AstraZeneca Opening Lecture
Douglas B. Kell
Metabolomics,
machine learning and modelling in systems biology: towards an understanding of
the language of cells
School of Chemistry, The University of
Manchester, Faraday Building, Sackville St, PO Box 88, MANCHESTER M60 1QD, UK dbk@man.ac.uk
http://dbk.ch.umist.ac.uk http://www.mib.ac.uk/ Tel: +44 161 306 4492
Progress in Systems Biology – or
in “understanding complex systems” – depends on new technology 1-4, computational assistance 5-7 and new philosophy 8, but probably not in that order (pace 9).
Some developments include all three 10, 11. My lecture will represent
an overview encompassing developments and challenges in each of these areas.
1 Kell, D. B. (2004).
Metabolomics and systems biology: making sense of the soup. Curr. Op. Microbiol. 7, 296-307.
2 Goodacre, R.,
Vaidyanathan, S., Dunn, W. B., Harrigan, G. G. & Kell, D. B. (2004). Metabolomics
by numbers: acquiring and understanding global metabolite data. Trends Biotechnol. 22, 245-252.
3 O'Hagan, S., Dunn, W. B., Brown, M., Knowles, J. D.
& Kell, D. B. (2005).
Closed-loop, multiobjective optimisation of analytical instrumentation:
gas-chromatography-time-of-flight mass spectrometry of the metabolomes of human
serum and of yeast fermentations. Anal Chem 77, 290-303.
4 Kell, D. B., Brown, M., Davey, H. M., Dunn, W. B.,
Spasic, I. & Oliver, S.
G. (2005). Metabolic footprinting and Systems Biology: the medium is the
message. Nat Rev Microbiol, submitted.
5 Mendes, P. & Kell, D. B. (1998).
Non-linear optimization of biochemical pathways: applications to metabolic
engineering and parameter estimation. Bioinformatics
14, 869-883.
6 Ihekwaba, A., Broomhead, D. S.,
Grimley, R., Benson, N. & Kell, D. B. (2004).
Sensitivity analysis of parameters controlling oscillatory signalling in the
NF-kB pathway: the roles of IKK and IkBa. Systems
Biology 1, 93-103.
7 Brown, M., Dunn, W. B.,
Ellis, D. I., Goodacre, R., Handl, J., Knowles, J. D., O'Hagan, S., Spasic, I. & Kell, D. B. (2005). A
metabolome pipeline: from concept to data to knowledge. Metabolomics 1, 35-46.
8 Kell, D. B. & Oliver,
S. G. (2004). Here is the evidence, now what is the hypothesis? The
complementary roles of inductive and hypothesis-driven science in the
post-genomic era. Bioessays 26, 99-105.
9 Brenner, S. (1980). Nature, June 5 issue.
10 King, R. D., Whelan, K. E., Jones, F.
M., Reiser, P. G. K., Bryant, C. H., Muggleton, S. H., Kell, D. B. & Oliver,
S. G. (2004). Functional genomic hypothesis generation and experimentation by a
robot scientist. Nature 427, 247-252.
11 Nelson, D. E., Ihekwaba, A. E. C., Elliott,
M., Gibney, C. A., Foreman, B. E., Nelson, G., See, V., Horton, C. A., Spiller,
D. G., Edwards, S. W., McDowell, H. P., Unitt, J. F., Sullivan, E., Grimley,
R., Benson, N., Broomhead, D. S., Kell, D. B. & White,
M. R. H. (2004). Oscillations in NF-kB signalling control the dynamics of target gene expression. Science 306, 704-708.
Symposium 
Principles of Systems Biology
Plenary Lectures
P-L01 Dynamics
and Design of Cellular Reaction Networks
Reinhart Heinrich
Theoretical
Biophysics, Humboldt University, Invalidenstraße 42, Berlin D-10115, Germany,
Phone: +49/30/20938698, FAX: +49/30/20938813, e-mail: reinhart.heinrich@biologie.hu-berlin.de, Web: http://www.biologie.hu-berlin.de/~theorybp/
An overview is given about
different methods for the mathematical analysis of cellular reaction networks.
The lecture focuses on methods of nonlinear dynamics, metabolic control analysis and on methods for elucidating the evolutionary network
design. Applications concern glycolysis, kinase/phosphatase cascades and the Wnt-signal transduction pathway. It is shown how bifurcation analysis helps
to understand the emergence of complex dynamical behaviour of networks as
metabolic oscillations and their synchronization. It is demonstrated how metabolic
control analysis (MCA) provides a framework for
identifying key components exerting rate limitation in metabolic pathways or
playing a crucial role in determining amplitudes and duration of signalling
outputs. Calculation of control coefficients is also helpful for quantifying
the oncogenic or tumor suppressing effects of proteins, for example in the Wnt-pathway.
Moreover, MCA can be used for characterizing the robustness of pathway dynamics
against parameter perturbations. As demonstrated for glycolysis special
features of the structural design of metabolic networks, such as location of
ATP producing and ATP consuming steps, can be explained on the basis
of evolutionary optimisation principles. Similar analyses for signal
transduction pathways yield the result that kinase/phosphatase cascades should
exceed a critical length for transmitting information in fast and reliable way.
A new method for elucidating the structural design of metabolic systems is
introduced which is based on network expansion starting from certain seed
compounds. It allows to draw conclusions concerning the robustness of networks
against elimination of reactions as well as concerning the temporal order of
the emergence of metabolic pathways during evolution.
P-L02 Organizational
complexity
John Doyle
Control and Dynamical
Systems, Caltech, 1200 E Cal Blvd, Pasadena CA 91125, US,
Phone: 6263954808, FAX: (626) 796-8914, e-mail: doyle@caltech.edu
A surprisingly
consistent view on the fundamental nature of complex systems can now be drawn
from the convergence of three distinct research themes. First, molecular
biology has provided a detailed description of much of the components of
biological networks, and the organizational principles of these networks are
becoming increasingly apparent. It is now clear that much of the complexity in
biology is driven by its regulatory networks, however poorly understood the
details remain. Second, advanced technology is creating engineering examples of
networks where we do know all the details and that have complexity approaching
that of biology. While the components are entirely different, there is striking
convergence at the network level of the architecture and the role of protocols,
layering, control, and feedback in
structuring complex system modularity. Finally, there is a new mathematical
framework for the study of complex networks that suggests that this apparent
network-level evolutionary convergence both within biology and between biology
and technology is not accidental, and follows necessarily from the requirements
that both biology and technology be efficient, robust, adaptive, and
evolvable. This talk will describe qualitatively in as much detail as time
allows these features of biological systems and their parallels in technology,
using hopefully familiar and concrete examples. The aim is to be accessible to
biologists, and not to depend critically on the mathematical framework. A
crucial insight is that both evolution and natural selection or engineering
design must produce high robustness to uncertain environments and components in
order for systems to persist. Yet this allows and even facilitates severe
fragility to novel perturbations, particularly those that exploit the very
mechanisms providing robustness, and this “robust yet fragile'” (RYF) feature
must be exploited explicitly in any theory that hopes to scale to large
systems. If time permits, we will briefly discuss how this view of “organized
complexity” contrasts sharply with the view of “emergent complexity” that is
favored among researchers who draw their inspiration from models and concepts
popular in physics, such as lattices, cellular automata, spin glasses, phase
transitions, criticality, chaos, fractals, scale-free networks, self-organization,
and so on.
P-L03 Computational
approaches to cellular rhythms
Albert Goldbeter
Unité de
Chronobiologie théorique, Université Libre de Bruxelles, Campus Plaine, CP 231,
Boulevard du Triomphe, Brussels B-1050, Belgium, Phone: +32/2/6505772,
FAX: +32/2/6505767, e-mail: agoldbet@ulb.ac.be, Web: http://www.ulb.ac.be/sciences/utc/
Oscillations arise
in genetic and metabolic networks as a result of various modes of cellular
regulation. In view of the large number of variables involved and of the
complexity of the intertwined feedback processes that generate oscillations, computational
models and numerical simulations are needed to clarify the molecular mechanism
of cellular rhythms. Computational approaches to two major examples of cellular
rhythms will be examined. (1) Intercellular communication by pulses of cyclic AMP (cAMP) in Dictyostelium
cells provides insights into the function of pulsatile patterns of hormone
secretion. Dictyostelium discoideum amoebae aggregate in a wavelike manner
after starvation, in response to pulses of cAMP emitted with a periodicity of
several minutes by cells behaving as aggregation centers. A model shows that
sustained oscillations in cAMP originate from the coupling between a negative
feedback loop involving cAMP-induced receptor desensitization and a positive
feedback loop due to the activation of cAMP synthesis by extracellular cAMP.
The model provides an explanation for the frequency encoding of pulsatile
signals of cAMP and for the origin of cAMP oscillations in the course of
development. (2) Among biological rhythms those with a circadian (close to 24h)
period are conspicuous by their ubiquity and by the key role they play in
allowing organisms to adapt to their periodically varying environment. In
all organisms studied so far circadian rhythms originate from the negative
autoregulation of gene expression. Computational models of
increasing complexity will be presented for circadian oscillations in the
expression of clock genes in Drosophila and mammals. When incorporating the effect of
light, the models account for phase shifting of circadian rhythms by light
pulses and for their entrainment by light-dark cycles. Stochastic simulations
permit to test the robustness of circadian oscillations with respect to
molecular noise. The model for the
mammalian circadian clock will be used to address the dynamical bases of
physiological disorders of the human sleep-wake cycle. The example of circadian
rhythms shows how computational models of genetic regulatory networks can be
used to address issues ranging from molecular mechanism to physiological
disorders.
P-L04 Fundamentals
and Applications of Metabolic Pathway Analysis
Stefan Schuster
Bioinformatics,
Jena University, Ernst-Abbe-Platz 2, Jena 07743, Germany,
Phone: +49-30-946450, FAX: +49-30-946452, e-mail: schuster@minet.uni-jena.de,
Web: http://pinguin.biologie.uni-jena.de/bioinformatik/
The topological
analysis of metabolic networks has attracted increasing interest in recent
years. Dynamic mathematical modelling of large-scale metabolic and regulatory
networks meets difficulties as the necessary mechanistic detail is rarely
available. In contrast, structure-oriented methods such as metabolic pathway
analysis only require network topology. In my talk, several concepts central to
this analysis are explained: basis vectors of the null-space, enzyme subsets,
elementary flux modes [1,2] and extreme pathways [3]. It is
shown that the concept of elementary modes is well-suited for determining
routes enabling maximum yields of bioconversions and properly describes
knockouts. Thus, it is well-suited for analysing redundancy and robustness
properties of living cells [4]. Another application is the assessment of the
impact of enzyme deficiencies in medicine. The advantages of elementary modes
in comparison to basis vectors of the null-space are outlined. The analysis is
illustrated by several biochemical examples, such as lysine synthesis in
Escherichia coli.
Metabolic pathway analysis for large, complex metabolic networks often meets
the problem of combinatorial explosion. One method for coping with this problem
is to set all intermediates that participate in more than a threshold number of
reactions to external status [5]. Another method is to vary the status of
metabolites in such a way that the number of elementary modes is minimized. By
both methods, networks can be decomposed into subnetworks.
1. T. Pfeiffer, I.
Sánchez-Valdenebro, J.C. Nuño, F. Montero, S. Schuster: METATOOL: For studying metabolic networks.
Bioinformatics 15 (1999) 251-257.
2. S. Schuster, D.A. Fell, T. Dandekar: A general definition of metabolic
pathways useful for systematic organization and analysis of complex metabolic
networks. Nature Biotechnol. 18 (2000) 326-332.
3. C.H. Schilling, B.O. Palsson: Assessment of the metabolic capabilities of
Haemophilus influenzae Rd through a genome-scale pathway analysis, J. Theor. Biol. 203
(2000) 249-283.
4. J. Stelling, S. Klamt, K. Bettenbrock, S. Schuster, E.D.
Gilles: Metabolic network structure determines key aspects of functionality and
regulation. Nature 420 (2002) 190-193.
5. S. Schuster, T. Pfeiffer, F. Moldenhauer, I. Koch, T. Dandekar: Exploring
the pathway structure of metabolism: Decomposition into subnetworks and
application to Mycoplasma pneumoniae, Bioinformatics 18 (2002) 351-361.
Symposium 
Guided General Discussion: Identifying issues concerning Systems Biology Principles
Chalk/Blackboard teaching
Notes
Symposium 
Principles of Systems Biology
Workshop Talks
&
Short Talks
P-W01 Expression dynamics of a cellular
metabolic network.
Dennis Vitkup
Biomedical
Informatics, Columbia Univeristy, 1150 St. Nicholas avenue, Russ Berrie
Pavilion, room 121G, New York NY 10032, USA, Phone: 212-851-5151,
FAX: 212-851-5149, e-mail: dennis.vitkup@columbia.edu
Towards the goal of understanding system
properties of the biological networks we investigate the global regulation of
gene expression in the Saccharomyces cerevisiae metabolic network. Our results demonstrate the predominance of
local gene regulation in the metabolism. The metabolic genes display
statistically significant co-expression on distances smaller than the average
network distance. Positive gene co-expression decreases monotonically with
distance in the metabolic network, while negative co-expression is strongest at
intermediate network distances. We find statistically higher co-expression in
the linear sections of metabolism compared to branched pathways. While the
majority of the traditionally defined metabolic pathways, when perturbed,
display significant of co-expression, only fractions of such pathways are
actually regulated together. Generally the structure of the metabolic
co-expression is different from the traditional pathway boundaries. Simple
topological motifs of the network show distinct co-expression patterns
highlighting important design principles of the metabolic dynamics.
P-S01 Smart
regulation of ammonium assimilation by Escherichia
coli: modularity, robustness, and
flux regulation.
Frank
J. Bruggeman
Dept of
Molecular Cell Physiology, Faculty Earth & Life Sciences & Biocentrum
Amsterdam, De Boelelaan 1085, Amsterdam NL-1081 HV, The
Netherlands, EU, Phone: +31/20/5987248, FAX: +31/20/5987229,
e-mail: frank.bruggeman@falw.vu.nl, Web: www.angelfire.com/scifi/frankb
Regulation of ammonium assimilation in E. coli is governed by two mechanisms: (i) by glutamine synthetase (GS)
and glutamate synthase (GOGAT) and (ii) by glutamate dehydrogenase (GDH). The
former system is active at low ammonium concentrations and the latter system
gradually takes over ammonium assimilation as function of an increase in the
ammonium supply. The net ammonium assimilation flux (Jn) is the sum of both mechanisms. A kinetic model of ammonium assimilation shall be introduced. It will be
analyzed in terms of: (i) the robustness of Jn and (ii) the regulation of the
individual ammonium-assimilation fluxes of GS/GOGAT (Jgs) and GDH (Jgdh).
The system will be dissected into mechanisms that guarantee nontrivial
robustness of Jn by ‘smart’ regulation of Jgs and Jgdh.
The regulation of Jgs shall be further analyzed in terms of the contributions
of different processes. Both types of analysis was carried out in terms of a
modular description of the network to facilitate understanding of this
complicated regulatory network, which involves feedback regulation, covalent
modification, parallel pathways, intracellular signalling via the two-component
mechanism and gene expression.
P-W02 An
adaptive system approach for the modelling of genetic regulatory networks.
Glucose metabolism study in the yeast.
Todor Vujasinovic and Sinisa Zampera
HeliosBiosciences
SARL, 8, rue Général Sarrail, Créteil 94010, France, Phone: +33/149813792,
FAX: +33/148985927, e-mail: sinisa.zampera@heliosbiosciences.com
We have used a dynamic neural network to model
the yeast glucose metabolism response to glucose deprivation in the culture medium.
Our aim was to produce a predictive rather than explicative model, in order to
address the question: 'which molecule of the network should we act upon to
obtain a given biological response?' The network was built from literature
analysis and KEGG data and includes 133 molecules (3 metabolites, 99 enzymes,
26 transcription factors, 5 signal transduction proteins, connected through 516 interactions). The
model was trained by DNA microarray data describing the gene expression response to the fermentation to respiration switch (De
Risi et al., Science (1997)278:680-6). The simulation provides a hierarchy of
the molecules classified in terms of relative distance to the biological
response to be obtained. The model has been applied to the prediction of a gene
knock-out response and the detection of the invalidated gene was within
acceptable error margins. We will present our model and results and more
specifically discuss the redundancy of biological regulatory mechanisms as
arguing towards the use of adaptive models, and the impact of the network
heterogeneity (scale-free structure) on the learning procedure and inferred
parameters.
P-S02 Design
Principles of Signal Transduction Pathways to attenuate Noise
Markus Kollmann,
Kilian Bartholome and Jens Timmer
Department of
Physics, University Freiburg, Hermann-Herder-Str. 3, Freiburg D-79104, Germany,
Phone: +49 761 203 5828, FAX: +49 761 203 5967, e-mail: markus.kollmann@physik.uni-freiburg.de, Web: http://webber.physik.uni-freiburg.de/~markus/
One of the great paradoxes in studying signal transduction pathways is their seemingly oversized topology. Even
in rather small signalling cascades like MAP kinase it is unclear why so many kinase reactions are involved and what
benefits multi-phosphorylation sites. Similarly one can show in bacterial
chemotaxis that the topology can be much more simplified to arrive at almost
perfect adaption. These facts give the impression that signalling pathways are
rather 'tinkered' than 'properly engineered' [1]. But the underlying assumption
within this view on signalling pathways is the concept of 'modularisation' on
one hand and moderate component tolerances on the other hand. Only these
assumptions allow us to investigate signalling networks ignoring strong
intra-cellular perturbations. In this work we show that the topology for
bacterial chemotaxis depends crucial on strength of intra-cellular
perturbations. We show that chemosensory pathways are not only designed to
transmit changes in ligand concentration to the flagella motor proteins under
the condition of almost perfect adaption but also to resist inter-cellular
noise. For the bacterium E.coli the magnitude of variations in concentration of signalling
proteins has been measured in detail [2,3] and can vary up to ten-fold between
individuals [2]. From the known strength of fluctuations we can interfere the
requirements on the topology to attenuate these variations. Under realistic
assumptions of variations in binding constants and stochastic noise effects we
show that the topology of chemotaxis pathways are not 'tinkered' but the
outcome of an evolutionary optimisation process.
[1] Alon U., (2003),
Science, 301
[2] Li M. & G. Hazelbauer, (2004), J.Bact., 186
[3] Elowitz M. et al., (2002), Science, 297
P-S03 On pathways and distances in metabolic networks
Esa Pitkänen 1, Ari Rantanen 1, Juho Rousu2 and
Esko Ukkonen 1
1 Department of Computer Science, University of
Helsinki, P.O.Box 68 (Gustaf Hällströmin katu 2b), Helsinki 00014, FINLAND,
Phone: +358/40/5314252, FAX: +358/9/1915 1120, e-mail: esa.pitkanen@cs.helsinki.fi
2 Department of Computer Science, Royal Holloway, University of
London
Recent 'small-world' studies of the global
structure of metabolic networks have been based on the shortest-path distance.
As this distance does not capture accurately the complexity of the underlying
biochemical processes, we propose new distance measures that are based on the
structure of feasible metabolic pathways between metabolites. We define a
metabolic pathway as a minimal set of metabolic reactions capable of converting
the source metabolites into the target metabolites. The metabolic distance is
defined as the number of reactions in a smallest possible pathway connecting
the sources to the targets. The production distance is defined as the minimum
number of successive reactions needed for such conversion, and is upper-bounded
by the first distance. These concepts are defined using an and-or graph
induced by the metabolic network.
We study the computational complexity and derive algorithms for evaluating the
distances. We also provide a linear-time algorithm for finding an upper bound
for the metabolic distance which itself is shown NP hard to evaluate. To test
our approach in practice, we calculated these and shortest-path distances in
two microbial organisms, S. cerevisiae and E. coli. The results show that
metabolite interconversion is significantly more
complex than was suggested in previous small-world studies. We also studied the
effect of reaction removals (gene knock-outs) on the connectivity of the S. cerevisiae network and found out that the network
is not particularly robust against such mutations.
Symposium 
Resumed General Discussion: Addressing the
issues concerning Systems Biology Principles
Reinhart Heinrich
John Doyle
Albert Goldbeter
Stefan Schuster:
Symposium 
Tools and Methods
(Part 1)
Plenary Lectures
T-L01 Quantitative
Proteomics: An Essential Component of Systems Biology
Ruedi Aebersold
Institute for
Molecular Systems Biology, ETH Zürich, Wolfgang-Pauli-Strasse 16, Zürich CH
8093, Switzerland, Phone: +41/16333170, FAX: +41/16331051,
e-mail: aebersold@biotech.biol.ethz.ch
Systems biology is the science of dynamic
networks of interacting biomolecules. It is based on the insight that such
networks have intrinsic properties determining their structure and function
that are not apparent from the analysis of the isolated components that
constitute the system and that are critical for an understanding of the
function and control of the system as a whole. Systems biology was made possible by the
availability of the complete genome sequence of the human and other species and by advances in
biology, engineering and computer science that have collectively catalyzed the
emergence of technologies for the systematic and quantitative measurement of
genomic and proteomic profiles and the integrative analysis of the obtained
results.
Most biological networks involve proteins. Proteomics, the systematic analysis
of proteins is therefore an important component of systems biology. In this
presentation we will discuss conceptual and technical advances in proteomics
and their application to the analysis of biological networks. Special emphasis
will be placed on techniques for the collection and accurate analysis of
quantitative proteomics data at high throughput.
T-L02 Modelling
and simulation of large-scale signal transduction networks
M. Bentele, H. Busch, I. Vacheva, R. Eils *
Division
Theoretical Bioinformatics, German Cancer Research Center (DKFZ), Heidelberg, *and Department for
Bioinformatics and Functional Genomics, Institute of Pharmacy and Molecular
Biotechnology (IPMB), University of Heidelberg
Mathematical modeling is required
for understanding the complex behavior of large signal transduction networks. Previous attempts to model signal
transduction pathways were often limited to small systems or based on
qualitative data only. We developed a mathematical modeling framework for
understanding the complex signaling behavior of programmed cell death
(apotosis). Defects in the regulation of apoptosis result in serious diseases
such as cancer, autoimmunity and
neurodegeneration. During the last decade many of the molecular mechanisms of
apoptosis signaling have been examined and elucidated. A systemic understanding
of apoptosis is, however, still missing. To address the complexity of apoptotic
signaling we subdivided this system into subsystems of different information
qualities. A new approach for sensitivity analysis within the mathematical
model was key for the identification of critical system parameters and two
essential system properties: modularity and robustness. Our model describes the
regulation of apoptosis on a systems level and resolves the important question
of a threshold mechanism for the regulation of apoptosis.
T-L03 Analysing
Networks of Biochemical Processes: Bioinformatics Meets Systems Biology
Shoshana
J. Wodak 1,2, Chris Lemer2, Jean Richelle 2,
Nicolas Simonis
2, Chris Orsi 1, Didier Croes 2, and Jacques
van Helden 2.
1 Structural Biology and Centre for
Computational Biology, Hospital for Sick Children, 555 University Avenue,
Totonto, M5G 1X8, Canada
2 Service de Conformation de Macromolecules
Biologiques, Centre de Biologie Structurale et Bioinformatique, Université Libre de Bruxelles, CP263, Blvd.
du Triomphe, 1050, Bruxelles, Belgium
The focus of biology has shifted
from the investigation of individual genes and proteins, to the study of large
complex networks featuring interactions between tens of thousands of molecular
and cellular components. Information on these networks is obtained from genome-scale experimental analyses, which
yield very large amounts of valuable but noisy data on biological processes
that are still poorly understood. Gaining understanding of these processes
remains the major goal. However, given the complexity of the underlying systems
this cannot be achieved without efficient means for handling this information,
-classically the task of Bioinformatics- and without sophisticated computational
approaches for interpreting it in terms of biological knowledge - the aim of Systems Biology-. This is hence where Bioinformatics and
Systems biology closely overlap.
We will illustrate this overlap
here with examples from our own work. We will briefly describe the aMAZE
workbench, which stores and handles information on various types of cellular
networks using the framework of graph theory. We will also show how a very
simple representation of metabolic networks as weighted graphs can go along way
towards building pathways from incomplete information and measuring functional
distances between genes. Lastly we will illustrate how simple bioinformatics
tools can help in analyzing time- and conditions- dependent aspects of the
transcriptional regulation of protein complexes in yeast.
T-L04 Global
Mapping of Synthetic Genetic Interactions in Yeast
Charlie Boone
Banting and Best
Department of Medical Research, University of Toronto, 112 College St, Toronto
ON M4T1K9, Canada, Phone: 416/946/7260, FAX: 416/978/8528,
e-mail: charlie.boone@utoronto.ca, Web: http://www.utoronto.ca/boonelab/
We are applying synthetic genetic array
analysis to the large-scale mapping of genetic interaction networks in yeast. A yeast genetic interaction
network containing ~1000 genes and ~4000 interactions was mapped by crossing
mutations in 132 different query genes into a set of ~5000 viable gene deletion mutants and scoring the resultant double mutant progeny
for a fitness defect. The average query gene showed ~30 synthetic genetic
interactions, indicating that the resulting genetic network is complex and may
contain as many as ~100,000 interactions. Connectivity of a gene in the network
is predictive of function because query genes tend to interact with genes of
related function. Moreover, cluster analysis revealed that subsets of genes
displaying similar patterns of genetic interactions may encode components of
the same pathway or complex. To investigate networks of essential genes, we
created promoter shut-off alleles for over two-thirds of essential yeast genes
and proof-of-principle screens show that these strains are compatible with SGA
analysis. To extend SGA analysis to synthetic dosage lethality (SDL) and
synthetic dosage suppression (SDS) screens, we constructed a plasmid-based
yeast array in which each strain expresses a unique, tagged yeast gene from the
inducible GAL promoter. We have completed a comprehensive analysis of yeast
genes that cause discernible growth defects when overexpressed, and have
applied SGA-based SDL/SDS analysis to a number of query genes. In an
application of the genetic network analysis, we showed that clustering
chemical-genetic profiles and genetic interaction profiles identifies target
pathways or proteins, providing a powerful means of inferring mechanism of drug action.
Symposium 
Guided General Discussion: Identifying issues concerning Systems Biology Tools and
Methods
Chalk/Blackboard teaching
Notes
Symposium 
Tools and Methods
Workshop Talks
&
Short Talks
T-W01 An
Integrated Interaction Network of Escherichia
coli for
Studying Genotype-Phenotype Relationship
An-Ping Zeng, Bharani Kumar and Hongwu Ma
Experimental
Bioinformatics/Genome Analysis, GBF-German Research Center for Biotechnology,
Mascheroder Weg 1, Braunschweig 38124, Germany, Phone: +49 531 6181188,
FAX: +49 531 6181751, e-mail: aze@GBF.de,
Web: http://genome.gbf.de/bioinformatics/index.html
Most recent theoretical studies have provided
us with a first transitory perception of the structure of molecular interaction
networks in biological systems. In particular, three non-disjoint molecular
interaction networks have been the focus of these studies: the metabolic
network, the protein–protein interaction network
and the transcriptional regulatory network. However these interaction networks
only contain information related to specific components like genes, proteins or
metabolites but not all the relevant correlated information. Ultimately a more
complete picture requires integrating the data obtained from all of these
approaches with modeling efforts at different levels of detail. An impressive
body of data is already available on E.
coli. We focus in this study on
the reconstruction and analysis of an integrated biological network which
incorporates metabolite protein interactions, transcriptional regulation,
protein-protein interactions and signal transduction. Along with these interactions, information related
with the newly discovered regulators, such as small RNA regulators are also added. First, we determined the metabolite
protein interactions. For this purpose we extracted the proteins expressed by
all the regulated genes found in an extended transcriptional regulatory network
of E. coli which contains more than
1050 transcriptional factors (Ma et al. 2004, Nucleic Acids Res. In print). The number of proteins
counts to 1283. Out of these, 556 proteins are involved in enzymatic reactions,
resulting in the formation of metabolites as output which couple with the
transcription factors thereby controlling the gene expression. Then the instances available for the signal
transduction system are also added to the network. Mostly the signal
transduction and the metabolite protein interactions act as feedback loops in
the network. Finally the protein-protein and metabolic interactions are added.
From the obtained integrated network further structure analysis on the network
is done in order to investigate the organizational structure, network motif and
regulation. A multi-layer hierarchical structure with the feedback loops from
the bottom layers to the top layers and interactions among the nodes within the
same layer illustrates the main feature of the integrated network. This
integrated network provides a solid basis for understanding the relation
between genotype and phenotype at systems level and is being used for a better
reverse engineering of biological networks from proteomic and microarray data.
T-S01 Oscillatory
mechanisms derived from phase and amplitude information
Sune Danø 1, Mads Madsen 2 and Preben
G. Sørensen 1
1 Department of Medical Biochemistry and
Genetics, University of Copenhagen, Blegdamsvej 3b, Copenhagen N DK-2200,
Denmark, Phone: +45 35 32 77 51, FAX: +45 35 35 63 10, e-mail: sdd@kiku.dk
2 Department of Chemistry, University of Copenhagen,
Universitetsparken 5, DK-2100 Copenhagen Ø, Denmark
Due to time-scale
separation, a dynamical system close to a bifurcation will evolve according to
the universal dynamics of that particular bifurcation. We have exploited this
fact to devise a novel approach for determining the oscillatory mechanism for
systems close to a supercritical Hopf bifurcation. In essence, the method works
by identifying the chemical components of the two dynamical modes associated
with the oscillatory dynamics: an activating mode and an inhibitory mode. There
is no need for prior knowledge of the network structure, the only information
required is measurements of the relative phases and amplitudes of the
oscillating substances. Hence, metabolomics and mRNA arrays are ideal sources
of data. The feasibility of the method is illustrated by its use for analysis
of glycolytic oscillations in yeast cells.
T-S02 Application
of modelling and simulation to drug discovery: The ErbB System
Bart Hendriks,
Gareth Griffiths, Jack Beusmans, Adrienne James, Julie Cook,
Jonathan Swinton and David De Graaf
Computational
Biology, Pathways, AstraZeneca, Mereside, Alderley Park, Macclesfield,
CHESHIRE SK10 4TG, ENGLAND, Phone: +44-1625-519391,
FAX: +44-1625-514463, e-mail: adrienne.james@astrazeneca.com
The implication of the ErbB family in the pathogenesis
of various cancers has made it a popular target for the development of targeted
anti-cancer therapies. ErbB dimerisation, trafficking, and activation are
complex processes, making it difficult to intuit how perturbations, such as drug intervention, will affect the system dynamics. We need
computational approaches to keep track of and to quantify this complexity.
AstraZeneca, in collaboration with the
Lauffenburger lab at Massachusetts Institute of Technology, have developed a
computational model implementing commonly accepted principles involved in ErbB
signal transduction. The current ErbB model is made up of Ordinary
Differential Equations (ODEs) and is based on detailed mechanisms of ErbB
receptor interactions and downstream signalling components. It contains ~300
species, ~400 parameters and more than 500 reactions. A major challenge in
dealing with models of this size is information management and model
visualization. Text mining software is used to capture kinetic constants and models are displayed graphically using TeraNodeTM
Design Suite. Parameter estimation and sensitivity analysis are being exploited
to assist model validation. The model is being used to predict the dynamics of
receptor phosphorylation in the context of different cell lines and ligand
environments. Recent work in our group has demonstrated that a deficiency in
internalisation is sufficient to explain the observed signalling phenotype of
the Gefitinib-responsive mutants found in NSCLC. Gefitinib ('Iressa'; ZD1839,
AstraZeneca, Wilmington, DE) is an ATP-competitive small molecule
inhibitor of ErbB1, approved for use in the treatment of non-small cell lung
cancer (NSCLC). About 80% of Gefitinib-responsive tumours in NSCLC carry
mutations in ErbB1. This model prediction has been experimentally validated
using a Gefitinib-responsive and non-responsive NSCLC cell line. The
Gefitinib-responsive cell line is shown to be deficient in the internalisation
of two ErbB1 ligands, EGF and TGFa. This work provides a mechanistic basis for
the link between the role of ErbB1 in oncogenesis and Gefitinib response
through decreased internalisation of ErbB1 and increased signalling to AKT.
T-S03 Combined
optimization technique for biological data fitting
Konstantin
N. Kozlov 1, Alexander
M. Samsonov 2 and John Reinitz 3
1 Department of Computational Biology, St.
Petersburg State Polytechnical University, Polytechnicheskaya st., 29, St.
Petersburg 195251, Russia, Phone: +7/812/5962831,
FAX: +7/812/5962831, e-mail: kozlov@spbcas.ru
2 The Ioffe Institute of the Russian Academy of Sciences,
St.Petersburg, 194021 Russia
3 Dept. of Applied Math and Statistics, The University at Stony
Brook, Stony Brook NY 11794-3600
Motivation.Development of the
organisms from embryo to the adult is one of the central unsolved problems of
biology. We are working on characterization problem of systems biology of
development in context of the segment determination gene network of a Drosophila embryo. While gene expression is evaluated at a time resolution of a few minutes
and a spatial resolution of one cell (see FlyEx
database), the regulatory parameters cannot be determined
experimentally, and are to be found as the solution of the inverse problem by
minimizing the deviation of the model output from the data. We apply a chemical
kinetic model describing the dynamics of the expression patterns of the
segmentation genes during the blastoderm stage by means of the system of highly
non-linear reaction-diffusion equations (Jaeger, J, et al., (2004), Nature, 430, 368).A random
search technique, being extremely computationally intensive, is sometimes the
only choice for finding the set of parameters that provides the best fit of
model to data. Therefore the main problem is to reduce the complexity of
finding the parameters of mathematical models.
Results. We developed the Combined Optimization Technique (COT) to reduce the
computational cost of solution of the inverse problem of modelling. COT
combines advantages of random search and gradient descent. Starting from an
arbitrary initial set of parameters, a rough approximation of a minimum is
found by the random search, namely, Simulated Annealing (SA), while the final
solution is given by Optimal Steepest Descent Algorithm (OSDA), developed
earlier (Kozlov, K, et al., (2003), Techn.
Physics, 48, 6), and successfully applied as the local optimizer in(Gursky, V, et al., (2004), Phys. D,
197, 286). The dependence of COT convergence of the initial approximation and
quality criterion is investigated and the strategy of transition from SA to
OSDA is studied here. COT demonstrated high accuracy in reconstruction of model
parameters and the 30% total performance benefit in a two-gene network. Further study is performed currently to increase the
speed up by application of new automated tuning methods for the OSDA part of
COT. Acknowledgments. The support of the
study by the NIH Grants RR07801, TW01147, the CRDF GAP Awards RBO685, RBO1286
is gratefully acknowledged.
T-S04 Systematic
identification and characterisation of synthetic lethal interactions in the
metabolic network of yeast
Balázs Papp 1,
Richard Harrison 1, Daniela Delneri 1, Csaba Pál 2 and
Stephen Oliver 1
1 Faculty of Life Sciences, Michael Smith
Building, The University of Manchester, Oxford Road, Manchester M13 9PT, United
Kingdom, Phone: + 44 161 275 1565, FAX: + 44 161 275 5082,
e-mail: pappb@ramet.elte.hu,
Web: http://ramet.elte.hu/~pappb
2 Theoretical Biology and Ecology Modelling Group, Hungarian Academy
of Sciences and Eötvös Loránd University,Pázmány Péter Sétány 1/C, H-1117
Budapest, Hungary
To what extent and why do the effects of
mutations depend on the genetic background? Do deleterious mutations act
synergistically? What is the mechanistic basis of genetic interactions and how
does it depend on the environment? Answers to these questions are relevant not
only to functional genomics, but also to problems such as the evolution of
sexual reproduction and how deleterious mutations are eliminated from the
population.
Owing to the huge number of potential gene combinations, progress in answering these questions is, however,
limited by the lack of efficient genome-scale experimental mapping of
genetic interactions. To overcome this difficulty, we propose a combination of in
silico and in vivo studies to screen for synthetic lethal
relationships in the yeast metabolic network.
First, we apply flux balance analysis (FBA) to the genome-scale metabolic model of S.
cerevisiae (Forster et al. 2003) to search for candidate gene pairs showing
synthetic lethal interactions. Next, we use laboratory experiments to validate
the model’s predictions. Our preliminary results suggest that i) FBA is able to
predict synthetic lethal interactions, ii) many of the interactions are
environment specific, iii) although the density of interactions do not differ
significantly between nutrient poor and nutrient rich growth conditions, we
observe twice as many genes participating in synthetic lethal interactions in
nutrient poor environment and iv) only about 20% of synthetic lethal gene pairs
can be explained by the presence of gene duplicates (isoenzymes), this
fraction, however, is significantly higher than the 2% previously reported for
non-metabolic genes (Tong et al. 2004). The implications of these findings for
genetic robustness and phenotypic plasticity are also discussed.
References
Forster, J, et al. (2003) Genome Res 13, 244.
Tong, AH, et al. (2004) Science 294, 2364.
Symposium 
Resumed General Discussion: Addressing the issues
concerning Systems Biology Tools and Methods
Rudi
Aebersold
Charlie
Boone
Roland Eils
Ursula
Kummer
Jacky Snoep
Shoshana
Wodak
Symposium 
Tools and Methods
(Part 2)
Plenary Lectures
T-L05 The
Silicon Cell approach to building detailed kinetic models of biological systems
Jacky L. Snoep 1 and Hans
V. Westerhoff 2
1 Department of Biochemistry, Stellenbosch
University, Private Bag X1, Stellenbosch 7602, South Africa,
Phone: +27218085844, FAX: +27218085863, e-mail: jls@sun.ac.za
2 Department of Molecular & Cell Physiology, Vrije Universiteit,
De Boelelaan 1085, NL-1081 HV, Amsterdam, The Netherlands
With the rapid developments in the “omics”
fields, the level of detail at which individual cellular components (e.g.
mRNAs, enzymes, metabolites) can be described has increased dramatically.
Although most of this information is qualitative and often far from complete,
for a number of systems, notably metabolic- and signal transduction-pathways, quantitative kinetic information is available to build detailed computer models.
We foresee that such detailed kinetic models will become available for a
growing number of (parts of) cellular systems and here propose an approach for
the construction of kinetic models such that they can be merged. Such combined
models could ultimately describe a complete cell.
The Silicon Cell approach (http://www.siliconcell.net) emphasizes to use
kinetic parameter values that are determined experimentally for each of the
isolated reaction steps in the system. A clear distinction is made between
model construction (on the basis of the characteristics of the isolated
components) and model validation (measurements on the complete system).
Validated models are collected in a model repository such as JWS Online
(http://jjj.biochem.sun.ac.za; http://jjj.bio.vu.nl), where they can be
interrogated using a web browser and from where they can be downloaded in
standardized format (i.e. SBML). We will illustrate the
approach using an existing detailed model for yeast glycolysis (Teusink et al., 2000) and introduce basic concepts of enzyme kinetics and
metabolic control analysis.
Teusink B, Passarge J, Reijenga CA,
Esgalhado E, Van der Weijden CC, Schepper M, Walsh MC, Bakker BM, Van Dam K,
Westerhoff HV, Snoep JL (2000) Can yeast glycolysis
be understood in terms of in vitro kinetics of the constituent enzymes? Testing
biochemistry. Eur J Biochem 267: 5313-5329
T-L06 Mathematical
modelling: Choosing the right simulation method
Ursula Kummer 1, Jürgen Pahle 1 and Marko Marhl
2
1 Bioinformatics and Computational Biochemistry,
EML Research, Schloss-Wolfsbrunnenweg 33, Heidelberg D-69118, Germany, EU,
Phone: +49/6221/533225, FAX: +49/6221/533298, e-mail: ursula.kummer@eml-r.villa-bosch.de,
Web: http://www.eml-research.de/english/Research/BCB/
2 University of Maribor,Faculty of Education,Koroška cesta 160
,SI-2000 Maribor, Slovenia
The mathematical representation of biochemical
systems is a central theme in systems biology. Different methods for the
modelling and simulation in this context have been developed so far. These
methods are shortly reviewed in this talk. Often, the choice of the method is
crucial for the success of the scientific investigation. However, this choice
is often done rather arbitrarily or using some heuristics. Given the increasing
demand for the usage of mathematical tools even by biologists which might not
be expert mathematicians, more rational criteria are necessary to avoid
misleading results. Two of the most common simulation methods are the numerical
integration of ordinary differential equations (ODEs) and stochastic
simulations using e.g. the Gillespie
algorithm. We have investigated the correlation between the divergence of
biochemical systems with the need to employ stochastic methodology. Our results
show that calculating the divergence of the system is one possible way to find
a rational basis for the decision between different simulation methods.
Symposium 
Unicellular Organims
(part 1)
Plenary Lectures
U-L01 Mathematical modeling of stress response in
yeast
Edda Klipp
Vertebrate
Genomics, Max Planck Institute for Molecular Genetics, Ihnestr. 73, Berlin
14195, Germany, Phone: +49 30 8413 9316, FAX: +49 30 8413 9322,
e-mail: klipp@molgen.mpg.de,
Web: http://www.molgen.mpg.de/~ag_klipp
The investigation of biological systems has
accelerated due to joined efforts in experimental investigation, data analysis
and modeling. Yeast is a valuable model system to study the molecular biology,
the physiology and the dynamics of cellular stress adaptation. The processes
involved in signaling and response and housekeeping constitute a highly
interconnected network. The understanding of its performance demands for an
integrated investigation of both the subsystems and the network properties
based on the knowledge of the pathway structures and experimental data.
Stress response is mediated by signaling pathways. These pathways consist of a
set of typical elements: receptors, G proteins, MAP kinase cascades and so on. We will show approaches to model the dynamics
of these elements
Mathematical modeling approach will be represented to analyze the response to
external signals. The dynamics of the reaction network is described with sets
of ordinary differential equations.
The following dynamic aspects of signal transduction will be considered:
Propagation and amplification of the signal: The kinetics of protein-protein interactions
determines how fast the signal is transmitted and which level of activation is
reached for downstream elements.
Adaptation to stress: The regulatory structure of a pathway decides
whether this pathway becomes susceptible or refractory after a first stress.
The contribution of different instances of feedback regulation on signal
termination will be analyzed.
Integration and separation of signals: Various signal pathways use common elements to
transfer different signals. We will investigate the dependence of cross
activation on network structure and kinetics.
U-L02 Hiding behind the Population Average - Cell
Cycle Dynamics of Energy Metabolism during the Lifelines of Individual Yeast
Cells
Matthias
Reuss
Institute for
Biochemical Engineering, University of Stuttgart, Allmandring 31, Stuttgart D 70569,
Germany, Phone: 0049 711/685-4573, FAX: 0049 711/685-5164,
e-mail: reuss@ibvt.uni-stuttgart.de, Web: www.ibvt.uni-stuttgart.de
A fundamental goal of systems biology is to
attain a systems-level understanding of the behaviour of single cells and cell
populations. To this end it it of major importance to assess the influence of
heterogeneity present in real cell populations on the resulting behaviour
observed at the population level; i.e. to what extent does the behaviour of the
individual cell during its lifeline differ from the population average which is
measured by most assay methods? The differential regulation of energy metabolism during the cell cycle via a cyclic AMP (cAMP)-dependent protein kinase cascade in the yeast Saccharomyces cerevisiae constitutes a characteristic example of
this kind. The present work focuses on the central role of the second messenger
cyclic AMP (cAMP) in coordinating energy metabolism and cell division via a
protein kinase A (PKA)-dependent signaling cascade. Experiments performed in
synchronous and continuous yeast cultures have demonstrated distinct cell cycle
dynamics of cAMP and its associated regulatory effects of energy metabolism
[1]. These results are incorporated into a mathematical model comprising
submodules for metabolism (glycolysis and storage carbohydrates), cell growth,
cell cycle progression and cAMP signal transduction, most of which exhibit mutual feedback effects.
Estimation of model parameters is performed on the basis of own and published
experimental data. This guarantees the intimate connection of ex0,periments and
model development characteristic of the system biology approach.
The integrated single cell model yields a dynamic description of the
cAMP-dependent regulation of metabolism and cell cycle progression during the
different cell cycle phases. The chosen modular approach is potentially
transferable to systems of medical importance, e.g. when modeling tumor cell behaviour. Moreover, the model can also serve as a basis for
a segregated description of heterogeneous cell populations, an issue of major
importance in the operation of large-scale bioprocesses.
