioning the myocardium: from cellular
physiology to clinical cardiology.
Physiol Rev

83,

1113
-
51.

Zupke, C. and
Stephanopoulos, G. (1994) Modeling of Isotope Distribution and
Intracellular Fluxes in Metabolic Networks Using Atom Mapping Matrices.
Biotechnol Prog

10,

489
-
498.




Figure legends



Figure 1: TOP: Overview of the systems approach to studying
cardiac hypoxia. In each
iteration, candidate genes for regulation of the hypoxia response are hypothesized
by the constraint
-
based and network models and then validated by experiment.


BOTTOM:
Initial construction of the model can incorporate several ge
nome
-
wide datasets of varying cell
-
type specificity and at different levels in the
biological hierarchy.


Figure 2: Automated cardiac phenotyping methods. A computer
-
controlled system
anesthetizes flies and mounts them on a microscope slide
, then uses image process
algorithms to find and measure the heart of each fly under controlled hypoxia
stimulus.


Figure 3: Time
-
course of hypoxia tolerance in flies and mammalian heart. Flies can
survive up to 4 hours of total anoxia, whereas heart cel
ls begin to die from
coronary occlusion in under an hour.


Figure 4: Integrating
the
metabolic model with a protein interaction network.

Fluxes
simulated by
flux
-
balance analysis can be linked to enzymes in the protein
interaction map.
Phenotype values
from the genetic screen can be included as
node values as well.
Statistical algorithms can
extract well
-
connected modules
enriched with high
-
scoring nodes, which can be further validated with
experiment.