U-L03 Knowledge
and data requirements for systems analysis of cellular networks
Jörg Stelling
Systems Biology,
Max Planck Institute DCTS, Sandtorstr. 1, Magdeburg D-39106, Germany,
Phone: +49-391-6110-475, FAX: +49-391-6110-503, e-mail: stelling@mpi-magdeburg.mpg.de
Systems biology aims at understanding complex
biological networks through a combination of comprehensive measurements and
(quantitative) mathematical modeling. At present, however, it is largely
unclear, which knowledge and data will be required for establishing realistic
mathematical models. Related to this, it is of equal importance to rigorously
ask to what extent the already available data allow for meaningful model
development. This talk will address these questions by relying on two examples
from metabolic and regulatory network analysis. Steady-state analysis of
metabolic networks can start from well-known structural features (e.g. reaction
stoichiometries and reversibilities) alone. With this very limited knowledge it
is possible to predict network functionalities and control schemes when cellular objectives such as efficiency and robustness
are taken into account. In contrast, dynamic is a hallmark of cellular
regulation and has to be adequately captured in dynamic mathematical models.
Frequently, the associated kinetic parameters are unknown, which could constrain the number of
biological systems ‘ready’ for dynamic modeling to very few cases. Systematic
analysis for the example of a complex network in yeast cell cycle regulation, however, showed that detailed predictive mathematical
models could already be developed based on the known regulatory interactions
and a limited set of ’traditional’ experimental data to estimate kinetic
parameters. Hence, both examples demonstrate an unexpectedly high degree of
information that one can extract from already available biological data and
knowledge. More generally, these studies suggest strategies for efficiently
linking future experimental and theoretical approaches to cellular networks,
and how robustness can help in model development.
Stelling, J. et al.
(2002), Nature 420: 190.
Stelling, J. et al. (2004),
Cell 118: 675.
Symposium 
Unicellular Organisms
(Part 2)
Plenary Lectures
U-L04 In Vivo Operation of Metabolic Pathways
Uwe Sauer, Lars Blank, Eliane Fischer, Lars Küpfer, Annik Perrenoud and Nicola Zamboni
Institute of
Biotechnology, ETH Zurich, ETH Honggerberg, Zurich CH 8093, Switzerland,
Phone: +41-1-6333672, FAX: +41-1-6331051, e-mail: sauer@biotech.biol.ethz.ch,
Web: http://www.biotech.biol.ethz.ch/sauer/
Data on intracellular mRNA, protein, or metabolite concentrations
reveal the composition of metabolic networks. In contrast to such compositional
information, molecular fluxes through intact metabolic networks link genes and
proteins to higher, system-level functions that result from interactions
between the components (1). Thus, fluxes are the functional output of the
integrated biochemical and genetic interactions within such networks, and are
key data for metabolic systems biology. Since such system-level in vivo
activities cannot be measured directly, they must be inferred indirectly from
measurable quantities. The currently most reliable approach is metabolic flux analysis based on 13C-labeling experiments (2, 3).
Recent examples of such network-based intracellular flux quantifications
unraveled novel pathways (4) and unexpected reactions (5), thereby questioning
the ability of well-known ‘textbook’ pathways to portray flux through complex
metabolic networks. Based on large-scale flux data from 13C-labeling
experiments in microtiter plates (6), we investigate the global regulatory
structure and functional design principles of bacterial metabolism. In
particular, we address the question of whether or not such flux responses to
gene deletions can be predicted with any certainty. For this purpose,
we use metabolic models of various complexity, whose predictions are compared
to the experimental results.
1. Hellerstein, M.K. In vivo measurement
of fluxes through metabolic pathways: The missing link in functional genomics
and pharmaceutical research. Annu. Rev. Nutr. 23, 379-402 (2003).
2. Wiechert, W. 13C metabolic flux analysis. Metabolic Eng. 3, 195-206 (2001).
3. Sauer, U. High-throughput phenomics: experimental methods for mapping
fluxomes. Curr. Opin. Biotechnol. 15, 58-63 (2004).
4. Fischer, E. & Sauer, U. A novel metabolic cycle catalyzes glucose oxidation and anaplerosis in
hungry Escherichia coli. J. Biol. Chem. 278, 46446-46451 (2003).
5. Sauer, U., Canonaco, F., Heri, S., Perrenoud, A. & Fischer, E. The
soluble and membrane-bound transhydrogenases UdhA and PntAB have divergent functions in
NADPH metabolism of Escherichia coli. J. Biol. Chem. 279, 6613-6619 (2004).
6. Fischer, E., Zamboni, N. & Sauer, U. High-throughput metabolic flux
analysis based on gas chromatography-mass spectrometry derived 13C
constraints. Anal. Biochem. 325, 308-316 (2004).
U-L05 Simplicity
in Biology
Uri Alon
Department of
Molecular Cell Biology, Weizmann Institute, Herzel 1, Rehovot 76100, Israel,
Phone: 972-8-934-4448, FAX: 972-8-934-4125, e-mail: urialon@weizmann.ac.il,
Web: www.weizmann.ac.il/mcb/UriAlon
Cells are matter that dances. Elaborate
structures spontaneously assemble, perform biochemical miracles and vanish
effortlessly when their work is done. Moreover, these molecular machines can
encode and process information virtually without errors, despite the fact that
they are under strong thermal noise and embedded in a dense molecular soup. How could this be? Are
there special 'laws of nature' that apply to biological systems that can help
us to understand why they are so different from non-living matter?
Recent discoveries suggest that one can, in fact, formulate general laws that
apply to biological network design. Since it has evolved to perform functions,
biological circuitry is far from random or haphazard. It has a defined style.
This is the style of objects that must function, and characterizes both
biological and engineered systems. Although evolution works by random tinkering
, it converges again and again onto a defined set of circuit elements called
network motifs that obey general design-principles.
The goal of this talk is to highlight the design-principles of biological
networks. The main message is that biological design contains an inherent
simplicity. Although it evolved to function and did not evolve to be
understandable to us, simplifying principles may make biological design
comprehensible.
U-L06 Stochastic
activation of the response regulator PhoB by noncognate histidine kinases
Lu Zhou,
Gérald Grégori, Jennifer Masella-Blackman, J. Paul Robinson
and Barry L. Wanner
Biological
Sciences, Purdue University, 915 W. State Street, West Lafayette IN 47907-205,
USA, Phone: (765) 494-8034, FAX: (765) 494-0876, e-mail: blwanner@purdue.edu
Two-component systems (TCS) are the most
prevalent gene regulatory mechanism in bacteria. A typical TCS is comprised of a
histidine kinase (HK) and a partner response regulator (RR). Specific environment
signals lead to autophosphorylation of different HKs, which in turn act as
phosphoryl donors for autophosphorylation of their partner RRs. Nonpartner
HKs and RRs also interact, giving rise to cross regulation among TCSs in
response to diverse signals.
PhoR (HK) and PhoB (RR) constitute the TCS for detection of environmental
(extracellular) inorganic phosphate (Pi). The PhoR/PhoB TCS controls the expression
of a large number of genes for acquisition of alternative phosphorus sources,
including phoA, which encodes the non-specific phosphohydrolase bacterial
alkaline phosphatase (Bap). Cross activation of PhoB by the nonpartner HK CreC is now a
classic example of cross regulation among TCSs. A systematic search for other
cross talking HKs revealed five additional HKs that activate (phosphorylate)
PhoB (J. M. B. and B. L. W., unpublished data).
Examination of cross activation of PhoB by these non-partner HKs by flow cytometry at the single-cell level revealed a bimodal,
“all-or-none,” distribution pattern for expression of a phoAp-gfp (green fluorescent protein) reporter fusion. Although
the basis of the observed stochastic behavior is unclear, it seems to reflect
an inherent property of TCSs. We propose that cells exploit the stochastic
character of TCSs to achieve nongenetic (epigenetic) diversity within
genetically homogeneous cell populations in order to facilitate adaptation to
environmental changes.
U-L07 Metabolome
analysis and systems biology
Masaru Tomita
, Inst. Adv.
Biosci., Keio Univ. and HMT, 5322, Endo, Fujisawa 252-8520, JAPAN,
Phone: +81-466-47-5099, FAX: +81-466-47-5099, e-mail: mt@sfc.keio.ac.jp,
Web: http://www.iab.keio.ac.jp/index.html.en
Institute for Advanced Biosciences of Keio
University has recently developed a novel technology for high-throughput
metabolome analysis. The technology is based on capillary electrophoresis
electrospray ionization mass spectrometry (CE/MS) and it can simultaneously quantify a large amount
of cellular metabolites ranged from 70 to 1,000 molecular weights. Using this
technology, we have been conducting a major project on computer modeling of E.coli, funded by New Energy and
Industrial Technology Development Organization (NEDO) of the Ministry of
Economy, Trade and Industry of Japan. The E.coli modeling project has two main
goals: (A) to construct a static model of entire metabolic pathways, and (B) to
construct a dynamic/simulation model of the primary energy metabolism.
Our approach to constructing entire pathway model consists of three steps: (1)
Top down modeling from genomic information, (2) Bottom up modeling from
metabolome analysis, and (3) Closing the gap by bioinformatics.
For a dynamic model of primary energy metabolism, we are currently collecting a
large amount of metabolome, transcriptome, and proteome data in a systematic
manner with various different culture conditions and many different single gene destructive mutants. Those data are then used to construct a
computer model using E-Cell System, a software package we have developed for
biological simulation.
.
Symposium 
Guided General Discussion: Identifying issues concerning the Systems Biology of
unicellular organisms
Symposium 
Unicellular Organisms
Workshop Talks
&
Short Talks
U-W01 Modelling
evolution of prokaryotic genomes: an integrative approach
Guillaume Beslon and
Carole Knibbe
Computer Science
Department, INSA Lyon, Bat. Blaise Pascal, Villeurbanne F-69621, France,
Phone: +33/4/72438487, FAX: +33/4/72438518, e-mail: guillaume.beslon@insa-lyon.fr,
Web: http://prisma.insa-lyon.fr/~gbeslon
Each
living system results from an evolutionary process. Thus, understanding the
system involves prior understanding of its evolutionary story: How did the
complex features we observe now emerge? Moreover, living organisms are
made up of multiple organisation levels that are all involved in evolution. The
phenotype depends on the genotype since DNA encodes the complex protein networks that achieve survival and
reproduction functions. But these networks may also influence the genome level. For instance, in bacterial genomes,
genes involved in a same biological process form clusters, showing that the
metabolic level may indirectly influence the genetic level. However, exploring
long term evolution simultaneously on these different organisation levels is
obviously impossible. That is why there is a need for In Silico models
in which virtual organisms evolve during thousands of generations.
With this objective in mind, we have defined the AEVOL model of bacterial
genome evolution. In AEVOL, each virtual bacterium competes for reproduction in
a virtual environment. Each of them owns a double-strand binary genome, uses it
to produce a proteome and expresses a phenotype. The sequence is parsed to
detect the virtual genes and a genetic code is used to translate them into
proteins. A fuzzy logic framework is used to compute the functional
capabilities of the proteins and to combine them in an interaction network in
order to compute the organism's global capabilities. All organisms are then
compared to the environment and the fittest ones are selected for reproduction.
While an organism reproduces, its genome is replicated with eventual random
local errors, large scale rearrangements and lateral transfer.
In AEVOL, the lifetime reproductive success of an organism depends on its
functional capabilities but not on its genomic structure. However, some
specific genome organisations emerge depending on various parameters (environmental
features, mutation rates, transfer rates, ...). Thus, AEVOL enables us to
explore how functional and genetic organisation level are linked by the
evolution (e.g. how gene clusters emerge in bacterial
genomes). Our aim is to use AEVOL to investigate how complex biological
networks (e.g. gene regulation networks, metabolic networks, ...) appear and
are modified during evolution. Indeed, it is well known that the topological
properties of complex networks are closely dependent on their history.
U-W02 Signal processing in bacterial
chemotaxis
Victor Sourjik
ZMBH, University
of Heidelberg, Im Neuenheimer Feld 282, Heidelberg D-69120, Germany,
Phone: +49-6221-546858, FAX: +49-6221-545894, e-mail: v.sourjik@zmbh.uni-heidelberg.de
Chemotaxis
in Escherichia coli is one of the most-studied model
systems for signal transduction. Receptor-kinase complexes organized in clusters
at the cell poles sense chemoeffector stimuli and transmit signals to flagellar
motors by phosphorylation of a diffusible response regulator protein.
Despite the apparent simplicity of the signal transduction pathway, its high
sensitivity, wide dynamic range, and integration of multiple stimuli remain
poorly understood. We use an in-vivo assay based on fluorescence resonance
energy transfer (FRET) to monitoring in real time
changes in the intracellular pathway activity upon chemoeffector stimulation.
Using FRET allows a quantitative analysis of signal amplification and
integration by the receptor clusters, as well as a quantitative analysis of the
adaptation system. In addition, we use fluorescent protein fusions to
characterize the cell-to-cell variation (noise) in the
expression of chemotaxis proteins. Presentation will compare our experimental
data to the predictions made by several recent computer models of chemotaxis.
U-W03 Combining experimental data and in
silico analysis to model the metabolic and regulatory network of Lactobacillus
plantarum
Bas Teusink 1, Christof Francke 2,
Anne Wiersma 1, Frank van Enckevort 3,
Arno Wegkamp 4, Jeroen Hugenholtz 4,
Eddy Smid 4 and Roland Siezen 5
1 NIZO food research, Wageningen Centre for Food
Sciences, Kernhemseweg 2, Ede NL-6718 ZB, The Netherlands, EU,
Phone: +31/318/659674, FAX: +31/318/650400, e-mail: Bas.Teusink@nizo.nl, Web: www.nizo.nl 2 WCFS p/a CMBI Radboud
University Nijmegen, Toernooiveld 1, Nijmegen, The Netherlands 3
NIZO food research, p/a CMBI Radboud University Nijmegen, Toernooiveld 1,
Nijmegen, The Netherlands 4 WCFS, p/a NIZO food research 5
CMBI Radboud University Nijmegen, Toernooiveld 1, Nijmegen, The Netherlands
We have
sequenced the complete genome of Lactobacillus plantarum WCFS1 (PNAS
USA 2003;100:1990). Lactobacillus plantarum is a versatile lactic acid bacterium that is important
in many food and feed fermentation processes. After prediction of gene function, which is an ongoing process, focus
is now on the development and improvement of methods and tools to go from
genome sequence to gene annotation, to pathway reconstruction and to prediction
of phenotype. Important aspects are how and where to incorporate and use
experimental (omics) data, and how and to what extent parts of the process can
be automated.
We have set up different bioinformatics tools and experimental techniques in
the area of functional genomics. Regulatory networks are being studied by motif
searches and promoter/operon predictions, as well as by transcriptome analysis.
Tools are in place for visualization of transcriptome data on metabolic maps
and on genome-maps (Microbial Genome Viewer: www.cmbi.ru.nl/MGV; encyclopedia
of L. plantarum: www.lacplantcyc.nl). Although regulated pathways can be
identified in this way, it remains difficult to understand the impact of the
observed regulation for overall metabolism. For better interpretation and
integration of omics data, a genome-scale model of L. plantarum was
developed. We have reconstructed the metabolic network of L. plantarum:
Our current network comprises 710 genes (23% of the genome), and 600 reactions.
The properties of the metabolic network are being investigated within the
framework of constraint-based modeling, and compared with physiological data
from continuous fermentations.
In the Netherlands we have set up a consortium for Systems Biology of Lactic
Acid Bacteria, in casu Lactococcus lactis. Within this consortium we
want to use Systems Biology to understand the physico-chemical and biological
constraints that limit the growth rate under a set of defined conditions. We
are seeking European partners for collaboration in future Systems Biology
programs, such as SysMO. For more information, see cmbi4.cmbi.kun.nl/~teusink/SBNL_LAB.
U-S01 Modelling fission yeast morphogenesis
Attila Csikasz-Nagy 1, Bela Gyorffy 1,
Wolfgang Alt 2, John J. Tyson 3 and
Bela Novak 1
1 Department of Agricultural Chenmical
Technology, Budapest University of Technology and Economics, Szt Gellert ter
4., Budapest 1111, Hungary, Phone: +36/1/4632910, FAX: +36/1/4632598,
e-mail: csikasz@mail.bme.hu,
Web: http://www.cellcycle.bme.hu/ 2 University of Bonn, Bonn,
Germany 3 Virginia Tech, Blacksburg, USA
Because
of its regular shape, fission yeast is becoming an increasingly important
organism to study cellular morphogenesis. Genetic studies have identified a
great number of proteins that are important to regulate shape changes during
the cell cycle. Most of these proteins interact with either microtubules or actin,
underlining the essential roles these cytoskeletal structures play in cellular
morphogenesis. Here we present a simple model for fission yeast morphogenesis
that describes the interplay between these cytoskeletal elements. An essential
assumption of the model is that actin polymerization is a self-reinforcing
process: filamentous-actin promotes its own formation from globular-actin
subunits via regulatory molecules. Microtubules stimulate actin polymerization
by delivering a component of the autocatalytic actin-assembly feedback loop. We
show that the model captures all the characteristic features of polarized
growth in fission yeast during normal mitotic cycles. We show that all the
major classes of morphogenetic mutants (monoipolar, orb and tea) are natural
outcomes of the model. We categorize the types of growth patterns that can
exist in our model and compared them with experimental observations.
U-S02 Metabolic quorum sensing: onset of
density-dependent oscillations
Silvia De Monte 1,Francesco d'Ovidio 2,Sune Danø 3
& Preben Grae Sørensen 4
1 Dept. of Biology, École Normale Supérieure,
46, rue d'Ulm, Paris F-75005, France, Phone: +33/(0)1/44322342,
FAX: +33/(0)1/44323885, e-mail: demonte@biologie.ens.fr, Web: http://www.fys.dtu.dk/~silvia 2 École Normale Supérieure,
Paris, France 3 Dept. of Medical Biochemistry and Genetics,
University of Copenhagen, Denmark4 Dept. of Chemistry, University of
Copenhagen, Denmark
Populations
of oscillating units coupled by diffusion through a homogeneous medium are studied as a
model for cells in a CSTR. In particular, we focus on the dependence of the
collective behaviour on the density of the cell suspension. Both the classical
Kuramoto model and the recent results on "coupling by quorum sensing"
(Garcia-Ojalvo, Elowitz and Strogatz (2004) PNAS 101,10955) indicate that, by
diluting the suspension, the cells should keep their oscillatory behaviour
while desynchronising.
A different scenario could however take place due to the delay introduced in
the coupling by the presence of a medium. In this case, the dilution of the
suspension results into the suppression of oscillations at both population and
individual levels. Such density-dependent phenomenon may be seen as a metabolic
analogous of quorum sensing in bacteria: the amplitudes of the individual
metabolic oscillations can provide each individual cell with information on the
population density and average state of the population.
U-S03 Integration of software tools for the in silico design of metabolic
pathways using flux balance analysis
Ana Sofia Figueiredo 1, Pedro Fernandes 1,
Pedro Pissarra 2 and António Ferreira 3
1 Bioinformatics
Unit, Instituto Gulbenkian de Ciencia, Rua da Quinta Grande, Oeiras 2781-901,
Portugal, Phone: +351 21 4407900, FAX: +351 21 4407970, e-mail: sofiafig@igc.gulbenkian.pt
2 Biotecnol SA, Taguspark, Edificio Inovacao IV nº809, Oeiras 2780-920,
Portugal, Phone: +351214220520 Fax:+351214220529
3 Departamento de Quimica e Bioquimica, Faculdade de Ciencias da Universidade
de Lisboa, Campo Grande, Lisboa 1749-016, Portugal, Phone: +351217500076 Fax:
+351217500994
The systems biology approach, where
one can envision the cell as a whole is a step in the direction of narrowing
the gap between the rate of data generation and the speed of analysis. This
embraces the much desired goal of understanding the role of the metabolic
pathways of a determined metabolic network.
This study describes the use of several software tools, integrated to perform
the simulation of a specific metabolic pathway by Flux Balance Analysis (FBA).
This simulator receives as input the stoichiometry, the thermodynamic and
capacity constraints of the metabolic network, and also an objective function.
The stoichiometry and the thermodynamic constraints are represented in the SBML format (Systems Biology Markup Language), whereas all the other
information is represented in a plain text file. The SBML file is parsed using
libSBML, which is a library that can be embedded into an application to read,
write and manipulate files in the SBML format. The text file is parsed using
FLEX, a lexical analyser that generates a C/C++ program that recognizes
specific lexical patterns in the text. With this information, one can construct
a Linear Programming (LP) problem. To solve it, GNU lp-solve is used. It uses
the simplex algorithm and sparse matrix methods for simple LP problems. The
solution provided is a possible flux distribution on the network, that maximises the objective
function. In this work, data from a batch fermentation process (where the host
system is Escherichia Coli strain BL21) is incorporated in the model
definition. In the experiment, Acetate secretion, Oxygen Uptake Rate (OUR),
Carbon Evolution Rate (CER) and Biomass production for wild type (wt) and
mutant (mt) strains were determined. The mutant was engineered with a plasmid
to express the human recombinant interkeukin 4(IL-4) using pRT as a promotor.
The analysis of the flux distribution for wt and mt is performed for the
maximisation of ATP production, incorporating as capacity constraints the different
data obtained from the experiment. A discrete time analysis was performed,
using the same variables, and assuming a steady state for each time sample.
Comparing FBA results for wt and mt, the induction of protein in the host
system decreases the capacity of producing ATP. The sensitivity of the system
to variations in glucose uptake was also performed. It was shown that, in normal conditions,
mt and wt were robust to variations with an amplitude of 2% and 20%. When the carbon
source is residual, the system shows a higher sensitivity to glucose
variations.
U-S04 Uncovering
the control of the respiratory clock in yeast
Douglas
B. Murray and
Hiroaki Kitano
Systems Biology
Institute,, Keio University School of Medicine, 9S3, Shinanomachi Research
Park, 35 Shinanomachi, Shinjuku-ku, Tokyo 160-8582, Japan,
Phone: +81-3-5363-3078, FAX: +81-3-5363-3079, e-mail: dougie@symbio.jst.go.jp
Continuously growing yeast cultures tend to auto-synchronise producing a robust respiratory oscillation (tau circa 40 min). Recently we have
carried out Affymetrix analyses that revealed the majority (>90%) of the
transcriptome oscillates within this timeframe [1]. Here we analyse this data
using a “Fourier focussing” technique in order to derive transcripts that are
closely coupled to the oscillation. The method involved dividing the amplitude
calculated by fast Fourier transformation by the mean of the amplitude for
three oscillation cycles. This ratio equated to the noise of the
transcript’s oscillation; where a perfect sine-wave generates a ratio of one
and random data generates a ratio approaching zero. When the ratio was
calculated for the yeast transcriptome and plotted, the resulting curve showed
two gradients. The intercept of these gradients was used as a noise threshold
(ratio of ~0.15; ~1500 genes). The strongly coupled transcripts above this
threshold and phenotypic events were then used to construct a “clock face”. A
network diagram was then constructed using high quality BIND and
transcriptional regulatory networks within Cytoscape [2]. The resulting network
consisted of ~1000 transcripts containing the most highly conserved aspects of
the eukaryotic process, e.g., ribosome, proteasome, DNA synthesis, autophagy, cyclins, amino acid biosynthesis, carbon
metabolism, stress response, respiration, etc. Furthermore two transcriptional
sub-graphs out of phase with each other were identified. CIN5, YAP6, YAP1, PHD1
and ROX1 comprised the core of the sub-graph whose transcripts peaked during
the low respiratory phase and MET4 and RAD59 comprised the sub-graph whose
transcripts peaked during the high respiration phase. The cultures
synchronisation mechanism revolves around the production of acetaldehyde and
hydrogen sulphide [3], which feed into and out of this network via ALD5/ADH2
and SUL2/MET3 respectively. It is concluded that these networks regulate the
respiratory clock within yeast. It is also postulated that this network may
form the centre of an energetic “bowtie” common to all eukaryotes because of
its high conservation among all eukaryotes.
[1] Klevecz RR, Bolen J, Forrest G,
Murray D.B.
(2004) Proc Natl Acad Sci USA. 101:1200-5
[2] Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N,
Schwikowski B, Ideker T. (2003) Genome Res. 13:2498-504.
[3] Murray DB, Klevecz RR, Lloyd D. (2003) Exp Cell Res. 2003 Jul
1;287(1):10-5.
Symposium 
Resumed General Discussion: Addressing the issues concerning
the Systems Biology of unicellular organisms
Uri Alon
Edda Klipp
Matthias Reuss
Uwe Sauer
Jörg Stelling
Masaru Tomita
Barry Wanner
Symposium 
Multicellular Organisms
Plenary Lectures
M-L01 Pharmacogenomics: a holistic approach
to drug organism interaction
Michel Eichelbaum
Dr. Margarete Fischer-Bosch-Inistitue
of Clin. Pharmacology, Auerbachstr. 112, Stuttgart D-70376, Germany,
Phone: +49/711/81013700, FAX: +49/711/859295, e-mail: michel.eichelbaum@ikp-stuttgart.de,
Web: www.ikp-stuttgart.de
For all
major classes of drugs a substantial proportion of patients will not respond,
respond only partially or develop toxicity when standard dose of the particular
drug are administered. It has become apparent
during the last three decades that mutation of genes encoding for proteins
involved in drug disposition and action of drugs can contribute substantially
to heterogeneity in the efficacy and toxicity of therapeutic agents.
Drug metabolizing enzymes, in particular cytochrome P450 enzymes play a pivotal
role in the elimination process of most drugs. Variability of drug metabolism
is responsible for the pronounced interindividual differences in plasma concentrations
when patients receive the same dose of a drug. As a consequence variability in
drug action and side effects / toxicity ensues. For many phase 1 and phase 2
enzymes catalysing the biotransformation of drugs mutations have been
identified. In the case of mutations leading to a loss of function
administration of a standard dose of a drug will lead to very high plasma
concentrations resulting in exaggerated response, side effects or toxicity. On
the other hand gene amplification of enzymes resulting
in ultrarapid metabolism of drugs has been identified as a mechanism of poor
response. Moreover, in the case of prodrugs which require bioactivation loss of
enzyme function is associated with a loss in efficacy therapeutic.
But even if the dose is individualized in order to achieve the same plasma
concentrations substantial variability in therapeutic response and side effects
will still be observed because concentrations at this site of action can vary
substantially. It is increasingly recognized that transfer of drugs in and out
of the cells is not a passive process depending on the physiochemical
properties, lipophilicity and protein binding but involves active transfer by
transport proteins. Because of their localization
in intestinal, hepatic and renal epithelial cells, these transport proteins are
important for the absorption and elimination of drugs. They play an important
role in the targeting drugs to organs because they are localized in blood-organ
barriers such as the blood-brain and blood-placenta barrier. Moreover, the same
concentration of a drug at the site of action does not necessarily mean
identical response because mutations at drug targets (receptors,
neurotransmitter transporters, signaling pathways) can profoundly alter the
response.
M-L02 Systems biology of receptor tyrosine
kinase signaling
Boris Kholodenko
Pathology,
Anatomy and Cell Biology, Thomas Jefferson University, 1020 Locust St,
Philadelphia PA 19107, USA, Phone: 1 215 503-1614, FAX: 1 215
923-2218, e-mail: Boris.Kholodenko@jefferson.edu, Web: http://www.cellnetworks.org
Extracellular
information received by plasma membrane receptors, such as G-protein coupled receptors (GPCRs) and receptor
tyrosine kinases (RTKs), is encoded into complex temporal and spatial patterns of phosphorylation and topological
relocation of signaling proteins. Processing and integration of this
information through MAPK cascades leads to important
cellular decisions ranging from proliferation to growth arrest, differentiation
or apoptosis. We applied a combined experimental and computational modeling
approach is applied to the EGFR signaling network. We explored kinetic and molecular factors that control the time course of phosphorylation responses,
including transient versus sustained temporal activation patterns and
oscillations in protein phosphorylation state.
Quantitative analysis of signal transduction is confronted by a
combinatorial explosion in the number of feasible molecular species presenting
different states of signaling networks that include receptors and scaffold
proteins with multiple binding domains. We show that a mechanistic description
of a highly combinatorial network generated by various phosphorylation and
binding forms of receptors and scaffolds may be drastically reduced using a
“domain-oriented” approach, referred to as a macro-model of the network.
Modeling of the spatial aspects of GPCR- and RTK-induced signaling emphasizes
the importance of receptor-mediated membrane relocation of cytosolic proteins.
We demonstrated that the spatial separation of kinases and phosphatases in MAPK
cascades may cause precipitous spatial gradients of activated MEK and ERK with
high concentration near the cell surface and low in the perinuclear area. The
results suggest that there are additional (besides diffusion)
mechanisms that facilitate passing of signals from the plasma membrane to
distant targets. They may involve endocytosis, scaffolding and active transport of signaling complexes by molecular motors.
We hypothesize that ligand-independent waves of receptor activation or/and
traveling waves of phosphorylated kinases spread the signals over long distances.
In addition to mechanistic modeling, an integrative modular approach to
inferring the structure of cellular signaling and gene networks was developed. We demonstrate how
dynamic connections leading to a particular module (e.g., an individual
gene/protein or a cluster) can be retrieved from experimentally measured
network responses to perturbations influencing other modules.
M-L03 Computational systems biology of
neuronal signalling
Nicolas Le
Novere
, EMBL-EBI, Wellcome-Trust Genome Campus, Hinxtin
CB10 1SD, UK, Phone: +44(0)1223 494 521, FAX: +44(0)1223 494 468,
e-mail: lenov@ebi.ac.uk, Web: http://www.ebi.ac.uk/compneur/
The
latest decade witnessed a drastic modification of our understanding of the
signal processing by neurons. It is now
clear that signalling pathways are interconnected, and that metabolic and gene regulatory networks should be taken into
account. Moreover, the topology of subcellular compartments cannot be ignored
any longer, nor considered as frozen. If we want to understand the mechanisms
of neuronal signalling, the integration of very heterogenous information is
necessary, ranging from kinetic data to geometrical description. That process
require large quantity of numerical data. In other term, Neurobiology ought to
set its clock to post-genomic time, and move to Systems Biology.
The
modeling of neuronal function is a tricky business. First of all, the models
have to span several rank orders both in the spatial and temporal dimensions, from the
conformational transitions of receptor for neurotransmitters (micro-seconds,
nanometers) to the electrical phenomena (seconds, millimeters). This in turn
required the use of several different algorithms (stochastic, ordinary and
partial equations, cable theory etc.). Secondly, the precise topology of
sub-cellular compartments and the location of the biological objects are
absolutely crucial for their proper function. A change of the geometry or the
density of a patch of receptors will change the post-synaptic potential. The
shape of a dendritic spine will strongly affect the impact of this potential on
the integrated signal. And
molecules such as the calcium calmoduline kinase II translocate between
subcompartments as their phosphorylation state is modified. Although some
functional modules can be defined, paving the way for distributed simulation,
most of the time the portions of program handling the various algorithms have
to share memory, in order to continuously update the concentration, location
and state of biological objects.
From
detailled simulations of a very restricted compartment, such as the
post-synaptic densities, to models of the whole cell, Different methods have
been used to simulate neuronal function. We shall provide several examples, and
discuss the advantages and pitfalls of the various approaches.
M-L04 Signal transduction and cancer – generation of high quality quantitative data
Ursula Klingmüller
Theodor Boveri
Group Systems Biology of Signal Transduction, German Cancer Research Center, Im
Neuenheimer Feld 280, Heidelberg 69120, Germany, Phone: +49 6221 42 4481,
FAX: +49 6221 42 4488, e-mail: u.klingmueller@dkfz.de
Cellular responses are controlled
by the activation of multiple signaling pathways that form complicated networks
of cross-talk and synergisms. Deregulation of the tight control of signal transduction leads to diseases such as the
onset of cancer. Although the core components of
many signaling pathways have been identified, very little is known how
information is processed through these pathways and how cellular decisions are
regulated. By combining experimental data with mathematical modeling systems
biology opens novel possibilities to analyze and predict the dynamic behavior
of signaling pathways deregulated in cancer. However, a major bottle-neck
currently is the lack of reliable quantitative data. We developed methods to
advance the established techniques of immunoprecipitation and immunoblotting to
highly accurate and quantitative procedures. As a consequence, reliable and
accurate time-resolved data of phosphorylated and total protein levels can be acquired and absolute concentrations of proteins can
be determined. This permits the generation of quantitative data for
systems-level analysis and thereby will facilitate the prediction of targets for
efficient intervention that could be used for the development of novel
anti-cancer therapies.
Symposium 
Guided General Discussion: Identifying issues concerning
the Systems Biology of Multicellular organisms
Michel Eichelbaum
Boris Kholodenko
Ursula Klingmüller
Nicolas Le Novere
Symposium 
Multicellular Organisms
Workshop Talks
&
Short Talks
M-W01 Computer
simulation analysis of ErbB signaling for understanding of cellular
transformation mechanism
Kaori Ide 1,
Takeshi Nagashima 1, Yoshiki Yamaguchi 1,
Takashi Naka 2, Shuhei Kimura 3,
Atsushi Suenaga 1, Makoto Taiji 1 and Mariko Hatakeyama 1
1 Bioinformatics Group, RIKEN Genomic Sciences
Center, 1-7-22 Suehirocho, Tsurumiku, Yokohama 230-0045, Japan,
Phone: +81-45-503-9302, FAX: +81-45-503-9158, e-mail: marikoh@gsc.riken.jp,
Web: http://big.gsc.riken.jp/big/Research
2 Kyushu Sangyo Univ. 3 Tottori Univ.
Deregulation
of ErbB receptors is implicated in various kinds of human cancers. We made
mathematical models of the ligand-induced ErbB signal transduction pathways based on
the experimental data using the ErbB receptor expressing CHO cells. The
coexpression of the different ErbB receptors induced changes in biochemical
dynamics as well as in protein-protein
interaction patterns. Interestingly, the coexpression of different ErbBs
induced specific biological architectures such as integration of pathways (e.g.
activation of B-Raf in addition to Raf1 activation) that enable to amplify an
initial signal. And those changes in signaling pathways affect the following
gene expression profile.
On the other hand, protein-protein interaction (PPI) dynamics plays an
important role to stimulate or attenuate signal transduction pathway and define
the later cellular state. We apply molecular dynamics (MD) simulation of
signaling proteins, such as Grb2, PI3K and Shc, for understanding of regulatory
mechanism of signal transduction pathways. MD simulation well predicted the
binding affinities of PPI and binding properties. Such a molecular knowledge
can be applied for precise mathematical modeling of signal transduction
network.
M-W02 Integration of signal transduction and cytokine expression in T
lymphocytes
Thomas Höfer
Department of
Theoretical Biophysics, Humboldt University Berlin, Invalidenstrasse 42, Berlin
10115, Germany, Phone: +49/30/2093/8592, FAX: +49/30/2093/8813,
e-mail: thomas.hoefer@biologie.hu-berlin.de
In the
course of an immune response, cytokines received and expressed by T cells play
a pivotal role in regulating their proliferation, effector function, and
differentiation into memory cells. We have studied the regulatory networks
involved in cytokine signaling at different temporal and spatial scales. Intracellular signal transduction is mediated by
cycles of reversible phosphorylation and nuclear transport of Stat transcription factors. By a
combination of mathematical modeling and experimentation, we have identified
critical control steps of the interferon/Stat1
network. Based on this analysis, we were able to design a Stat1 mutant whose
altered kinetic parameters render it a more
potent transcriptional activator than the wildtype protein. Thus
signaling pathways may not be optimized solely for efficient relay and
amplification. Rather, the sensitivity of the Stat1 pathway can be adapted to
its input signal. The Stats commonly induce their activating cytokines and
cytokine receptors, which results in autocrine feedback loops. A kinetic model
for the growth factor interleukin-2 shows how multistationarity caused by
autocrine signaling can underlie the discrete decision of a cell on its future
proliferation. The system exhibits complex dynamics of intercellular
communication which explain seemingly paradoxical findings on the observed
functions of interleukin-2 as a T cell activator and a suppressor of immune
responses. Interestingly, the feedback regulation of the expression of signal
transduction components is a frequent motif in T cell cytokine signaling. We
will discuss the implication of this design for developing mechanistic models
of gene-regulatory
networks.
M-S01 Inferring
feedback mechanisms in cellular transformation due to oncogenic RAS
Nils Blüthgen 1, Christine Sers 2,
Jana Keil 2, Szymon M. Kielbasa 1,
Reinhold Schäfer 2 and
Hanspeter Herzel 1
1 Theoretical Biology, Humboldt University
Berlin, Invalidenstr. 43, Berlin 10115, Germany, EU,
Phone: +49/30/20938496, FAX: +49/30/20938801, e-mail: nils@itb.biologie.hu-berlin.de,
Web: http://itb.biologie.hu-berlin.de/~nils/
2 Insitute of Pathology, Charite, Berlin
Intracellular
signaling cascades display distinct activation profiles in response to various
stimuli. Such activation patterns are strongly influenced and shaped by
feedback loops. Different feedback loops can act in a cell context- and
stimulus-dependent manner and produce a variety of temporal activation profiles,
including oscillations and hysteresis. The MEK-ERK cascade plays an
important role in cell-cycle regulation, differentiation and in cell
transformation caused by oncogenic RAS. This cascade is
regulated by several positive and negative feedback loops and is essential for
signal transmission due to many different stimuli.
While post-translational feedback loops have been subject to extensive
mathematical modeling, feedbacks that are mediated by transcriptional control are still poorly understood.
Using a combination
of time-course experiments, mathematical modeling and bioinformatic analysis we investigate the effect of
transcriptional feedback regulation in cellular transformation following
induction of oncogenic RAS. In fibroblasts
harboring an inducible RAS oncogene, we monitor the phosphorylation of ERK1,2
by Western Blot analysis. In addition, we analyze the expression profiles of
RAS target genes with microarrays in a time-resolved manner. The
phosphorylation of ERK shows a biphasic response upon constant induction and an
oscillatory response after brief induction of RAS. We find that several dual
specific phosphatases are expressed with similar kinetics. A bioinformatic
analysis unveils two ERK-dependent transcription factors that control this battery of phosphatases. Together with
the transcription factors, these phosphatases constitute a negative feedback
for ERK-activity. Mathematical modeling and experimental interference shows
that we can explain the biphasic and oscillatory dynamics as a result of
phosphatase activation.
M-S02 REGULATION
of MAPK signalling determining cell fate in PC-12 cells - a step beyond
biochemistry
Silvia D. Santos, Eli Zamir, Peter Verveer and
Philippe Bastiaens
Cell Biology and
Biophysics, EMBL, Meyerhofstrasse-1, Heidelberg D-69117,
Germany, Phone: +39/6221/387406, FAX: +39/6221/387242, e-mail: santos@embl.de
Mitogen
activated protein kinase (MAPK)
cascades participate in a wide array of cellular transduction programmes
including cell growth and division, movement, differentiation and cell death. A
paradigm system to study how the activity of these cascades produces different
cell responses is the PC-12 cells system. In these cells the classical ERK
pathway is activated by both EGF and NGF, giving rise to cellular opposite
fates – division and differentiation, respectively. We believe different
biochemical topology may be the key determining these specific responses. We
are therefore interested in measuring reaction states of main components of this
pathway, to analyze how the kinases are spatially organized and biochemically
connected. We are using polychromatic fluorescence activating cell sorting
(FACS) with phospho-labelled antibodies, which detect the active state of
network components. By applying systematic perturbations of activities and
subsequent read out on multiple reaction states at steady-state we are able to
retrieve information on the network topology. Single cell measurements are
being performed and RNAi and pharmacological inhibitors
used for the perturbations. Moreover, response coefficients for each kinase,
before and after perturbations will be calculated and first order connectivity maps built.
In addition, by using fluorescence resonance energy transfer (FRET) imaging with multiple optical
sensors, reaction states of kinases and their spatial information are being determined
simultaneously in one cell. Fusion proteins of GFP mutants and pathway kinases
allow the detection of protein-protein interactions and molar ratios of
phospho-proteins can be detected, by using phospho-antibodies against
phospho-residues on active kinases.
M-S03 Mathematical
modeling of neuronal response to neuropeptides: Angiotensin II signaling via
G-protein coupled receptor
Thomas Sauter 1, Rajanikanth Vadigepalli 2 and
James Schwaber 2
1 Institute for System Dynamics and Control
Engineering, University of Stuttgart, Pfaffenwaldring 9, Stuttgart D-70550,
Germany, Phone: +49/711/6856611, FAX: +49/711/6856371, e-mail: sauter@isr.uni-stuttgart.de, Web: http://www.isr.uni-stuttgart.de/~sauter/
2 Daniel Baugh Institute, T. Jefferson University, Philadelphia, PA
In neurons G-protein coupled
receptors (GPCRs) are involved in the alteration of neuronal activity
(neuromodulation) via cascades of interacting proteins. The complex dynamic
behavior of these networks, e.g. the integration of different signals, cannot
be understood by intuition alone. Mathematical modeling provides an appropriate
tool to decipher this complexity. Angiotensin II and AT1 receptor dependent
signaling was investigated as an examples that use GPCR signaling pathways
(Gq). AT1 signals via a wide variety of intracellular signaling molecules,
involving (1) G-protein mediated stimulation of phospholipase C (PLC), with
subsequent Ca2+ mobilisation; (2) Jak/STAT pathway; (3) transactivation of
tyrosine kinase pathways.
Relevant signaling outputs are modified gene expression
patterns and modified neuronal activity via changes in membrane ionic
currents and firing rate.
New data that was collected recently [Fernandez et al.,
Hypertension Jan.2003:56-63] showed that Angiotensin II can elicit stimulating
and suppressive effects in the same neurons in dependency of the basal Ca2+
level. We have built a detailed mechanistic model of Angiotensin II signaling
that captures both the stimulating and suppressive effects. This ODE model
includes the AT1 mediated activation of PLC and PKC, and IP3 and channel
mediated variation of the cytosolic Ca2+ level after Angiotensin II stimulation
(adapted from [Mishra and Bhalla, Biophys. J., 83:1298-1316, 2002]). Based on
in silico simulations of this model, we hypothesize that the observed
biological variability is based on cell-to-cell variation in the dynamics of
the Na-Ca exchanger.
Furthermore, a Hodgkin-Huxley model approach is used to investigate the
function of cell signaling in altering the firing behavior of NTS neurons in
response to various baroreceptor stimuli. Angiotensin II was found to activate
neuronal firing in low firing NTS neurons.
In summary, detailed mathematical is a
valuable tool to understand and investigate neuronal response to neuropeptides
and furthermore to link signal transduction
to the electrophysiological behavior of neurons.
Symposium 
Resumed General Discussion: Addressing the
issues
Michel Eichelbaum
Boris Kholodenko
Nicolas Le Novere
Ursula Klingmüller
NovoNordisk Closing Lecture
Highlights of SysBio2005: From genes to whole organs
Vertical integration using mathematical simulation
Denis Noble
University
Laboratory of Physiology, Park Road, Oxford OX1 3PT, UK
Biological modeling of cells,
organs and systems has reached a very significant stage of development.
Particularly at the cellular level, there has been a long period of iteration
between simulation and experiment (Noble 2002a). We have therefore achieved the
levels of detail and accuracy that are required for the effective use of models
in systems biolocal research and in drug development. To be useful in this way, biological models must
reach down to the level of proteins (receptors, transporters, enzymes etc), yet
they must also reconstruct functionality right up to the levels of organs and
systems (Noble 2002b). I will illustrate these points with reference to both
the proceedings of this course and models of the heart (Noble 2002c). The
lecture will use this work to illustrate some fundamental principles of
Systems Biology.
Noble
D (2002a) The Rise of Computational Biology. Nature Reviews Molecular Cell
Biology, 3, 460-463
Noble
D (2002b) Modelling the heart: insights, failures and progress. Bioessays 24,
1155-1163
Noble
D (2002c) Modelling the heart: from genes to cells to the whole organ. Science
295, 1678-1682
P-S01 Smart regulation of ammonium assimilation by Escherichia coli: modularity, robustness, and flux regulation.
Frank
J. Bruggeman, Fred C. Boogerd and Hans
V. Westerhoff
Dept of
Molecular Cell Physiology, Faculty Earth & Life Sciences & Biocentrum
Amsterdam, De Boelelaan 1085, Amsterdam NL-1081 HV, The
Netherlands, EU, Phone: +31/20/4447248, FAX: +31/20/4447229,
e-mail: frank.bruggeman@falw.vu.nl,
Web: www.angelfire.com/scifi/frankb
Regulation of ammonium assimilation in E. coli is governed by two mechanisms: (i) by glutamine synthetase (GS)
and glutamate synthase (GOGAT) and (ii) by glutamate dehydrogenase (GDH). The
former system is active at low ammonium concentrations and the latter system
gradually takes over ammonium assimilation as function of an increase in the
ammonium supply. The net ammonium assimilation flux (Jn) is the sum of both mechanisms. A kinetic model of ammonium assimilation shall be introduced. It will be
analyzed in terms of: (i) the robustness of Jn and (ii) the
regulation of the individual ammonium-assimilation fluxes of GS/GOGAT (Jgs)
and GDH (Jgdh). The system will be dissected into mechanisms that
guarantee nontrivial robustness of Jn by ‘smart’ regulation of Jgs and Jgdh.
The regulation of Jgs shall be further analyzed in terms of the
contributions of different processes. Both types of analysis was carried out in
terms of a modular description of the network to facilitate understanding of
this complicated regulatory network, which involves feedback regulation,
covalent modification, parallel pathways, intracellular signalling via the
two-component mechanism and gene expression.
P-S02 Design Principles of Signal Transduction
Pathways to attenuate Noise
Markus Kollmann, Kilian Bartholome and Jens Timmer
Department of
Physics, University Freiburg, Hermann-Herder-Str. 3, Freiburg D-79104, Germany,
Phone: +49 761 203 5828, FAX: +49 761 203 5967, e-mail: markus.kollmann@physik.uni-freiburg.de,
Web: http://webber.physik.uni-freiburg.de/~markus/
One of the great paradoxes in studying signal transduction pathways is their seemingly oversized topology. Even
in rather small signalling cascades like MAP kinase it is unclear why so many kinase reactions are involved and what
benefits multi-phosphorylation sites. Similarly one can show in bacterial
chemotaxis that the topology can be much more simplified to arrive at almost
perfect adaption. These facts give the impression that signalling pathways are
rather 'tinkered' than 'properly engineered' [1]. But the underlying assumption
within this view on signalling pathways is the concept of 'modularisation' on
one hand and moderate component tolerances on the other hand. Only these
assumptions allow us to investigate signalling networks ignoring strong
intra-cellular perturbations. In this work we show that the topology for
bacterial chemotaxis depends crucial on strength of intra-cellular
perturbations. We show that chemosensory pathways are not only designed to
transmit changes in ligand concentration to the flagella motor proteins under
the condition of almost perfect adaption but also to resist inter-cellular
noise. For the bacterium E.coli the magnitude of variations in concentration of signalling
proteins has been measured in detail [2,3] and can vary up to ten-fold between
individuals [2]. From the known strength of fluctuations we can interfere the
requirements on the topology to attenuate these variations. Under realistic
assumptions of variations in binding constants and stochastic noise effects we
show that the topology of chemotaxis pathways are not 'tinkered' but the
outcome of an evolutionary optimisation process.
[1] Alon U., (2003), Science, 301
[2] Li M. & G. Hazelbauer, (2004), J.Bact., 186
[3] Elowitz M. et al., (2002), Science, 297
P-S03 On pathways and distances in metabolic
networks
Esa Pitkänen 1, Ari Rantanen 1, Juho Rousu 2 and Esko Ukkonen 1
1 Department of
Computer Science, University of Helsinki, P.O.Box 68 (Gustaf Hällströmin katu
2b), Helsinki 00014, FINLAND, Phone: +358/40/5314252,
FAX: +358/9/1915 1120, e-mail: esa.pitkanen@cs.helsinki.fi
2 Department of Computer Science, Royal Holloway, University of London
Recent 'small-world' studies of the global
structure of metabolicnetworks have been based on the shortest-path distance.
As this distance does not capture accurately the complexity of the underlying
biochemical processes, we propose new distance measures that are basedon the
structure of feasible metabolic pathways between metabolites. We define a
metabolic pathway as a minimal set of metabolic reactions capable of converting
the source metabolites into the target metabolites. The metabolic distance is
defined as the number of reactions in a smallest possible pathway connecting
the sources to the targets. The production distance is defined as the minimum
number of successive reactions needed for such conversion, and is upper-bounded
by the first distance. These concepts are defined using an and-or graph induced
by the metabolic network.
We study the computational complexity and derive algorithms for evaluating the
distances. We also provide a linear-time algorithm for finding an upper bound
for the metabolic distance which itself is shown NP hard to evaluate. To test
our approach in practice, we calculated these and shortest-path distances in
two microbial organisms, S. cerevisiae and E. coli. The results show that
metabolite interconversion is significantly more complex than was suggested in
previous small-world studies. We also studied the effect of reaction removals
(gene knock-outs) on the connectivity of the S. cerevisiae
network and found out that the network is not particularly robust against such mutations.
P-P01 The use of accurate mass and
time tags to measure yeast’s glycolytic proteome.
Ronald Aardema, Henk L. Dekker, Jaap
Willem Back, Leo J. de Koning, Luitzen de Jong and Chris
G. de Koster
Biomacromolecular
Mass Spectrometry, Swammerdam Institute of Life Sciences, UvA, Nieuwe
Achtergracht 166, Amsterdam 1018 WV, The Netherlands, EU, Phone: +31 20 525 5669,
FAX: +31 20 525 6971,
e-mail: aardema@science.uva.nl
Within the Vertical Genomics project, focus
lies on the glycolysis of Saccharomyces
cerevisiae. This pathway will be studied
with respect to promoter activities, mRNA levels, protein concentrations, enzyme activities, metabolite concentrations and
metabolic fluxes, under well defined growth conditions.
The high-throughput proteomics method, presented here is based on a combination
of stable 1 5N isotope
labeling of a reference culture to produce isotopically labeled internal
standards for calibration purposes and the use of accurate mass and time tags (AMT tags) to quantify 1 4N/ 1 5N isotope ratio’s
of glycolytic peptides obtained by tryptic digestion of whole yeast cell
lysates.
Peptides of 14 glycolytic enzymes are readily detected. However,
nanoLC-QTOF-MS/MS measurements and data processing with our in-house developed
Virtual Mass Spectrometry Lab software package have shown the need to invoke
normalized LC retention times of glycolytic peptides. To increase coverage of
all enzymes of interest simple strong cation exchange chromatography is used.
Once unique peptides of the proteins of interest are found (using MS/MS), the
accurate mass is calculated and stored in a database together with the
normalized retention time during reversed phase chromatography. This retention
time in combination with high mass accuracy will be used to obtain unique AMT
tags. With these validated AMT tags, differences in protein concentration can
be measured between the 1 5N
labeled glycolytic subproteome of the reference culture and yeast culture grown
under varying experimental conditions with LC-FTICR-MS analysis.
The data will be correlated with data from other hierarchic levels of the
pathway such as the transcriptome, enzyme activities and the metabolome to gain
insight in higher order regulation of glycolysis.
P-P02 Metabolic footprinting: its role in
systems biology
Marie Brown, Rick Dunn, Julia Handl and
Douglas Kell
Department of
Chemistry, The University of Manchester, PO Box 88 Sackville Street, Manchester
M60 1QD, England, Phone: +44/161/200/4414, FAX: +44/161/200/4556,
e-mail: m.brown-3@postgrad.manchester.ac.uk
Of the 6000 genes sequenced in the yeast Saccharomyces cerevisiae
genome fewer than 50% have function that is confidently known. One
approach to determine gene function is the metabolome-based analysis of single gene knockout
mutants. Metabolic footprinting is concerned with the analysis carried out
where the extra-cellular metabolites secreted into the culture medium are
analysed and provides a rapid and non-invasive methodology (Allen et al. 2003).
Analytical technologies employed include chromatography-mass spectrometry, direct injection mass spectrometry and Fourier
Transform-Infra Red spectroscopy (Dunn et al. 2004).
A variety of methods as detailed below are then
used to obtain information about the gene knockouts and determine which show similar patterns of metabolic
changes. Generally gene functionality can be determined by calibration of genes
of known function with those of unknown function or by comparison of the
metabolic profiles of mutant with wild-type strains.
• standard chemometric methods including Principal
Components Analysis
• a machine learning approach, genetic programming
• a new multiobjective clustering method
• integration of results with high confidence protein-protein and genetic
interactions
• the use of contraints-based modelling software to compare experimental and in
silico results (Förster et al. 2003)
The challenges encountered in this work, e.g.
the collection and instrument analysis of biological data and the selection of
appropriate data pre-processing and analytical methods, are typical of those
encountered in many ‘omic’ studies. From this data hypotheses can be tested
experimentally, results integrated with those obtained from proteomic and
transcriptomic studies and incorporated into existing models to gain more biological
understanding of gene function. These are first steps in trying to build towards
achieving the core aim of system biology, the wish to understand the whole and
build quantitative models that will allow the investigation of dynamic systems.
Acknowledgement The authors would like to thank the BBSRC for their financial
support.
Allen, J.,
et al. (2003). Nature Biotechnol., 21, 692
Dunn, W., et al., (2004), TrAC, In Press
Förster, J., et al. (2003), Genome Res., 13, 244
P-P03 Genetic
network model for the AP-1 system
David Camacho and
Roland Eils
iBIOS, DKFZ, Im Neuenheimerfeld 580, Heidelebrg WB 69120,
Germany, Phone: 49 (0) 6221 42 2720, FAX: 49 (0) 6221 42-3620,
e-mail: d.camacho@dkfz.de,
Web: www.dkfz.de/ibios
The AP-1 system is a dimer composed by
different transcription factor families. Activation of signal transduction pathways as cellular and oxidative stress, DNA damage, antigen binding in lymphocytes, and cytoskeleton
rearrangement converge to it. Therefore AP-1 system is centrally involved in
different phenotypic responses such as differentiation, proliferation, cell
arrest, and neoplastic development. Here, we model the combinatorial genetic
network that is derived from the activation of this AP-1 transcription factor
as an additional component of cellular information processing.
While AP-1 network architecture was reconstructed by scanning of the literature
we establish it dynamics by correlating cellular stimulus with experimental
transcription expression profiles of their related genes and their different
cellular phenotypic responses. To reproduce the activation over time of the
mentioned transcriptions factors we model them with nodes of Continuous Time
Recurrent Neural Networks (CTRNN). CTRNN have been shown to approximate the
trajectories of any smooth dynamical system. For evolving the CTRNN, a simple
variation of a Genetic Algorithm is utilized in order to achieve with the
multi-objective optimisation problem.
Our modelling approach successfully can reproduce the dynamics of the AP-1
system. Together with additional experimental data, the genetic network model
of AP-1 will be used to clarify and predict the central role of AP-1 as a key
player in the signalling system of the cell.
P-P04 Pathways to analysis of microarray data
R.
Keira Curtis and Antonio Vidal-Puig
Clinical
Biochemistry, University of Cambridge, Box 232, Addenbrooke's Hospital, Hills
Road, Cambridge CB2 2QR, UK, Phone: +44 1223 336781, FAX: +44 1223
330598, e-mail: rkc24@cam.ac.uk,
Web: http://www.clbc.cam.ac.uk
Microarrays are increasingly used to profile
genome-wide gene expression. A common aim in the analysis of gene expression data
is the integration of array data with biological annotations, in order to
identify pathways (such as metabolic or signalling pathways), or functions
(such as those from the gene ontology) that are co-ordinately regulated.
Several methods of pathway analysis quantify the overrepresentation of pathway
annotations in genes displaying a particular expression pattern. These include calculating a
‘z score’, based on the observed number of matches between a gene list and a
pathway and the expected matches, and ‘gene set enrichment analysis’ (GSEA),
which involves comparing a pathway to a ranked list of microarray data. We
compare these two methods of analysis on the same microarray dataset in order
to investigate their differences. Broadly, the same pathways are indicated as
overrepresented according to the two contrasting methods, but GSEA tends to
find more significantly regulated pathways.
P-P05 Multiscale
modelling of a cell
Gianni De
Fabritiis and
Peter Coveney
Centre for
Computational Science, Chemistry Department, University College London, 20
Gordon street, London WC1H 0AJ, UK, Phone: +442076795300,
FAX: +442076795300, e-mail: g.defabritiis@ucl.ac.uk,
Web: http://www.iac.rm.cnr.it/~gianni
The relevant information of the functioning of
a cell covers several orders of magnitudes in time and length scales. For such
a reason, simpler models based on ordinary (stochastic) differential equations
(ODEs, SDEs) are used to simulate the emergent macroscopic behaviour, rather
than more fundamental molecular descriptions. This approach is based on a
vertical integration of models where the coefficients of the ODEs/SDEs are
representative of the molecular information [1]. At a more detailed level,
molecular dynamics (MD) and Brownian dynamics (BD) simulations can be used to
compute these coefficients which are fed into the differential equations. In
this case, the computational requirements are much higher and the complexity of
the codes is such that appropriate coupling frameworks have to be used. Systems
Biology Workbench (SBW) furnish a way to expose these models as 'remote
procedures', which can be interfaced and coupled to describe higher level cell
functions. However, due to the large computational cost, a possible interesting
approach is to prepare such simulations in a way that they are exposed as GRID
services [2] which can be accessed by ODE cell models to explore functional
effects of microscopic mutations. With RealityGrid middleware for GRID
computing, the computational resources are not assigned in advance and the
effort required to expose the simulation as a GRID service is minimum. In
particular, we look at a coupled molecular dynamics model and fast methods for
the calculation of the chemical potential [3] which open a larger window on
equilibrium and non-equilibrium MD simulations and has potentially interesting
implications for the simulations just mentioned. In perspective, it would interesting
to integrate these GRID services (such as BD and MD simulations, ODE solvers,
etc) in order to explore the macroscopic functional effects at several levels
of descriptions of a cell.
[1] Integrative Biology project, www.integrativebiology.ox.ac.uk;
[2] RealityGrid project, www.realitygrid.org;
[3] R. Delgado-Buscalioni and P.V.Coveney, Phys. Rev. E 67, 046704 (2003);
G. De Fabritiis, R. Delgado-Buscalioni and
P.V.Coveney, publishing J. Chem. Phys. (2004)
(www.integrativebiology.ox.ac.uk/publications/JChemPhyspaper.pdf)
P-P06 A
genetical genomics approach to gene network inference
Alberto de
la Fuente, Bing Liu and
Ina Hoeschele
Virginia
Bioinformatics Institute, Virginia Polytechnic Institute and State University,
1880 Pratt Drive, Blacksburg VA 24061, USA, Phone: 001 540 231 1791,
FAX: 001 540 231 2606, e-mail: alf@vbi.vt.edu
One of the challenges in systems
biology is to discover the interaction structure of biochemical systems. We
here propose to use a systematic approach to infer gene networks using several types of experimental data and consisting of several steps. The
experimental data used in this approach consist of gene expression profiles and
a detailed genomic marker map of multiple individuals in a segregating
population; genetics and genomics combined. The steps are:
(1) Genome-wide QTL analysis of gene expression profiles to identify eQTL ('expression' quantitative
trait loci) confidence regions
(2) Identification of regulatory candidate genes in each eQTL region
(3) Directional links are established from regulatory candidate genes to
genes affected by the eQTL
(4) Statistical validation and refinement of the inferred network
structures using Structural Equation Modeling.
The rationale behind this approach is that each gene's expression level can be
treated as a quantitative phenotypic trait and used in conventional QTL analysis
to identify chromosomal locations (QTLs) affecting it. In step 1 we thus find
chromosomal locations (eQTLs) having causal influences on the expression of
particular genes. In step 2 we identify genes located in the eQTL regions. The
genes located in these eQTL regions are potential regulators of the genes whose
expression is affected by these eQTLs. We can then propose causal links between
regulator (located in the eQTLs) genes and the regulated genes (affected by the
eQTLs)(step 3). The QTL analysis may suggest an encompassing network from all
identified regulatory interactions, as well as candidate links which may be
indirect and, hence, eliminated. In the final step, alternative models are
tested using Structural Equation Modeling, a statistical technique for
evaluating linear interaction models. In this technique, models are scored
based on how well their model-implied covariance matrix matches the observed
covariance matrix. In my presentation I will illustrate and evaluate this
strategy using a series of simulated data.
P-P07 A
dynamic model of cAMP signal transduction in yeast
Dirk Müller 2, Helena Diaz-Cuervo
1,
Luciano Aguilera-Vazquez 2, Klaus Mauch 2 and
Matthias Reuss 2
1 Centro de
Investigacion del Cancer, Universidad de Salamanca/CSIC, Campus Miguel de
Unamuno, Salamanca 37007, Spain, Phone: 0034/923/294805,
FAX: 0034/923/294795, e-mail: helenadc@usal.es,
Web: www.cicancer.org
2 Institute of Biochemical Engineering, University of Stuttgart,Allmandring 31,
D-70569 Stuttgart, Germany
Yeast cells possess a number of
signaling pathways to integrate information about nutrient supply with cellular
growth and proliferation. The signaling route mediated by cyclic AMP (cAMP) and protein kinase A (PKA) regulates a large number of targets both at the
posttranslational and transcriptional level in response to changes in carbon
source availability. By affecting both metabolic processes and the cell cycle machinery, it also serves to coordinate cell growth and division.
Measurements were performed in synchronous cultures and in oscillating
continuous cultures of Saccharomyces cerevisiae to analyze the cell-cycle dynamics of cAMP and energy metabolism. A modular single-cell model integrating cAMP signaling
with descriptions of the cell cycle machinery and central carbon metabolism is
currently under development. This single-cell model will permit to simulate
cellular behavior resulting from the joint action of metabolic and signaling
networks during the yeast cell cycle. The present contribution focuses on two
models of cAMP signal transduction, which can be used as
exchangeable submodules in the integrated model. On the basis of an extensive
literature survey, two dynamic models of the cAMP signalling pathway were
developed, both of which provide a comprehensive description of the current
knowledge, but differ in the level of detail. They account for stimulation of
adenylate cyclase via Ras and a GPCR system, cAMP destruction by
phosphodiesterases, (in)activation of PKA, and for the negative feedback
exerted by PKA on its own activity. Results of the above-mentioned experiments
were employed in combination with literature data and stability constraints to
estimate model parameters. As a starting point, protein levels determined in a
genome-wide analysis [1] served as
estimates of the initial values of model species. Simulation results of both
the small-scale model (20 reactions) and the large-scale model (400 reactions)
will be presented and compared to experimental findings. The models provide a
basis to address open questions regarding the underlying network structure and
dynamic behavior of this signaling pathway. Plus, they can serve as a tool to
identify suitable experimental conditions to efficiently discriminate between
alternative hypotheses. Future work aims at incorporating spatial information and transcriptional regulation of key components of
the cAMP pathway into the model.
[1] Ghaemmaghami, S., et al. (2003) Nature 425(6959): 737-741.
P-P08 Metabolic quorum sensing:
experiments with S. cerevisiae
Francesco d'Ovidio 1, Silvia De Monte 2,
Sune Danø 3 and Preben Graae Sørensen 4
1 , École
normale supérieure, 24, rue Lhomond, Paris F-75231, France, EU, Phone: +33
1 44322223, FAX: +33 1 44322223, e-mail: dovidio@lmd.ens.fr
2 École Normale Supérieure, Paris, France
3 Dept. of Medical Biochemistry and Genetics, Univ. of Copenhagen, Denmark
4 Dept. of Chemistry, Univ. of Copenhagen, Denmark
Populations of yeast cells in a homogeneous environment are able to synchronise their
metabolic activity. Such a phenomenon can be viewed as a possible step in the
development of multicellularity. The mechanism by which such collective
periodic oscillations arise is however not yet fully understood.
In order to explore the nature of cell coupling, we performed experiments on
yeast cells suspensions at different cell density. Using Stuart-Landau
formalism, the microscopic state of the population is deduced by studying the
response of the metabolic activity to perturbations and forcing with
acetaldehyde.
A transition to incoherent oscillations was expected for low densities. The
analysis of our experiments instead point to a different scenario, where each
individual cell stops to oscillate when the cell density falls below a critical
value. Thus, our experiments suggest that metabolic oscillations can sustain a
quorum sensing-like mechanism in eukariotes.
P-P09 Phylogenetic
analysis based on structural information of metabolic networks
Oliver Ebenhöh, Thomas Handorf and
Reinhart Heinrich
Department of
Biology, Humboldt University Berlin, Invalidenstr. 42, Berlin 10115, Berlin,
Phone: +49 30 2093 8382, FAX: +49 30 2093 8813, e-mail: oliver.ebenhoeh@rz.hu-berlin.de
In this work we propose a method to extract
phylogenetic information based on structural properties of the metabolic
systems of different organisms. Such information on the relatedness of
different species is commonly depicted in the form of phylogenetic trees.
In the literature there are many examples of the reconstruction of phylogenetic
trees. Regularly, the data source on which such calculations are based upon are
similarity measures between genomic, tRNA or protein sequences. In this way, hypotheses on the evolution of specific
enzymes or genes can be developed.
Rather than focussing on single enzymes or genes, in this work we propose to
use as a data source information regarding the complete metabolism of an
organism. Based on metabolic pathway information of over 160 species contained
in the KEGG database, we identify the number and types of organisms in which
each reactions occurs. This information allows to draw conclusions on which
pathways are widely conserved, and therefore can be assumed to have appeared
early during the evolutionary history of metabolic pathways, and which pathways
are specific to certain organisms and can therefore be considered to have
appeared later.
Also based on the pathway information from the KEGG database, we define various
distance measures, use these to calculate phylogenetic trees, and compare their
structure. Already simple distance measures such as the number of reactions
which are different in two organisms result in trees in which related
organsisms are generally grouped together. A more elaborate definition of the
distance measure is based on the principles of network expansion presented by
Handorf et al. This method describes the emergence of complex metabolic
networks from simple structures. We claim that the proposed measure based on
expansion processes more realistically reflects the evolutionary distance
between metabolic networks since it is based on elementary mutation steps
altering the network structures.
P-P10 Modelling of Drosophila segmentation gene expression with and without usage of attractors
Vitaly
V. Gursky 1, Johannes Jaeger 2, Konstantin N. Kozlov 3, John Reinitz 2 and Alexander
M. Samsonov 1
1 Theoretical Department, Ioffe Physico-Technical Institute, 26
Polytekhnicheskaya, St. Petersburg 194021, Russia, Phone: +7 812 247 9352,
FAX: +7 812 247 1017, e-mail: gursky@math.ioffe.ru
2 Department of Applied
Mathematics and Statistics, Center for Developmental Genetics, Stony Brook
University, Stony Brook, NY 11794-3600, USA
3 Department of
Computational Biology, Center for Advanced Studies, St. Petersburg State
Polytechnic University, 29 Politekhnicheskaya st., St. Petersburg 195251,
Russia
We have found recently that explicit
representation of nuclear structure of Drosophila embryo is not necessary for modelling pattern formation in Drosophila segment determination (Gursky, V. V., et al. (2004) Physica
D, v.197, p.286). This conclusion follows from the fact that correct pattern
dynamics can be obtained in both spatially discrete and spatially continuous
models. In particular, the successive nuclear divisions occurring in the embryo
during this developmental time appears to be not connected to pattern formation
in segmentation gene expression, because different schemes for mitosis lead to qualitatively
the same results. One of important features of the considered models (both
discrete and continuous) is that solution stays far from steady state
(attractor) at all times. We study the discrete model modified by an assumption
that final expression patterns are close to an actual attractor of the
dynamical system. Parameter values (elements of a matrix of interactions
between genes and rates for synthesis, decay, and diffusion of gene products) in the equations are found by fitting numerical
solution to expression data. Advantages and drawbacks of using attractors to
model gene expression patterns are shown in this specific system. We also
demonstrate that in terms of attractors mitosis can play a dynamical role,
providing a natural selector in multiple steady states which are potentially
possible for the gene circuit.
Acknowledgments: The support of the study by the NIH Grants RR07801, TW01147,
the CRDF GAP Awards RBO685, RBO1286 is gratefully acknowledged.
P-P11 Discovering
compound mode of action with CutTree
Kristofer Hallén 1, Johan Björkegren 2 and Jesper Tegnér 3
1 ,
Linköpings universitet, IFM, Linköping 581 83, Sweden,
Phone: +46 13 286801, FAX: +46 13 137568, e-mail: hallen@ifm.liu.se, Web: http://www.ifm.liu.se/~hallen
2 CGB, Karolinska
Institutet, S-171 77 Stockholm, Sweden. Clinical Gene Networks AB, Karolinska
Science Park, Fogdevreten 2B, S-171 77 Stockholm, Sweden
3 IFM, Linköping University,
S-581 83 Linköping, Sweden. Clinical Gene Networks AB, Karolinska Science Park,
Fogdevreten 2B, S-171 77 Stockholm, Sweden
Defining the primary targets of chemical
compounds and thereby improving the selection of which targets to be further
evaluated in animal models and clinical trials (Phase I- III) has for long time
been a key issue for pharmaceutical companies because improving the accuracy of
this selection process will be extremely cost-effective. To reveal the
mechanisms of action of a drug, it is necessary to identify
the primary affected genes, the PAGs. We have developed an algorithm, CutTree,
which distinguishes the PAGs from the secondary downstream effects of a
chemical compound. CutTree provides an integral experimental design of cellular
perturbation experiments and whole-genome expression measurements. Unlike previous methods (Gardner et al.,
Lum et al.) CutTree does not depend on the availability of whole-genome
deletion libraries or a complete map of the gene network architecture. The efficacy of CutTree, compared to
strategies depending on whole-genome deletion libraries, increases with the
number of primary affected genes. An experimental validation reveals that
CutTree discovers 80 % (4/5) of the primary targets of galactose in yeast using 17 samples of micro-array data. This success rate is almost
identical to our predicted CutTree accuracy of 70 % in the case of five primary
targets.
1 Gardner,
T, et al., (2003), Science 301, 102.
2. Lum, P, et al., (2004), Cell 116, 121.
P-P12 Scopes:
A new concept for the structural analysis of metabolic networks
Thomas Handorf, Oliver Ebenhöh and
Reinhart Heinrich
AG
Theoretische Biophysik, Biologie, HU-Berlin, Invalidenstr. 42, Berlin DE 10117, Germany,
Phone: +49/30/20938325, FAX: +49/30/20938813, e-mail: Thomas.Handorf@biologie.hu-berlin.de
In this work we present a new method for the
mathematical analysis of large metabolic networks. Based on the fact that the
occurrence of a metabolic reactions generally depends on the existence of other
reactions a series of metabolic networks is constructed. The algorithm
iteratively calculates subsequent network generations by adding in each step
those reactions whose substrates can be provided by the network of the previous
generation. In our case, the reactions which are eligible to be added during
the process are taken from the KEGG database. The course and the final result
of this network expansion is strongly dependent on the initial substrates. We
define the set of compounds which can be reached by a network expansion
starting from some initial substrates as the scope of these substrates.
The concept of scopes can be used to answer the question which compounds can in
principle be synthesized from a given set of substrates using a specified set of
biochemical reactions. Moreover, we can identify reactions which are essential
for these syntheses by analyzing the robustness of the scopes against deletion
of reactions. The method can also be used to identify small chemical building
blocks from which a majority of all compounds can be synthesized. We could
show, for example, that almost half of the cellular compounds can be
synthesized from the building blocks carbon dioxide, ammonia, sulfate and
phosphate together with water. An analysis of the expansion process revealed
crucial metabolites which can influence the expansion process dramatically. In
this way we could show that common cofactors such as NAD+, ATP, and Coenzyme-A facilitate
the incorporation of a large number of reactions in subsequent generations.
The scope also represents a functional aspect of metabolic network itself. In
this way, we can answer the question to which extend different organisms can
use the available chemical resources. This information can be used to determine
evolutionary advantages of certain organisms in different chemical
environments.
We expect that the expansion process shows features characteristic for the
evolution of metabolic systems. We may argue that the dependence of reactions
on substrate availability implies a certain temporal order in which the
different reactions may have appeared during evolution. This gives reason to
assume that the history of the evolution of metabolism can be deciphered from
current metabolic networks.
P-P13 Inferring
gene regulatory relationships from time series microarray data based on
the trend of expression changes
Feng He and An-Ping Zeng
Department of
Genome Analysis, GBF - German Research Center for Biotechnology, Mascheroder
Weg 1, Braunschweig 38124, Germany, Phone: +49/531/6181188,
FAX: +49/531/ 6181751, e-mail: aze@gbf.de
The
availability of time series expression data opens up new possibilities to study
the regulatory and dynamic relationships among genes via reverse engineering.
The current approach used in literature for reverse engineering of microarray
data mainly focuses on inferring the regulatory relationships between genes
from point-to-point expression values of gene pairs. Many of the important features of gene
expression such as noises and time-shift of expressions cannot be properly
considered by this conventional approach. To consider time-shifted and inverted
gene expression profiles, Qian et al (J.Mol.Biol.,314,1053, 2001) proposed a
local clustering method. However, this method still suffers from problems such
as large local noises and calculation of time-shift point-to-point expression
values. Here, we present a novel method to infer the regulatory relationships
which is based on extracting the main characteristics of trends of expression
changes between genes and is therefore more noise-tolerant.
The method includes two major steps: first calculate the trend score and trend
correlation coefficient among all the significantly expressed genes and then to
infer their relationships based on a comprehensive index of these parameters.
We applied this method to cell-cycle expression time series data of yeast. We
detected many additional correlated gene pairs beyond those resulting from
conventional correlation algorithms including local clustering analysis. We
then examined many of the inferred gene pairs which have significantly
correlated change-trends. Some of them are well-documented in literature which
supported our predictions. For instance, we found a significantly correlated
change-trend between the gene REB1(mainly as RNA polymerase II transcription factor or Pol I
transcription termination factor)and MF(ALPHA)2.This is confirmed (with a
p-value of 7.4e-4) by the genome-wide
location analysis of Lee et al (Science, 298,799, 2002). We also found many new
gene pairs with significantly correlated change-trends the functions of which
are unknown. Based on this we could suggest possible regulatory relationships
for experimental studies. Thus, this method is useful for inferring regulatory
relationships from time series microarray data and for extending the
co-expression network inferred by conventional clustering and local clustering
algorithms.
P-P14 Secondary
metabolites can create coexistence in the chemostat
Julia Heßeler 1, Julia K. Schmidt 2,
Udo Reichl 3 and Dietrich Flockerzi 3
1 Chair of Biomathematics, Eberhard-Karls University Tübingen, Auf der
Morgenstelle 10, Tübingen 72070, Germany, EU, Phone: +49/7071/2976843,
FAX: +49/7071/294322, e-mail: julia.hesseler@web.de
2 Chair of Bioprozess
Engineering, Otto-von-Guericke University, Magdeburg
3 Max Planck Institute of
Dynamics of Complex Technical Systems, Magdeburg
For
microbial species competing for one limiting resource in a chemostat,
mathematical models and in particular the competitive exclusion principle (CEP)
predicts survival of only one species in any case. Quantitative experimental
data from our model system related to the genetic disease Cystic Fibrosis
alludes to the coexistence of at least two competing species. We developed a
new mathematical model (extension of the classical chemostat) to comply with
the experimental phenomena by including species specific properties of the
microorganisms of concern.
We will present the mathematical tools and the analysis of the mathematical
model, consisting of a four-dimensional system of nonlinear ordinary
differential equations, as well as computed simulations for experimental data.
We found that the dynamic of the system changes in a fundamental way, if
interspecific competition is included; a Hopf bifurcation occurs for an
appropriate choice of parameters.
Experimental data serve as basis for the assumptions. These are a) one species
produces a secondary metabolite, b) the metabolite has a growth-inhibiting
effect, but can also be exploited as a secondary carbon source, c) some of the
species could compete directly (e.g. via toxin production), and d) a lethal
inhibitor could be introduced that cannot
be eliminated by one of the species and is selective for the stronger
competitor.
P-P15 Two
Numerical Model Analyses for the Movement of a Restriction Enzyme.
Noriko Hiroi, Akira Funahashi and
Hiroaki Kitano
Kitano Symbiotic Systems Project, JST, 6-31-15 Jingumae, M31 6A,
Shibuya-ku, Tokyo 150-0001, Japan, Phone: +81-3-5468-1661,
FAX: +81-3-5468-1664, e-mail: nhiroi@symbio.jst.go.jp
The
inner cellular environment restricts the dimension of molecular diffusion. The
environment is different from the idealized free reaction space condition. Defining fundamental reaction
formulas are essential for faithful modeling and simulation of intra-cellular
biochemical processes. In this study, we chose the EcoRV diffusion process to
investigate the dimension restricted reaction.
There are 3 kinds of hypothesis for the targeting mechanism of EcoRV: sliding,
hopping, and jumping. The first 2 is highly correlated process that means the
dimension of the molecular diffusion is restricted, and the last is
uncorrelated process.
In this study, we analyzed the EcoRV diffusion process by 2 different numerical
models to estimate which manner is realistic as the targeting mechanism of EcoRV.
First, we tried to test by theoretical method for restricted reaction if the
targeting process is correlated or not. This analysis suggested highly
correlated process might be included in the targeting process.
Next, we tried to express the more detailed behavior of EcoRV with the
stochastic model, containing an approaching, an association, and a sliding step
by random walk.
This model could find 2 kinds of answer for reconstructing the experimental
results [1].
One is the case the enzymes slide along on DNA or dissociate far away from the substrate.
The other is the case the enzymes are well restricted around the substrate DNA
but not always slide along the substrate DNA, and repeat hopping.
To explain the experimental results of with catenane [2], the latter case is
supported because dissociation/re-association by hopping is required for the
facilitation of the digestion of the target site in catenane.
In our stochastic model, the enzymes can slide less than 40 bp by random walk.
This result agrees with the work based on a diffusion model [3].
These numerical analyses suggest that the diffusion manner of the EcoRV does
not consist of long 'sliding'. At the same time, the diffusion manner is highly
correlated and does not allow frequent 'jumping'. The most major diffusion
process is suggested as 'hopping'. This study will be applied for the analysis
of the dimension restricted biochemical reactions, which have occurred
inner-cellular environment.
[1] Jeltsch A and Pingoud A. Biochemistry 1998 37: 2160-2169.
[3] Gowers DM and Halford SE. EMBO J. 2003 22: 1410-1418.
[4] Halford SE, and Marko JF. Nucleic Acids Res. 2004 32: 3040-3052.
P-P16 A
reductive approach to analyze stochasticity in intracellular networks
Tetsuya J. Kobayashi 1 and Kazuyuki Aihara 2
1 Aihara Laboratory, Institute of Industrial Science, University of
Tokyo, 4-6-1 KOMABA MEGURO-KU, TOKYO 153-8505, Japan,
Phone: +81/3/5452/6695, FAX: +81/3/5452/6695, e-mail: tetsuya@sat.t.u-tokyo.ac.jp
2 Aihara Complexity
Modelling Project, ERATO, JST., Shibuya-ku, Tokyo, 151-0064
The
stochasticity in behaviors of intracellular chemical reaction networks has been
attracting much attentions from both theoretical and experimental viewpoints.
However, roles of stochasticity in intracellular networks are not understood
enough. From the theoretical viewpoint, one of the important issues for
understanding stochasticity is how to evaluate fluctuation in molecular species
involved in intracellular networks. This issue is not yet solved mainly because
chemical reactions can work as the sources of fluctuation and the propagators
of fluctuation simultaneously and they are inseparable. In addition the
feedback structures in the intracellular networks makes it more difficult to
understand how the fluctuation of molecular species is determined by the
combination of the two roles of chemical reactions.
In this work, we develop a graph-based method named ``stochastic network analysis''
that were proposed by the authors to evaluate the fluctuation of intracellular
networks. The method is designed to facilitate us to elucidate how fluctuation
of a molecular species involved in an intracellular network is determined by
the reactions in the network and the structure of the network. The stochastic
network analysis is essentially composed of cumulant evolution equation and
stochastic network graph that is a graph representation of the cumulant
evolution equation. The stochastic network graph facilitates us to capture how
fluctuation is generated and propagates into an intracellular network. The
formulation of the cumulant evolution equation enables us to mathematically
characterize the generation and propagation of fluctuation, which is intuitively
understood with the aid of the stochastic network graph. In addition, by
applying the decomposition of sources of fluctuation and transformation of
stochastic network graphs, we can separate the roles of chemical reactions to
generate fluctuation and to propagate fluctuation. Furthermore, the theory of
signal flow graph and the contraction of stochastic
network graphs provide us with a mathematical representation of fluctuation of
a molecule in such a way that it is tightly associated to the structure of a
SNG. By using the stochastic network analysis, we analyze fundamental
mechanisms to attenuate fluctuation of molecular species, and demonstrate how
the stochastic network analysis facilitates our understanding of the
stochasticity in intracellular networks.
P-P17 Linlog
Modeling Approach: Theoretical Platform for System Biology
M.T.A. Penia Kresnowati, Wouter van Winden and Sef Heijnen
Department
Biotechnology, Delft Technical University, Julianalaan 67, Delft 2628 BC, The
Netherlands, Phone: +31-15-2785025, FAX: +31-15-2782355,
e-mail: m.t.a.p.kresnowati@tnw.tudelft.nl,
Web: http://www.bt.tudelft.nl/content/bpt/pdf/kresnowati.pdf
The
advances in experimental protocols for sampling and measuring various cell
components, i.e. intracellular and extracellular metabolites, amino acids and
proteins, genes and transcripts lead to the generation of enormous amounts of
data on biological systems. This raises the question of how to utilize these
data to extract information on the functioning of the biological system.
Modeling of the biological system offers a good tool to answer question. By
developing a good model of a biological system and validate it using available
data, we can enhance our understanding of what actually happens in the system
under condition where the data is collected. Once the model is validated, we
can use it to predict what is going to happen in our biological system under a
particular condition by performing in-silico simulations. The challenge is how
to develop a good model from available data.
In the last 10 years our group has been working on the development of an
experimental platform for metabolomics: short-term perturbation experiments,
sampling techniques, sample processing and LC-MS analysis of metabolite levels.
On the theoretical side, we have introduced the linlog kinetic model, a simple-standardized-approximative
way to model the kinetics of a biological system. The linlog kinetic model has
been shown to have a good approximation quality using relatively few model
parameters (1,2). My presentation will focus on the method development for
kinetic parameter estimation from transient metabolite data using the linlog
kinetic model (3). The method developed has been tested in a case study and has
been shown to be simple, robust towards error in the data originating from
sampling or measurement inaccuracy, and can accommodate various type of
perturbations, either in metabolite levels of in enzyme activities.
References: 1.Visser, D. et al. 2003.
Metabol.Eng. 5(3), 164-176
2.Wu, L. et al. 2004. Eur.J.Biochem. 271, 3348-3359
3.Kresnowati, M.T.A.P. et al. 2004. Metabol. Eng., submitted
P-P18 Knowledge
discovery by integrated analysis of metabolic and regulatory networks
Hong-Wu Ma and
An-Ping Zeng
Experimental
Bioinformatics, German Research Center for Biotechnology, Mascheroder Weg 1,
Braunschweig 38124, Germany, Phone: 49/531/6181460,
FAX: 49/531/6181751, e-mail: hwm@gbf.de,
Web: http://genome.gbf.de/bioinformatics/
Recent
studies on genome scale biological networks have
revealed several inspiring structural and functional properties of these
complex networks at systems level. However, relatively little has been done on
an integrated analysis of these different networks. We recently reconstructed
the transcriptional regulatory network (TRN) of E. coli which contains 1278 genes. Among
them, about 400 genes code for nearly 300 metabolic enzymes. We then mapped
these regulated enzymes on the reaction graph of the genome-based metabolic
network of E. coli. This allowed us to identify about 100 missing links
(unregulated reactions, most of their neighbors are regulated). These missing
links turned out to be promising candidates for discovering new regulatory
interactions. Using the arginine metabolism system as an example, two genes (argA
and argG) in the arginine synthesis pathway are predicted to be
regulated by the transcription factor ArgR because all the other genes in this
pathway are repressed by it. This prediction is further strengthened by the
finding that both argA and argG have the binding site for ArgR.
This result promotes us to develop a new strategy for reverse engineering of
TRN from metabolic phenotype, namely the reconstruction of a reasonable TRN to
make metabolic pathways work properly under different conditions. Through this
strategy, more regulatory interactions can be discovered for other functional
systems.
On the other hand, integrating regulatory information into the metabolic
network can improve the annotation of metabolic enzymes. For the arginine
metabolism system, three enzymes in the arginine uptake pathway are found to be
wrongly annotated in KEGG but rightly in Ecocyc. Furthermore, we found that
among the three pathways from glutamate to ornithine, only the genes in the
longest pathway are co-regulated with the genes in the pathway for converting
ornithine to arginine. This implies that in reality the longest pathway rather
than the shortest pathway are used for arginine synthesis. We further show that
the longest pathway is also the most energy consuming pathway. This challenges a basic
concept of flux balance analysis which predicts
metabolic fluxes often by assuming a maximal energy or biomass production as
the optimization objective. In summary, an integrated analysis of metabolic and
regulatory networks is important for understanding both cellular metabolism and
its regulation and essential for systems biology.
P-P19 Modelling
and simulation of dynamic signals in cells
Thomas Millat and
Olaf Wolkenhauer
Department of
Computer Science, University of Rostock, Albert-Einstein-Str. 21, Rostock MV
18184, Germany, EU, Phone: +49/381/498/3337, FAX: +49/381/498/3336,
e-mail: thomas.millat@uni-rostock.de,
Web: http://www.sbi.uni-rostock.de
The
transduction of external and internal signals from cell membrane and other cell organelles to the nucleus is a
dynamic process, enabling the cell to react and to adapt to changes in their environment. Feedback
loops inside the complex biochemical networks leads to a complex spectrum of
response signals, e.g. reversible and irreversible switches, stable and
instable oscillations.
With systems of nonlinear coupled ordinary differential equations we
investigate dependence of the dynamic response on parameters, missing or
additional feedback loops.
The role of fluctuations in biochemical signalling networks we study in the
framework of Monte-Carlo simulations techniques and compares the results with
corresponding ODE-modells.
P-P20 Systems
analysis of yeast glucose sensing system
Hisao Moriya and
Hiroaki Kitano
Keio Univ. Research
Park 9S3, The Systems Biology Institute, 35 Shinano-machi, Shinjyuku-ku, Tokyo
180-8582, Japan, Phone: +81/3/5363/3078, FAX: +81/3/5363/3079,
e-mail: hisaom@symbio.jst.go.jp,
Web: http://www.symbio.jst.go.jp/symbio2/
In the
budding yeast S.cerevisiae, expression
of glucose transporter genes (HXTs) are regulated
by glucose. Especially, the system that induces the expression of HXTs
is called “glucose induction system”. In this system, glucose sensors (Rgt2 and
Snf3) ultimately regulate the activity of a transcriptional repressor (Rgt1),
through mediator proteins (Mth1 and Std1) and ubiquitin ligase (SCFGrr 1).
We recently obtained evidence that Rgt2 glucose sensor activates casein kinase I (Yck1 and Yck2) on the plasma
membrane, then
the activated casein kinase I phosphorylates Mth1 and Std1. This
phosphorylation triggers the ubiquitination and degradation of Mth1 and Std1
(1). In this condition, Rgt1 is phosphorylated and lost its DNA binding activity (2). In addition to these,
genome-wide
microarray analysis revealed that the whole regulatory network structure of
this system including the auto-regulation (3). From these knowledge, we made a
mathematical model using biochemical process diagram editor software
CellDesigner (4) and simulator softwares, and studied about the possible
characteristics of the glucose sensing system.
References:
1. Moriya & Johnston, Proc. Natl.
Acad. Sci. USA, 2004
2. Kim JH. et al., Mol. Cell. BIol. 2003
3. Kaniak A. et al., Eukaryote. Cell, 2004
4. Funahashi A. et al., Biosilico, 2003
P-P21 Investigating
the structure of integrated biological networks
Venkata Gopalacharyulu Peddinti 1, Erno Lindfors 2 and
Matej Oresic 3
1 Quantitative Biology and Bioinformatics, VTT Biotechnology, Tietotie 2,
P.O-1500, Espoo 02044VTT, Finland, Phone: +358/9/456/4493,
FAX: +358/9/455/2103, e-mail: ext-gopal.peddinti@vtt.fi,
Web: http://sysbio.vtt.fi
2 E-mail: erno.lindfors@vtt.fi
3 Principal Investigator,
E-mail: matej.oresic@vtt.fi
Theory
of complex networks provides an intuitive setting for studying biological
relationships at the cellular level and beyond in the topological context [1].
It has already been suggested that the topological properties of networks
relate to biological function [2]. However, due to diversity of biological
relationship types even at the molecular level, studying the biological network
topology at one level only (e.g. metabolic networks) may miss much of important
information about cross-talk across multiple pathways and potential feedback
loops via regulatory networks. Our aim is to develop a framework to study
topology of multiple biological networks.
As a
start, we mapped KEGG, TransFac, TransPath, MINT, and BIND databases using XML
[3], with example shown in Figure1. We developed a Java-based tool that
allows parallel retrieval across multiple databases, incl. metabolic pathways,
protein-protein
interactions, signalling and regulatory networks. The results are then visually
displayed as a network (Figure2). Edge attributes contain information
about type of relationship, possibly quantitative or semantic information (such
as is located in in case of linking a protein with a complex entity such
as cell organelle).
Information
in integrated network form is a starting point for deeper topological and
functional mining. Presently, we are primarily interested in the concept of
similarity, i.e. how to define a distance metric across multiple networks. We
applied various distance metrics and nonlinear mappings into lower-dimensional
space such as Sammon’s mapping and self-organizing
maps, of which results will be presented.
We are
also extending our approach to more complex entities by combining existing
pathways with the automated text mining. We believe our approach will provide a
powerful framework for context-based mining and modelling of biological systems.
References
[1] Strogatz, S.H. (2001) Nature 410, 268 - 276.
[2] Jeong, H., Tombor, B., Albert, R., Oltvai, Z.N. and Barabási, A.-L. (2000)
Nature 407, 651 - 654.
[3] Gopalacharyulu, P.V., Lindfors, E., Bounsaythip, C., Wefelmeyer, W. and
Oresic, M. (2004) in: W3C Workshop on Semantic Web for Life Sciences,
Cambridge, MA.
P-P22 An in silico model for the optimization of
threonine production in Escherichia coli
Juan-Carlos Rodriguez 1, Jerome Maury 2, Christophe Chassagnole
2, Josep Centelles 1, Nic Lindley 2 and
Marta Cascante 1
1 Department of Biochemistry and Molecular Biology, University of
Barcelona, Marti i Franques 1, Barcelona 08028, Spain,
Phone: +34934021217, FAX: +34934021219, e-mail: augusto@bq.ub.es, Web: http://www.bq.ub.es/bioqint/recerca.html
2 Laboratoire de
Biotechnologie-Bioprocédés, UMR CNRS 5504, UMR INRA 792, Institut National des
Sciences Appliquées, 135 avenue de Rangueil, 31077 Toulouse cedex 04,FRANCE
Threonine
is an essential amino acid for birds and mammals and so there is considerable
interest in its economic industrial production for a variety of uses. Its
industrial production is insured by Escherichia
coli overproducer strain. We have
adapted and improved pre-existing dynamic model of central metabolism and
threonine pathway by Chassagnole et al.
(Biochem. J., 2001, 356,415-423 and Biotechnol. Bioeng., 2002, 79(1):53-73) to
create a global in silico model of
such a production strain.
The aim of this work was to obtain strain improvements in the threonine
production by DNA recombinant techniques directed
to the controlling enzymes in the threonine synthesis. A study of the
modification effect of the key branch points of the network (PEP and G6P node)
provides us enough information to predict changes on threonine production. The
model has been implemented with experimental results (enzyme activities, flux measurements and intracellular metabolite
concentrations) from a bioreactor cultivation.
The first part of the modeling procedure was the estimation of kinetic parameters (Michaelis and inhibition constants
and rates maximum for each enzyme) by comparison between simulations and
experiments. Then, we simulated the “in
vivo” behavior of threonine production for identification of metabolic
targets for genetic modifications. Finally, we checked the efficiency in
threonine production of the optimized strains.
P-P23 Kinetic
models of phosphorylation cycles: the role of protein-protein interactions
Carlos Salazar and
Thomas Höfer
Theoretical
Biophysics, Humboldt University, Invalidenstr. 42, Berlin 10115, Germany,
Phone: +49/30/2093 8694, FAX: +49/30/2093 8813, e-mail: carlos.salazar@rz.hu-berlin.de,
Web: http://www.biologie.hu-berlin.de/~theorybp/index.php?goto=cs
Activation-inactivation
cycles catalysed by kinases and phosphatases are a core component of cellular
signal transduction. Usually
phosphorylation reactions have been mathematically treated assuming
Michaelis-Menten kinetics and mostly disregarding the role of phosphatases. In
signaling networks, however, the substrates of many reactions are kinases and
phosphatases themselves, and the balance between activating and deactivating
enzymes constitutes a key aspect to be considered. Therefore, we have developed
a systematic and general analysis, which, in particular, resolves the kinetics
of kinases and phosphatases equally. The model accounts for the formation of
enzyme-substrate complexes and the subsequent modifications of a single or
several phosphorylation sites in the target protein.
Two main parameters shape the stimulus-response curves in the model: the degree
of saturation of kinase and phosphatase with the target protein and the
extent of inhibition exerted by binding of the respective reaction products to
the enzymes. By changing these parameters, the response curves were compared
with respect to the stimulus sensitivity measured through the response
coefficient. As result, well-structured “phase diagrams” of the phosphorylation
cycles were obtained. There are two regions of high sensitivity corresponding
to the parameter ranges of zero-order ultrasensitivity, on the one hand, and
strong product inhibition with low enzyme saturation, on the other. Moreover,
higher stimulus sensitivities are particularly observed in the phase diagram of
proteins modified at several residues. In these proteins, a bistable response
can arise in the region of zero-order ultrasensitivity when the phosphorylation
sites are cooperatively modified.
We also computed in such phase diagrams the response time to reach a
steady-state phosphorylation level of the substrate after a stimulus change.
The result demonstrates that this transition time also strongly depends on the
kinetic design of the phosphorylation cycle. The most
influential parameter is the substrate saturation of the enzymes: saturation
cycles are generally slower than unsaturated ones. Finally we addressed the
effect of the order of phosphate processing on multiply phosphorylated
proteins. Response curves in a random phosphorylation are shallower but with a
faster kinetics compared to the sequential case.
Salazar C. and Höfer T. (2003), J. Mol. Biol. 327, 31. Salazar C. and Höfer T. (2004), FEBS Lett., in press
P-P24 Modelling
transient dynamics of osmo-stress response in Yeast
Jörg Schaber 1,
Bodil Nordlander 2 and
Edda Klipp 1
1 Vertebrate Genomics, Max Planck Institute for Molecular Genetics,
Ihnestr. 63, Berlin 14196, Germany, Phone: +34 30 804093 19, FAX: +34
30 804093 21, e-mail: schaber@molgen.mpg.de,
Web: www.molgen.mpg.de/~schaber
2 Göteborg University
A
special focus of System Biology has been the modelling of cellular signalling
processes. Especially the MAP kinase signalling cascade that is
widely conserved among eukaryotic cells has been studied intensively in theory
and practice. Theoretical studies of MAP kinase cascades often concentrate on
their steady state properties, like, e.g. ultrasensitivity or oscillations. It has been noted that for MAP kinases to exhibit ultrasensitivity a
distributive double phosphorylation is one possible mechanism. For few systems
it was proved that this is actually the case. In many systems, e.g. in yeast, MAP
kinase cascades are neither the only components of a signalling cascade nor are
they isolated from other cellular compounds. It is rather the case the MAP cascades
are usually preceded by other signalling constructs, e.g. G-protein complexes or phospho-relay
systems and are often recruited to scaffold complexes where a distributive
phosphorylation has not been demonstrated, yet. Thus, on the one hand it is
possible that ultrasensitivity is achieved by other parts of the signalling
cascade than the MAP cascade and one the other hand steady state properties
might not be so important when it is the transient dynamics that are decisive
as, e.g., in stress response. From
the practical point of view it is desirable to have models with as few
reactions as possible because data to fit parameters is often scarce and models
are usually underdetermined. Addressing these issues we will present a modelling
study of the Sln1 branch of the HOG pathway in Yeast that consists of a
three-compound phospho-relay system and a three-step MAP kinase cascade. We fit
several possible model structures of the Sln1 branch to measured data of
transient Hog1 activation upon several osmotic shocks. We explore the influence
of possible model structures on the resulting fit with special emphasis on a)
the effect of model structure, e.g. processive versus distributive
phosphorylation, on transient dynamics, b) the effect of scaffolding and c)
steady state properties. Moreover, we examine the performance of several global
optimisation algorithms to fit the measurements to the model parameters.
P-P25 Inferring
regulatory networks from experimental data
Christian Spieth, Felix Streichert, Nora Speer and
Andreas Zell
Centre for
Bioinformatics, University of Tübingen, Sand 1, Tübingen D-72076, Germany,
Phone: +49/7071/29/78987, FAX: +49/7071/29/5091, e-mail: spieth@informatik.uni-tuebingen.de
Standard
methods for the analysis of microarray data are often emphasizing on the
identification of single genes within the process of interest only and thus
neglecting important information like the time dependencies hidden in the data
sets. From a systems biology point of view it is therefore necessary to develop
a new class of analytical methods. One major aspect that has to be addressed by
these new methods is to understand the regulatory mechanisms within a cell, i.e. the structure of regulatory
networks and the corresponding kinetic parameters. Those methods must be able to
cope with the high complexity of the regulating system and the ambiguity in the
data. Additionally, they have to comply with biological constraints like the
stoichiometry of biochemical reactions or known interactions of system
components.
Over the last year, we developed a software framework that aims to infer gene networks from microarray data and also
metabolic systems from experimental data. The framework comprises known models
for simulating regulatory networks like linear weight matrices, non-linear
S-systems and arbitrary differential equations as well as models that were
developed in our research group like pseudo-linear weight matrices. For the
estimation of the model parameters, we developed new algorithms primarily based
on evolutionary algorithms (EAs) including evolutionary strategies, genetic
algorithms and genetic programming or on direct heuristics. Our developments
include massively parallel EA implementations, hybrid optimization approaches
combining EAs with local parameter searches and also multi-objective
exploration of alternative network structures. Additionally, these algorithms
automatically incorporate biological knowledge from public databases like KEGG
to meet certain constraints to include a priori information of network
topologies. Further on, we are experimenting with extended solution
representations to increase the performance of the simultaneous optimization of
parameters and topologies. The framework was successfully used on the inference
problem of gene and metabolic networks and the project’s website (http://www.jcell.org)
gives a comprehensive list of recent publications.
Acknowledgement :The project is supported by the National Genome Research
Network (NGFN) of the BMBF in Germany under contract no
0313323.
Spieth, C., et al., (2004), LNCS, 3102, 461
Spieth, C., et al., (2004), CEC 2004, 152
Spieth, C., et al., (2003), LNCS, 3005, 102
P-P26 First
steps towards a multi-dimensional iron regulatory network
Yevhen Vainshtein 1, Martina Muckenthaler 2,
Alvis Brazma 3 and Matthias W. Hentze 1
1 Gene Expression, EMBL, Meyerhofstr. 1, Heidelberg 69117, Germany,
Phone: +49/6221/387-8139, FAX: +49/6221/387-8306, e-mail: Yevhen.Vainshtein@embl.de
2 Department of Pediatric
Oncology, Hematology and Immunology University of Heidelberg, Germany; European
Molecular Biology Laboratory, Heidelberg, Germany
3 European Molecular Biology
Laboratory, European Bioinformatics Institute, Hinxton, Great Britain
Iron
homeostasis is central to many biological processes and thus imbalances in
mammalian iron metabolism are associated with frequent iron overload and iron
deficiency disorders. The iron regulatory network is a complex
multi-dimensional network combining organ to organ communication (mediated by
iron hormones), cellular interactions and biochemical pathways. This network
can be addressed by several endogenous (e.g.
cell cycle) and exogenous (e.g.
infection) stimuli.
In order to define and ultimately visualize a complete iron regulatory network
we produce and collect specific signature profiles derived from cell based and
whole animal experiments and correlate them with biochemical, phenotype and
physiological information. Gene expression profiles recorded on IronChips form the basis for our in silico iron regulatory network. IronChips
are cDNA based microarrays that sensitively and accurately measure expression
changes of genes involved in iron uptake, storage and recycling, as well as
from genes involved in a number of interlinked pathways (e.g. NO metabolism,
redox pathways, stress responses as well as acute phase and immunity).
Systematic IronChip analysis of
selected tissues derived from knock-out mice with defects in central players of
iron metabolism inform us about the relationship of these genes within the
network.
IronChip data evaluation and
construction of such a multi-dimensional iron regulatory network requires the
development of sophisticated algorithms. The recently developed IronChip
Analysis Tools (ICAT) offers a solution for the automated data analysis
associated with this specialized array platform.
P-P27 Nutrient
starvation in baker’s yeast,
and the implication of protein degradation for Vertical Genomics
Karen van Eunen, Jildau Bouwman, Sergio Rossell, Rob
J.M. Spanning, Barbara M. Bakker and Hans V. Westerhoff
Molecular Cell
Physiology, Biocentrum Amsterdam, Vrije Universiteit Amsterdam, De Boelelaan
1085, Amsterdam NL-1081 HV, The Netherlands, EU, Phone: +31(0)204446966,
FAX: +31(0)204447229, e-mail: karen.van.eunen@falw.vu.nl
In the Vertical Genomics program
gene expression, protein synthesis, protein degradation and posttranslational modification
will be related to metabolic fluxes and metabolite concentrations. This will be
done by quantitative dynamic measurements, regulation analysis and kinetic modeling. Glycolysis in Saccharomyces cerevisiae is a good model system to test these
relations, because it is one of the few pathways for which kinetic properties
of the enzymes are known sufficiently to calculate the flux from the enzyme activities. Furthermore yeast can be brought under
well-defined steady-state and transient conditions for a detailed quantitative
analysis.
In this project different perturbations will be studied and nutrient starvation
is one of them. During industrial production of beer and baker’s yeast the
cells often undergo periods of nutrient starvation. Starvation leads to a loss
in fermentative capacity, due to a degradation of glycolytic enzymes,
modification of the palette of expressed transporters, and due to altered
topogenesis in some of the transporters. Fermentative capacity is an important
characteristic for the application of baker’s yeast in the dough. During
washing, packaging and storage, the baker’s yeast is subjected to complete
starvation, while in the dough the cell is subjected to a nutrient-rich
environment and has to start fermentation. Knowledge about the regulations at
the various levels of the cellular regulation hierarchy should guide
engineering towards improved fermentation capacity after a period of
starvation. Therefore the dynamics of glycolytic genes and proteins during
starvation will be measured and correlated to the response of the glycolytic
flux and metabolite concentrations. Within the Vertical Genomics program the
starvation project will be a test case to quantify the importance of protein
degradation in the regulation of fluxes.
P-PoP1 New
parameter estimation method with possible application in systems biology
Ioan Grosu
Bioengineering/Exact
Sciences, University of Medicine and Pharmacy "Gr.T.Popa", Str. Universitatii Nr. 16, Iasi 700 115,
Romania, Phone: +40 232 211 810, FAX: +40 232 211 820, e-mail: igrosu@umfiasi.ro, Web: http://www.umfiasi.ro
Parameter
estimation is a research topic of large interest : from science and engineering
to biology and medicine. One of the major problems is that the error function
has multiple minima. Using our previous results on synchronization of chaotic
systems we obtained promissing results[1] : a global minimum for the error
function. We want to apply these general ideas to the specific problems of
systems biology[2,3] with the focussed goal of improving the current
software[2,3].
[1].Grosu,I.,(2004),International Journal of Bifurcation and Chaos 14(6),2133
[2] Moles C.G.,et al (2003) ,Genome Research,p.2467, Oct.2004,http://www.genome.org/cgi/doi/10.1101/gr.126503
[3] Kremling A,et al ,(2004),GenomeResearch,p.1773
http://www.genome.org/cgi/doi/10.1101/gr.1226004
P-PoP2 Effects
of noise in metabolic flux analysis
Visakan Kadirkamanathan, Steve Billings, Sarawan Wongsa,
Jing Yang and
Philip Wright
Department of
Automatic Control & Systems Engineering, The University of Sheffield,
Mappin Street, Sheffield S1 3JD, United Kingdom, EU, Phone: +44 114
2225618, FAX: +44 114 2225661, e-mail: visakan@sheffield.ac.uk, Web: http://www.shef.ac.uk/acse/people/v.kadirkamanathan/
The
quantification of metabolic fluxes are paramount to identify the cause-effect
relationship between genetic modifications and resulting changes in metabolic
activities. Given that intracellular fluxes are nonmeasurable quantities,
careful experimental methods and associated estimation methods are needed to
determine them. In metabolic flux balance, intracellular fluxes are calculated
from measured extracellular fluxes and stoichiometric equations, coupled with
13C labelling experiment that provides additional information about the
intracellular fluxes. The measurements are typically noisy and here, the aim of
the study is to analyse the effects of noise in the estimation accuracy and to develop
methods that compensate for these directly.
The flux estimation methods investigated belong to three different types. In
the seminal work on metabolic flux analysis, the flux estimation problem was
posed as a classical least squares problem. This and other variants of least
squares methods form the first of these types. Second class of methods are
based on computational intelligence methods such as evolutionary algorithms.
The choice of evolutionary algorithms is motivated by the need for nonlinear
optimisation methods that are required in the presence of noise. Finally, the
third class of algorithms are based on stochastic approximation type methods.
Results will be presented for a small metabolic network consisting of the Cycle
Pentose Phosphate pathway demonstrating the advantages of methods that attempt
to compensate for noise directly.
P-PoP3 A new dynamic complexity reduction
method for biochemical reaction networks
Dirk Lebiedz 1, Jürgen Zobeley 2,
Julia Kammerer 1 and Ursula Kummer 2
1 ,
Interdisciplinary Center for Scientific Computing (IWR), Im
Neuenheimer Feld 368, Heidelberg D-69120, Germany, Phone: +49/6221/548250,
FAX: +49/6221/548884, e-mail: lebiedz@iwr.uni-heidelberg.de,
Web: http://reaflow.iwr.uni-heidelberg.de/~Dirk.Lebiedz
2 EML Heidelberg
Modeling
and simulation of the dynamical behavior in cellular systems are often hindered
by size and complexity of the underlying biochemical reaction networks. With
the aim to facilitate the identification of dynamical key features of such
systems and to aid a functional separation into subsystems we present a new
dynamical complexity reduction method based on the concept of a time scale
decomposition (TSD). The TSD-based method relies on the fact that biochemical
processes typically take place on time scales differing by up to several orders
of magnitude. In contrast to most existing complexity reduction methods -
either based on the evaluation of structural network properties only or on
specific dynamical assumptions (e.g. steady state) - our approach is
independent of any specification for the dynamical behavior. Processes being
sufficiently fast compared to the actual time scale of interest - which
definitely changes in the course of the dynamical simulation - are assumed to
be relaxed. Thus, the effective number of ordinary differential equations which
need to be solved in order to obtain the time evolution of the reaction network
in the context of a deterministic, homogeneous modeling formalism is reduced in
an automated way. The reduced system dynamics at a given state is confined to a
so called Intrinsic Low-dimensional manifold (ILDM) of the full state space. An analysis
of the reaction species participating in the active processes provides valuable
insight into the nature of the interactions that are responsible for the system
dynamics on a specific time scale and may allow the identification of
functional couplings within the network. The capabilities of the new dynamic
complexity reduction method are illustrated in a study of the complex dynamics
of a Peroxidase-Oxidase (PO) reaction network model. As a result, we managed to
reduce the dimension of the active state space substantially in the course of
the simulation, even in the difficult case of complex oscillatory system
dynamics. For the first time a potential decomposition of the PO network into
subsystems is shown to depend sensitively on the specific dynamics of the
system.
Reference:
J. Zobeley, D. Lebiedz, J. Kammerer, A. Ishmurzin, U. Kummer, A new
time-dependent complexity reduction method for biochemical systems, submitted
to Transactions on Computational Systems Biology (2004)
P-PoP4 Determination
of in vivo non-steady-state fluxes and kinetic information using stable isotope labeling and metabolite pool size
data: theory and application
Junli Liu 1, Alisdair R. Fernie 2 and
David F. Marshall 1
1 Computational Biology Programme, Scottish Crop Research Institute,
Invergowrie, Dundee DD2 5DA, UK, Phone: 01382 568500, FAX: 01382
562426, e-mail: jliu@scri.sari.ac.uk
2 Department of Lothar
Willmitzer, Max-Planck-Institut für Molekulare Pflanzenphysiologie, Am
Mühlenberg 1, 14476 Golm, Germany.
Incubation of plant material in
labeled isotopes often reaches neither isotopic nor metabolic steady state
within the constrained time span of experimental measurements. This can be
problematic since at a non-steady state, flux balance cannot be established. For this reason here we develop a
mathematical approach to calculate fluxes within systems that do not
approximate steady state. The approach is based on mass balance of all forms of molecules and employs trapezoid Euler’s
numerical method to deal with differential equations. The general method for
deriving non-steady-state fluxes is established, and construction of
overdetermined systems for non-steady states is proposed. Determination of
fluxes based on GC-MS analysis is analysed in detail. We show that
non-steady-state fluxes can be derived based on time-dependent metabolite pool
size and specific labeling data. Moreover, when a number of data points are
available, the approach is able to predict the dependence of flux on pool size
of substrate and as such to reveal which enzymes do not follow conventional
Michaelis-Menten type kinetics. Therefore, regulatory enzymes can be identified
for further study. In addition, kinetic parameters can be estimated on the basis of this approach for
Michaelis-Menten type kinetics. As a first example of the application of this
approach we apply it to the analysis of primary metabolism in tuber discs
isolated from wild type potato plants. We demonstrate that this approach is
able to identify irrational flux combinations that violate mass balance, and as
such allows a structural analysis of the metabolic network under evaluation.
Despite the challenges presented by these findings certain fluxes including the
interchange between glucose-6-P and fructose-6-P could be
readily calculated. We discuss the potential of the approach both in analysing
fluxes of metabolic networks and more generally within kinetic modeling
strategies.
P-PoP5 An
adaptive system approach for the modelling of genetic regulatory networks.
Glucose metabolism study in the yeast.
Sinisa Zampera and Todor Vujasinovic
, Helios
BioSciences SARL, 8 rue Général Sarrail, Créteil 94010, France, EU,
Phone: +33/1/49 81 37 92, FAX: +33/1/48 98 59 27, e-mail: todor.vujasinovic@heliosbiosciences.com
We have used a dynamic neural
network to model the yeast glucose metabolism response to glucose deprivation in the culture medium.
Our aim was to produce a predictive rather than explicative model, in order to
address the question: “which molecule of the network should we act upon to
obtain a given biological response?” The network was built from literature
analysis and KEGG data and includes 133 molecules (3 metabolites, 99 enzymes,
26 transcription factors, 5 signal transduction proteins, connected through 516 interactions). The
model was trained by DNA microarray data describing the gene expression response to the fermentation to respiration switch (De
Risi et al., Science (1997)278:680-6). The simulation provides a hierarchy of
the molecules classified in terms of relative distance to the biological
response to be obtained. The model has been applied to the prediction of a gene
knock-out response and the detection of the invalidated gene was within
acceptable error margins. We will present our model and results and more
specifically discuss the redundancy of biological regulatory mechanisms as
arguing towards the use of adaptive models, and the impact of the network
heterogeneity (scale-free structure) on the learning procedure and inferred
parameters.
T-S01 Oscillatory
mechanisms derived from phase and amplitude information
Sune Danø 1, Mads Madsen 2 and Preben
G. Sørensen 1
1 Department of Medical Biochemistry and Genetics, University of
Copenhagen, Blegdamsvej 3b, Copenhagen N DK-2200, Denmark, Phone: +45 35
32 77 51, FAX: +45 35 35 63 10, e-mail: sdd@kiku.dk
2 Department of Chemistry,
University of Copenhagen, Universitetsparken 5, DK-2100 Copenhagen Ø, Denmark
Due to
time-scale separation, a dynamical system close to a bifurcation will evolve
according to the universal dynamics of that particular bifurcation. We have
exploited this fact to devise a novel approach for determining the oscillatory
mechanism for systems close to a supercritical Hopf bifurcation. In essence,
the method works by identifying the chemical components of the two dynamical modes
associated with the oscillatory dynamics: an activating mode and an inhibitory
mode. There is no need for prior knowledge of the network structure, the only
information required is measurements of the relative phases and amplitudes of
the oscillating substances. Hence, metabolomics and mRNA arrays are ideal
sources of data. The feasibility of the method is illustrated by its use for
analysis of glycolytic oscillations in yeast cells.
T-S02 Application
of modelling and simulation to drug discovery: The ErbB System
Bart Hendriks, Gareth Griffiths,
Jack Beusmans, Adrienne James, Julie Cook, Jonathan Swinton and
David De Graaf
Computational
Biology, Pathways, AstraZeneca, Mereside, Alderley Park, Macclesfield,
CHESHIRE SK10 4TG, ENGLAND, Phone: +44-1625-519391,
FAX: +44-1625-514463, e-mail: adrienne.james@astrazeneca.com
The
implication of the ErbB family in the pathogenesis of various cancers has made
it a popular target for the development of targeted anti-cancer therapies. ErbB dimerisation, trafficking,
and activation are complex processes, making it difficult to intuit how
perturbations, such as drug intervention, will affect the system
dynamics. We need computational approaches to keep track of and to quantify
this complexity. AstraZeneca, in collaboration with the Lauffenburger lab at Massachusetts Institute
of Technology, have developed a computational model implementing commonly
accepted principles involved in ErbB signal transduction. The current ErbB
model is made up of Ordinary Differential Equations (ODEs) and is based on detailed
mechanisms of ErbB receptor interactions and downstream signalling components.
It contains ~300 species, ~400 parameters and more than 500 reactions. A major
challenge in dealing with models of this size is information management and
model visualization. Text mining software is used to capture kinetic constants and models are displayed
graphically using TeraNodeTM Design Suite. Parameter estimation and
sensitivity analysis are being exploited to assist model validation. The model
is being used to predict the dynamics of receptor phosphorylation in the
context of different cell lines and ligand environments. Recent work in our
group has demonstrated that a deficiency in internalisation is sufficient to
explain the observed signalling phenotype of the Gefitinib-responsive mutants
found in NSCLC. Gefitinib ('Iressa'; ZD1839, AstraZeneca, Wilmington, DE) is an
ATP-competitive
small molecule inhibitor of ErbB1, approved for use in
the treatment of non-small cell lung cancer (NSCLC). About 80% of
Gefitinib-responsive tumours in NSCLC carry mutations in ErbB1. This model
prediction has been experimentally validated using a Gefitinib-responsive and
non-responsive NSCLC cell line. The Gefitinib-responsive cell line is shown to
be deficient in the internalisation of two ErbB1 ligands, EGF and TGFa. This
work provides a mechanistic basis for the link between the role of ErbB1 in
oncogenesis and Gefitinib response through decreased internalisation of ErbB1
and increased signalling to AKT.
T-S03 Combined
optimization technique for biological data fitting
Konstantin N. Kozlov 1, Alexander M. Samsonov 2 and
John Reinitz 3
1 Department of Computational Biology, St. Petersburg State Polytechnical
University, Polytechnicheskaya st., 29, St. Petersburg 195251, Russia,
Phone: +7/812/5962831, FAX: +7/812/5962831, e-mail: kozlov@spbcas.ru
2 The Ioffe Institute of the
Russian Academy of Sciences, St.Petersburg, 194021 Russia
3 Dept. of Applied Math and
Statistics, The University at Stony Brook, Stony Brook NY 11794-3600
Motivation.Development of the organisms from
embryo to the adult is one of the central unsolved problems of biology. We are
working on characterization problem of systems biology of development in
context of the segment determination gene network of a Drosophila embryo. While gene expression is
evaluated at a time resolution of a few minutes and a spatial resolution of one cell (see FlyEx database), the regulatory parameters
cannot be determined experimentally, and are to be found as the solution of the
inverse problem by minimizing the deviation of the model output from the data.
We apply a chemical kinetic model describing the dynamics of the
expression patterns of the segmentation genes during the blastoderm stage by
means of the system of highly non-linear reaction-diffusion equations (Jaeger, J, et al., (2004), Nature, 430, 368).A random search
technique, being extremely computationally intensive, is sometimes the only
choice for finding the set of parameters that provides the best fit of model to
data. Therefore the main problem is to reduce the complexity of finding the
parameters of mathematical models.
Results. We developed the Combined
Optimization Technique (COT) to reduce the computational cost of solution
of the inverse problem of modelling. COT combines advantages of random search
and gradient descent. Starting from an arbitrary initial set of parameters, a
rough approximation of a minimum is found by the random search, namely, Simulated
Annealing (SA), while the final solution is given by Optimal Steepest
Descent Algorithm (OSDA), developed earlier (Kozlov, K, et al., (2003), Techn. Physics, 48, 6), and successfully applied as the local optimizer in(Gursky, V, et al., (2004), Phys. D, 197, 286). The dependence of COT convergence of the initial approximation and
quality criterion is investigated and the strategy of transition from SA to
OSDA is studied here. COT demonstrated high accuracy in reconstruction of model
parameters and the 30% total performance benefit in a two-gene network. Further study is performed currently
to increase the speed up by application of new automated tuning methods for the
OSDA part of COT. Acknowledgments: The support of the study by the NIH
Grants RR07801, TW01147, the CRDF GAP Awards RBO685, RBO1286 is gratefully
acknowledged.
T-S04 Systematic
identification and characterisation of synthetic lethal interactions in the
metabolic network of yeast
Balázs Papp 1, Richard Harrison 1,
Daniela Delneri 1, Csaba Pál 2 and
Stephen Oliver 1
1 Faculty of Life Sciences, Michael Smith Building, The University of
Manchester, Oxford Road, Manchester M13 9PT, United Kingdom, Phone: + 44
161 275 1565, FAX: + 44 161 275 5082, e-mail: pappb@ramet.elte.hu, Web: http://ramet.elte.hu/~pappb 2 Theoretical Biology and Ecology
Modelling Group, Hungarian Academy of Sciences and Eötvös Loránd University,Pázmány
Péter Sétány 1/C, H-1117 Budapest, Hungary
To what
extent and why do the effects of mutations depend on the genetic background? Do
deleterious mutations act synergistically? What is the mechanistic basis of
genetic interactions and how does it depend on the environment? Answers to
these questions are relevant not only to functional genomics, but also to
problems such as the evolution of sexual reproduction and how deleterious
mutations are eliminated from the population.
Owing to the huge number of potential gene combinations, progress in answering these
questions is, however, limited by the lack of efficient genome-scale
experimental mapping of genetic interactions. To overcome this difficulty, we
propose a combination of in silico and in vivo studies to screen
for synthetic lethal relationships in the yeast metabolic network.
First, we apply flux balance analysis (FBA) to the
genome-scale metabolic model of S. cerevisiae (Forster et al. 2003) to
search for candidate gene pairs showing synthetic lethal interactions. Next, we
use laboratory experiments to validate the model’s predictions. Our preliminary
results suggest that i) FBA is able to predict synthetic lethal interactions,
ii) many of the interactions are environment specific, iii) although the
density of interactions do not differ significantly between nutrient poor and
nutrient rich growth conditions, we observe twice as many genes participating
in synthetic lethal interactions in nutrient poor environment and iv) only
about 20% of synthetic lethal gene pairs can be explained by the presence of
gene duplicates (isoenzymes), this fraction, however, is significantly higher
than the 2% previously reported for non-metabolic genes (Tong et al. 2004). The
implications of these findings for genetic robustness and phenotypic plasticity
are also discussed. References: Forster, J, et al. (2003) Genome Res 13, 244; Tong, AH, et al. (2004)
Science 294, 2364.
T-P01 Genome-scale
analysis of Streptomyces coelicolor A3(2) metabolism
Irina Borodina 1, Preben Krabben 2 and
Jens Nielsen 1
1 Center for Microbial Biotechnology, Denmark Technical University,
Søltofts Plads, Kgs. Lyngby DK-2800, Denmark, EU, Phone: +45/45252659,
FAX: +45/45884148, e-mail: ib@biocentrum.dtu.dk
2 Department of Biochemical
Engineering, University College London, Torrington Place, London, UK
Streptomyces coelicolor A3(2) is by far the best genetically studied Streptomyces strain
and has become a model organism for Streptomyces species. Release of the
S. coelicolor A3(2) genome sequence has further expanded the knowledge
of this organism and enabled the application of genome-wide analysis techniques
like DNA arrays for transcriptome analysis and
proteomics studies. We believe that integration of the genome-wide data through
the use of mathematical models will enhance the extraction of information about
the molecular mechanisms governing different processes in S. coelicolor
A3(2). In this context it has recently been shown particularly valuable to use
metabolic genome-scale models.
The metabolic network of the Streptomyces model organism - S.
coelicolor A3(2) was reconstructed at the genome-level. The reconstruction
was based on annotated genes, physiological and biochemical information. The
network includes 823 biochemical conversions and 151 transport reactions, accounting for a total of 974
reactions. 700 of the reactions in the network are unique whereas the remaining
reactions are iso-reactions. The number of metabolites in the network is 500.
Seven hundred sixty nine (769) open reading frames (ORFs) were included in the
model, which corresponds to 14% of the ORFs with assigned function in the S.
coelicolor A3(2) genome. Flux balance analysis was used for studies of the
reconstructed metabolic network and for assessing its metabolic capabilities
for growth and polyketides production. The reactions essentiality was studied
for growth on 63 carbon sources and 2 nitrogen sources; hereby the core of the
"real" essential genes was identified. Furthermore, we illustrated
how reconstruction of a metabolic network at the genome level could be used to
fill gaps in genome annotation.
The ongoing project concentrates on generation of large-scale data (gene expression, proteomics and metabolomics) in S.
coelicolor A3(2) and its regulatory mutant and on the integrated analysis
of these datasets in combination with metabolic model.
T-P02 Relational
learning of biological networks
Cyril Combe 1, Florence d'Alché-Buc 2,
Vincent Schachter 3 and Stan Matwin 4
1 Genoscope - LaMI UMR 8042, Genopole, 2 rue Gaston Crémieux, Evry 91000,
France, Phone: +33/6/60/84/9064, FAX: +33/1/60/87/2514, e-mail: cyril.combe@gmail.com 2 Maison Genopole des Sciences de la
Complexité - 93 rue Henri Rochefort, Evry 91000, France 3 Genoscope, 2 rue Gaston Crémieux,
Evry 91000, France 4 SITE -
University of Ottawa, 800 King Edward Avenue, Ottawa, Ontario, K1N 6N5, Canada
The last
few years have seen a lot of different approaches for the reconstruction of
biological networks. Those approaches essentially differ about the kind of data
used and about the models of biological networks considered. The data used can
be either numerical like time series generated by micro array experiments, or
symbolic, like annotation databases or ontologies extracted from scientific
articles by text mining algorithms. The models considered usually have graph
structures and involve different kind of objects, like genes, proteins,
metabolites or reactions. In order to deal with heterogeneous data as for
representing highly relational models, it appears appropriate to infer
biological networks with relational learning techniques.
As a proof of concept, we used the Inductive Logic Programming [1] system
Progol to learn the concept of gene regulation based on gene
expression data. The approach is the following:
-We consider a known gene network with associated expression data and we
represent both in first order logic
-We learn a first order logic definition of gene regulation with Progol
-We try to discover potential regulations with the definition of regulation
outputted by Progol
In order to represent expression data in a compact way, we discretized it in
terms of expression levels, of variation directions and of time. The time discretization
uses the notion of time intervals. We empowered the system with predicates able
to capture relations between intervals, inspired by a formalism introduced by
Allen in [2], which constitutes a new approach to deal with time series in ILP.
The system has been successfully tested on artificial datasets generated by
different kinds of dynamic systems. It has also been tested on real datasets,
one related to the SOS DNA Repair network of E. Coli and
one related to the cell cycle of the Yeast. Current work
follows three complementary directions:
-We try to use other sources of information like the Gene Ontology database or
metabolic datasets
-We are working on new models combining gene regulatory networks and metabolic
pathways
-We are currently investigating methods combining first order logic and
probabilities (see [3])
[1] Muggleton, S., et al., Inductive
Logic Programming: Theory and Methods (1994), JLP, 19-20, 629
[2] Allen, J., et al., Actions and Events in Interval Temporal Logic (1994),
Rochester TechReport, 521, 1
[3] De Raedt, L., et al., Probabilistic Inductive Logic Programming (2004),
ALT, 15, 19
T-P03 A
new Information System to manage and analyse information on biochemical
interactions
Holger Dach 1, Juliane Fluck 1,
Kai Kumpf 1 and Rainer Manthey 2
1 Department of Bioinformatics, Fraunhofer Institute for Algorithms and
Scientific Computing, Schloss Birlinghoven, Sankt Augustin 53754, Germany, EU,
Phone: +49-(0)2241-142549, FAX: +49-(0)2241-142656, e-mail: Holger.Dach@scai.fraunhofer.de,
Web: http://www.scai.fraunhofer.de/bio.0.html?&L=1
2 Institute for Informatics
III at the University of Bonn
We are
developing a relational database system for representing biochemical objects
and their functions.
A major challenge of molecular biology is to understand how biochemical objects
interact. There is a huge amount of publications, published every month, which
provide new data on those interactions. We are able to automatically extract
these information but the crucial question is:
Based on the bulk of existing data, what is the information gain we can draw
from textmining results?
In order to approach this problem, we are developing a system for representing
and analysing biochemical interactions.
In a second step we will try to draw useful conclusions by applying reasoning
methods, based on classification co-occurences.
T-P04 Reduced
order modeling of global regulation - redox regulation in Escherichia coli
Michael Ederer 1, Thomas Sauter 1 and Ernst Dieter Gilles 2
1 Institute for System Dynamics and Control Engineering, University of
Stuttgart, Pfaffenwaldring 9, Stuttgart D-70569, Germany, EU,
Phone: +49/711/685/6296, FAX: +49/711/685/6371, e-mail: ederer@isr.uni-stuttgart.de, Web: http://www.isr.uni-stuttgart.de
2 Max-Planck-Institute for
Dynamics of Complex Technical Systems, Sandtorstr. 1, D-39106 Magdeburg,
Germany, EU
The
amount of measurement data necessary to identify and validate detailed
mathematical models of global regulatory systems makes it difficult to build
such models. Nevertheless, modeling global regulation is helpful to understand
the role of the involved regulatory proteins.
Therefore, there is a need for a modeling methodology capturing the key aspects
of global regulation, without requiring the knowledge of the kinetic properties of all involved compounds.
As a case study we consider the global redox regulation of E. coli. With decreasing oxygen availability E. coli adapts its
metabolism in four distinct phases from pure respiration to pure mixed acid
fermentation. When operating in one phase the redox state as reflected by the
concentration of NADH is kept constant, even if oxygenation changes (Alexeeva
S. et al., (2003) J Bacteriol, 185, p204).
This indicates that despite the multitude of the actual signals of involved
transcription factors (e.g. Fnr: O2, ArcAB: quinones) the underlying redox
state of the cell is the relevant overall input signal for a controller that is able to
stabilize the redox state at different setpoints. This simplified control structure is the basis for a reduced order
model. The model is divided in a model of the controlled system, i.e.
metabolism, and the controller, i.e. regulation.
The metabolism is approximated by a simple phenomenological model where all
compounds except NADH and ATP are assumed to be quasistationary. The inputs
to the model of metabolism are the concentrations of key enzymes (control
inputs) and the environmental conditions (disturbances).
The regulation model uses the NADH and ATP concentrations (controller inputs)
in order to compute the activities of the transcription factors via sigmoid
characteristic curves. The transcription factors influence gene expression (controller output).
The resulting ODE model of metabolism and regulation contains only a relatively
small number of dynamical state variables (ATP, NADH & enzyme
concentrations) and kinetic parameters.
Simulation studies show good qualitative agreement with experimental data. For
further validation of the model mutant strains have to be considered. An
extension of the model for the study of other global regulatory systems will
clarify the generalizability of this approach.
The introduced approach concentrates on the modelling of principal dynamic
variables. It may be appropriate for the modelling of large regulated metabolic
networks.
T-P05 Technical
variance, quality control and scaling: necessary steps towards meta-analyses on large
expression databases
Martin Eisenacher 1, Harald Funke 2,
Thomas Vogl 3, Christoph Cichon 4,
Kristina Riehemann 1, Clemens Sorg 1 and
Wolfgang Koepcke 5
1 Integrated Functional Genomics (IFG) / IZKF Muenster,
Westfaelische-Wilhelms-Universitaet Muenster, von-Esmarch-Str. 56, Muenster
D-48149, Germany, Phone: +49 / 251 / 83-52207, FAX: +49 / 251 /
83-55651, e-mail: eisenach@uni-muenster.de,
Web: http://ifg-izkf.uni-muenster.de
2 Institute of Vascular
Medicine, Hospital of Friedrich-Schiller-University, D-07743 Jena
3 Institute of Experimental
Dermatology, Westfaelische Wilhelms-Universitaet, D-48149 Muenster
4 Institute of Infectiology,
Center for Molecular Biology of Inflammation, Westfaelische
Wilhelms-Universitaet, D-48149 Muenster
5 Institute of Medical
Informatics and Biomathematics, Westfaelische Wilhelms-Universitaet, D-48149
Muenster, Germany
Objectives: High-density oligonucleotide
expression arrays (GeneChips) enable the measurement of tens of thousands of
genes in parallel. For every measured transcript (represented by a probe set on
the chip) a Signal value is produced, which is ideally linear proportional to
the mRNA concentration in the original sample. As the target measurement of a
chip experiment, Signal values of course reflect the influence of biological
variance sources, which are actually to be enlightened by the experimental
design. Unfortunately technical variance sources disturb the Signal values in
several ways. Even chips of good quality show influences of technical variance
sources and have to be scaled for further analyses.
Result: As a fundamental framework for own
analyses an SPLUS library for the interaction with the GeneChip technology was
implemented. The first result of our work was the description and
quantification of the occurring technical variance and insights about the main
source of technical variance. To do so, a set of quality criteria was evaluated
on several real-world data sets. The second result was the assessment of
scaling methods towards their capabilities for the compensation of technical
variance effects.
Conclusions: The presented work is an important
step to make meta-analyses on large expression databases reasonable.
Meta-analyses on publicly available expression data analogous to meta-analyses
on publicly available sequence data like BLAST give insights about the
expression behavior on a global scale. In addition, these meta-analyses open
the field of system biology in mammalian organisms.
T-P06 Genomic
rearrangements : influence of the genetic context on chromosomal dynamics
Emilie Fritsch, Jean-luc Souciet, Serge Potier
and Jacky de Montigny
laboratoire
de microbiologie et génétique, institut de botanique, 28 rue Goethe, Strasbourg
67000, France, Phone: 0033390242023, FAX: 0033390242028, e-mail: fritsch@gem.u-strasbg.fr
Chromosomal
rearrangements such as duplications, insertions and deletions contribute to the
genome plasticity and represent a key event in
evolution. They can also be involved in oncogenesis in pluricellular organisms.
In our laboratory, a genetic screening based on a particular allele of URA2
gene in Saccharomyces cerevisiae was created thus allowing a
spontaneous and positive selection of rearrangements such as duplications, Ty1
insertions or deletions.
In order to test the influence a context might have on appearance of
chromosomal rearrangements, we selected revertants in Saccharomyces
cerevisiae S288c haploid or diploid strains. Using molecular and genetic
approaches, we determined the rearrangement responsible for the reversion event
in 13 revertants. We first observed that the frequency of diploid or haploid
revertants obtained in S288c context was higher than the one obtained in FL100.
Moreover, no insertion of Ty1 transposon was observed in S288c whereas in FL100
this event occurred in half of the revertants. This suggests that FL100 and
S288c strains have a different behaviour regarding revertants selections.
Therefore genetic context is important for obtaining different chromosomal
rearrangements both concerning the frequency and particularly the type of
selected rearrangement.
S288c will facilitate the study of the mechanisms involved in chromosomal
rearrangements appearance. The knowledge of the sequence should allow easier
characterization of the chromosomal sites involved in these rearrangements such
as the duplication sites. The S288c context has also been used for a collection
of systematic genes deletion mutants that will help us to construct more easily
isogenic S288c mutants in order to study the genes whose product is involved in
mechanisms responsible for the rearrangements.
T-P07 CellDesigner2.0: A process diagram editor for
gene-regulatory and biochemical networks.
Akira Funahashi 1, Naoki Tanimura 2,
Yukiko Matsuoka 1, Naritoshi Yoshinaga 2 and
Hiroaki Kitano 1
1 ERATO-SORST Kitano Symbiotic Systems Project, JST, Shibuya-ku Jingumae 6-31-15 M-31
6A, Tokyo 150-0001, JAPAN, Phone: +81-3-5468-1661,
FAX: +81-3-5468-1664, e-mail: funa@symbio.jst.go.jp
2 Mizuho Information &
Research Institute, Inc. , 2-3 Kanda Nishikicho, Tokyo 101-8443, Japan
Systems
biology is characterized by synergistic integration of theory, computational
modeling, and experiment. Though software infrastructure is one of the most
critical components of systems biology research, there has been no common
infrastructure or standard to enable integration of computational resources. To
solve this problem, the Systems Biology Markup Language (SBML) [1]
and Systems Biology Workbench (SBW) [2] have been developed. A number of
simulation and analysis software packages already support SBML and SBW, or are
in the process to support it.
An identification of logic and dynamics of gene-regulatory
and biochemical networks is a major challenge of systems biology. We believe
that such network building tools and simulation environments using standardized
technologies play an important role in software platform of systems biology. As
one of the approaches, we have developed CellDesigner [3], which is a process
diagram editor for gene-regulatory and biochemical networks.
The aim of the development of CellDesigner is to supply a process diagram
editor with standardized technology for every computing platform so that it
benefits the users as many as possible. By using the standardized technology,
created model can be easily used with other applications which use standardized
technology, thus it reduces efforts of users to create a model for each
editing/simulation/analysis tools. The main features of standardized technology
which CellDesigner supports are "Graphical representation",
"Model description", and "Application integration
environment".
CellDesigner supports biologist to easily create and simulate gene-regulatory
and biochemical networks using solidly defined and comprehensive graphical
representation [4]. CellDesigner is SBML compliant, and SBW-enabled software so
that it can import/export SBML described models, and can integrate with other
SBW-enabled simulation/analysis packages. CellDesigner runs on Windows, MacOS
X, Linux and other UNIX platforms. The current release version of
CellDesigner2.0 is freely available from http://www.celldesigner.org/.
[1] Hucka, M., et al., Bioinformatics, 19(4),
524-531, 2003.
[2] Sauro, H., et al., Omics, 7, 355-372, 2003.
[3] Funahashi, A., et al., BioSilico, 1, 159-162, 2003.
[4] Kitano, H., BioSilico, 1, 169-176, 2003.
T-P08 Simulation
of epidermal homeostasis including barrier formation
Niels Grabe 1 and Karsten Neuber 2
1 Systems Biology
Group, Centre for Bioinformatics, Bundesstr. 43, Hamburg 20146, Germany,
Phone: +40/42838/7341, FAX: +40/42838/7352, e-mail: grabe@zbh.uni-hamburg.de, Web: www.zbh.uni-hamburg.de
2 Department of Dermatology,
Martinistr. 52, University Clinic Hamburg, Germany
Epithelial
homeostasis is a complex self organizing system of special interest since 85%
of cancers arise in this tissue. Although computational models should lead to a
better understanding of the processes, there is, for example, still no model
reproducing a horizontally layered epidermis emerging from a limited set of
stem cells on an undulating basal membrane. We
propose such a model integrating temporal, spatial and functional aspects. In principle, the
model treats epidermal homeostasis as a steady granular flow and allows the free movement, division and
removal of keratinocytes. Cycling stem cells generate cycling transit
amplifying cells giving rise to keratinocytes committed to differentiation.
Differentiation is dependent on extracellular calcium levels. Following the literature, our model
has a transepidermal water flow transporting particles towards the surface and
leading to particle loss. The epidermal barrier is formed by the secretion of
lipids enclosed in lamella in deeper layers. An intact barrier traps calcium
ions forming a calcium gradient. This gradient automatically results in a
correct layering of differentiated cells. The model is implemented as an
interactive simulation environment, which allows one to modify parameters and
follow properties such as 2D morphology and particle flow. Charts show
gradients, differentiation and kinetic parameters. Given the success modelling a
small set of physical processes, we now intend to extend the methodology to
include more effects and permit the study of other skin
processes.
T-P09 Modelling
protein motions for systems biology
Benjamin A Hall and
Mark Sansom
Department of
Biochemistry, University of Oxford, South Parks Road, Oxford OX1 3QU, United
Kingdom, EU, Phone: 00441865275273, FAX: 00441865275182,
e-mail: hall@biop.ox.ac.uk,
Web: http://sansom.biop.ox.ac.uk
Structural
biology provides atomic resolution descriptions of proteins and related
biological molecules. To link structures to systems descriptions, methods for
predicting the large-scale conformational dynamics of proteins are needed. Such
methods will enable us to describe the dynamic properties of individual
components (i.e. nodes) of complex biological networks.
In this study we have used the Gaussian Network model to define domains of a
protein in relationship to their predicted movements,
by analysis of the eigenvectors and cross correlation of motions over the whole
protein. We have used this approach to examine protein components of complex
membrane transport systems from bacteria. Twisting and fraying
motions could be identified in barrel-like structures, including TolC, BtuB,
and OmpA. Opening, twisting and rocking motions could be identified in
clam-like periplasmic binding proteins such as BtuF and GluR0. In larger
proteins, such as ABC transporters, the relationships between opposing
transmembrane and nucleotide binding domains could be seen. In toroidal proteins,
motions were identified showing that the domains could act in a fashion
comparable to simpler ring structures such as sugars. GNM provides a valuable
tool in identifying such motions and a strong starting point for linking
systems biology with structural biology.
T-P10 Speeding
up the central metabolism in Pichia pastoris
Franz Hartner 1, Lars Blank 2, Alexander Kern 1, Uwe Sauer 2 and Anton Glieder 1
1 Institute of Molecular Biotechnology, Graz University of Technology,
Petersgasse 14/2, Graz A-8010, AUSTRIA, Phone: +43/316/873/4077,
FAX: +43/316/873/4071, e-mail: franz.hartner@tugraz.at
2 Institute of
Biotechnology, ETH Zurich, 8093 Zurich, Switzerland
Pichia pastoris is an excellent host for the
production of heterologous proteins and has major advantages for the production
of eukaryotic proteins when compared to the common host Saccharomyces
cerevisiae. Especially its capability of introducing posttranslational
modifications with similarity to that of higher eukaryotes, its rapid growth to
high cell densities and its folding capacity for eukaryotic proteins makes P.
pastoris well suited for the ever growing market of recombinant proteins.
From a metabolic point of view are the two yeasts well distinct. First, P.
pastoris can utilize methanol as sole energy and carbon source. Second, the metabolism of P.
pastoris under glucose excess conditions is highly respiratoric,
thus this yeast is Crabtree negative. And third, the electron transport chain of P. pastoris is made up of two
alternative oxidases. Therefore, in addition to its biotechnological
importance, P. pastoris is an excellent model organism for eukaryotic
energy metabolism.
We isolated and cloned the alternative oxidase (AOD) of Pichia pastoris.
It enables adaptation to a wide variety of environmental constraints and has
been found in every higher plant and most fungi tested. In yeasts occurrence of
alternative oxidase is very common, but almost exclusively present in species
known to be Crabtree negative.
Manipulating the electron transport chain in a respiratory yeast will directly
influence the entire metabolism through the system-wide high integration of
redox co-factors such as NAD(P)H and the energy currency ATP. Thus,
respiration is an excellent target for research but should also be a major
target for the engineering of industrial microorganisms.
We therefore analysed respiration modified P. pastoris strains under
different defined environmental conditions using quantitative physiology and 1 3C based metabolic flux analysis. This baseline study should end up
in a metabolic model which will be very useful for the inclusion of more
accurate physiological data into the modeling of fermentation processes.
T-P11 Software
components for analysis of DNA microarray and quantitative proteomics data
Sergii Ivakhno and
Olexander Kornelyuk
Department of
Protein Engineering, Institute ofMolecular Biology and Genetics of NAS of
Ukraine, Yakuba Kolosa 8V 24, Kyiv 03148, Ukraine, Phone: 38 044 403 32
49, FAX: 380 44 407 14 43, e-mail: ivakhno@ukr.net
Although
DNA microarray and LC-MS/MS with stable isotope
labeling, such as SILAC (stable isotope labeling by amino acids in cell
culture) and ICAT (isotope coded affinity tags) produce complementary results,
they have not yet been integrated into one software package. As transcription
and translation are mostly uncoupled processes in eukaryotes, a number of
intervening steps, like selective mRNA degradation and stimuli-dependent
translation can contribute to a big difference between mRNA and protein concentration in a single gene.
Furthermore, the data on the dynamics of regulatory post-translational protein
modifications (e.g. phosphorilation) can be combined with the mRNA profiling information
to obtain the snapshot of entire signaling cascades. Our software components
take into account conceptual similarities between ICAT- and SILAC-based
quantitative proteomics and DNA micorarray data. They provide three types of
modules: components for analysis of DNA microarray data, components for
analysis of quantitative proteomics data and components for comparison of
quantitative proteomics and DNA microarray data. All components are programmed
in Java language with some data modules encoded in XML. Components for separate
analysis of quantitative proteomics and DNA microarray data share a number of
similarities, and both of them offer several tools and algorithms: importing of
DNA micorarray or tandem-mass spectrometry data; basic
statistical methods for comparison of quantitative proteomics or DNA microarray
experiments; clustering and classification; visualization tools. The distinct
feature of our software components are tools designed for direct comparison of
DNA microarray and quantitative proteomics data. These tools include new data
structures for storage of DNA microarray/quantitative proteomics data, modified
algorithms for analysis of mRNA/proteins dynamics in time course experiments
with time shift for uncoupled mRNA/protein expression, methods for mapping mRNA
splice variants to corresponding protein isoforms based on their expression
values. A number of image representation algorithms based on relevance networks
are under development, which will allow mutual representation of DNA
microarray/quantitative proteomics data. Bayesian networks for classification
of clinical and other samples for DNA microarray/quantitative proteomics data
are also being incorporated into software components.
T-P12 Systemic
models for metabolic dynamics and regulation of gene expression – easy access, retrieval and search for publicly
available gene expression data
Per Harald Jonson and M.
Minna Laine
Bioinformatics
services, CSC, The Finnish IT Center for Science, P. O. Box 405, Espoo
FI-02101, Finland, Phone: +358 9 457 2263, FAX: +358 9 457 2302,
e-mail: Per.Harald.Jonson@csc.fi, Web: http://www.csc.fi/molbio/
Studies
of gene expression using microarrays result in huge
amounts of data. Many scientific journals require data to be publicly available
at the time of publication, but generally no specific requirements are made
regarding what kind of data and in which format they must be submitted. The
major repositories today are ArrayExpress (AE), GEO and SMD. The suggested
standards like MIAME, and for presenting expression data in XML form, MAGE-ML
are not in full use. Therefore, most sources offer the data in unique formats.
The kind of data offered also varies from images and raw data to normalized
log-ratios. In practice researchers need to search for, download and modify
each dataset extensively before it can be utilized. This makes data retrieval,
evaluation and comparison cumbersome. We believe that the images and the raw,
untreated data should always be made available.
The aim of the project is to bring together data scattered in different sources
for easy access, search and retrieval, and to offer this data in a suitable
form for modelling and other data analysis purposes. The project is carried out
in collaboration with two university groups modelling the yeast gene expression data. Therefore, our primary
interest is in publicly available expression data for Saccharomyces
cerevisiae.
The project is linked to a larger development plan for a systems biology
platform at CSC. In a parallel project ongoing at CSC, a repository and an
analysis environment is being built for microarray data. Our collection of
yeast gene expression measurements will be added into that database. A
comprehensive database will enable researchers to easily retrieve previous
experiments for analysis, comparison with private data and modelling.
At present we are aware of 136 published datasets containing more than 4500
individual slides. Of these about 300 slides (20 publications) are only
available on individual Internet pages. Only 8 publications can be found in AE,
but AE contains more than half of the available slides. About 2000 slides are
unique to AE, whereas SMD and GEO contain about 250 unique slides each.
The gene expression data will be linked with supporting information to enable
retrieval of relevant data for model building. The system should enable
extensive queries – e.g. data from cells growing under iron limitation, cells
having a deletion in HXT1, and so on.
Acknowledgements: We thank Tekes, the
National Technology Agency of Finland for financial support
T-P13 NMR spectroscopy in systems biology: methods for metabolomics and
fluxomics
Paula Jouhten 1, Minna Perälä 1,
Eija Rintala 1, Laura Ruohonen 1,
Perttu Permi 2, Merja Penttilä 1 and
Hannu Maaheimo 1
1 VTT Biotechnology, VTT Technical Research Centre of Finland, Tietotie
2, P.O. Box 1500, Espoo FIN-02044, Finland, Phone: +358/9/19159934,
FAX: +358/9/19159541, e-mail: paula.jouhten@vtt.fi
2 Institute of
Biotechnology, P.O. Box 65, 00014 University of Helsinki, Finland
Systems
biology approach aims to finding links between different levels of cell
function along with determination of links between components on a single
level. Since all the levels interact constantly, the analysis of only
transcriptome and proteome levels provides quite incomplete information of the
system level function. NMR spectroscopy has numerous
advantages in analyses of the metabolome and the fluxome, the complement of the
metabolic fluxes.
The strength of NMR methods in metabolomics is their versatility. 1H NMR is superior in being unbiased
and more sophisticated methodology and detection of other nuclei can be used in
targeted analyses. We are currently using NMR methods both in rapid metabolite
profiling and in identification and quantification of the metabolites. Varian’s
cryogenic probe has provided a huge increase in sensitivity to the analyses. A
spectra library that is being built will aid identification of the yeast metabolic intermediate signals. The
metabolome profile NMR spectra are converted to multivariate data sets and
multivariate data analyses such as principal component analysis (PCA) are being
made.
Carbon-13 labelling experiments are currently the only means to obtain direct
information on the metabolic fluxes in the system. One of the most effective
carbon-13 tracer protocols is metabolic flux ratio (METAFoR) analysis (Szyperski et al.,
1999) where a fraction (10%) of uniformly labelled carbon source is used. Flux
ratios can be determined measuring a single two-dimensional 1H- 1 3C HSQC NMR spectrum
of hydrolysed biomass. We have extended the method to the eukaryotic organism Saccharomyces
cerevisiae, with compartmentalised metabolism, under glucose repressing conditions (Maaheimo et
al., 2001), and also included the glyoxylate shunt to the formalism.
METAFoR analysis provides a global profile of the flux state of the system and
the ratios can be used as constraints in metabolic flux analysis (MFA) (Fischer
et al., 2004). Experiments that employ positionally labelled carbon
source molecules possess higher information content. A phosphorus-31 NMR based
method, 1H- 3 1P
HSQC-TOCSY, for detection of positional fractional enrichments in sugar
phosphate intermediates has been developed.
Fischer, E., et al., (2004), Anal.
Biochem., 325, 308; Maaheimo, H., et al., (2001), Eur. J.
Biochem., 268, 2464; Szyperski, T., et al., (1999), Metab. Eng., 1,
189
T-P14 Autonomous
oscillations in Saccharomyces
cerevisiae during
batch cultures on trehalose
Matthieu Jules, Jean-Marie Francois and
Jean-Luc Parrou
Molecular
Physiology of lower Eucaryotes, INSA Toulouse, 135, Avenue de Rangueil, Toulouse
F-31077, France, EU, Phone: +33/664001212, FAX: +33/561558400,
e-mail: matthieujules@yahoo.fr
In a
previous work, we have shown that exogenous trehalose assimilation in Saccharomyces
cerevisiae is purely oxidative and takes
place by two independent pathways (Jules, M, et al., 2004, AEM, 70, 2771). The main one relies on the
periplasmic hydrolysis of trehalose by the acid trehalase Ath1p, while the
second route requires the coupling of the trehalose uptake (Agt1p) and its
intracellular hydrolysis (Nth1p).
Recently, we reported for the first time autonomous oscillations in discontinuous cultures (batch) of S. cerevisiae growing on trehalose as the sole carbon
source. This unexpected oscillatory behaviour was examined from online gases
measurements data (Mass Spectrometry) using Fast Fourier Transformation
(FFT). This robust mathematical analysis, coupled
to phase portrait diagrams, highlighted the existence of two types of
oscillations as well as basic information about their properties. The
first type of oscillations was found to be linked to the cell cycle since (i) the periods were
fractions of the generation time, (ii) the oscillations were accompanied by
transient increase in the budding index, mobilisation of storage carbohydrates
and fermentative activity, and (iii) the occurrence of this type of oscillation
was dependent on the specific growth rate (use of mutants exhibiting distinct,
and constant or variable specific growth rates). The second type are
sustained, short-period, respiratory oscillations and are independent of
the specific growth rate. These two types of oscillations were remarkably
identical to the cell cycle related and short-term oscillations that are
usually observed in aerobic glucose-limited continuous cultures in the dilution rate range of 0.03 to 0.15
h-1 (Beuse, M, et al., 1998, J. Biotechnology, 61, 15; Lloyd, D, et al., 2003,
FEMS Yeast Res., 3, 333). Another originality of this work was to find that
contrary to previous considerations in continuous cultures, these two types of
oscillations can take place consecutively and/or simultaneously during batch
cultures on trehalose. In addition, mobilisation of intracellular trehalose
emerged as a key parameter for sustainability of ultradian oscillations
(subject of in silico investigations). Altogether, batch cultures on
trehalose could be an excellent device to further investigate the molecular
mechanisms that underlie autonomous oscillations (Jules, M, et al., 2004, Eur.
J. of Biochem., submitted).
T-P15 Automated
construction of genetic networks from mutant data
Peter Juvan 1, Gad Shaulsky 2 and Blaz Zupan
1
1 Artificial Intelligence Laboratory, Faculty of Computer and Information
Science, Trzaska 25, Ljubljana 1000, Slovenia, Phone: +386/1/4768 267,
FAX: +386/1/4768 386, e-mail: peter.juvan@fri.uni-lj.si,
Web: http://www.ailab.si
2 Department of Molecular
and Human Genetics, Baylor College of Medicine, Houston, USA
Geneticists often use mutations to investigate
biological phenomena 2. Mutations cause changes of organism's
phenotype and may reveal which genes participate in a certain biological
process and how. To represent these functional interactions between genes, a
genetic regulatory networks are an often used formalism 1.
We have developed a system called GenePath for automated construction of
genetic regulatory networks from mutant data. GenePath considers classical
genetic data where a phenotype is observed for a set of single or double
mutations. Prior knowledge, expressed through relations between genes (possibly
extracted from the relevant literature) can also be included. GenePath employs
a set of logical patterns of the type “IF there exist a set of experiments that
involves genes A, B, ... THEN a certain relation between these genes can be inferred”.
These relations are then used to propose genetic networks. An important feature
of GenePath is the ability to explain each relation from the constructed
network by reporting on the logic that was used to infer it together with the
corresponding experiments.
GenePath thus formalizes genetic data analysis 3, facilitates
the consideration of all the available data in a consistent manner, and allows
for the examination of the large number of possible consequences of planned
experiments. We will illustrate through an example the advantage of using an
automated approach and report on recent extensions of GenePath that include (1)
handling of uncertainties in genetic data by allowing to assign confidence to
experiments and background knowledge, (2) assistance in experiment planning by
proposing a set of the “cheapest” experiments to assert new gene-to-gene relations, (3)
interactive what-if analysis, which enables the user to on-the-fly test
alternative hypothesis about the organism's regulatory mechanisms, and (4)
handling of cyclic pathways through detection of genes that are involved in
such network and appropriate visualizations. GenePath is implemented as a
web-based application that is available at
<Ahref="http://www.genepath.org" target="_blank">http://www.genepath.org.
References: (1) Altman, R.B. et al., (2001), Curr Opin
Struct Biol., 11(3), 340. (2). Avery,
L. et al., (1992), Trends Genet., 8(9), 312.
(3). Zupan, B. et al., (2003), Bioinformatics, 19(3), 383.
T-P16 An
integrative framework for modeling signaling pathways
Robert Modre-Osprian, Marc Breit, Visvanathan Mahesh,
Gernot Enzenberg and Bernhard Tilg
Institute for
Biomedical Signal Processing and Imaging, UMIT, Eduard-Wallnöfer-Zentrum I,
Hall in
Tirol A-6060, Austria, Phone: ++43 50 8648 3819, FAX: ++43
50 8648 3850, e-mail: robert.modre@umit.at,
Web: http://imsb.umit.at/
Objective: Much of the effort to understand
signal transduction has been aimed at
identifying the molecules that participate in signaling cascades and at
qualitatively characterizing the activities and interactions of these
molecules. Improving our understanding of pathways typically involves an
iterative interplay between experimental and modeling strategies. The intention
of the proposed framework is to enable such an iterative interplay. The formation
of protein complexes based on
protein-protein interactions was investigated and used to define the
requirements for that framework.
Methods: The framework includes environments for pathway visualization
and simulation, as well as a knowledgebase providing biological data
(components and interactions), modeling data (chemical-reaction network of the
components and its rate constants), simulation data (sensitivity analysis), and
experimental data (connectivity map of the components and RNA interference experiments). The formation of
protein complexes was investigated using a recently produced model of the
TNFalpha-mediated NF-kappaB pathway [1]. An integrated approach of proteomic
pathway mapping based on established components provided experimental
information [2].
Results: Modeling, simulation, and sensitivity analysis with respect to
rate constants and initial concentrations were performed. Modified mathematical
models with new added proteins and modified protein-complexes were
reconstructed based on given experimental data. Experimental data of systematic
single gene expression perturbations using
RNA interference, which in some respects mimics pharmacological treatment, were
used to validate the behavior of the produced mathematical models.
Conclusion: The proposed framework might allow one to simulate
experimentally observed input-output relationships of signaling pathways and
might support the iterative interplay between experimental and modeling
strategies. In addition, the efforts of modelers and quantitative
experimentalists will have to be tightly integrated in order to understand how
the components in a signaling cascade work in concert.
References: [1] Cho KH
etal. Experimental design in systems biology based on parameter sensitivity
analysis with Monte Carlo simulation. Simulation. 2003;79(12):726-39.
[2] Bouwmeester T etal. A physical and functional map of the human
TNFalpha/NF-kappaB signal transduction pathway. Nat Cell Biol. 2004;6(2):97-105.
T-P17 Data
visualization for gene selection and modeling in cancer bioinformatics
Minca Mramor, Gregor Leban and Blaž Zupan
Artificial
Intelligence Laboratory, Faculty of Computer Science and Informatics, Tržaška
25, Ljubljana 1000, Slovenija, EU, Phone: +386/1/4768 299,
FAX: +386/1/4768 386, e-mail: minca.mramor@fri.uni-lj.si,
Web: www.ailab.si
Microarray technology enables simultaneous
investigation of expression of thousands of genes and has become a major tool
for genomic studies of human cancer. Gene expression studies may
contribute to more effective tumor classification and provide insights into
pathogenesis, diagnosis, prognosis, therapeutic targets and clinical outcome of
tumors. While the goal of such studies - finding genes that can differentiate
between groups of outcomes - is clear, the analysis of experimental data sets
is complex and difficult. Modeling of such data belongs to the domain of
supervised learning methods. Such methods that are most often applied in cancer
bioinformatics include support vector machines and artificial neural networks.
Although these methods may find accurate predictive models, their
interpretation is difficult and thus their contribution in discovery of new
knowledge rather limited. In contrast to complex computational models (1), we
claim that simple data visualization techniques may be used to clearly expose
discrimination between different types of tumors.
In research reported here we used the data sets from the Cancer Program
of the Broad Institute (http://www.broad.mit.edu/cgi-bin/cancer/datasets.cgi). We evaluated
the predictive power of expression of genes using a standard, well-known non-myopic
measure called ReliefF. Our experiments show that in most cancer microarray
data sets, there are at most few tens of genes which have distinctively highest
predictive value. We have also observed that their functional annotation
related them to the biological processes in observed cancers.
To search for visualizations where data points belonging to distinct
outcomes are well separated we used a method called VizRank. VizRank is able to
automatically rank visual projections according to estimated degree of
separation of outcomes, and provides heuristics to search only through the
space of most promising visualizations (2). To visualize the expression
data, we used either scatterplot or radviz. For most of the data sets, VizRank
found radviz visualizations where separation of groups with different outcomes
was clear and obvious. This method provides for a simple, understandable, and
significatly less sophisticated classification model than prevailing techniques
used in current cancer bioinformatics practice.
References:
1.Statnikov, et. al. (2004) Bioinformatics, in print 2.Leban G, et al., (2004) Bioinformatics, in print
T-P18 Accelerating
the construction of genome-scale metabolic models: a
test case for Lactococcus lactis
Richard A. Notebaart 1, Frank H.J. van Enckevort 2, Bas Teusink 3 and Roland J. Siezen 4
1 CMBI, Centre for Molecular and Biomolecular Informatics, Radboud
University, Toernooiveld 1, Nijmegen NL-6525ED, The Netherlands, Phone: +31/24/3652346,
FAX: +31/24/3652977, e-mail: R.notebaart@cmbi.ru.nl
2 NIZO food research, Ede, Food Valley, The Netherlands 3 Wageningen Centre for Food Sciences,
Wageningen, The Netherlands 4
CMBI, Centre for Molecular and Biomolecular Informatics, Radboud University,
Toernooiveld 1, Nijmegen NL-6525ED, The Netherlands
In the
past few years, a large number of sequenced and annotated genomes have become
available and many more genome-sequencing
projects are still in progress. The availability of the genomic information of
a species allows genome-scale metabolic reconstruction based on its annotation
and on existing metabolic information from other species.
The process of metabolic reconstruction is extremely time-consuming. We have
developed (and benchmarked for Lactoccocus lactis IL403) a method to
benefit from available well-curated metabolic databases to accelerate this
time-consuming process. First we created an in-house database of orthologous
genes for a selected number of species. The procedure for constructing the
database of orthologous genes is based on a combination of COG 1 and
INPARANOID 2, resulting in an improved orthology definition. Then we
selected orthologous hits with genes of L. lactis to retrieve metabolic
information from manually curated metabolic databases of E. coli, L. plantarum and B. subtilis. We used both public data from EcoCyc (http://ecocyc.org/) and
LacplantCyc (http://www.lacplantcyc.nl/) and proprietary data from SimPheny
models (Genomatica Inc.).
The number of automatically assigned gene-reaction
associations, using Pathologic software 3 and Genbank information
for L.lactis (2400 genes), was 483. With the described method we
identified 207 extra putative gene-reaction associations. Using SimPheny models
of E. coli, L. plantarum and B. subtilis we were able to
assign 336 likely and 183 putative gene-reaction associations for the
construction of a L. lactis SimPheny model. The total assignment of
gene-reaction associations by manual curation is still in progress.
The results show that we can accelerate the metabolic reconstruction by taking
optimal advantage of already existing, manually curated databases and models.
1Tatusov, R.L., et
al., (1997), Science, 278, 631; 2Remm
M, et al., (2001), JMB, 314, 1041; 3Karp P.D., et al., (2002), Bioinformatics, 18, 225
T-P19 Application
of yeast genomic strategies to link biologically active compounds to their
cellular targets
Ainslie
B. Parsons 1, David Williams 2, Satoru Ishihara 3, Yoshi Ohya 3, Raymond Andersen 2, Timothy Hughes 1 and
Charles Boone 1
1 Department of Molecular and Medical Genetics, University of Toronto,
112 College St, Toronto ON M5G 1L6, Canada, Phone: +1 416 214 9471,
FAX: +1 416 978 8598, e-mail: ainslie.parsons@utoronto.ca
2 Department of Chemisty,
University of British Columbia, Vancouver, Canada
3 Department of Integrated
Biosciences, University of Tokyo, Kashiwa, Japan
Bioactive
compounds can be valuable research tools and drug leads, but it is often difficult to identify
their mechanism of action or cellular target. To address this, we have recently
developed a genomic platform in the budding yeast Saccharomyces cerevisiae that we
believe is an effective and broadly applicable method for deciphering the
cellular pathways and proteins affected by inhibitory compounds (Parsons et al., (2004) Nat Biotechnol,
22:62-9). Here, comparing chemical-genetic interaction profiles, where the
complete set of yeast viable mutants are tested for hypersensitivity to a
target-specific compound, with a compendium of global genetic interaction
profiles provides a powerful key for deciphering the pathways and targets of
the growth-inhibitory compounds. In concert with chemical-genetic profiling we
are currently employing additional yeast genomic tools such as expression
profiling, drug induced haploinsufficiency analysis, and mapping of
drug-resistant mutants using Synthetic Genetic Array (SGA) methodology to
identify targets of novel anti-fungal compounds derived from natural product
extracts. In particular, Papuamide B (Pap B), a high molecular weight cyclic
lipopeptide, is intriguing because our analysis suggests that it may function
similarly to Caspofungin, an echinocandin-like cyclic lipopeptide that disrupts
the yeast cell wall by inhibiting 1,3 beta-glucan synthesis. The
chemical-genetic interaction profiles for Pap B and Caspofungin are highly
similar, containing numerous common genes, many of which are associated with in
cell wall organization and biogenesis, cytokinesis, and chitin metabolism. However,
unlike Caspofungin, Pap B does not inhibit beta-1,3-glucan synthesis in
vitro, suggesting it functions through a different but functionally related
target. To gain further insight into the mode of action of Pap B, we are
applying our SGA mapping methodology to identify Pap B resistant mutants. We
have identified one Pap B resistant strain as a cho1 mutant. CHO1
encodes a non-essential enzyme required for the synthesis of phosphatidylserine
(PS), one of four major phospholipids found in yeast cell membranes. Further
analysis is suggestive of an interaction between Pap B and PS and we are
currently examining whether Pap B may be entering the cell via PS and then
acting on an intracellular target or if Pap B could be exerting its effect
directly through PS.
T-P20 SCIpath
- an integrated environment for systems biology analysis and visualisation
Manish Patel
Academic
Oncology, University College London, RF Hospital & Medical School, Rowland
Hill Street, London NW3 2PF, UK, Phone: +44/207/7940500/5499,
FAX: +44/207/7943341, e-mail: mpatel@medsch.ucl.ac.uk,
Web: www.ucl.ac.uk/oncology/MicroCore/microcore.htm
The
SCipath toolkit is a SBML-aware suite of analysis programs for microarray and proteomics data
analysis that is open source and programmed exclusively in Java. SCIpath
provides a flexible and extensible environment for the interpretation of
functional genomics data through visualisation. The first version of the
application (downloadable from the SCIpath website), implements four tools that
include functionalities such as editing and visualisation of custom-made
pathways, visualisation and manipulation of microarray data alongside pathway
data and a fuzzy clustering algorithm for gene expression data. The current tools provide a
simple yet powerful way of graphically relating large quantities of expression
data from multiple experiments to cellular pathways and biological processes in
a statistically meaningful way. SCIpath is an expandable toolkit with the aim
of promoting understanding of systems biology by bringing together different
aspects of cellular biology through visualisation and analytical techniques.
This objective is realised through the integration of applications, written
independently by other programmers, in a plug-in style interface that enables
the end-user to manipulate different types of data from different applications
to build innovative visualisations and analytical software components. By using
a structured framework that controls the messaging between applications in a
service-oriented fashion, the SCIpath design enables the user and programmer to
port the functionalities of different applications so that biological data from
multiple applications can be visualised intuitively, thereby enhancing the
understanding of the biological system.
T-P22 Glycolytic
oscillations in spatially ordered interacting cells
Jana Schütze and
Reinhart Heinrich
Group of
Theoretical Biophysics, Humboldt-Universität zu Berlin, Invalidenstr. 42,
Berlin D-10115, Germany, Phone: +49/30/2093-8381,
FAX: +49/30/2093-8813, e-mail: jana.schuetze@rz.hu-berlin.de
Synchronisation
of glycolytic oscillations in populations of yeast cells have been intensively analysed
experimentally [1] as well as theoretically [2]. There is evidence that the
individual cells communicate by exchanging products of glycolysis as
acetaldehyde, for which the plasma membrane is permeable. There are, however, a number of
discrepancies between theoretical predictions and experimental results. For
example, in existing models synchronisation occurs much slower than observed in
populations of yeast cells [2]. In extension to previous models which
considered stirred cell suspensions we study oscillations in spatially ordered
cells. We aim to reproduce data of experiments where glucose is added to starved cells in a
limited region of the cell layer initiating in this way a wave resulting from
the propagation of glycolytic oscillations [3]. For the generation of
oscillations in the individual cells a simple model of glycolysis is used
containing an autocatalytic step. Cells are embedded in an extracellular medium
in which the added glucose and the extracellular product can diffuse. The model
takes into account special kinetic properties of glucose carriers in yeast
cells. Intercellular coupling takes place via diffusion of the end product between neighbouring
cells. For a small number of linearly ordered cells, the oscillations can be
studied by using bifurcation analysis. We could find very complex oscillatory
states already for three interacting cells with a uniform glucose input. The
model is extended by considering linear arrangements of many interacting cells.
Moreover, we studied regular and irregular two-dimensional spatial arrangements. When glucose injection is
confined to a limited number of neighbouring cells, we observe the formation of
propagating waves of glycolytic oscillations over the whole arrangements. The
characteristics of wave formation are studied in dependence on the diffusion of
glucose in the extracellular medium, the uptake of glucose by the individual
cells, as well as on the permeability of the membrane for the coupling
substance. It is shown that the existence of waves depends crucially on the
strength of the coupling.
[1] Richard,
P. et al. (1996), Eur. J. Biochem. 235, 238;
[2] Wolf, J. and Heinrich, R. (2000),
Biochem. J. 345, 321 [3] Mair, T.
et al. (2001), Faraday Discuss. 120, 249
T-P23 Database
Support for Yeast Metabolomics Data Management
Irena Spasic, Warwick Dunn and Douglas Kell
School of
Chemistry, University of Manchester, Sackville Street, Manchester M60 1QD,
United Kingdom, Phone: +441612004414, FAX: +441612004556,
e-mail: i.spasic@manchester.ac.uk,
Web: http://dbk.ch.umist.ac.uk/
The
enormous flood of omics data brings the need for well-designed and -curated
databases, which can store, handle and disseminate large amounts of data
efficiently and readily lend themselves to data mining techniques used to
extract hidden patterns from the data. The extracted facts can be further
explored in simulation-based analyses providing predictions to be tested in
vivo/vitro. Such a framework consisting of modelling, machine learning and
simulation can help achieve the systems biology objective (to understand the
way in which the heterogeneous parts of biological systems combine to form the
whole).
The first step is to develop and standardise the omics models and populate them
with curated data. The proteomics field has significantly advanced in such
efforts: HUPO has driven the Protein Standards Initiative in order to develop
an exchange standard for proteomics data (MIAPE) based on the PEDRo schema.
Similarly, MAGE-ML represents an emerging standard for transcriptomics. Some
attempts have been made in metabolomics, e.g. in ArMet, a data model for plant
metabolomics (www.armet.org).
We developed a similar approach for yeast metabolomics with an emphasis on metabolomic
footprinting as a strategy for functional genomics. Nonetheless, the general
schema is applicable to a wide range of metabolomics experiments. The core
schema (modelled in UML) consists of abstract classes, which can be specialised
in order to embrace different types of experiments, results, organisms, etc. It
is designed to capture information about the overall experimental cycle,
including growth, sample preparation and analytical experiments. Storage of the
information about specific conditions, protocols and parameters used in wet
experiments (i.e. the metadata) is needed to interpret the experimental results
and support their comparison and reproducibility. In addition, metabolomics
experiments in the post-genomic era often need to be extended beyond the
traditional wet experimental framework. In order to process the vast amount of
metabolomics data, data mining experiments (or dry experiments) need to be
performed in silico to extract knowledge.
Our metabolomics database model has been implemented as a relational database
and an XML schema. In both cases, flexibility has been supported by using a
modular approach where different metadata modules (implemented as separate XML
schemas) can be plugged into the overall metabolomics schema (in both
relational and XML versions).
T-P24 Fokker-Planck
equations for IP3 mediated Calcium dynamics
Rüdiger Thul and
Martin Falcke
Abteilung
Theorie, Hahn-Meitner Institut, Glienickerstrasse 100, Berlin D-14109, Germany,
Phone: +49 30 8062 3198, FAX: +49 30 8062 2098, e-mail: thul@hmi.de
The
stochastic behavior of release channels on the membrane of the endoplasmic reticulum has proven
essential for intracellular Ca 2+ dynamics. This holds in particular
for inositol-1,4,5-trisphosphate (IP3) mediated Ca 2+
release through IP3 receptor channel clusters. Recent studies have
revealed two states with an increased probability. In a deterministic model,
these states correspond to stable stationary states. While one is a state of
low Ca 2+ concentration with no channels open, the second represents
a state with high Ca 2+ concentration known as a Ca 2+
puff. The impact of fluctuations in such a system is reflected in transitions
from one stationary state to the other. Puffs are pivotal for the intracellular
Ca 2+ dynamics because a concerted action of several puffs can
initiate vital Ca 2+ waves. Therefore, we derive a master equation
and corresponding Fokker-Planck equations for the stochastic dynamics of an IP3
receptor channel cluster. It incorporates the strong localization of the
channel clusters on the membrane of the endoplasmic reticulum and the ensuing
large Ca 2+ concentrations at an open cluster. We employ our ansatz
to compute the stochastic fraction of the puff frequency. Our results shed new
light on the stochastic aspect of puff occurrence.
T-P25 Global
transcriptional response of Saccharomyces cerevisiae to ammonium, alanine, or glutamine limitation
Renata Usaite, Birgitte Regenberg and Jens Nielsen
Center for
Microbial Biotechnology, BioCentrum-DTU, Technical University of Denmark,
Soeltofts Plads, Building 223, Kgs Lyngby DK-2800, Denmark, EU,
Phone: +45/4525/2673, FAX: +45/4588/4148, e-mail: ru@biocentrum.dtu.dk, Web: http://www.biocentrum.dtu.dk/index_eng.html
Cell
physiology and changes on the gene expression level under nitrogen limitation of the haploid yeast Saccharomyces cerevisiae strain
CEN.PK-113-7D were studied using nitrogen limited chemostat cultures and global transcription analysis
tools. S. cerevisiae was grown separately at alanine, ammonium, and glutamine
limitation in aerobic chemostat cultures with a dilution rate (D) of 0.2 h-
1. In order to make nitrogen limitation studies more comprehensive,
transcription data was analyzed together with results obtained from glucose limitation studies. Genome scale
metabolic model knowledge and various bioinformatics tools were applied in the
transcription data analysis in order to study cell physiological response to
the specified limitations. Results showed that central nitrogen metabolism was
affected not only due to nitrogen limitation, but it also depended on the
limited nitrogen source. Genes with altered expression level under every
nitrogen limitation were involved in amino acid metabolism, protein biosynthesis, cell organization, ion
homeostasis, transcription regulation, energy production and various transport systems. Alanine limitation produced the most
drastic differences on the yeast metabolism. Group of genes involved in
nitrogen catabolic repression, amine group, or pyruvate metabolism had changed
expression, including strongly affected alanine aminotransferases coding genes YDR111C
and YLR089C. Transcription analysis results were consistent with
physiological observations and indicated that yeast’s ability to grow under
various nitrogen limitations was acquired by altering regulation of central
nitrogen metabolism and through propagated changes in other metabolic pathways,
crucial to cell growth and development.
T-P26 Identification
of the C-terminal signal peptides for GPI modification and prediction
of the cleavage sites.
Yu Zhang, Thomas Skoet Jensen, Ulrik de
Lichtenberg and Soeren Brunak
Center for
Biological Sequence Analysis, BioCentrum, The Technical University of Denmark,
Building 208, Kgs. Lyngby DK-2800, Denmark, Phone: (+45) 4525 2427,
FAX: (+45) 4593 1585, e-mail: yu@cbs.dtu.dk,
Web: http://www.cbs.dtu.dk
Glycosylphosphatidylinositol
(GPI)-anchored proteins represent a subclass of cell surface proteins found in
all eukaryotic cells. Knowledge of a protein's GPI
modification is very valuable, since it defines its subcellular localization
and limits the range of possible cellular functions. The modification takes
place in the Endoplasmatic Reticulum (ER) and involves enzymatic removal of a
C-terminal signal peptide (much like N-terminal signal
peptides), followed by addition of the sugar moeity to the new C-terminal amino
acid residue of the protein.
A number of sequence based methods have been developed for the prediction of
GPI-anchored proteins, most of which rely on defining a C-terminal consensus
sequence for GPI modification. However, only about 40 proteins with
experimentally verified cleavage sites are known to date. Many of the
prediction tools currently available are therefore based on protein examples
for which the GPI signal sequence and the cleavage site are either inferred
from homology or based on predictions by previous in silico methods.
Here we present a new sequence based prediction tool of GPI-anchored proteins,
based exclusively on experimentally verified data extracted from Swiss-Prot.
The method is based on artificial neural networks - a machine-learning method
able to capture non-linear context dependent patterns. The predictive
performance of the method is better than all other methods currently available.
T-P27 The
Genevestigator gene function discovery engine
Philip Zimmermann, Matthias Hirsch-Hoffmann, Lars Hennig
and Wilhelm Gruissem
Plant
Biotechnology/Bioinformatics, ETH Zurich, Universitätsstrasse 2, Zurich 8092,
Switzerland, Phone: ++41/1/6322244, FAX: ++41/1/6321079,
e-mail: philip.zimmermann@ipw.biol.ethz.ch,
Web: http://www.pb.ethz.ch/~zimmerph/
Despite
its short history, the era of gene expression profiling has already acquired an
expertise that matches up to the virtue of many older sciences. From the
beginning, hopes to rapidly discover the function of genes fostered the
development of ever better analysis tools. Nevertheless, progress in
computational tools to rapidly identify gene function is still facing a
prevailing gap between wet lab biologists and the bioinformatics world.
Moreover, the genome-wide study of gene expression in higher organisms has furthermore
increased the level of complexity, outstripping the ability of many scientists
to analyze and interpret this bulk of data. This discrepancy challenges the
scientific community to increase the pace of developing computational analysis
capabilities targeted to non-bioinformatics users.
Epitomizing the gap between data generation and data analysis, the availability
of tools for biologists to easily query large microarray datasets to retreive
the properties of genes of interest is currently scarce. Genevestigator [1] is
an Arabidopsis microarray database with a suite of tools for gene function
discovery. User-friendly web-based query forms allow to retrieve the expression
patterns of genes of interest in chosen contexts, such as plant development,
organs, and responses to stresses or mutations. Further tools allow to
indentify marker genes expressed specifically in these categories or responding
to particular conditions. The Genevestigator suite of tools is being completed
by a metabolic and regulatory pathway map that will allow users to match genes
of interest to these pathways and to identify genes involved in different
biological processes. The goal of Genevestigator is to drive lab research by
providing contextual information about the expression of genes, allowing
targeted design of new experiments.
As of December 2004, 1500 plant scientists have registered and more than 50
distinct users query the database daily. Owing to the strong interest from
other communities, we look for partners to help extend Genevestigator to other
organisms. In fact, a multi-organism and comparative genomics platform could further
contribute to fulfilling the hope for an accelerated discovery of gene
function.
1. Zimmermann et al., 2004. Plant Physiol 136: 2621-2632.
T-PoP1 1/f
Noise in Ion Channel: A Theory Based on Self-Organised Criticality
Jyotirmoy Banerjee 1, Mahendra
K. Verma 2 and Subhendu Ghosh 1
1 Department of Biophysics, University of Delhi South Campus, Benito
Ju8arez Road, New Delhi 110021, India, Phone: 91-11-26887005,
FAX: 91-11-26885270, e-mail: subhog@vsnl.com
2 Department of Physics,
Indian Institute of Technology, Kanpur, U.P. India
The aim
of this work is to investigate the noise profile of Voltage Dependent Anion Channel
(VDAC). VDAC from mitochondria of rat brain was isolated and purified using
standard procedure, and reconstituted into a planar lipid bilayer (Banerjee & Ghosh, 2004). Single-channel currents were recorded under different voltage
clamp conditions across the membrane. Power
Spectrum analyses of the current-time traces were done. It was found by the
slope measurement of the power spectrum (spectral density versus frequency
log-log plot) that the open state noise of single-channel VDAC follows power
law and the noise is of 1/f nature. Moreover, this 1/f nature of the open
channel noise is maintained throughout the range of electric potential -45 to
+45mV. Having obtained this result on open channel noise we looked into the
origin of the 1/f pattern in ion channels. It is
interesting to note that 1/f noise has been observed in a wide variety of
systems ranging from earthquake, turbulence to electrical circuits and now in
biological membrane system. We have used the concept of Self-Organised
Criticality (SOC) and proposed a model to explain the existence of power law
(1/f) in channel noise. It is being proposed that in this process of passive
diffusion transport through VDAC, the 1/f noise arises out of
defects in the passage of ions across the membrane. In doing so an analogy has
been drawn between the sand pile avalanches (Bak et al, 1987) and the sudden
opening of the VDAC. Although the theory proposed by us is based on the
experimental data (bilayer electrophysiology) on VDAC, it should hold for ion
channels in general.
Reference:
Banerjee, J. & Ghosh, S. (2004) Biochemical &
Biophysical Research Communications, 323, 310.
Bak, P., Tang, C. and Wiesenfeld, K. (1987) Phys. Rev. Lett., 59, 381.
T-PoP2 Single
cell mechanics and mechano signal transduction using a micro-force loading device
Hao Zhang, Zhiqing Feng, Ning Fang,
Vincent Chan and Kin Liao
Bioengineering,
Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798,
Singapore, Phone: +65-67905835, FAX: +65-67916905, e-mail: askliao@ntu.edu.sg
The
interplay between molecular forces within a biomolecule, between biomolecules
and sub-cellular structures and biochemical processes is vital in maintaining
proper functioning of a living system. However, how exactly the molecular
machines of a living system are organized, regulated, and challenged by various
forces are not fully understood at the present. A general strategy to reach
such an understanding is to disturb the otherwise dynamically equilibrium
living system by externally applied force and study its effects on various
levels of cellular structures.
In this study we have developed an integrated analytical system for studying
cellular mechanics, in which a loading device with a force resolution of 0.01
micro newton and displacement resolution of 1 nm constructed (Fig. 1). The
loading device is in assembly with a confocal force microscope such that
mechano-signal transduction at the micrometer scale (through
real-time imaging of deformation of a single cell, Fig. 1) and at the nanometer
scale (through real time visualization of the deformation of green fluorescence
protein labeled cytoskeleton), can be quantitatively
interpreted. At the cellular level, interpretations of experimental data were
also carried out by using continuum mechanics modeling (finite element method,
Fig. 2) of a single cell, and at the molecular level, using molecular dynamics
simulation (Fig. 3).
Applying the novel nano-analytical tools coupled with modeling at multiple
length scales for studying the complex mechano-biochemical-driven cellular
processes at the molecular/cellular level will provide new systematic insights
into the physical principles of mechanics-driven functions, e.g., injury,
organogenesis, etc, and powerful design principles for novel diagnostic and
therapeutic strategies.
T-PoP3 Connectivity
matrix for describing all the atom-level connectivities in a given metabolic
network and its use for analysis of the network structure
Jun Ohta
, Okayama
University Graduate School of Medicine and Dentistry, 2-5-1 Shikatacho, Okayama
700-8558, Japan, Phone: +81/86/235/7124, FAX: +81/86/235/7126,
e-mail: jo25@md.okayama-u.ac.jp,
Web: http://www.metabo-info.org
Atom is
the smallest node in metabolic networks. Therefore, atom-level consideration is
important in the understanding of metabolic networks. Atom-level connectivity in metabolic networks is classified into
inter-metabolite connectivity through enzymatic reactions and intra-metabolite
connectivity through chemical bonds. Evolutional co-emergence of a new
metabolite and a new enzymatic reaction responsible for the metabolite is a reasonable
idea, suggesting potential correlation and probably complicated relationship
between inter- and intra-metabolite connectivities in the network structure. On
the other hand, membrane transport is common biological phenomena and
stoichiometry is one of the most important characteristics in metabolic
networks. Here, I present a data format for describing both inter- and
intra-metabolite atom-level connectivities including information of
compartmentation and stoichiometry. In the present format, each atom is
expressed as a row vector composed of 4 integers indicating metabolite species,
position of atom in the metabolite, atom species, and compartmentation.
Information about connectivity itself is also expressed as a row vector
composed of 3 integers. A vector for inter-metabolite connectivity includes 1
(connectivity-type number), reaction number, and stoichiometry number which
indicates how many times the connectivity appears in the reaction, whereas a
vector for intra-metabolite connectivity includes 0 (connectivity-type number)
as the 1st component, and bond-type number as the 2nd component. Thus,
conversion of atom va1 to atom va2 through reaction vr is expressed as (va1,
va2, vr), a row vector formed by the combination of 3 row vectors. Using this
format, all the atom-level connectivities in a network are expressed as a
matrix, connectivity matrix (CM), each row of which corresponds to one
connectivity. Using a database, IMAC (www.metabo-info.org), CM for a given network
can be prepared. Structural analyses of the network can be performed on GNU
Octave using the CM obtained and m-files for Octave scripts or functions. At
present, calculation of the paths between 2 specific atoms and calculation of
net balance of the reaction sequence constituting such a path can be performed
using m-files for Octave. The present format is expected to contribute to the
understanding of structure-function relationship of metabolic networks.
T-PoP4 Using
SRS to develop and populate an information layer for the EMI-CD modeling
platform
Dan Staines, Daniel Flint and Thure Etzold
SRS Development
Group, LION Bioscience Ltd., 80-82 Newmarket Road, Cambridge CB5 8DZ, United
Kingdom, Phone: +44/1223/224700, FAX: +44/1223/224701, e-mail: daniel.staines@uk.lionbioscience.com,
Web: http://www.lionbioscience.com/
The
European Modeling Initiative Combating Complex Diseases consortium (EMI-CD; http://www.molgen.mpg.de/~EMI-CD/)
aims to develop a modeling and data integration platform for complex and
heterogeneous data sources, that can generate hypotheses and models for disease
processes based on in silico predictions. LION (http://www.lionbioscience.com/)
is developing software to construct and automatically populate an information
layer, supplying exhaustive knowledge of the relevant biological objects
(genes, proteins, protein complexes, organelles, cells, etc.).
The information layer is based on the SRS data integration platform (Etzold,
T., et al., in "Bioinformatics: Managing Scientific Data”, Lacroix,
Z. and Critchlow, T. (eds), Morgan Kaufman, 2003), which can integrate data
from diverse sources. Interaction with the analysis and modeling layers of the
EMI-CD platform will use the SRS loader technology combined with a web services-based
interface to provide the required data in an accessible, coherent and flexible
manner.
A
central source of data for this project is Reactome (http://www.reactome.org/),
which has been incorporated into SRS, but additional sources include
sequence-based databases (e.g., RefSeq, Ensembl), structural databases (e.g.,
PDB, CATH), model-based databases (e.g., KEGG) and knowledge-based
databases (e.g., GeneCards). Further data sources will be added as the
project progresses. Cross-reference and related information between these data
sources will also be used to generate a network of information sources. In
order to support specific simulation experiments, mechanisms using the original
sources will be developed to generate data sets relevant to specific diseases,
or disease areas.
It is of
central importance to the information layer that the data is up-to-date and of
high quality and internal consistency. An automated mechanism, based on SRS
Prisma, will be developed to carry out automatic updating of data from remote
repositories, further data processing, and quality assurance using tailor-made
tests to check the quality of any new data.
The
EMI-CD consortium consists of three academic partners and two SMEs, and is
funded by the European Commission within its FP6 Programme, under the thematic
area "Life sciences, genomics and biotechnology for health", contract
number LHSG-CT-2003-503269.
U-S01 Modelling
fission yeast morphogenesis
Attila Csikasz-Nagy 1, Bela Gyorffy 1, Wolfgang Alt 2, John
J. Tyson 3 and Bela Novak 1
1 Department of Agricultural Chenmical Technology, Budapest University of
Technology and Economics, Szt Gellert ter 4., Budapest 1111, Hungary,
Phone: +36/1/4632910, FAX: +36/1/4632598, e-mail: csikasz@mail.bme.hu, Web: http://www.cellcycle.bme.hu/
2 University of Bonn, Bonn,
Germany
3 Virginia Tech, Blacksburg,
USA
Because
of its regular shape, fission yeast is becoming an increasingly important
organism to study cellular morphogenesis. Genetic studies have identified a
great number of proteins that are important to regulate shape changes during
the cell cycle. Most of these proteins interact with either microtubules or actin,
underlining the essential roles these cytoskeletal structures play in cellular
morphogenesis. Here we present a simple model for fission yeast morphogenesis
that describes the interplay between these cytoskeletal elements. An essential
assumption of the model is that actin polymerization is a self-reinforcing
process: filamentous-actin promotes its own formation from globular-actin
subunits via regulatory molecules. Microtubules stimulate actin polymerization
by delivering a component of the autocatalytic actin-assembly feedback loop. We
show that the model captures all the characteristic features of polarized
growth in fission yeast during normal mitotic cycles. We show that all the
major classes of morphogenetic mutants (monoipolar, orb and tea) are natural
outcomes of the model. We categorize the types of growth patterns that can
exist in our model and compared them with experimental observations.
U-S02 Metabolic
quorum sensing: onset of density-dependent oscillations
Silvia De Monte 1,
Francesco d'Ovidio 2,
Sune Danø 3
and Preben Grae Sørensen 4
1 Dept. of Biology, École Normale Supérieure, 46, rue d'Ulm, Paris
F-75005, France, Phone: +33/(0)1/44322342, FAX: +33/(0)1/44323885,
e-mail: demonte@biologie.ens.fr,
Web: http://www.fys.dtu.dk/~silvia
2 École Normale Supérieure,
Paris, France
3 Dept. of Medical
Biochemistry and Genetics, University of Copenhagen, Denmark
4 Dept. of Chemistry,
University of Copenhagen, Denmark
Populations
of oscillating units coupled by diffusion through a homogeneous medium are studied as a
model for cells in a CSTR. In particular, we focus on the dependence of the
collective behaviour on the density of the cell suspension. Both the classical
Kuramoto model and the recent results on "coupling by quorum sensing"
(Garcia-Ojalvo, Elowitz and Strogatz (2004) PNAS 101,10955) indicate that, by
diluting the suspension, the cells should keep their oscillatory behaviour
while desynchronising.
A different scenario could however take place due to the delay introduced in
the coupling by the presence of a medium. In this case, the dilution of the
suspension results into the suppression of oscillations at both population and
individual levels. Such density-dependent phenomenon may be seen as a metabolic
analogous of quorum sensing in bacteria: the amplitudes of the individual
metabolic oscillations can provide each individual cell with information on the
population density and average state of the population.
U-S03 Integration of software tools for the in silico design of metabolic
pathways using flux balance analysis
Ana Sofia Figueiredo 1, Pedro Fernandes 1,
Pedro Pissarra 2 & António Ferreira 3
1 Bioinformatics Unit, Instituto Gulbenkian de Ciencia, Rua da Quinta
Grande, Oeiras 2781-901, Portugal, Phone: +351 21 4407900, FAX: +351
21 4407970, e-mail: sofiafig@igc.gulbenkian.pt 2 Biotecnol SA, Taguspark, Edificio
Inovacao IV nº809, Oeiras 2780-920, Portugal, Phone: +351214220520
Fax:+351214220529 3
Departamento de Quimica e Bioquimica, Faculdade de Ciencias da Universidade de
Lisboa, Campo Grande, Lisboa 1749-016, Portugal, Phone: +351217500076 Fax:
+351217500994
The systems biology approach, where
one can envision the cell as a whole is a step in the direction of narrowing
the gap between the rate of data generation and the speed of analysis. This
embraces the much desired goal of understanding the role of the metabolic
pathways of a determined metabolic network.
This study describes the use of several software tools, integrated to perform
the simulation of a specific metabolic pathway by Flux Balance Analysis (FBA).
This simulator receives as input the stoichiometry, the thermodynamic and
capacity constraints of the metabolic network, and also an objective function.
The stoichiometry and the thermodynamic constraints are represented in the SBML format (Systems Biology Markup Language), whereas all the other
information is represented in a plain text file. The SBML file is parsed using
libSBML, which is a library that can be embedded into an application to read,
write and manipulate files in the SBML format. The text file is parsed using
FLEX, a lexical analyser that generates a C/C++ program that recognizes
specific lexical patterns in the text. With this information, one can construct
a Linear Programming (LP) problem. To solve it, GNU lp-solve is used. It uses
the simplex algorithm and sparse matrix methods for simple LP problems. The
solution provided is a possible flux distribution on the network, that maximises the objective
function. In this work, data from a batch fermentation process (where the host
system is Escherichia coli strain BL21) is incorporated in the model
definition. In the experiment, Acetate secretion, Oxygen Uptake Rate (OUR),
Carbon Evolution Rate (CER) and Biomass production for wild type (wt) and
mutant (mt) strains were determined. The mutant was engineered with a plasmid
to express the human recombinant interleukin 4(IL-4) using pRT as a promotor.
The analysis of the flux distribution for wt and mt is performed for the
maximisation of ATP production, incorporating as capacity constraints the different
data obtained from the experiment. A discrete time analysis was performed,
using the same variables, and assuming a steady state for each time sample.
Comparing FBA results for wt and mt, the induction of protein in the host
system decreases the capacity of producing ATP. The sensitivity of the system
to variations in glucose uptake was also performed. It was shown that, in normal
conditions, mt and wt were robust to variations with an amplitude of 2% and 20%. When the carbon
source is residual, the system shows a higher sensitivity to glucose
variations.
U-S04 Uncovering
the control of the respiratory clock in yeast
Douglas B. Murray and
Hiroaki Kitano
Systems Biology
Institute,, Keio University School of Medicine, 9S3, Shinanomachi Research
Park, 35 Shinanomachi, Shinjuku-ku, Tokyo 160-8582, Japan,
Phone: +81-3-5363-3078, FAX: +81-3-5363-3079, e-mail: dougie@symbio.jst.go.jp
Continuously growing yeast cultures tend to auto-synchronise producing a robust respiratory oscillation (tau circa 40 min). Recently we have
carried out Affymetrix analyses that revealed the majority (>90%) of the
transcriptome oscillates within this timeframe [1]. Here we analyse this data
using a “Fourier focussing” technique in order to derive transcripts that are
closely coupled to the oscillation. The method involved dividing the amplitude
calculated by fast Fourier transformation by the mean of the amplitude for
three oscillation cycles. This ratio equated to the noise of the transcript’s oscillation; where a perfect sine-wave
generates a ratio of one and random data generates a ratio approaching zero.
When the ratio was calculated for the yeast transcriptome and plotted, the
resulting curve showed two gradients. The intercept of these gradients was used
as a noise threshold (ratio of ~0.15; ~1500 genes). The strongly coupled
transcripts above this threshold and phenotypic events were then used to
construct a “clock face”. A network diagram was then constructed using high
quality BIND and transcriptional regulatory networks within Cytoscape [2]. The
resulting network consisted of ~1000 transcripts containing the most highly
conserved aspects of the eukaryotic process, e.g., ribosome, proteasome, DNA synthesis, autophagy, cyclins, amino acid biosynthesis, carbon
metabolism, stress response, respiration, etc. Furthermore two transcriptional
sub-graphs out of phase with each other were identified. CIN5, YAP6, YAP1, PHD1
and ROX1 comprised the core of the sub-graph whose transcripts peaked during
the low respiratory phase and MET4 and RAD59 comprised the sub-graph whose
transcripts peaked during the high respiration phase. The cultures
synchronisation mechanism revolves around the production of acetaldehyde and
hydrogen sulphide [3], which feed into and out of this network via ALD5/ADH2
and SUL2/MET3 respectively. It is concluded that these networks regulate the
respiratory clock within yeast. It is also postulated that this network may
form the centre of an energetic “bowtie” common to all eukaryotes because of
its high conservation among all eukaryotes.
[1] Klevecz RR, Bolen J, Forrest G,
Murray D.B. (2004) Proc Natl Acad Sci
USA. 101:1200-5
[2] Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N,
Schwikowski B, Ideker T. (2003) Genome Res. 13:2498-504. [3] Murray DB, Klevecz RR, Lloyd D. (2003)
Exp Cell Res. 2003 Jul 1;287(1):10-5.
U-P01 Neisserial
DNA uptake sequences: biased distribution and influence on
transformation
Ole Herman Ambur, Stephan Frye, Tonje Davidsen,
Hanne Tuven and Tone Tønjum
Institute of
Microbiology, University of Oslo, Rikshospitalet, Sognsvannsveien 20, Oslo
N-0027, Norway, Phone: +47 23074064, FAX: +47 23074061, e-mail: o.h.ambur@labmed.uio.no
DNA uptake sequences (DUS) are short oligomers that occur in
exceptionally high numbers throughout the genomes of, among others, Neisseria and Haemophilus influenzae. DUS are required for efficient transformation
of DNA in these naturally competent bacterial species 1.
DUS are present both outside and within
annotated coding regions. A previous report has shown that DUS were the most
abundant 9-10 mers within coding regions in the genomes of the human pathogens N. meningitidis,
N. gonorrhoeae and H. influenzae
2. More importantly, a significantly higher density of DUS was found
within genes involved in DNA repair, recombination, restriction-modification and replication
than in any other group of genes. DUS in genome maintenance genes might provide a mechanism for facilitated recovery from DNA damage after
genotoxic stress.
DUS outside genes very frequently occur as
inverted repeats and are likely to be involved in transcriptional termination1.
In order to fully determine the extent of inverted DUS repeats, all the DUS of
the N. meningitidis MC58 genome were collected and investigated. Furthermore, a genome scanner for
terminators was applied to identify intrinsic terminators. The two sets of data
show that about half the N. meningitidis genes were terminated
intrinsically and that half of these terminators contain DUS, making up a
number of DUS that amounts to about half of the total number of DUS in the N. meningitidis
genome. In summary, DUS seems to hold dual functions: mediating binding and
uptake of transforming DNA with a bias towards genome maintenance genes as well as being an
element in transcriptional termination. Currently, the biased distribution of
intergenic DUS as well as the functional role of DUS in the transformation
process are being assessed.
References
1. Smith, H.O., et al., (1999) Res.Microbiol. 150, 603-616
2. Davidsen, T., et al., (2004) Nucl. Acids Res. 32:1050-1058
U-P02 Gene
expression and adaptive responses of in situ fermentation
Herwig Bachmann, Michiel Kleerebezem and Johan
E. van Hylckama Vlieg
NIZO food
research, Kernhemseweg 2, Ede 6718 ZB, The Netherlands, Phone: ++31 318
659 668, FAX: ++31 318 650 400, e-mail: Herwig.Bachmann@nizo.nl, Web: www.nizo.nl
Lactic acid bacteria
(LAB) belong to the most important microorganisms used in food fermentation. In
recent years several genomes of LAB have been sequenced and the big task ahead
in the post-genomic era is to elucidate the function, responses and interaction
of genes and their products under various environmental conditions. Lactococcus
lactis is an important component in dairy starter cultures and is one of
the best characterised LAB species. Numerous studies carried out under
laboratory conditions have resulted in detailed knowledge on the physiology and
molecular biology of L. lactis. Yet, little is known about its response
to complex environments during in situ fermentation. Recombinant in
vivo expression technology (RIVET) allows studying the genomic response of
an organism in situ(1). We will apply RIVET to L. lactis which
will help to increase our understanding of the genomic response during e.g. the
ripening of cheese. Currently adaptions are applied to the RIVET tools which
will allow the high throughput screening of obtained target sequences under
various conditions in vitro. This will lead to large functional datasets
which will contribute to the further understanding of the cellular processes
and responses of lactic acid bacteria. Eventually knockout and/or
overexpression studies of relevant targets will allow their further evaluation.
References :1. Bron PA et al. J.Bacteriol. 2004 Sep;186:5721
U-P03 Metabolic
functions of duplicate genes in Saccharomyces cerevisiae
Lars M. Blank 1, Lars Küpfer 2 and Uwe Sauer 2
1 Department of Biochemical and Chemical Engineering, University
Dortmund, Emil-Figge-Str. 66, Dortmund 44227, Germany,
Phone: +49/2317/7557383, FAX: +49/2317/7557382, e-mail: lars.blank@bci.uni-dortmund.de,
Web: http://btwww.bci.uni-dortmund.de/
2 Institute of
Biotechnology, ETH Zürich, 8093 Zürich, Switzerland
The role
of duplicate genes in Saccharomyces cerevisiae is still not known, despite
exhaustive genome data of hemiascomycetous yeast
nowadays available. In contrast to previous works that favored isolated
functions we here propose a dispersed array of functions for the occurrence of
duplicate genes in the yeast metabolism. These include back-up activity, hence
complementation of a genetic dysfunction by a duplicate gene,
regulatory function, enhanced enzyme synthesis by parallel transcription on
multiple gene copies and a specialized role of enzymes within a duplicate
family. The various functions were analyzed with genome scale phenotype testing
and C 1 3 constraint based flux analyses on 5 different growth conditions. In
a highly integrative approach computational modeling was used to overcome the
lack of an exhaustive library of complete duplicate gene family mutants. For
our analysis we used a reconciled full genome scale metabolic model that was
validated with in vivo data for the case of singleton gene mutants. It was such
possible to predict the phenotype for the deletion of complete duplicate gene
families and to analyze the inherent network topology.
Interestingly, for a significant number of duplicate genes no particular function
could be attributed and phylogenetic clustering suggests that a fraction of
these are “snapshots in time” and might be lost during evolution. This study,
for the first time, quantifies the role of duplicate genes in the yeast
metabolism and highlights that duplicate genes cover a wide spectrum of
functions, which may explain the high number of duplicate genes present in S.
cerevisiae.
Acknowledgment Lars M. Blank gratefully acknowledges
financial support by the Deutsche Akademie der Naturforscher Leopoldina (BMBF-LPD/8-78).
U-P04 Experimental
manipulation and mathematical modeling of arginine biosynthesis in Escherichia
coli
Marina Caldara 1, K. Verbrugghe 2, L. De
Vuyst 3, M. Crabeel 1, G. Dupont 4,
A. Goldbeter 4 and R. Cunin 1
1 Microbiology, Vrije Universiteit Brussel, Pleinlaan 2, Brussel 1050,
Belgium, Phone: +32/02/6291343, FAX: +32/02/6291345, e-mail: mcaldara@vub.ac.be 2 IMDO, VUB, Pleinlaan2, Brussel,
Belgium
3 IMDO,VUB, Pleinlaan2,
Brussel, Belgium 4
Theoretical Chronobiology, ULB,B-1050, Brussel, Belgium
The
different elements of the arginine pathway and the principal regulatory
mechanisms operating at genetic and enzymatic levels in E.coli are identified, allowing to integrate them in
an initial mathematical model, using the XPPAUT software. The first enzyme of
the pathway is N-acetylglutamate synthase, (NAGS) which acetylates L-glutamate
using Acetyl-coA. NAGS is genetically repressed by the arginine repressor argR
in the presence of arginine and feedback-inhibited by arginine. Carbamyl
phosphate synthase (CPS) is at a branch point between the pryrimidine and
arginine biosynthetic pathways: it provides carbamyl phosphate (CP) for both.
It is subject to a cumulative genetic repression by arginine and pyrimidines
and its activity is inhibited by UMP and activated by ornithine. CP is utilized
by ornithine transcarbamylase (OTC) to form citrulline for arginine
biosynthesis and by aspartate transcarbamylase (ATC) in the first step of
pyrimidine synthesis. OTC is subject to genetic repression by arginine. ATC is
repressed by pyrimidines and its activity is subject to feedback-inhibition by CTP
and UTP and to activation by ATP.
Monitoring these enzymatic activities in different balanced growth conditions
showed that OTC activity is strongly repressed by the presence of arginine in
the medium, but is increased by the presence of uracil; similarly, uracil
increases the NAGS activity. This must reflect a physiological derepression,
since this effect is not seen in an argR- strain.
The hypothesis that uracil limits the synthesis of CP and its availability for
arginine synthesis is supported by measurements of the cellular pools of UMP,
ornithine, citrulline and arginine, using HPLC and LC-MS, and by the properties
of a strain in which a very low constitutive synthesis of CP has been
engineered.
U-P05 Comparative
metabolomics of Saccharomyces yeasts
Robert Davey 1, G Lacey 1, DA MacKenzie
1, M Defernez 1, FA Mellon 1,
K Huber 2, V Moulton 2 and IN Roberts
1
1 National Collection of Yeast Cultures, Institute of Food Research,
Colney Lane, Norwich NR4 7UA, UK, Phone: +44 1603 255000, FAX: +44
1603 458414, e-mail: robert.davey@bbsrc.ac.uk
2 School of Computing
Sciences, University of East Anglia, Norwich NR4 7TJ, UK
Comparative
analysis of gene sequence and gene content
datasets from the genomes of closely-related yeast species reveals clear evidence of reticulate
evolution (Holland et al 2004, Savva et al 2004). We are extending this work to
examine network-like behaviour in “metabolic footprinting” datasets (Allen et
al 2003, Lacey et al 2004). Networks derived from genomic and metabolomic data
will be compared using combinatorial methods. The development of new
dissimilarity measures will draw on previous work utilising similarity theory
(Collins et al 2000, Dress et al 2002). This will reveal links between
genotypic and metabolic variation and guide hypothesis-driven research into
molecules and mechanisms responsible for the observed differences.
Understanding such links will enable the results of comparative analyses to
inform metabolic modelling.
References
Allen J, Davey HM, Broadhurst D, Heald JK, Rowland JJ,
Oliver SG, Kell DB (2003) High-throughput classification of
yeast mutants for functional genomics using
metabolic footprinting. Nature Biotechnology 21, 692-696
Collins L, Moulton V, Penny D (2000) Use of RNA secondary structure for studying the
evolution of RNase P and RNase MRP. Journal of Molecular Evolution 51, 194-204
Dress A, Huber K, Moulton V (2002) An explicit computation of the injective
hull of certain finite metric spaces in terms of their associated Buneman
complex. Advances in Mathematics 168, 1-28
Holland B, Huber K, Moulton V, Lockhart P (2004) Using consensus networks to
visualize contradictory evidence for species phylogeny, Molecular Biology and
Evolution, 21, 1459-1461
Lacey G, Aroso M, MacKenzie DA, Fuller L, James S, Bond C, Defernez M, Mellon F, Roberts IN
(2004) Genome rearrangement and comparative metabolomics. Consortium for
Post-Genome Science 2nd Conference: “Genomes to Systems”, 1-3 September 2004,
Manchester, UK
Savva G, Davey R, Dicks J and Roberts IN (2004) A maximum likelihood framework
for phylogenetic analysis of gene content datasets derived from comparative
genome hybridisation experiments using microarrays.
Bioinformatics (in preparation)
U-P06 Metabolic
network analysis in six microbial species
Tobias Fuhrer, Eliane Fischer and Uwe Sauer
Department of
Biology, ETH Zürich, Institute of Biotechnology, HPT E55, Wolfgang-Pauli-Str.
16, Zürich CH-8093, Switzerland, Phone: +41 1 633 67 09, FAX: +41 1
633 10 51, e-mail: fuhrer@biotech.biol.ethz.ch,
Web: http://www.biotech.biol.ethz.ch/sauer/
The
structurally conserved and ubiquitous pathways of central carbon metabolism
provide building blocks and cofactors for the biosynthesis of cellular
macromolecules. The relative use of pathways and reactions, however, varies
widely between species and conditions and some are not used at all. Based on
stoichiometric models of the central carbon metabolism, we here identify the
network topology of glucose metabolism and its in vivo
operation by quantification of intracellular carbon fluxes from 1 3C-tracer experiments. Specifically,
we investigated Agrobacterium tumefaciens, two pseudomonads, Sinorhizobium
meliloti, Rhodobacter sphaeroides, Zymomonas mobilis and Paracoccus
versutus, which grow on glucose as the sole carbon source, represent
fundamentally different metabolic life styles (aerobic, anaerobic,
photoheterotrophs and chemoheterotrophs), and are phylogenetically distinct
(firmicutes, gamma-proteobacteria and alpha-proteobacteria). When compared to
the model bacteria Escherichia coli and Bacillus subtilis, metabolism in the investigated species differed significantly in
several respects: i) the Entner-Doudoroff pathway was the almost exclusive
catabolic route, ii) the pentose phosphate pathway exhibited exclusively
biosynthetic functions, in many cases requiring also flux through the non-oxidative branch, iii) all
aerobes exhibited fully respiratory metabolism without significant overflow
metabolism, and iv) all aerobes used the pyruvate bypass of the malate
dehydrogenase reaction to a significant extent. Exclusively, Pseudomonas
fluorescens converted most glucose extracellularly to gluconate and
2-keto-gluconate. Overall, the results suggest that metabolic data from model
species with extensive industrial and laboratory history are not representative
for microbial metabolism, at least not quantitatively.
U-P07 Retrograde
response to mitochondrial dysfunction is separable from Tor1/2 regulation of
retrograde gene expression.
Sergio Giannattasio 1, Zhengchang Liu 2 and Ronald Butow 2
1 Istituto di Biomembrane e Bioenergetica, Consiglio Nazionale delle
Ricerche, Via Amendola 165/A, Bari I-70126, Italy, EU,
Phone: +390805443316, FAX: +390805443317, e-mail: s.giannattasio@ibbe.cnr.it 2 Department of Molecular Biology,
University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd.,
Dallas, TX 75390-9148, USA
To obtain insight into how
retrograde (RTG) and target of rapamycin (TOR) signaling pathways integrate in
response to specific stimuli in yeast cells, rapamycin sensitivity of the expression of the RTG target
gene CIT2 and of two nitrogen catabolite repression (NCR) sensitive genes, GLN1 and DAL5, has
been analyzed in rho+ and rho0 cells. Rapamycin treatment
of rho+ cells caused a 60-, 6- and 8-fold increase in the expression
of CIT2, GLN1 and DAL5 expression, respectively. In rho0 cells CIT2
expression is upregulated 13-fold compared with rho+ cells.
Rapamycin treatment of rho0 cells increased CIT2 expression to the
same level observed in rapamycin treated rho+ cells. Mutations of
LST8, a negative regulator of the RTG pathway and a component of TOR1 and TOR2
complexes, upregulated CIT2 and GLN1 but not DAL5 expression, which was induced
by rapamycin treatment in wild type and lst8 mutant cells. Rapamycin-activated
DAL5 expression was virtually abolished by deletion of either GLN3 or GAT1 GATA
transcription factors. Subcellular localization analysis of Gln3p-GFP and
Gat1p-GFP showed that Gat1p-GFP is completely excluded from the nucleus in
lst8-5 mutant cells. The role of GATA factors in the RTG-dependent retrograde
response was also studied. In both rho+ and rho0 wild
type and gln3delta yeast cells, CIT2 expression was not affected by GLN3
deletion in rho+ cells with or without TOR inhibition by rapamycin,
whereas GLN3 deletion virtually abolished CIT2 upregulation due to retrograde
response in rho0 cells. TOR inhibition by rapamycin restored CIT2
expression at similar levels both in respiratory competent and deficient wild
type and mutant strains. Reintegration of GLN3 by transformation of rho0
gln3delta strain with pRS416-GLN3-GFP restored CIT2 expression. Thus the
retrograde response in respiratory deficient cells follows a different pathway
from that in TOR-regulated retrograde gene expression.
Acknowledgment. This work was supported by
grants GM 22525 from the NIH and I-0642 from The Robert A. Welch Foundation.
U-P08 Unravelling
new metabolic metworks in LAB via the thioredoxin system
Mariela Hebben-Serrano 1, Eddy Smid 1 and Willem
M. de Vos 2
1 Processing, Nizo
Food Research, PO Box 20, Ede BA-6710, The Netherlands, Phone: +31 318 659
647, FAX: +31 318 650 400, e-mail: Mariela.Hebben@nizo.nl, Web: www.nizo.nl 2 Wageningen Centre of Food and Sciences,P.O. Box 557,6700
AN Wageningen
Co-factors
are produced by several micro-organisms and are often also essential components
in human diet. The biosynthesis of these metabolites involves complicated
metabolic pathways which represent a challenge for genetic engineering.
Furthermore, identifying all the pathways where co-factors play a role in the
cell metabolism requires an integrated approach. Hence, functional
genomics-based approaches are being used to evaluate the effect of modulating
production of a co-factor on the overall metabolism and functionality of lactic acid bacteria (LAB). The two LAB
strains: Lactococcus lactis (2.3 Mb) and Lactobacillus plantarum (3.3
Mb) are used as the model bacteria, because of their available genome sequence, their scientific paradigm status,
and the vast experience with these LAB strains in both practical application
and biotechnological development.
The co-factor of choice is the redox-mediating thiol, thioredoxin. This
oxido-reductase is involved in a broad spectrum of cellular processes such as
DNA synthesis, protein folding, stress-response, and detoxification.
In food this component can act as an anti-oxidant for stabilization of starter
cultures and food proteins. Nevertheless, to be functional thioredoxin needs
first to be reduced by thioredoxin reductase (TRXB) in a NADPH dependent
reaction. Due to the essential role the reductase plays in the thioredoxin
system, TRXB was further investigated.
Genetic modifications in the thioredoxin reductase (trxB) gene have lead to engineered strains in LAB will
either increased levels of thiol or with no thiol at all. At the same time,
enzyme assays have been optimized to quantify the thioredoxin and thioredoxin
reductase activity. At the moment we have a construct that can over express the
reductase activity approximately six times more than in the wild type.
Furthermore, performed phenotypic characterization suggests that the over
expression of thioredoxin reductase leads to better resistance towards diamide,
a disulfide stress agent.
Currently, we are monitoring the mRNA response of these genetically modified
LAB strains under different oxidative stress conditions. In the near future, we
will study the impact of these modifications at the metabolome level. The data
of these studies will be linked using metabolic-modelling tools such as
SimphenyTM. Undoubtedly, this functional genomics approach will
contribute to our understanding of the crucial role of the thioredoxin system
in Lb. plantarum.
U-P09 The
regulatory circuitry of arabinases in Bacillus subtilis
José M. Inácio and
Isabel de Sá-Nogueira
Microbial
Genetics, Instituto de Tecnologia Quimica e Biologica, Av. da Repuplica, Apt
127, Oeiras 2781-901, Portugal, EU, Phone: +351 21 4469525, FAX: +351
21 4411277, e-mail: jinacio@itqb.unl.pt
Hemicellulases
are a diverse group of enzymes capable of hydrolyze hemicellulose, the
second-most abundant renewable biomass polymer in nature. This fraction of
plant cell walls comprises a complex mixture of xylans, arabinans, galactans
and mannans. Enzymes responsible for degrading plant cell wall polysaccharides
have many applications in different fields such as food technology, nutritional
medical research, mobilization of plant biomass, plant biochemistry and organic
synthesis. The saprophytic endospore-forming Gram-positive bacterium Bacillus
subtilis produces two major enzymes,
alpha-L-arabinofuranosidases (EC 3.2.1.55) and endo-alpha-1,5-arabinanases (EC
3.2.1.99), involved in degradation of the homoglycane arabinan, which are
capable of releasing arabinosyl oligomers and arabinose from plant cell walls.
In B. subtilis, arabinan is degraded by at least two
extracellular hemicellulases, AbnA and YxiA. The resulting products, arabinose,
arabinobiose, arabinotriose, and arabinooligosaccharides, are transported by
different systems. Arabinose enters the cell mainly through the AraE permease,
and the uptake of arabinose oligomers occurs most likely via AraNPQ, an ABC
type transporter. These latter products might be further digested
intracellularly by AbfA and Xsa. In previous studies our group showed that
expression of the arabinases genes is regulated at the transcriptional level:
(i) induction in the presence of arabinose and arabinan, (ii) repression by the
transcription factor AraR, (iii) carbon catabolite repression by glucose, and (iv) temporal regulation (1). Here, we integrate our current
knowledge concerning the regulatory network of arabinases with ongoing studies
on the functional analysis and biochemical characterization of these enzymes.
(1) Raposo, M. P., et al. (2004) J Bacteriol 186(5):1287-96.
U-P10 Extending
life by alternative respiration?
Alexander Kern, Franz Hartner and
Anton Glieder
Institute of
Molecular Biotechnology, Graz University of Technology, Petersgasse 14, Graz
A-8010, Austria, Phone: +43/3168734077,
FAX: +43/3168734071, e-mail: a.kern@tugraz.at
Alternative
oxidase transfers electrons directly from the ubiquinol pool in mitochondria to
oxygen, allowing cell respiration in presence of complex III and IV inhibitors
like antimycin A or cyanide. Electron transfer by alternative oxidase is not
coupled with proton transfer across the mitochondrial membrane,
thereby uncoupling the supply of small metabolic intermediates by the central
metabolic pathway from energy production in the cell. Alternative oxidase
is present in mitochondria of plants, many fungi and a few, mostly
Crabtree-negative yeasts, but not in Pichia angusta (Hansenula
polymorpha) and Saccharomyces cerevisiae. Alternative oxidase has multiple functions in different organisms. It
is involved in stress answers, in programmed cell death, maintenance of the
cellular redox balance, and also citric acid accumulation in Aspergillus
niger.
We isolated the alternative oxidase gene from the methylotrophic yeast Pichia
pastoris in order to study its effects on the cellular energy content,
respiratory activity, its protective role against oxidative stress. Our results
indicate the importance of an exact regulation of the alternative oxidase due
to its impact on many cellular functions, especially cell viability.
U-P11 Systematic
computational modelling reveals a key operating principle of TOR signalling in yeast
Lars Kuepfer 1, Matthias Peter 2,
Jörg Stelling 3 and Uwe Sauer 1
1 Institute of Biotechnology, ETH Zurich, HPT D78, Zurich CH 8093,
Switzerland, Phone: 0041/1/6333758, FAX: 0041/1/6331051,
e-mail: kuepfer@biotech.biol.ethz.ch,
Web: http://www.biotech.biol.ethz.ch/sauer/
2 Institute of Biochemistry,
ETH Zurich, CH-8093 Zurich
3 MPI for Complex Technical
Systems, D-39106 Magdeburg
The TOR (target of rapamycin) pathway is a highly
conserved signalling route that couples nutrient availability to cell growth.
Many of its components and their interactions are characterized, but, as a
typical cellular regulatory circuit, the operating principles are not yet fully
understood, in particular owing to conflicting hypotheses and experimental
data. For TOR signaling in Saccharomyces cerevisiae, we report a quantitative analysis that systematically casts molecular
hypotheses into a family of computational models and evaluates these against
experimental observations. In contrast to the prevailing view of a de novo
assembly of type 2A phosphatase complexes (PP2As), this approach
proposed competitive binding of the phosphatases as a key signaling mechanism.
Subsequent experimental analysis confirmed this prediction, thereby settling,
for instance, apparent discrepancies between rapid PP2A activation an much
slower substrate dephosphorylation. The comparative computational biology
approach is broadly applicable, and can help unravelling the operation of
incompletely characterized pathways.
U-P12 Dynamic
on-line investigation of lactic acid bacteria
Ann Zahle Larsen 1, Lars Folke Olsen 1 and Frants
Roager Lauritsen 2
1 Celcom, Biochemistry and molecular biology, University of Southern
Denmark, Campusvej 55, Odense M 5230, Denmark, Phone: +45/65502486,
FAX: +45/6550 2467, e-mail: zahle@odense.kollegienet.dk
2 Department of Chemistry,
Copenhagen University,Universitetsparken 5, DK-2100 Copenhagen
Lactic
acid bacteria (LAB) play a fundamental role in industrial food fermentation
processes. In addition, their relatively simple sugar metabolism and high
degree of auxotrophy present them as ideal model organisms for studying
glycolysis and pyruvate derived pathways. Many studies have been conducted on
LAB, but so far there are no unambiguous answers to the question of what
controls the rate of glycolysis in this simple organism. A common feature of
the investigations is that they maintain a constant pH during experiments of
typically 6.5, thus counteracting the large acid production by the glucose consuming bacteria. Restraining
pH allows for experimental simplifications, but also removes one of the key
regulating factors of the system. We therefore find it crucial to be able to
measure key metabolites continuously and quantitatively, during the large pH
drop (from 6.7 to 4) normally observed during a standard milk fermentation
(Oestlie, H ,et al., 2003, Int J Food Microbiol, 87, 17). By correcting
for pH changes instead of preventing them, we obtain data, which reflects the
natural parameter space of the system. This allows us to
investigate glycolysis under the physiological conditions, which govern the
glycolysis of a LAB performing its primary function, namely acidification.
In our laboratory we focus on development and use of high throughput
quantitative on-line methods for continuous monitoring of relevant biological
compounds during sugar fermentation. These methods include Membrane Inlet Mass
Spectrometry of gasses and volatiles, fluorescence spectroscopy of NADH and
amperometric glucose electrodes. The methods have in common the ability to
generate almost continuous data without any sample treatment or undesired
perturbations to the system. These methods produces data well suited for
kinetic modelling. They ensure a time scale that
allows the results of minor perturbations to be studied in detail and exclude
regulation on other levels than that of the enzymes. This way we hope to reveal
the still elusive regulatory network governing glycolysis of this simple
organism on an enzymatic level. Using the systems biology approach, the
investigation and the resulting model can then hopefully be extended and merged
with knowledge of proteomics, genomics etc. obtained under similar conditions
in other laboratories.
U-P13 A
dynamic model of cAMP signal transduction in yeast
Dirk Mueller 1, Helena Díaz Cuervo 2,
Luciano Aguilera-Vázquez 1, Klaus Mauch 1 and Matthias Reuss 1
1 Institute of Biochemical Engineering, University of Stuttgart,
Allmandring 31, Stuttgart D-70569, Germany, Phone: +49-(0)711-685-7532,
FAX: +49-(0)711-685-5164, e-mail: mueller@ibvt.uni-stuttgart.de,
Web: www.ibvt.uni-stuttgart.de
2 Centro de Investigación
del Cáncer, Universidad de Salamanca, Spain
Yeast
cells possess a number of signaling pathways to integrate information about
nutrient supply with cellular growth and proliferation. The signaling route
mediated by cyclic AMP (cAMP) and
protein kinase A (PKA) regulates a large number
of targets both at the posttranslational and transcriptional level in response
to changes in carbon source availability. By affecting both metabolic processes
and the cell cycle machinery, it also serves to
coordinate cell growth and division.
Measurements were performed in synchronous cultures and in oscillating
continuous cultures of Saccharomyces cerevisiae to analyze the cell-cycle
dynamics of cAMP and energy metabolism. A modular single-cell model
integrating cAMP signaling with descriptions of the cell cycle machinery and
central carbon metabolism is currently under development. This single-cell
model will permit to simulate cellular behavior resulting from the joint action
of metabolic and signaling networks during the yeast cell cycle. The present
contribution focuses on two models of cAMP signal transduction, which can be used
as exchangeable submodules in the integrated model.
On the basis of an extensive literature survey two dynamic models of the cAMP
signaling pathway were developed, both of which provide a comprehensive
description of the current knowledge, but differ in the level of detail. They
account for stimulation of adenylate cyclase via Ras and a GPCR system, cAMP
destruction by phosphodiesterases, (in)activation of PKA, and for the negative
feedback exerted by PKA on its own activity. Results of the above-mentioned
experiments were employed in combination with literature data and stability
constraints to estimate model parameters. As a starting point, protein levels
determined in a genome-wide analysis [1] served as estimates of the initial values of model
species. Simulation results of both the small-scale model (20 reactions) and
the large-scale model (400 reactions) will be presented and compared to
experimental findings.
The models provide a basis to address open questions regarding the underlying
network structure and dynamic behavior of this signaling pathway. Plus, they
can serve as a tool to identify suitable experimental conditions to efficiently
discriminate between alternative hypotheses. Future work aims at incorporating
spatial information and transcriptional regulation of
key components of the cAMP pathway into the model.
[1] Ghaemmaghami, S., et al. (2003)
Nature 425:737-741.
U-P14 Natural
sweetening of food products: engineering Lactococcus lactis for glucose production
Wietske A. Pool 2, Ana
R. Neves 1, Jan Kok 2, Helena Santos 1 and Oscar P. Kuipers 2
1 ,
Instituto de Tecnologia Química e Biológica, Rua da Quinta Grande,
6, Oeiras 2780-156, Portugal, Phone: +351214469824,
FAX: +351214428766, e-mail: arn@itqb.unl.pt
2 Molecular Genetics, University
of Groningen, Haren, The Netherlands
Lactic
acid bacteria (LAB) are used worldwide in the production of fermented dairy
products. Lactococcus lactis is generally recognized as a model
organism, characterized by the simplicity of its metabolism and the
availability of a wide range of genetic tools as well as the availability of a
complete genome sequence. These features make L.
lactis a suitable object for metabolic engineering strategies aiming at the
improvement of food quality and human health. Production of glucose from lactose to be used as a
natural sweetening requires engineering the catabolism of glucose. A strain
that ferments only the galactose moiety should be impaired in glucose transport and phosphorylation. Hence, a double mutant
carrying specific deletions of the glucose PTS (ptnABCD) and glucokinase
(glk) was made. However, this strain could still grow on glucose.
Determination of enzymatic activities coupled to in vivo NMR studies of glucose metabolism suggested the
presence of a second PTS system with a clear preference for alpha-glucose. A
DNA microarray experiment revealed up-regulation
of the genes coding for the cellobiose-PTS (ptcBAC). Indeed, deletion of
ptcBA in the double mutant strain abolished growth on glucose. The
lactose plasmid was introduced in triple knock-out mutant and in a control strain and lactose metabolism studied using in
vivo NMR. In the control stain, both moeities were completely converted to
lactate and a transient accumulation of galactose was observed. In contrast, in
the mutant the glucose moiety is fully recovered as free glucose. Therefore, a L.
lactis strain can be used to produce glucose as a natural sweetener in milk
products was obtained. We showed that deletion of the genes coding for
glucokinase, mannose-PTS and cellobiose-PTS is required and sufficient to
completely block glucose metabolism and we identified the cellobiose-PTS as a
glucose transporter in L. lactis.
U-P15 Adaptative
response of the central metabolism in Escherichia coli to quantitative modulations of a single enzyme: glucose-6-phosphate dehydrogenase
Cécile Nicolas, Fabien Létisse and Jean-Charles Portais
Laboratory
Biotechnology-Bioprocessing, Institut National des Sciences Appliquées, 135
avenue de Rangueil, Toulouse F31077, France, Phone: +33 561 559 399,
FAX: +33 561 559 689, e-mail: cecile.nicolas@insa-toulouse.fr
Microorganisms have an efficient
capacity for adapting their metabolism in response to genetic or environmental
changes, and understanding metabolic robustness has become an emergent issue.
Part of the robustness originates from the network organization of metabolic
systems, where the interplay between all available biochemical reactions
provides alternative mechanisms for compensating the perturbations. Recently, 1 3C-Metabolic Flux Analysis ( 1 3C-MFA)
has been applied to E. coli knock-out mutants lacking key enzymes to
determine the phenotypic effects of structural changes in the metabolic
network, providing further evidences for compensatory phenomena. The aim of our
on-going work is to understand how the central metabolism in E. coli
responds to quantitative alterations at a specific key-point of the metabolic
network. The glucose-6-phosphate dehydrogenase (G6PdH),
a key enzyme in the central metabolism for which the effects of deleting the
gene (zwf) has been already described (Zhao, J, et
al.,(2004),Metab Eng., 6, 164) was chosen as the target. To this aim we
have generated a set of expression mutants, i.e. mutants having each a fixed
level of expression of the zwf gene. Four different levels of
expression, leading respectively to G6PdH activity 2;2,9;5,7 and 14 times
higher than in the WT strain, have been obtained. For each mutant,
transcriptomics analysis will be carried out and compared to both the zwf-
and WT strains to detect changes in the network structure, and the distribution
of fluxes will be measured using 1
3C-MFA. The flux maps obtained for the various strains will be compared to evaluate
the quantitative response of the central metabolic network to imposed and
increased G6PdH activity. Metabolic control analysis will be applied to provide insights onto the sensitivity
of the measurable metabolic fluxes to G6PdH activity. Combination of
transcriptomics and fluxomics approaches will provide information on the nature
and extent of the compensatory mechanisms. Because the activity of a single
enzyme is tuned at different levels in knock-out and expression mutants, this
investigation provides a situation that mimicks gene-level regulation of
metabolism.
U-P16 Stress
induced by weak organic acids in Saccharomyces
cerevisiae
Rick Orij, Jarne Postmus, Gerco van Eikenhorst,
Stanley Brul and Gertien Smits
Molecular
Biology & Microbial Food Safety, University of Amsterdam, Nieuwe Achtergracht 166, Amsterdam
NL-1018 WV, The Netherlands, EU, Phone: +31/20/525/5027,
FAX: +31/20/525/7056, e-mail: orij@science.uva.nl
Sorbic
acid, a weak organic, is the most widespread food preservative used in the
industry. Yeast and other fungi are, to a certain extent, able adapt to this acid and resume growth in the
presence of the highest concentrations allowed in foods. This can result in
substantial economic losses. Quite a lot is know how yeast responds to sorbic acid stress at the genetic
level from transcript data and using the yeast knockout collection, but we
still do not understand why the cell arrests growth, and why, after adaptation,
it can resume growth. Therefore, to understand the mechanisms of growth
limitation and adaptation we study yeast at the level of gene expression, protein composition, but mainly at the metabolic
level. By calculating energy generating capacity, we try to map the cost
and benefit of the various aspects of the stress response. In practice this
means we determine metabolic fluxes, ATP/ADP
ratios and ultimately try to construct a mathematical model of the response to
the stress. We think that the combination of physiological, protein and
gene-expression data will provided us with a more complete understanding of the
mechanisms underlying the stress and adaptation.
U-P17 Adaptation
of yeast glycolysis to temperature changes
Jarne Postmus 1, Jildau Bouwman 2,
Rick Orij 1, Stanley Brul 1 and
Gertien Smits 1
1 Molecular Biology & Microbial Food Safety, University of Amsterdam, Nieuwe Achtergracht 166, Amsterdam NL-1018,
The Netherlands, Phone: +31/20/5255027, FAX: +31/20/5257056,
e-mail: jpostmus@science.uva.nl
2 Department of Molecular
& Cell Physiology, De Boelelaan 1085, Amsterdam
Organisms
are able to respond to their environment to maintain homeostasis. A
well-studied environmental parameter is temperature, which exerts a complex
combination of effects on the cell. We have studied the relation of growth rate
and temperature of S. cerevisiae under well-defined conditions
and focused on the effect on glycolysis. We selected a temperature from the
supra-optimal side of the temperature growth rate graph and shifted a culture
from the reference temperature to a higher temperature (38ºC) to examine
quantitatively the effects on glycolytic flux during adaptation. We observed a twofold
increase in glycolytic flux. Now we want to find out how this flux increment is
regulated; is it regulated metabolically, through temperature effect on the
glycolytic enzyme reaction rates, through variations in metabolite
concentrations, or is it regulated hierarchically, at the level of mRNA levels,
protein levels, through the expression of alternative
iso-enzymes. We will use regulation analysis to determine this for all enzymes
in the glycolytic pathway, and will then fully study interesting cases on all
levels, from gene to function...and back.
U-P18 Progressive
adaptation of Lactococcus lactis to stress
Emma Redon, Pascal Loubière and
Muriel Cocaign-Bousquet
Laboratoire
Biotechnologie Bioprocédés, Institut National des Sciences Appliquées, 135
avenue de Rangueil, Toulouse 31077, France, Phone: +33 5 61 55 94 18,
FAX: +33 5 61 55 94 00, e-mail: redon@insa-tlse.fr, Web: http://www.insa-toulouse.fr/lbb
The
bacteria Lactococcus lactis, recently sequenced and recognized as the
model organism for the study of lactic acid bacteria, is encountered in
various environments (industrial dairy media, natural vegetal ecosystems, human
digestive tract) in which it is submitted to multiple growth-limited stresses
(nutrient starvation, thermal, oxidative, acidic or osmotic stresses).
Surprisingly, mechanisms of adaptation against these adverse environments are
poorly characterized and fundamental knowledge is still missing. As no
alternative sigma factor involved in stress adaptation has been identified in
its genome, regulation of response towards environment may differ in L. lactis
from what is observed in the Gram+ model Bacillus subtilis.
To study the L. lactis response to adverse growth conditions, a vertical
approach ranging from transcription to phenotype is followed. During controlled
cultures (pH, temperature, medium composition), a stress is progressively
imposed to observe and analyse the dynamic adaptation of L. lactis.
Throughout the culture, whole-genome messengers expression is measured and
integrated in a global physiologic analysis based on particular metabolic
indicators (metabolic rates, enzyme concentrations, amino acid consumptions).
In order to evaluate the mRNA turnover impact on the overall regulation, mRNA
half-lives are also measured and analysed with the transcriptome data (Even,
S., et al., (2002), Mol Microbiol, 45, 1143). This complete approach allows
to quantify the various determinants of the response regulation and identify
the major bottlenecks (Even, S., et al., (2003), Microbiology, 149, 1935).
One of the major stress investigated in L. lactis was glucose starvation (Redon, E., et al., (2004), J Bacteriol, submitted). After a rapid exponential phase associated to nutrient excess and a
short deceleration phase due to decreasing glucose concentration, succeeded a
stationary phase characterized by glucose starvation. The transcriptional
response triggered early in the fermentation, notably before glucose
exhaustion, and varied slightly once starvation was established. 704 genes (30%
of genome) were shown to be involved in carbon starvation adaptation. Their
functional analysis allowed to identify different types of response, providing
a deep understanding of the mechanisms involved in stress adaptation.
U-P19 Evolutionary
conservation and divergence of fungal promoter sequences
Riccarda Rischatsch, Sylvia Voegeli and
Peter Philippsen
Applied
Microbiology, Biozentrum Basel, Klingelbergstr. 50-70, Basel CH-4056, Switzerland,
Phone: +41 61 267 14 89, FAX: +41 61 267 14 81, e-mail: Riccarda.Rischatsch@unibas.ch
The
recently sequenced and fully annotated genome of the filamentous fungus Ashbya gossypii
revealed striking similarity to the baker's yeast Saccharomyces
cerevisiae. 90% of A. gossypii genes share homology and also a substantial
degree of synteny (gene order conservation) with S.
cerevisiae. Although both organisms originate from the same ancestor (carrying
about 5000 protein coding genes), the evaluation of
synteny was complicated by the fact that their evolutionary paths included not
only about 300 translocations and inversions but also a whole genome
duplication in the S. cerevisiae
lineage followed by loss of 4000 genes. As a consequence the alignment of the
A. gossypii genes with homologous S.
cerevisiae genes results in many synteny clusters in which one A. gossypii
chromosomal region aligns with two chromosomal regions of S. cerevisiae. The
clusters themselves are made of gene regions displaying relaxed (incomplete)
and stringent (complete) synteny. The latter is found in many small regions of
up to eleven genes which, very importantly, are not interrupted by end points
of rearrangements. Thus, these regions are particularly suitable for investigations
of evolutionary conservation and divergence of syntenic sequences which started
diverging over 100 million years ago. In the past, most studies of syntenic
regions looked into conservation and divergence of open reading frames (ORFs)
and the proteins they encode. We have started an investigation of evolutionary
selection regarding size and sequence of inter-ORF regions. A detailed
discussion of the subject will be presented taking into account DNA-binding
sites of transcription factors, transcription start and terminator sites and
inter-ORF lengths discerning between bidirectional or unidirectional promoters
and pure terminator-bearing inter-ORFs.
U-P20 A
Systems Biology approach for the optimization of recombinant protein production in E. coli
Eugénio Ferreira and Isabel Rocha
Dept. Eng.
Biologica, Universidade do Minho, Campus de Gualtar, Braga 4710-057, Portugal,
Phone: +351 253 604 407, FAX: +351 253 678 986, e-mail: ecferreira@deb.uminho.pt, Web: http://www.deb.uminho.pt/ecferreira
Escherichia coli has been the organism of choice
for the production of many recombinant proteins with high therapeutic value.
However, while the research on molecular biology has allowed the development of
very strong promoters, there are still several phenomena associated with this
process that have hampered the full use of that promoter strength, namely the
aerobic acetate production associated with high specific growth rates. The
presence of acetate is known to reduce both biomass yield on the chosen carbon
source and protein productivity while totally
inhibiting growth when present at high concentrations due to its toxic effect.
While there have been several studies covering the recombinant protein
production process with the bacterium Escherichia coli, including genome-scale
analysis of the transcriptome, proteome, fluxome or metabolome, there has been
a lack of an integrative approach that is able to combine genomic and
physiological information about those processes with high-throughput analysis.
Also, the existence of genome-scale models that cover both stoichiometry and
regulation of some pathways has not been taken into account in genome-scale
data analysis and for the consequent formulation of hypothesis and development
of new strategies for improving the performance of the process.
In our group, a high-cell density fed-batch process for recombinant protein
production in E. coli is being
studied, giving particular relevance to acetate production. A systematic
approach is being used, by first compiling the existing knowledge about this
phenomenon, extending existing genome-scale models to accommodate that
knowledge, derive hypothesis in silico that are then tested by using
genome-scale analysis of the omes. A reliable fermentation process was
developed to be able to reproducibly study this phenomenon in different strains
in order to reduce external variances to a minimum.
U-P21 Some
properties and partial purification of ICandida Guilliermondii NP-4/I and
IParamcium Multimcronucleatum/I glutaminase
Ara H. Tamrazyan, Misak A. Davtyan and Susanna
A. Karapetyan
Department of
Biochemistry, Yerevan State University, Charents str. 8, Yerevan 375025,
Armenia, Phone: +374 9 361770, FAX: +3741 429888, e-mail: aratam2000@yahoo.com
Have
been investigated the activity of enzyme glutaminase in yeasts Candida guilliermondii and in aerob
parameciums Paramecium
multimicronucleatum.
The investigations show that the general activity of glutaminase is
concentrated in the mitochondrias (about 85%), this mitochondrial glutaminase
is phosphate depended. Also there is a low level of glutaminase activity in the
nucleus (about 15%), and in the mycrosomal fraction (1%), this glutaminases
were phosphate dependent.
The investigations show two isoensymes of glutaminase in yeasts Candida guilliermondii. The phosphate
independed isoglutaminase was located in the nucleic fraction of cells (about
30%). Mithochondrial (70%) and mycrosomal (2%) glutaminase of yeasts Candida guilliermondii were phosphate
dependent. pH optimum for yeasts glutaminase was 8,5.
Have been investigated the effects of different activators and inhibitors on
the mitochondrial glutaminase of yeasts Candida
guilliermondii and parameciums Paramecium
multimicronucleatum.
The mithochondrial glutaminase of Paramecium
multimicronucleatum have been partrial purified by the method of the
ionexchange chromatography (using gradiative elution by different
concentrations of NaCl. Have been shown, that more than 80% of total
glutaminase activity of Paramecium
multimicronucleatum are located in the mithochondrias of the cells. Two
isoensymes of the glutaminase have been revealed in the mithochondrias of Paramecium multimicronucleatum. The
total activity of glutaminase I have been 45.26 ± 1.51 mM ammonium (specific
activity - 22.86 ± 1.12 mM ammonium/ prot.). The total activity of glutaminmase
II have been 13.61 ± 1.02 mM ammonium (specific activity - 17.01 ± 1.08 mM
ammonium/ prot.). The activity of the glutaminase I makes about 75% ot total glutaminase
activity of mithochondria. The degree of enzyme purification has been 26 (72
%). Km for glutamine of the glutaminase I has been 4.3 X 10 - 4
M, for glutaminase II - 5.5 X 10- 4 M. Ki for glutamic
acid of glutaminase I was 1.5 X 10- 5 M, for glutaminase II - 2.5 X 10- 6 M. Have been shown
that glutaminase I is phosphate dependent, while glutaminase II is phosphate
independent, it is positively regulated by citrate. ATF, tiroxine,
dexametazone, adrenaline and bicarbonate have show the positive effect on
activation of both mitochondrial isoglutaminases.
U-P22 Unusual
group II introns in bacteria of the Bacillus cereus group
Nicolas Tourasse, Fredrik Stabell, Lillian Reiter and
Anne-Brit Kolstø
Biotechnology
Center of Oslo, University of Oslo, P.O. Box 1125 Blindern, Oslo 0349, Norway,
Phone: +47 22 84 05 36, FAX: +47 22 84 05 01, e-mail: nicolat@biotek.uio.no
A
combination of sequence and structure analysis and RT-PCR experiments was used
to characterize the group II introns in the complete genomes of two strains of
the pathogen B. cereus. While B. cereus ATCC 14579 harbors a single
intron element in the chromosome, B.
cereus ATCC 10987 contains three introns in the chromosome and four in its
208-kb pBc10987 plasmid. The most striking finding is the presence in B. cereus ATCC 10987 of an intron
(B.ce.I2) located on the reverse strand of a gene encoding a putative cell surface protein which appears to be correlated to strains of
clinical origin. Because of the opposite orientation of B.ce.I2 the gene is
disrupted. Even more striking is that B.ce.I2 splices out of an RNA transcript corresponding to the opposite DNA strand, a situation never observed before.
All other intragenic introns studied here are inserted in the same orientation
as their host genes and splice out of the mRNA in vivo setting the flanking
exons in-frame. Noticeably, B.ce.I3 in B. cereus ATCC 10987 represents the
first example of a group II intron entirely included within a conserved
replication gene, namely the alpha subunit of DNA polymerase III. Another
striking finding is that the observed 3 prime splice site of B.ce.I4 occurs 56
bp after the predicted end of the intron. This apparently unusual splicing
mechanism may be related to structural irregularities in the 3 prime terminus.
Finally, we also show that the intergenic introns of B. cereus ATCC 10987 are transcribed with their upstream genes and
do splice in vivo.
U-P23 The
effect of oxygen tension on yeast glycolysis
Isil Tuzun, Klaas Hellingwerf and M.
J. Teixeira de Mattos
Department of
Microbiology, SILS-UVA, Nieuwe Achtergracht 166, Amsterdam 1018
WV, The Netherlands, Phone: +31/20/5256424, FAX: +31/20/5257056,
e-mail: tuzun@science.uva.nl
The
catabolism of S. cerevisiae ranges from a highly efficient,
respiratory mode with carbon dioxide as sole end product, to a fully
fermentative mode that results in high specific rates of sugar conversion to
carbon dioxide and ethanol. Fully respiratory growth is only found when oxygen
is saturating and glucose availability is limited, while fully
fermentative growth is found under strictly anaerobic conditions. Intermediate
situations, characterized by a mixed respiro-fermentative growth, are found
either when the oxygen availability is limited, or when well-aerated cultures
are subjected to glucose excess. The transition between respiratory and
fermentative catabolism is known to involve various levels of the cellular
regulation hierarchy. However, the relative importance of the various
regulation mechanisms is not known let alone quantitated.
In this project the regulation of these transitions is investigated in a
time-resolved manner at the metabolic, the enzyme activity, the proteomic and
the transcriptomic levels. To induce these transitions, S.cerevisiae
CEN.PK 113-7d is cultivated in steady-state glucose-limited chemostat cultures under various oxygen regimes (fully
aerobic, hypoxic, microaerobic and anaerobic). After quantification of the
oxygen availability for a given experimental set-up, the experiments will focus
on a quantitative description of the cell in steady state (a snapshot) with
respect to gene expression, enzyme activity,
metabolite concentrations and specific metabolic activity. Quantification of
these regimes necessitates calibration of the experimental set-up.
Subsequently, the changes in cellular make-up and activity will be monitored (a
movie) when it adapts from one steady state to the other. This approach will
allow for the discrimination between cause and effect during adaptation. The
use of well-defined steady state conditions will yield a data set that is
sufficiently quantitative for modeling of the response.
U-P24 Vertical
genomics in baker’s yeast: adaptation of respiring
cells to anaerobic sugar-excess conditions.
Joost van den Brink, Pascale Daran-Lapujade, Han de
Winde and Jack Pronk
Biotechnology,
Delft University of Technology, julianalaan 67, delft 2628 BC, The Netherlands,
EU, Phone: +31/15/2787466, FAX: +31/15/2782355, e-mail: j.vandenbrink@tnw.tudelft.nl
The primary role of bakers’ yeast (Saccharomyces cerevisiae)
in the leavening of bread dough is the production of carbon dioxide via the
alcoholic fermentation of sugars. Within this process an important parameter is
the fermentative capacity, defined as the specific rate of carbon dioxide (and
ethanol) production immediately upon introduction of yeast into dough (3).
However, alcoholic fermentation is highly undesirable during the industrial
production of bakers’ yeast, as it reduces the biomass yield on the
carbohydrate feedstock. Industrial bakers’ yeast production is therefore,
performed in aerobic, sugar-limited fed-batch cultures. Hence, conditions
during production of bakers’ yeast differ drastically from the dough
environment, which is anaerobic and initially contains an excess of sugars.
This project undertakes a Systems Biology approach to ultimately understand and
control the regulatory mechanisms (from gene to flux) that govern the induction of
fermentative capacity. This project is part of a larger IOP-Genomics program
“Vertical Genomics”, which is a collaboration between six research groups from
different universities (TUDelft, UvA and VUA).
The dynamic control of both fermentative capacity and fermentative activity
will be simulated and analysed in detail. Aerobic glucose-limited chemostat cultures will be used to simulate a baker’s yeast production
process. Chemostat cultures are preferred over fed-batches, as the culture
conditions can be defined and tightly controlled. The dynamic dough environment
is simulated by shifting the chemostat to anaerobic batch conditions, followed
by addition of a glucose pulse. To understand at which cellular ‘level(s)’ the
fermentative capacity, and specifically glycolysis, is regulated (i.e. transcriptional or
post-transcriptional control) the experimental approach includes analysis at
various regulatory levels, i.e.
transcript level (micro-array/ qPCR), protein levels (protein chips/ MS), enzyme activity assays and in vivo carbon-fluxes with
stoichiometric modelling. Diverse datasets will be aligned, compared and
integrated, and corellations between regulatory levels will be mapped.
References: 1. Boer V.M., Winde J.H. de, Pronk J.T.
and Piper M.D.W. (2003) J. Biol. Chem, 278, 3265 – 3274
2. Daran-Lapujade P,
Jansen ML, Daran JM, van Gulik W, de Winde JH, Pronk JT. (2004) J Biol Chem.
279, 9125-38
3. Hoek W.P.M. van, Dijken
J.P. van and Pronk J.T. (2000) Enzyme Microb.Technol. 26, 724 – 736
U-P25 LacplantCyc:
in silico reconstruction of the metabolic pathways of Lactobacillus
plantarum.
Frank H.J. van Enckevort 1, Bas Teusink 2, Christof Francke 3 and Roland
J. Siezen 2
1 CMBI, Centre for Molecular and Biomolecular Informatics, Radboud
University, Toernooiveld 1, Nijmegen NL-6525ED, The Netherlands,
Phone: +31/24/3653358, FAX: +31/24/3652977, e-mail: frankve@cmbi.ru.nl, Web: www.lacplantcyc.nl
2 NIZO food research, Ede,
Food Valley, The Netherlands 3
Wageningen Centre for Food Sciences, Wageningen, The Netherlands
Lactobacillus plantarum is a versatile
and flexible lactic acid bacterium (LAB) that is important in
industrial food fermentation processes. It is also one of the LAB species that
are important as probiotics in health-promoting food products. We have
sequenced and annotated the genome of L. plantarum WCFS1 (PNAS USA 2003;
100:1990-1995), which is one of the largest Lactobacillus genomes. L.
plantarum serves as a model organism for genome annotations and comparisons
with other LAB (lactobacilli, lactococci, streptococci). This work describes a
reference database of the metabolic network, based on the Lactobacillus genome
annotation.
LacplantCyc is a pathway / genome database (PGDB) describing the entire
genome as well as its biochemical pathways, reactions, and enzymes. This
database was automatically generated from the annotated L. plantarum
WCFS1 genome and the MetaCyc database (Nucleic Acids Research 2004; 32:D438-42)
using the PathoLogic software from Pathway Tools (Bioinformatics 2002;
18:S225-32). LacplantCyc was subsequently curated manually: transporters were
added, new pathways were created and others updated.
LacplantCyc adds an extended dimension to the genome of L. plantarum,
providing researchers with a helpful tool for the analysis of the genomic,
proteomic, and metabolic information of the organism. Visualization of the data
sets in different levels of detail is extremely important to help interpreting
these data from a biological viewpoint. Once the connections between genes,
proteins and reactions in a metabolic map have been defined, high-throughput
transcriptome data can be projected on metabolic maps. An additional aim during
this process is to formulate which information is useful for improved pathway
reconstruction. This should speed up the reconstruction of the metabolic
networks of other LAB of which the complete genome has been sequenced. In this
way LacplantCyc is envisioned to become the reference Gram-positive PGDB for
LAB. It is the first well-curated metabolic pathway database for Gram-positive
bacteria in general, and lactic acid bacteria in particular. LacplantCyc is available at http://www.lacplantcyc.nl.
Supported by
the Netherlands Organisation of Scientific Research (NWO) BioMolecular Informatics
Programme, grant 050.50.206.
U-P26 HIGH-THROUGHPUT
screening of Saccharomyces cerevisae
knockout library: method development and stoichiometric profiling
Vidya R. Velagapudi 1, Christoph Wittmann 1,
Thomas Lengauer 2, Priti Talwar 2 and
Elmar Heinzle 1
1 Biochemical Engineering, Saarland University, Im Stadtwald, Geb-2,
Saarbrucken D-66123, Germany, Phone: +49/681/302/3590,
FAX: +49/681/302/4572, e-mail: v.mangadu@mx.uni-saarland.de
2 Max-Planck-Institute for
Informatics, Saarland University, Germany.
The present work aims at the method development
for high-throughput screening and stoichiometric profiling of Saccharomyces cerevisiae mutants. For selected gene deletion mutants, we studied kinetic and stoichiometric profiles (growth rate, respiration, yields of
biomass and ethanol) at miniaturized scale by cultivation in 96-well microtiter
plates. Cultivation included on-line sensing of dissolved oxygen by immobilized
fluorescence sensors. Loss of ethanol and water due to evaporation during
cultivation was corrected by using a dynamic model. Cultivation at this
micro-scale displayed growth profile, substrate consumption and product
formation patterns highly similar to conventional shake flask cultivation.
Furthermore, we have investigated the effect of shaking rate on oxygen
limitation. Mutants, cultivated on glucose, fructose and galactose
showed substrate specific differences in specific growth rates and yields.
Comparative phenotypic profiles in different environments allow a detailed
classification of mutants. First experiments involved tracer studies with 1-13C
labeled glucose, fructose and galactose. Labeling patterns of mutants were
different on the different carbon sources, giving a first impression of
differences in the underlying fluxes. We propose the utilization of the
developed methodology for large-scale quantitative screening of yeast deletion
mutants.
U-P27 A
Sysytems
Biology Strategy For Understanding The Genome-wide Control Of Growth Rate And
Metabolic Flux In Yeast
Jian Wu, Nianshu Zhang, Andy Hayes, Douglas Kell, Stephen Oliver and Jian Wu
Faculty of Life
Science, The university of Manchester, Oxford Road, Manchester M13 9PT, UK,
Phone: 0044 161 275 1579, FAX: 0044 161 275 1505, e-mail: jian.wu@manchester.ac.uk
Baker’s
yeast, Saccharomyces
cerevisiae, was the first eukaryotic organism to have its genome sequenced. This makes it a central ‘model’
organism in modern biology, and it is important to know what controls the rate
at which it can grow. Moreover, yeast is an industrially significant organism
from many points of view. Some industries (e.g. baker’s yeast production) wish
to maximise growth rates, others (e.g the antifungal industry) to minimise
them. We are engaged in the development of both top-down and bottom-up
genome-wide models for the control of the maximum specific growth rate in S.
cerevisiae. Genes with high flux-control
coefficients will be identified via haploinsufficiency measurements in
turbidostats, and will form the components of a coarse-grained model in the
top-down approach. Intra- and extra-cellular metabolome transcriptome and
proteome measurements in selected genetically defined strains will be used,
iteratively, to validate models. Flux-balance modelling will also be used to
define specific modulations likely to be most discriminatory between competing
models. The result will be the first example in which the controls on growth
rate and metabolic fluxes are established on a genome-wide scale.
U-PoP1 Modeling
and analyses of Mycobacterium tuberculosis metabolism
Asawin Meechai 1, Supapon Cheevadhanalak 2 and
Sakarindr Bhumiratana 3
1 Department of Chemical Engineering, King Mongkut's University of
Technology Thonburi, Prachautid, Bangkok 10140, Thailand,
Phone: 66-02-4709616, FAX: 66-02-8729118, e-mail: asawin.mee@kmutt.ac.th
2 School of Bioresources and
Technology, King Mongkut's University of Technology, Thonburi, Bangkok, 10140,
Thailand
3 National Center of Genetic
Engineering and Biotechnology (BIOTEC), Prathumtani, 12120, Thailand
Tuberculosis (TB)
disease remains a serious problem that threatens human health around the world.
TB prevention and treatment has been hindered by multi-drugs resistant TB
(MDR-TB) mainly emerged from the lengthy drug treatment, and the lack of discipline of TB
patients. This lures us to seek for new drug targets for further drug design
and development. In this study, we reconstructed the species-specific metabolic
pathways network of Mycobacterium tuberculosis H32Rv (MTB) from its
complete genome sequence information and from
various literatures. We employed 2 different approaches to analyze and model
the network to identify potential anti-TB enzyme targets. The first approach
was based on the analysis of network topology using elementary flux mode analysis. METATOOL software was used to
determine all possible elementary routes for the synthesis of mycolic acid, an
important component of cell envelope of MTB. Enzymes that were present in every
elementary route were considered the key drug targets for TB because the lack
of these enzymes would not lead to synthesis of mycolic acid. It was found that
the enzymes, inhA, accD6, kasA, kasB, cmaA2, pcaA, fabD-Pacp, accD4, accD5,
accD3, desA1, desA2, desA3, dcb, mmaA1, mmaA2, cmaA1 and acrA1 are potential
drug targets for TB disease. The second approach used the metabolic network
information to build a genome-scale metabolic model of MTB using flux balance
analysis. This MTB model integrated 473 genes and 419 metabolites with 601
biochemical reactions. The model was used to perform in silico single
gene knockout. From a total of 21 genes, which
were previously identified as drug targets, 18 cases led to the fatality of the
in silico MTB, indicating an 85% model accuracy. From the in silico
gene knockout experiments of all remaining genes, we identified a list of 91
essential genes whose protein products are promising targets for TB drugs.
We found that most targets identified by both approaches are in good agreement.
Among those, a few will be chosen for further experimental validation and
future steps in the development of new drugs for multi-drug-resistant (MDR)
strains that have caused millions of deaths worldwide.
U-PoP2 SOME
CHANGES IN THE COMPOSITION OF NUCLEAR COMPONENTS DURING CEREAL SEEDS
GERMINATION.
Liya A. Minasbekyan and
Poghos H. Vardevanyan
Department of
Biophysics, Yerevan State University, str.A.Manougian,1, Yerevan 375025,
Armenia, Phone: (374)+1 / 57-10-61, FAX: (374)+1 / 55-46-41,
e-mail: minlia@ysu.am, Web: hhtp://www.ysu.am
Most
eukaryotic genomes are packaged into two constitutive regions of chromatin:
euchromatin and heterochromatin. Heterochromatin represents significant
fraction of most eukariotic genome and its function remain unknown. Nuclear pore
complexes (NPC), including in nuclear membrane also appear to control the spatial orientation and transcriptional activity of
chromatin. In yeast, NPC influence telomeric silencing via fiber-forming proteins emanating
from their nucleoplasmic face. Recent studies have also shown that tethering a
gene to the NPC can prevent its repression by
creating a boundary between the gene and surrounding silenced heterochromatin
Our investigations shown changes in some physico-chemical characteristics of
cereal seeds DNA, chromatin and nuclear membrane during genome activation. We have
suggested that the high methylated region of repeated DNA commonly lies in
heterochromatin region adjacent to the nuclear envelope (NE).
Recent studies have support that the NPC through association with the
underlying chromatin regulates gene expression. Particularly have been obtained
the changes in the DNA, RNA,
protein and phospholipid content of the NE, soluble
nuclear fraction and chromatin during germination of cereal seeds embryos and
under influence of exogenous gibberellin A3. The characteristic trait for
growing seed nucleus is a rising of protein and phosphatidic acid in nuclear membrane
content to the third day of growing. The chromatin separated on the euchromatin
and heterochromatin constitutive regions, and subsequently obtained parameters
of thermal denaturation and the level of DNA methylation. Have been revealing
changes in all abovementioned parameters during seed germination and under
influence of gibberellic acid A3.
The goal of this work is to show correlation between genome activation and the
change in content of the nuclear membrane during genome expression.
Nevertheless much about the relationship between chromatin organization and the
NE remains to be discovered.
U-PoP3 Differentiation
in a genetic network with duplicate repressors: simulating evolutionary
pathways based on Lac mutational data
Frank Poelwijk, Daniel Kiviet and Sander Tans
AMOLF, Kruislaan
407, Amsterdam 1098
SJ, the Netherlands, Phone: +31(0)20 6081266, FAX: +31(0)20 6684106,
e-mail: tans@amolf.nl
To add
and alter functions, gene regulatory networks must be able
to make new interactions and break old ones. A much-debated concept for such
network “re-wiring” is the differentiation of duplicate regulatory components.
However, we still lack a quantitative insight in the underlying evolutionary
pathways at a mutation-by-mutation level. Here, extensive mutational data of
the Lac system is integrated into a computer simulation scheme to
systematically study such pathways. In a network with duplicate repressors and
their binding sites, recognition is initially indiscriminate and must evolve
towards independent binding. We find the paths to be surprisingly short and
without detours, with the fitness increasing rapidly and almost immediately. In
an alternative scenario, where a new operator must emerge from a random
sequence, chances of success are much reduced because prolonged neutral drift
is required. The fitness landscape as revealed by the rapid pathways appears
diverse, with steep cliffs but also smooth regions. These landscape features,
together with a cooperation between repressor copies appear to be key for the
system’s evolvability. The presented approach can be more broadly applied to
render network evolution simulations more experimentally relevant, and provides
key insight for new experiments where natural as well as artificial networks
are shaped by directed evolution.
Multicellular/mammalian
Posters
M-S01 Inferring
feedback mechanisms in cellular transformation due to oncogenic RAS
Nils Bluthgen 1, Christine Sers 2,
Jana Keil 2, Szymon M. Kielbasa 1,
Reinhold Schaefer 2 and Hanspeter Herzel 1
1 Theoretical Biology, Humboldt University Berlin, Invalidenstr. 43,
Berlin 10115, Germany, EU, Phone: +49/30/20938496,
FAX: +49/30/20938801, e-mail: nils@itb.biologie.hu-berlin.de,
Web: http://itb.biologie.hu-berlin.de/~nils/
2 Insitute of Pathology,
Charite, Berlin
Intracellular
signaling cascades display distinct activation profilesin response to various
stimuli. Such activation patterns are strongly influenced and shaped by
feedback loops. Different feedback loops can act in a cell context- and
stimulus-dependent manner and produce a variety of temporal activation
profiles, including oscillations and hysteresis. The MEK-ERK
cascade plays an important role in cell-cycle regulation, differentiation and
in cell transformation caused by oncogenic RAS. This
cascade is regulated by several positive and negative feedback loops and is
essential for signal transmission due to many different stimuli.
While post-translational feedback loops have been subject to extensive mathematical
modeling, feedbacks that are mediated by transcriptional control are still poorlyunderstood.
Using a
combination of time-course experiments, mathematical modeling and bioinformatic analysis we investigate the effect of
transcriptional feedback regulation in cellular transformation following
induction of oncogenic RAS. In fibroblasts harboring an inducible RAS oncogene, we monitor the
phosphorylation of ERK1,2 by Western Blot analysis. In addition, we analyze the
expression profiles of RAS target genes with microarrays in a time-resolved
manner. The phosphorylation of ERK shows a biphasic response upon constant
induction and an oscillatory response after brief induction of RAS. We find
that several dual specific phosphatases are expressed with similar kinetics. A
bioinformatic analysis unveils two ERK-dependent transcription factors that
control this battery of phosphatases. Together with
the transcription factors, these phosphatases constitute a negative feedback
for ERK-activity. Mathematical modeling and experimental interference shows
that we can explain the biphasic and oscillatory dynamics as a result of
phosphatase activation.
M-S02 Regulation
of MAPK signalling determining cell fate in PC-12 cells - a step beyond
biochemistry
Silvia D. Santos, Eli Zamir, Peter Verveer and
Philippe Bastiaens
Cell Biology and
Biophysics, EMBL, Meyerhofstrasse-1, Heidelberg D-69117,
Germany, Phone: +39/6221/387406, FAX: +39/6221/387242, e-mail: santos@embl.de
Mitogen
activated protein kinase (MAPK)
cascades participate in a wide array of cellular transduction programmes
including cell growth and division, movement, differentiation and cell death. A
paradigm system to study how the activity of these cascades produces different
cell responses is the PC-12 cells system. In these cells the classical ERK pathway
is activated by both EGF and NGF, giving rise to cellular opposite fates –
division and differentiation, respectively. We believe different biochemical
topology may be the key determining these specific responses. We are therefore
interested in measuring reaction states of main components of this pathway, to
analyze how the kinases are spatially organized and biochemically connected. We
are using polychromatic fluorescence activating cell sorting (FACS) with
phospho-labelled antibodies, which detect the active state of network
components. By applying systematic perturbations of activities and subsequent
read out on multiple reaction states at steady-state we are able to retrieve
information on the network topology. Single cell measurements are being performed
and RNAi and pharmacological inhibitors used for the
perturbations. Moreover, response coefficients for each kinase, before and
after perturbations will be calculated and first order connectivity maps built.
In addition, by using fluorescence resonance energy transfer (FRET) imaging with multiple optical
sensors, reaction states of kinases and their spatial information are being determined
simultaneously in one cell. Fusion proteins of GFP mutants and pathway kinases
allow the detection of protein-protein interactions and molar ratios of
phospho-proteins can be detected, by using phospho-antibodies against
phospho-residues on active kinases.
M-S03 Mathematical
modeling of neuronal response to neuropeptides: Angiotensin II signaling via
G-protein coupled receptor
Thomas Sauter 1, Rajanikanth Vadigepalli 2 and
James Schwaber 2
1 Institute for System Dynamics and Control Engineering, University of
Stuttgart, Pfaffenwaldring 9, Stuttgart D-70550, Germany,
Phone: +49/711/6856611, FAX: +49/711/6856371, e-mail: sauter@isr.uni-stuttgart.de,
Web: http://www.isr.uni-stuttgart.de/~sauter/ 2 Daniel Baugh Institute, T. Jefferson University,
Philadelphia, PA
In
neurons G-protein coupled receptors (GPCRs) are
involved in the alteration of neuronal activity (neuromodulation) via cascades
of interacting proteins. The complex dynamic behavior of these networks, e.g.
the integration of different signals, cannot be understood by intuition alone.
Mathematical modeling provides an appropriate tool to decipher this complexity.
Angiotensin II and AT1 receptor dependent signaling was investigated as an
examples that use GPCR signaling pathways (Gq). AT1 signals via a wide variety
of intracellular signaling molecules, involving (1) G-protein mediated
stimulation of phospholipase C (PLC), with subsequent Ca2+ mobilisation; (2)
Jak/STAT pathway; (3) transactivation of tyrosine kinase pathways. Relevant signaling
outputs are modified gene expression patterns and modified
neuronal activity via changes in membrane ionic currents and firing rate.
New data that was collected recently [Fernandez et al., Hypertension
Jan.2003:56-63] showed that Angiotensin II can elicit stimulating and
suppressive effects in the same neurons in dependency of the basal Ca2+ level.
We have built a detailed mechanistic model of Angiotensin II signaling that
captures both the stimulating and suppressive effects. This ODE model includes
the AT1 mediated activation of PLC and PKC, and IP3 and channel mediated
variation of the cytosolic Ca2+ level after Angiotensin II stimulation (adapted
from [Mishra and Bhalla, Biophys. J., 83:1298-1316, 2002]). Based on in silico
simulations of this model, we hypothesize that the observed biological
variability is based on cell-to-cell variation in the dynamics of the Na-Ca
exchanger.
Furthermore, a Hodgkin-Huxley model approach is used to investigate the
function of cell signaling in altering the firing behavior of NTS neurons in
response to various baroreceptor stimuli. Angiotensin II was found to activate
neuronal firing in low firing NTS neurons.
In summary, detailed mathematical is a valuable tool to understand and
investigate neuronal response to neuropeptides and furthermore to link signal transduction to the
electrophysiological behavior of neurons.
M-P01 Control
of the ATP/ADP ratio in pancreatic beta
cells
Charles Affourtit and
Martin D. Brand
, MRC
Dunn Human Nutrition Unit, Hills Road, Cambridge CB2 2XY, United Kingdom,
Phone: +44/1223/252803, FAX: +44/1223/252805, e-mail: ca@mrc-dunn.cam.ac.uk
Pancreatic
beta cells respond to rising blood glucose concentrations by increasing
their oxidative metabolism, which leads to an increased ATP/ADP ratio, closure of KATP channels, depolarisation of the
plasma membrane potential, influx of calcium and the eventual secretion of
insulin (Rutter, G, (2001), Mol. Aspects
Med., 22, 247). Such a signalling mechanism implies that the ATP/ADP ratio in
beta cells is flexible, which is in contrast to other cell types (e.g. muscle)
that maintain a stable ATP/ADP poise whilst respiring at widely varying rates.
To determine whether this difference in flexibility is accounted for by
mitochondrial particularities, we are currently performing a top-down metabolic
control analysis to assess
quantitatively how the ATP/ADP ratio is controlled in mitochondria isolated
from rat skeletal muscle and cultured beta cells. We have defined the
experimental system to contain two explicit intermediates, the protonmotive
force and the external ATP/ADP poise, through which electron transfer,
phosphorylation, proton leak, and ATP-consuming (i.e. added hexokinase)
reactions interact. The elasticities of these processes to both intermediates
are determined by the multiple-modulation method, and control values are
calculated from the elasticities (Cornish-Bowden, A, et al., (1994), Biochem.
J., 298, 367).
Preliminary measurements of oxygen-uptake activity and membrane potential
suggest several differences between the mitochondrial energetics of muscle and
beta cells. For example, the basal proton leak in muscle mitochondria is
reduced, at every membrane potential measured, by ATP as well as ADP. Neither
nucleotide affects proton leak in the presence of carboxyatractyloside,
suggesting the effects are mediated by the adenine nucleotide translocator. In
beta cell mitochondria, however, proton leak is not affected by ATP or ADP.
Furthermore, the activity of the respiratory chain in beta cell mitochondria is
lowered significantly when the organelles are depleted of adenine nucleotides, which
is not the case in muscle. Also of potential relevance is the observation that
the mitochondrial sample obtained from beta cells, unlike that from muscle,
exhibits hexokinase activity. Data will be presented to reveal the extent to
which the observed differences between muscle and beta cell mitochondria are
reflected by the way the ATP/ADP ratio is controlled in these systems.
This work is supported by the Medical
Research Council.
M-P02 Regulation
of the INF-Gamma/JAK/Stat1 signal transduction pathway
Stephan Beirer 1, Thomas Meyer 2,
Uwe Vinkemeyer 2 and Thomas Höfer 1
1 Theoretical Biophysics, Humboldt University Berlin, Invalidenstrasse
42, Berlin D-10115, Germany, Phone: +49/30/2093-8694,
FAX: +49/30/2093-8813, e-mail: s.beirer@biologie.hu-berlin.de
2 Signal Processing Group,
Forschungsinstitut Molekulare Pharmakologie, Berlin, Germany
The
STATs (Signal Transducers and Activators of Transcription) constitute a family
of seven transcription factors regulating a multitude of cellular functions
like immune reactions, growth, proliferation, differentiation and apoptosis. In
response to stimulation by extracellular signaling proteins, such as cytokines
and growth factors, the STATs are transiently activated by phosphorylation
through receptor-bound Janus kinases (JAKs) and accumulate rapidly in the
nucleus, where they switch on expression of their target genes. Here we present
a detailed mathematical model of IFN-Gamma/JAK/Stat1 signaling, which is able
to reproduce the behavior of this pathway in a quantitative manner for wildtype
Stat1 and three Stat1 mutant proteins under various exeprimental conditions.
Parameters of the model were adjusted to independent experiments of wildtype
Stat1 and mutant proteins. The import mutant shows no nuclear import of
phosphorylated Stat1 and is dephosphorylated significantly slower. A second
mutant is rapidly exported to the cytoplasm both in the phosphorylated and
unphosphorylated form. Using data from experiments with these proteins,
parameters for dephosphorylation and transport rates could be fitted. With this parameter
set we were able to reproduce quantitatively the phosphorylation kinetics and
subcellular distribution timecourses under stimulation with IFN-Gamma and for a
number of pharmacological protocols using transport and kinase inhibitors.
Using the model we analyze the properties of Stat1 signal transduction by computing the
control of the individual pathway steps on the
amplitude and duration of the nuclear phosphorylation signal for different
stimulation patterns. The analysis reveals that for weak stimulation conditions
the amplitude is primarly governed by receptor-driven processes, DNA binding and the nuclear phosphatase. Under strong stimulation the phosphatase retains its high negative
effect. The most significant control is now exerted by nuclear import and
export of inactive Stat1. The residence time of phopshorylated Stat1 in the
nucleus is negatively influenced by the nuclear phosphatase, too. DNA binding
and receptor half-life exert significant positive control on the signal
duration. We also found an intrinsic response time of the system, during which
variations of ligand occupancy of the receptor are not transduced to the
nuclear compartment.
M-P03 Comprehensive
analysis of the cancer Tyrosine Kinome & Phosphatome
Martin Bezler 1, Christian Mann 1, Detlev
T. Bartmus 1, Pjotr Knyazev 1,
Tatjana Knyazeva 1, Sylvia Streit 2 and
Axel Ullrich 2
1 Molecular Biology, Max-Plack-Institute of Biochemistry, Am Klopferspitz
18, Martinsried 82152, Germany, Phone: +49/089/8578/2508,
FAX: +49/089/8578/2454, e-mail: bezler@biochem.mpg.de
2 Singapore Oncogenome
Laboratory (SOG), Centre of Molecular Medicine, Institute of Molecular and Cell
Biology, Singapore
Protein
tyrosine kinases and phosphatases play crucial roles in the regulation of
cellular processes like proliferation, differentiation, motility and survival.
Receptor type tyrosine kinases (RTKs) function as signal transmitters across the plasma
membrane by integrating a multitude of extracellular
stimuli. Protein tyrosine phosphatases (PTPs) are the negative regulators of
these processes through the removal of phosphate groups from tyrosine residues
in RTKs and downstream signal proteins. Disturbances in this tightly controlled
system have been shown to be causally connected to a variety of
pathophysiological phenomena such as cancer. The
largest group of genes with oncogenic potential belong to the PTK family and
PTPs have by definition a tumor suppressor function which however, is largely
unproven.
Cancer represents a genetic disease that begins with a series of damages in the
genome of one cell in the form of DNA modifications such as point mutations,
rearrangements or sequence amplification affecting regulatory genes such as
PTKs. For example a point mutation in the transmembrane domain encoding region
has been found to cause constitutive activation of the neu RTK and to induce
cancer in rats 1. Moreover, mutations in the FGFR3 gene have been associated with bladder and cervix
cancer in humans 2. Breast cancer patients with high expression of
the FGFR4 gene and a single nucleotide change showed accelerated disease
progression and reduced overall survival 3.
The aim of our study is the analysis of all PTKs, PTKLs as well as PTPs and
DUSPs by direct sequencing of cDNA from cancer cell lines and primary tumors in
order to identify so far unknown oncogenic or tumor suppressing mutations.
Therefore we designed gene–specific primers to generate overlapping PCR
fragments of about 900bp of each PTK/PTP. RNA was isolated from cancer cell lines and
tissues, reverse transcribed and the cDNA is used as template for PCR to
generate fragments for sequencing.
Identified genetic alterations will be verified in clinical tumor samples,
correlated clinico-pathological parameters and functionally characterized in
appropriate cell systems for their ability to influence
transformation-characteristic cell properties like contact inhibition, colony
formation in semi solid agar, migration, invasion and tumorigenesis in mice
1. Bargman et al., (1988) Proc Natl Acad Sci USA, 85,
5394. 2. Capellen et al.,
(1999) Nat Genet, 23, 18.
3. Bange et al., (2002)
Cancer Res, 62, 840.
M-P04 Sensitivity
analysis with respect to initial values of the TNFalpha mediated NF-kappaB
signalling pathway
Marc Breit, Gernot Enzenberg,
Visvanthan Mahesh, Robert Modre-Osprian and
Bernhard Tilg
IMSB, UMIT, EWZ
I, Hall in
Tirol A-6060, Austria, EU, Phone: +43-50-8648-3821,
FAX: +43-50-8648-67-3821, e-mail: marc.breit@umit.at,
Web: imsb.umit.at
Objective The objective of this work is to investigate the behaviour of the
TNFalpha mediated NF-kappaB signalling pathway. Starting from a mathematical model
of the pathway consisting of ordinary differential equations, we studied
derivatives of the solution of the model with respect to its initial values.
Methods The mathematical pathway
model presented by Cho etal in [1] was taken as basis to build up an improved
model of the pathway. The focus here lied on assembling new components into the
given model. For this purpose a comparison of the models of Cho etal [1] and
Schoeberl etal [2] was performed. Further more the model was checked against
the interaction-map presented by Bouwmeester etal [3]. A sensitivity analysis
regarding the initial values of the components was performed for both pathway
models, the original one and the extended one. The matlab function sens_ind [4]
was used to carry out this task.
Results The extended model
encompasses three additional proteins: MEKK3, FLIP and TRAF1. All together the
new model encompasses 8 additional components - three proteins and
corresponding complexes. Further more five inconsistencies in the original
model given by Cho etal [1] were eliminated. The sensitivity analysis shows
that the behaviour of the extended pathway depends strongly on the initial
values of the new added components.
Conclusion The sensitivity analysis of initial values showed that
further investigations regarding the new model will be necessary. Challenge is
a meaningful extension of the mathematical model. An open question to work on
is, how to integrate new components and reactions into the model.
References [1] Cho KH, Shin SY, Kolch W, Wolkenhauer O.
Experimental Design in Systems Biology Based on Parameter Sensitivity Analysis
with Monte Carlo Simulation: A Case Study for the TNFalpha Mediated
NF-kappaB-Signal Transduction Pathway. Simulation. 2003 Dec;79(12):726-39. [2]
Schoeberl B, Gilles ED, Scheurich P. A Mathematical Vision of TNF Receptor
Interaction. Proceedings of the International Congress of Systems Biology.
Pasadena, CA. 2001;158-67. [3] Bouwmeester T, etal. A physical and functional
map of the human TNF-alpha/NF-kappa B signal transduction pathway. Nat Cell Biol. 2004 Feb;6(2):97-105. Epub 2004 Jan 25. [4] Mollá GVM, Padilla
GR. Description of the MATLAB functions SENS_SYS and SENS_IND. 2002 Jun.
M-P05 A
domain-oriented approach to the reduction of combinatorial complexity in signal transduction networks
Holger Conzelmann 1, Julio Saez-Rodriguez 2, Thomas Sauter 1, Boris Kholodenko 3 and Ernst-Dieter Gilles 2
1 Institute for
System Dynamics and Control Engineering, University of Stuttgart,
Pfaffenwaldring 9, Stuttgart D-70182, Germany, EU,
Phone: +49/711/685/6296, FAX: +49/711/685/6371, e-mail: Conzelmann@isr.uni-stuttgart.de
2 Max-Planck Institute for Dynamics of Complex Technical Systems,
Sandtorstr. 1, 39106 Magdeburg, Germany 3 Department of Pathology, Anatomy and Cell Biology, Thomas
Jefferson University, 1020 Locust St., Philadelphia, PA 19107, USA
Signaling
networks play a crucial role in the regulation of a variety of cell functions.
A common feature of signaling pathways is the formation of multiprotein
signaling complexes. Receptors and scaffold proteins usually possess a number
of distinct domains and bind multiple partners. The number of feasible
multiprotein species grows exponentially with the number of binding domains and
can reach thousands or even millions [1].
We introduce a systematic approach, which allows reducing signal transduction models
considerably. A mechanistic description, which follows all possible states, is
substituted by a macro-description. This approach can be compared with thermodynamics.
At a microscopic level, a thermodynamic system should be described by position
and speed of each molecule. However, in most cases it would be sufficient to
know macroscopic properties like temperature, pressure and mass. The
macro-states in our approach follow the levels of occupancy of binding domains
[3,4]. These new quantitative indicators of the system (like degrees of
phosphorylation) are widely used in biology and have a higher significance than
the concentrations of each feasible multiprotein complex. This choice of states
corresponds to the view that molecular domains, instead of molecules, are the
fundamental elements of signal transduction networks [2]. In contrast to many
other model reduction methods our approach is independent of numerical values.
Qualitative biological knowledge about the domain-domain interactions is
sufficient to derive the model equations.
Our method is based on the system theoretical concept of observability. Using a
state space transformation, the complete
mechanistic model can be transformed to new coordinates (including the levels
of occupancy). Assuming that the levels of occupancy are the quantites of
interest, the transformed model equations can be separated into observable
states (states that influence the macro-states) and unobservable states (which
can be neglected). Applying this method to the adaptor molecule LAT (Linker for
activation of T-cells), the mechanistic model consisting of 36 differential
equations could be reduced to a 10-state model [4].
[1] Hlavacek et
al., 2003. The Complexity of Complexes in Signal Transduction, Biotechnol
Bioeng, 84(7):783-94 [2] Pawson et al., 2003 Assembly of cell regulatory
systems through protein interaction domains, Science 300:445-452. [3] Borisov et
al.,submitted [4] Conzelmann et al.,submitted
M-P06 Model
building in a systems biology company: the cell cycle and apoptosis
Cathy Derow, Chris Snell,
Christophe Chassagnole, John Savin and David Fell
, Physiomic
plc, The Magdalen Centre, The Oxford Science Park, Oxford OX4 4GA, U.K.,
Phone: +44/1865/784983, FAX: +44/8701/671931, e-mail: cderow@physiomics-plc.net,
Web: www.physiomics-plc.net
The work
flow structure of the company with respect to
model building will be discussed. This starts with literature research. It is
difficult to obtain the necessary quantitative data from the literature. For
this reason Physiomics plc has entered into a collaboration with the laboratory
of Marta Cascante to obtain good quantitative data
regarding a number of proteins involved in the cell cycle.
Literature research is followed by the creation of detailed maps to aid model
building. After this the programming to build the model takes place. The
company uses Jarnac for single-cell modelling and proprietary software for
modelling populations of cells. Then the model must be validated. This is done
using experimental data based upon which experiments are carried out with the
model. The model is then ready for use. The company’s aim is to use models of
biological systems to aid in drug discovery and development. The model was
recently used successfully to test data regarding cyclin-dependent kinase inhibitors being developed as
anti-cancer drugs. These processes can also
be aided by physiology-based pharmacokinetic (PBPK) modelling using PK-Sim the
software package Physiomics markets through a collaboration with Bayer
Technology services.
Recently literature research on apoptosis has been carried out in Physiomics
with the aim of adding modelling of apoptosis to the cell cycle model. This
research will be briefly discussed as well as the challenge of integrating
modelling of apoptosis with the existing model of the cell cycle.
M-P07 Na,K-ATPase
regulation via phospholemman phosphorylation
Claudia Donnet 1, Jia Li Guo 2, Amy Tucker
2 and Kathleen Sweadner 1
1 Laboratory of Membrane Biology, Massachusetts General Hospital /
Harvard Medical School, 55 Fruit St, Wellman #415, Boston MA 02114, USA,
Phone: 1 617 726-8560, FAX: 1 617 726-6529, e-mail: donnet@helix.mgh.harvard.edu
2 Cardiovascular Research
Center, University of Virginia Health System, Charlottesville, Virginia 22908
Many
hormones affect the Na+, K+ active transport, whose
signaling pathways involve kinases. PKG has been shown to affect the
Na,K-ATPase activity, however the mechanism is not known yet. Many attempts
have been made to characterize the effects of protein kinases on Na,K-ATPase function, however,
there is a lot of controversy since different researchers have found different
(even opposite) effects. The regulation seems to strongly depend on the tissue
under study, which led us to the idea that kinases may modulate Na,K-ATPase
activity via intermediary proteins. Recent evidence shows that phospholemman
(PLM), a single span transmembrane protein, can interact with Na,K-ATPase (1).
PLM is expressed in several tissues where the Na pump is known to play an key
regulatory role, such as heart and choroid plexus. PLM is known to be a
substrate for several kinases, having at least two known phosphorylation sites.
We propose that PLM could be the convergence point of different signaling
pathways of regulation of the Na pump.
To determine the functional effect of the presence of PLM, we generated a
PLM-knockout mouse (2). Heart sarcolemma from these mice had a much lower
Na,K-ATPase activity than wild type, while the level of expression of the pump
was only reduced by 15%. We have cloned and expressed human PLM in HEK cells,
which have negligible endogenous expression of PLM. Membranes isolated from
these cells were enriched in Na,K-ATPase by a procedure that removes many
contaminating proteins. Na,K-ATPase activity was determined on this membranes
as a function of Na concentration. Both Vmax and KNa were
significantly different between cells expressing PLM and mock transfected
controls. Incubation of choroid plexus membranes with ATP resulted in phosphorylation of PLM, which was
blocked by a non-specific kinase inhibitor,
suggesting the involvement of an endogenous membrane-bound
kinase. PKA is known to phosphorylate S68 and PKC to phosphorylate S63 and S68.
We obtained constructs for PLM mutants that constitutively mimick
phosphorylated states of these sites by replacing S residues by D residues (so
far: S63D, S68D, S63A, S68A). We are expressing these mutants in HEK cells to
characterize the effect of different states of phosphorylation on Na,K-ATPase
activity.
1 Feschenko M et al. (2003) J.
Neurosci. 23:2161-.
2 Li Guo J et al. (2004) Am. J.
Physiol. Heart Circ. Physiol. In the press.
M-P08 System
Properties of the Core Reactions of Apoptosis
Thomas Eißing 1, Carla Cimatoribus 1,
Frank Allgöwer 1, Peter Scheurich 2 and
Eric Bullinger 1
1 Institute for Systems Theory in Engineering, University of Stuttgart,
Pfaffenwaldring 9, Stuttgart D-70569, Germany, EU,
Phone: +49/711/685/7750, FAX: +49/711/685/7735, e-mail: eissing@ist.uni-stuttgart.de,
Web: http://www.ist.uni-stuttgart.de/~eissing/
2 Institute of Cell Biology
and Immunology, University of Stuttgart, Allmandring 31, 70569 Stuttgart,
Germany
Summary We present one qualitative and two
quantitative methods for discriminating biological models. Applying these to
the apoptotic core reactions not only helps to identify a suitable model
structure but enables several new and interesting insights into system
properties and performance of this signal transduction pathway.
Introduction Apoptosis is an important physiological process crucially
involved in the development and homeostasis of multicellular organisms.
Although the major signaling pathways leading from the extrinsic induction to
the execution of apoptosis have been unraveled during the past years, a
detailed mechanistic understanding of the complex underlying network remains
elusive. Previous modeling efforts focused on a descriptive behavior of large
parts of the pathway using data derived from population studies. However, new
data show that, within a single cell, the majority of caspases is activated
much faster than in cell populations.
Results Based on the current literature, we derive a differential
equation based model for the direct signal transduction pathway of receptor
induced apoptosis [1]. The new single cell data together with physiologically
motivated, theoretical considerations state a requirement for a bistable
behavior as a qualitative system property.
Bifurcation analyses show that bistability is only possible for parameter
values far away from those reported in literature, indicating the presence of
an additional control mechanism. We propose a new
model with a suitable additional control structure. This extended model
displays a bistable behavior and a fast caspase activation with kinetic parameters close to those reported in
literature.
Employing these two models as test cases, we present two quantitative
procedures for model discrimination based on the concept of robustness of
biological behavior. Taking into account the non-linearity of the systems, the
first method assesses the robustness of the bistable behavior with respect to
parameter variations. The second method evaluates the robustness of the
bistable threshold under the influence of the stochastic nature of reactions.
These discrimination criteria favor the extended model, in accordance with our
previous findings and novel experimental evidences.
Using a distributed input also allows us to reconcile the differences between
the observed kinetics of single cells and populations in terms of understanding
and modeling.
[1] Eißing et al. J. Biol. Chem.
279(35):36892-7.
M-P09 Meshfree
modelling of biological transport processes in complex domains
Martin Eigel and
Markus Kirkilionis
Department of
Mathematics, University of Warwick, -, Coventry CV4 7AL, United Kingdom,
Phone: +44 24 7657 4235, FAX: +44 24 7657 3133, e-mail: eigel@maths.warwick.ac.uk,
Web: http://www2.warwick.ac.uk/fac/sci/csc/people/markus_kirkilionis/group.html
Spatial
models of biological systems often require the representation of a hierarchy of
nested domains that can exhibit high geometrical complexity. Moreover,
dynamical changes of geometrical properties are common.
We are interested in analysing, modelling and simulating passive or mediated
transport processes in cells and
organelles. The focus lies on complex interactions between compartments, i.e.
domains and subdomains exchange substrates through their membranes, depending
on the presence of certain receptors and signals. Interface charactereristics
can likewise be connected to specific processes like concentrations and
gradients and change accordingly.
Systems of second order PDEs are used for describing the transport processes
under investigation. For the special requirements of biological models, classic
discretisations, based on a triangulation of the computational domain, often
cause technical problems, which esp. is true in 3D. Our research is based on
recent results in the theory of so called meshfree methods. This versatile approach
has gained much attention in the engineering community during the last years
for providing more flexible approximation space construction and nice adaptivity properties.
We strive to adopt and extend a class of meshfree methods to make it applicable
to the simulation of spatio-temporal processes in complex, nested domains.
One application is the simulation of a specific protein translocation pathway in the thylakoid, which
is the organelle in the plant cell's chloroplast where photosynthesis takes
place.
M-P10 Generating
conceptual models in Zebrafish zinc homeostasis: The first steps towards an
holistic view of zinc metabolism
Graham Feeney 1, Dongling Zheng 2,
Peter Kille 1 and Hogstrand Christer 2
1 Biosciences, Cardiff University, Main Builiding, Museum Avenue, Cardiff
CF10 3TL, UK, Phone: +44 (0) 2920876655, FAX: +44 (0) 2920874305,
e-mail: FeeneyGP@cf.ac.uk 2 Life Sciences, Kings College, London,
SE1 9NH
Zinc is
an essential metal ion, utilised by over 3000 proteins and therefore zinc
homeostatic control is an intrinsic requirement for
every metabolic pathway. Characterising the complex adaptive control systems
involved in zinc regulation, together with conceptual modelling of the
potential metabolic sequelae and their feedback pathways demands a Systems
Biology approach.
To date little work has been done to tie the role of individual zinc
transporters with the compartmental models of whole body zinc homeostasis, not
least because the epithelial sites of zinc exchange, the intestine, kidney and
biliary system, are not readily accessible for detailed characterisation. In
the teleost fish the gill is also a site of zinc exchange and this easily
accessible epithelial surface presents an ideal experimentally manipulatable
site for the clean and direct exploration of zinc handling in vivo without
confounding effects.
Building on the bedrock of physiological observations and mathematical
modelling of zinc homeostasis in teleost fish and humans, we are using multiple
approaches to inform the generation of conceptual models for zinc homeostasis
in the zebrafish:
1) Using molecular biology combined with comparative bioinformatics and genome mining, we have identified more than twenty
putative zebrafish zinc transporters each with its own profile of
tissue-specific expression and zinc-regulation. Parallel experiments in cell
culture have allowed the comparison of epithelial and non-epithelial zinc
handling.
2) Specific functional knockdown of the metal responsive transcription factor
MTF-1 mRNA using siRNA is revealing the role of this key zinc-regulator in the
transcriptome responses to changing zinc conditions.
3) Microarray characterisation of zinc responses inform the tentative models of
zinc homeostasis, allow the clear identification of novel zinc pathway genes
and provide a genome-wide view of the transcriptional response to zinc excess
and deficit.
Our Systems Biology approach has facilitated the formation of tentative
tissue-specific models for zinc handling that will be combined to form a
conceptual model for whole fish zinc homeostasis. In time such an approach will
facilitate the simulation of the effects of aquatic zinc pollution in fish.
Importantly, our approach is revealing the fundamental biology behind the
complex, essential and sometimes novel mechanisms in place to ensure that
sufficient zinc is supplied to meet metabolic requirements.
M-P11 Repression
of SOX6 transcriptional activity by SUMO modification
Fernandez-Lloris Raquel 1, Osses Nelson 1,
Jaffray Ellis 2, Shen LinNan 2,
Vaughan Owen Anthony 2, Girdwood David 2,
Bartrons Ramon 1, Rosa Jose Luis 1 and
Ventura Francesc 1
1 Physiological Sciences II, University of Barcelona, C/ Feixa Llarga
s/n, L'Hospitalet de Llobregat (BCN) 08907, Spain, Phone: +34687741911,
FAX: +441334462595, e-mail: rakylloris@yahoo.es
2 Centre for Biomolecular
Sciences, School of Biology, University of St. Andrews. North Haugh, St.
Andrews, KY 169 ST, Scotland, U.K.
SOX6, a
member of the SOX family of Sry-type HMG transcription factors, plays key
functions in several developmental processes, including neurogenesis, sex
determination and skeleton formation. In this report, we show that SOX6 is
covalently modified in vitro and in vivo by SUMO1, 2 and 3 on two consensus
sites (IK364NE and VK377DE). Mutation of both lysines to arginine abolished
SOX6 sumoylation and increased SOX6 transcriptional activity as well as
SOX6/SOX9 synergistic activation of the Col2a1 enhancer. SUMO dependent
repression of SOX6 transcription was demonstrated by Ubc9 overexpression
whereas siRNA to Ubc9, cotransfection of a catalytically inactive UBC9 or a
SUMO specific protease increased SOX6 transcriptional activity.
Immunofluorescence analysis showed a predominant diffused nuclear localization
of SOX6 when expressed alone. Coexpression of SOX6 with SUMO1 and/or SUMO2
results in the appearance of SOX6 in a punctate nuclear pattern that colocalized with PML. PML body
localization of SOX6 was abolished by mutations in SOX6 sumoylation sites.
Thus, SUMO modification of SOX6 alters its subnuclear localization and leads to
transcriptional repression.
ACKNOWLEDGEMENTS
R. Fernández-Lloris received a predoctoral fellowship from the Generalitat de
Catalunya and Nelson Osses is a recipient of a postdoctoral fellowship from the
Fundación Carolina. This research was supported by grants from the MCyT
(BMC2002-00737) and Generalitat de Catalunya (Distinció de la Generalitat a
joves investigadors) .
M-P12 Network
synchronization from population to cell level
Laurent Gaubert and Magali Roux-Rouquié
Laboratoire
d'Informatique de Paris 6 (LIP6), Université Pierre et Marie Curie, 8 rue du
Capitaine Scott, Paris 75015, France, Phone: +33/01/44 27 88 25,
FAX: +33/01/44 27 74 95, e-mail: laurent.gaubert@lip6.fr
Expression
profiles prepared from cell cultures and tissues, are used to infer molecular
networks underlying biological functioning. Assuming that each cell
instantiates a temporal version of the same functional network, expression
profiles measures a mixture over time of this network according to cellular
dynamics. To address these issues in network engineering, we studied the
relations when scaling models from individual to population cell levels.
In one individual cell, the list of RNA concentrations is a function x(t). An
expression profile made at time T will involve a population of cells, and
because of the divergent evolution of individual cells over time, the RNA
content at the population level won't exhibit the dynamic of the single functional
network.
As an example, we modeled the list of ages of cells as random variables, with
common distribution. Under some conditions, we show that RNA concentration is
approaching a quantity X(T) that is the image of x(t) by an integral operator.
Whatever the vector X(T) still represents a list of concentrations, it does not
behave as the vector x(t), and in some bad cases, the measurements can’t be
used to infer the network directly. We must then identify the kernel that
define the integral operator. But modeling it appears to be a hard task, as the
phenomenon of desynchronization inside a population of cell seems to be
unreachable. Nevertheless, under some conditions, we can avoid this step, and
directly build approximations of the initial curves of concentrations x(t).
Generalizing the model, we propose a numerical method, which could also lead to
the design of specific experiments.
M-P13 Impaired
gene expression in Sjogren's disease
Adi Gilboa-Geffen and Hermona Soreq
Biochemistry,
Silverman institute of life science, Givat Ram, Jerusalem 91904, Israel,
Phone: 972-2-658-5450, FAX: 972-2-658-6448, e-mail: adigil@pob.huji.ac.il
Sjogren’s
disease (SjS) involves salivary glands malfunctioning and disrupted cholinergic
signaling. Here, we report SjS-associated changes in the acetylcholine (ACh)
hydrolyzing enzyme, acetylcholinesterase (AChE) in acinar cells and secretory
ducts from SjS salivary glands. Both AChE mRNA and interleukin 1 (IL-1) levels
were reduced in acinar gland cells, suggesting an accompanying immunomodulary
decline. In contrast, secretory ducts showed increased AChE expression and
elevated IL-1 levels, demonstrating an inflammatory reaction associated with
reduced ACh signaling and relieved suppression of proinflammatory cytokines
production. Increasing severity of disease symptoms further involved AChE
increases in acinar cells and residing lymphocytes and larger fractions of
acini with polarized AChE-R mRNA distribution. To test relevance of AChE
distribution for ACh metabolism, we used salivary gland sections from
transgenic mouse lines overexpressing distinct splice variants of human AChE
yet with similar ACh hydrolytic activities. Significant increases in the
fraction of duct, but not acinar cells expressing nuclear “synaptic” AChE-S
were found in TgR mice with excess of the stress-induced hAChE-R protein.
Significant IL-1 decreases ,down to 60+6% of FVB/N controls, occurred in
acinar, but not duct cells, of TgR mice (p<0.001, Student’s t-test) as well
as in TgS mice overexpressing human AChE-S and murine AChE-R (p<0.005).
TgSin mice expressing genetically inactivated AChE-S presented a trend for
further IL-1 decline (p>0.3), altogether suggesting causal involvement of
cholinergic signaling. Our findings suggest contribution of the
stress-inducible AChE-R splice variant to the initiation and progression of the
compromised secretory features and immune profile of salivary glands from Sjogren’s
disease patients.
M-P14 Modeling
the synchronization of circadian oscillators in the suprachiasmatic nucleus
Didier Gonze 1, Samuel Bernard 2,
Christian Waltermann 2, Achim Kramer 3 and
Hanspeter Herzel 2
1 Unite de Chronobiologie Theorique, Universite Libre de Bruxelles, Bvd
Triomphe, Brussels 1050, Belgium, Phone: +32/2/6505770,
FAX: +32/2/6505767, e-mail: dgonze@ulb.ac.be,
Web: http://www.ulb.ac.be/sciences/utc/
2 Institute for Theoretical
Biology, Humboldt Universitat zu Berlin, Invalidenstrasse 43, D-10115 Berlin,
Germany
3 Laboratory of
Chronobiology, Institute of Medical Immunology (Charite), Humboldt Universitat
zu Berlin, Hessische Str. 3-4, D-10115 Berlin, Germany
In
mammals, the suprachiasmatic nucleus (SCN) constitutes the central pacemaker
which controls circadian rhythms in physiology and behavior. Individual SCN
cells exhibit sustained circadian oscillations with periods ranging from 20 to
28 h. These oscillations are generated by a molecular regulatory network based
on a negative feedback loop. On the tissue level, SCN neurons display a
significant degree of synchrony. Neurotransmitters have been shown to play a
crucial role in the coupling mechanism. Depending on the type of
neurotransmitters released by the cells, the SCN has been subdivided into two
regions, the ventro-lateral and the dorso-medial parts. Only the neurons
present in the first region are sensitive to light and convey the light signal to the dorso-medial region.
Furthermore when isolated from the ventrolateral part, dorso-medial cells get
out of phase. However, unexpectedly, the dorso-medial part is phase leading. We
present a mathematical model for the coupling of a population of ten thousand
circadian oscillators in the SCN. The cellular core oscillator is described by
a three-variable model relying on a negative feedback loop. The coupling is incorporated
through the global level of the neurotransmitter concentration. We first show
that such a global coupling is efficient to synchronize a population of
thousands cells. The synchronized cells can be entrained by a 24 h light-dark
cycle. The study of the interaction between two population of oscillators,
representing the two regions of the SCN, shows that the driven region can be
phase leading. An experimentally testable prediction of our model is that
synchrony is reached when the average neurotransmitter concentration brings the
cells outside their region of individual sustained oscillation.
M-P15 Modelling,
Enzyme kinetics & Fluorescence Imaging of the NF-kappaB Signalling Pathway
Adaoha EC. Ihekwaba 1, Rachel Grimley 2,
Neil Benson 2, David Broomhead 3 and Douglas
B. Kell 1
1 School of Chemistry, University of Manchester, Sackville Street,
Manchester M60 1QD, UK, Phone: +44/161/200/4414,
FAX: +44/161/200/4556, e-mail: a.ihekwaba@postgrad.manchester.ac.uk
2 Pfizer Central Research,
Ramsgate Road, Sandwich, Kent, CT13 9NJ, UK
3 School of mathematics,
University of Manchester, Sackville Street, Manchester, M60 1QD, UK
Analysis
of cellular signalling interactions is expected to create an enormous
informatics challenge, perhaps even greater than that of analysing the genome. We
have reconstructed a model of the NF-kappaB signalling pathway, containing 64
parameters and 26 variables, including steps in which the activation of the
nuclear factor kappaB (NF-kappaB) transcription factor is intimately associated
with the phosphorylation and ubiquitination of its inhibitor kappaB by a membrane-associated
kinase, and its translocation from the cytoplasm to the nucleus. We apply
sensitivity analysis to the model. This identifies those parameters in this
IkappaB-NF-kappaB signalling system (containing only induced IkappaBalpha
isoform) that most affect the oscillatory concentration of nuclear NF-kappaB
(in terms of both period and amplitude). We have also measured enzyme kinetics
of IKK2 with the substrates in this pathway and have studied the impact of
these data in the model. We observed these new data to have a profound effect
on the model. We also carried out fluorescence cell-based imaging and studied
the correlation of measured enzyme kinetics with the observed images.
Ihekwaba, A. E. C., et al. (2004).
Sensitivity analysis of parameters controlling oscillatory signalling in the
NF-kappaB pathway: the roles of IKK and IkappaBalpha. Syst. Biol. 1, 93
Nelson, D. E., et al (2004). Oscillations in NF-kappaB signaling control the dynamics of gene expression. Science 306, 704
M-P16 Towards
a systems biology of signal transduction by insulin and insulin-like growth factors.
Shaukat Mahmood, Jane Palsgaard,
Soetkin Versteyhe, Maja Jensen and Pierre De Meyts
Receptor Biology
Laboratory, Hagedorn Research Institute, Niels Steensens vej 6, Gentofte
DK-2820, Denmark, Phone: +45 44439339, FAX: +45 44438000,
e-mail: shma@novonordisk.com,
Web: http://www.hagedorn.dk/
Insulin
and the related insulin-like growth factors exert their effects by binding to and activating
separate but related membrane receptor tyrosine kinases, triggering a
complex intracellular network of signaling pathways. Disorders of these systems
lead to serious diseases like diabetes,
metabolic syndrome, small size and cancer.
Physiological and genetic evidence suggests that insulin is primarily a
metabolic regulator, while the related IGF-I and IGF-II are primarily growth
promoters. Despite extensive studies, the molecular basis of this specificity
of actions is still poorly understood, because:
- the ligands as well as the receptors are closely related
- their signaling network are largely overlapping
- each receptor has been shown to be able to mediate the effects of each ligand
in a given cellular context.
It is clear that new methodological approaches that include cellular imaging,
real-time kinetic analysis and network integrated
analyses are required to progress in understanding the combinatorial nature of
signaling specificity.
Our approach to this problem consists in combining multiple novel
methodologies:
• Real-time kinetic measurements of receptors and proximal signalling molecules
(IRS 1-6, PTP-1B) using FRET (fluorescence resonance energy transfer) and BRET (bioluminescence resonance
energy transfer).
• Confocal microscopy for live cellular imaging of signalling events using
fluorescent probes.
• Microarray gene profiling.
• siRNA interference with key signalling proteins.
Target cells explored include human preadipocytes, a rat beta cell line, human
myoblasts from normal and diabetic subjects, myoblasts from insulin receptor
knockout mice, and mouse cell lines devoid of IGF-II/M6P receptors.
The ultimate goal is to reverse engineer end point biological data and
correlate them by mathematical modelling with the kinetic aspects of receptor
binding of IGF-I and II as well as insulin analogues with different kinetics,
and real-time measurement of key signalling steps.
This program is
supported by the BIO+IT programme of Øresund IT Academy, Medicon Valley
Academy, and the Danish Ministry of Science, Technology and Innovation.
M-P17 BOOLEAN
analysis of the signaling network triggered by neurotrophic factors and
extracellular matrix in sensory neurons
Mikhail Paveliev, Maria Lume and Mart Saarma
Institute of
Biotechnology, University of Helsinki, Viikinkaari 9, Helsinki 00014, Finland,
Phone: +358/408332701, FAX: +358/919159366, e-mail: Mikhail.Paveliev@helsinki.fi
Neurotrophic
factors and laminins are important regulators of posttraumatic regeneration in
the nervous system. Cyclin-dependent kinase 5 (Cdk5) regulates cytoskeleton
mobility and mediates the effect of neurotrophic factors on axonal growth in
various types of neurons. Here we show that NGF, GDNF and neurturin activate
axonal growth in mature dorsal root ganglion neurons in the absence of Cdk5
activity as the effect of these neurotrophic factors was not affected by 50uM
of roscovitine. On contrary laminin-dependent outgrowth in the absence of
neurotrophic factors was fully blocked by roscovitine. GDNF- and
laminin-dependent types of axonal growth also have different sensitivity to src
inhibitor SU6656. We use Boolean networks formalism to analyze differential
contribution of neurotrophic factors- and laminin-triggered pathways to the
converging signaling network.
M-P18 A
topological analysis of the human transcription factor interacting network
Carlos Rodríguez-Caso 1, Miguel Ángel Medina 1 and
Ricard V Solé 2
1 Molecular Biology and Bichemistry, Universidad de Málaga, Campus de
Teatinos s/n, Málaga 29071, Spain, EU, Phone: +34/952137135, FAX: +34/952
131674, e-mail: caso@uma.es 2 ICREA-Complex Systems Lab,
Universitat Pompeu Fabra, Barcelona, Spain.
Protein-protein interactions are one of the bases of
regulation in biosignalling and gene expression regulation. Massive biological
data acquisition is allowing to study biological systems as whole. For this
purpose, it is necessary the use of systemic approaches, such as is the case of
graph theory. In this context, several protein interacting networks or so
called interactomes have been described in the last years.
In this work, we study the human protein-protein interacting transcription
factor network, obtained from data contained in a transcription database
(Transfac), using graph theory.
Clustering coefficient and diameter of the network show a small-world pattern and distributions for degree, betwenness
centrality and clustering suggested a scale-free behaviour, as it is occurs
with other biological networks. Topological overlap matrix and correlation
profiles were calculated and we could see that this transcriptional network is
also hierarchical and shows modularity. Modules are composed of factors with
both functional, structural features. Some modules include factors that share
function but not structural features. However, others showed similar structure
and common functionality or, indeed, phylogenetic origins. According to these
data, we suggest that two motors have modelled the transcription network. The
first one would be based on gene amplification and shuffling of certain domains
such as bHLH and Zn fingers motifs, capable to form dimers allowing for the
interaction of proteins, most probably due to the necessity of new regulatory
factors during multicellularity acquisition in the evolution. On the other
hand, proteins without these general binding domains could establish
connections by a random process yielding some benefit for the system. Some
features of this network can be derived from biological and structural
constrains, such as the grade of autolinks explained by the use of these kind
of domains that allow a regulation with a low cost of structure types.
We also show that it is possible to identify potential relevant elements in the
system using topological properties. For instance most of highly-connected
nodes are related with proliferative processes. On the other hand, nodes with
few connections and high betweenness centrality levels could be important due
to the neighbours they connect.
We conclude that using topological approaches
for the study of a not yet wholly-described system can give relevant
information about the nature of gene regulation.
M-P19 Molecular
dissection of the key LGS residues involved in the control of glycogen biosynthesis.
Susana Ros and
Joan J. Guinovart
Metabolic
engineering and diabetes therapy, Barcelona Science Park, c/ josep samitier 1-5, Barcelona
08028, Spain, Phone: +34934037163, FAX: +34934037114, e-mail: sros@pcb.ub.es, Web: http://www.pcb.ub.es
Glycogen
synthase (GS) catalyses the addition of glucose residues to the non-reducing end
of a nascent glycogen chain via alpha-1,4-glycosidic
bonds, using UDP-glucose as substrate. Two isoforms of mammalian GS have been
described; most tissues express the muscle form, whereas the liver isoenzyme (LGS) appears to be
tissue-specific. GS activity is highly regulated via phosphorylation and
allosteric effectors, mainly glucose 6-phosphate (Glc-6-P).
It is generally accepted that the reaction catalysed by GS is rate-limiting for
glycogen synthesis in all organs [1]. The importance of this enzyme in the
overall process of glycogen deposition is confirmed by the observation that
overexpression of GS in cultured hepatocytes increases glycogen accumulation
[2]. This is the consequence of the action of Glc-6-P produced by endogenous
glucokinase (GK). This metabolite causes the allosteric activation of the total
amount of LGS, through a conformational rearrangement that converts this enzyme
into a better substrate for protein phosphatases, which catalyze its dephosphorylation,
thus leading to an increase of ‘active’ LGS. Moreover, when GK is
overexpressed, the increase in Glc-6-P results in a higher degree of activation
of the endogenous GS, which also leads to the deposition of larger amounts of
glycogen. Finally, when both enzymes are overexpressed, there is a combination
of the two effects. Therefore GK and LGS share the control of hepatic glycogen biosynthesis, in which
the control exerted by LGS is in turn controlled by GK.
We aim to make the molecular dissection of the key LGS residues involved in the
control of glycogen biosynthesis. To this end, we are currently generating
adenoviruses of LGS where Ser to Ala mutations have been introduced
individually or in combination at phosphorylation sites of the enzyme. These
adenoviruses will provide further insight into the role of GS and GK in the
control of LGS.
REFERENCES: 1
Roach P. J., et al. (1998) J. Basic Clin. Physiol. Pharmacol. 9, 139
2 Gomis R., et al. (2000) Biochem J. 351,811
M-P20 Analysis
of the signaling network involved in the activation of T-Lymphocytes
Julio Saez-Rodriguez 1, Xiaoqian Wang 1, Birgit Schoeberl 2, Steffen Klamt 1, Jonathan Lindquist 3, Stefanie Kliche 3, Buckhart Schraven 3 and Ernst Dieter Gilles 1
1 Systems Biology
Group, Max-Planck-Institute for Dynamics of Complex Technical Syste,
Sandtorstr. 1, Magdeburg 39104, Germany, Phone: 0049-391-6110-479,
FAX: 0049-391-6110-552, e-mail: saezr@mpi-magdeburg.mpg.de,
Web: http://www.mpi-magdeburg.mpg.de/people/saezr/ 2 MIT/Merrimack Pharmaceuticals, Cambridge, USA
3 Institute of Immunology,
Otto-von-Guericke University, Magdeburg, Germany
Despite
the intense research and the considerable progress within the last years, how
T-lymphocytes are able to distinguish foreign agents among the myriads of
components of our own body is still not fully understood. T-cell reactivity has
to be exquisitely regulated since not only a decrease (since it weakens the
defense against pathogens), but also an increase (which can lead to autoimmune
disorders) can be dangerous.
The central sensor in the recognition process is the T-Cell receptor (TCR).
Upon binding of an antigen to the TCR, several signaling processes take place.
Additionally to the ligand itself, other signals from other cells of the immune
system are also sensed by different receptors on the T-cell membrane. The
resulting signaling network, of extreme complexity, assures that T-cells become
activated only when and where they should. We try to unravel this complexity by
a combination of analyses at different levels and with different tools. On one
hand, we perform a qualitative analysis of an extensive portion of the network.
On the other hand, we concentrate on a key process, namely the activation of
the MAPK Cascade and its dynamics. For this pathway, a
detailed mechanistic model is set up. The model is fitted by quantitative
measurements of activation levels of key proteins performed on naive T-cells
from transgenic mice. The experiments will also be extended to cases where the
network is perturbated (e.g. by knock-out mice, inhibitors, iRNA, etc.).
We try to address several key questions, such as the importance of the
antigen:TCR affinity, the topology of the TCR signal transduction network -in
particular feedbacks mechanisms (Reth, M., et al., Nat Rev Immunol.,(2004),
4(4):269),- and the role of transmembrane adaptor proteins like PAG (Horejsi,
V., et al. Nat. Rev. Immunol., (2004), 4(8):603). For this purpose the model is
decomposed into modules connected unidirectionally, which are subsequently analyzed.
Rewiring together the modules new insights into the whole cascade can be
obtained (Saez-Rodriguez, J., et al., IEEE CSM, (2004),
24(4):35).
M-P21 Flavo-di-iron proteins: role in microbial detoxification
by NO
Francesca
Maria Scandurra 1, Paolo Sarti 1, PierLuigi Fiori 2, Elena Forte 1, Alessandro Giuffrè 3, P. Rappelli 2, G. Sanciu 2,
Daniela Mastronicola 4, Miguel Teixeira 5 and
Maurizio Brunori 1
1 Department of Biochemical Sciences "A.Rossi-Fanelli",
University of Rome "La Sapienza", piazzale Aldo Moro n°5, Rome IT
00185, Italy, Phone: +39-06-49910944, FAX: +39-06-4440062,
e-mail: francescascandurra@yahoo.it
2 Departement of Biomedical
Scinces University of Sassari 3
CNR Institute of molecular Biology and Pathology University of Rome "La
Sapienza" 4 IFO, Cancer
Institute Regina Elena of Rome 5
Instituto de tecnologia Quimica e Biologica, Universidade Nova de Lisboa
The
flavo diiron proteins (FDP) are soluble flavoproteins discovered in facultative
and strict anaerobic prokaryotes. They are built by the two core domain: a
metallo beta-lactamase domain at the N-terminal region harboring a non heme
di-iron site and a flavodoxin-like domain. Recently it has been proposed that
this proteins have a role in NO metabolism [1]. Evidence for NO reductase
activity was provided by amperometric measurements on the ricombinant isolated
enzyme from Escherichia coli [Gomes]. It has been generally
believed that FDPs were restricted to the Archea and Eubacteria domains, until
genomic analysis [2] demonstrated that microaerobic pathogenic protists (like Giardia
lamblia, Entamoeba histolytica and Spironucleus barkhanus) have
genes coding for FDPs. The working hypothesis is that via lateral gene transer these pathogenic mireoorganism
acquired NO-reductase activity to counteract tha microbicidal action of NO
produced by the host macrophages.
We have direct evidence that the microaerbic eukaryotic parasite Trichomonas
vaginalis, the causative agent of trichomoniasis which seems to have at
least three genes coding tor FDP, degrades NO under anaerobic conditions
[Sarti]. The NO degradation activity is: maximal at low NO concentraton,
NADH-dependent, cyanide insensitive and is inhibited by O2. In
addition a protein band was immunodetect using
antibodies again the E. coli FDP. This acquired NO reductase activity
might be at the basis of mechanism of defense which helps T. vaginalis
to survive at high fluxes of NO produced by immune response of human host.
1)Gardner et
al (2002) J.Biol.Chem. 277, 8172-8177.
2)Gomes et al (2002)J.Biol.Chem 277, 25273-25276.
3)Andersson et
al (2003) Curr. Biol. 13, 94-104.
M-P22 Quantitative
modeling of EGFR-internalization as a mechanism of signaling specificity
Hannah Schmidt-Glenewinkel, Constantin Kappel and
Ivayla Vacheva
Theoretical
Bioinformatics, German Cancer Research Institute, Im Neuenheimer Feld 580,
Heidelberg D 69120, Germany, Phone: +49/6221/423609,
FAX: +49/6221/423610, e-mail: H.Schmidt-Glenewinkel@dkfz.de
Epidermal
Growth Factor binding to its receptor (EGFR) results in a variety of cellular
responses, communicated by phosphorylation cascades via receptor tyrosine
kinases (RTKs) and ultimate activation of transcription factors like Myc, Fos
and Elk. Signal attenuation is achieved by receptor internalization to
endosomes and eventual degradation. However, there is growing evidence that
intracellular trafficking to endosomes itself constitutes a mechanism of
regulating signal specificity. Suppression of endocytosis does
not lead to a uniform up-regulation of EGF-dependent signaling events and
numerous RTK-specific regulators affecting endosome sorting have been
identified [1].
We have developed a quantitative mathematical model, which explicitly incorporates
intermediate steps of transport and degradation of internalized receptor.
Specific key questions the model addresses are: Which are the crucial features
of receptor endocytosis mediating signal specificity? Which are the critical parameters
in the system that determine the fate of internalized EGFR? The existence of
phosphatase gradients in the cytosol has been observed
[4], our model investigates whether signaling endosomes are used to overcome
these phosphatase barriers and thus propagating the signal inside the cell.
Model development is driven by in vivo data retrieved from fluorescence
microscopy images on living cells [2]. Here, the phospohrylation state of EGFR
and downstream effectors such as Mek, Ras and Erk have been imaged under
conditions perturbating endocytosis and down-regulated phosphatase activity by
measuring fluorescence resonance energy transfer.
In order to deal with the problem of underdetermined systems, we are using a
sensitivity analysis of key parameters previously employed in our lab [3],
which reduces the effective dimension of the model by, e.g. identifying
correlated parameters.
Based on the model we have developed a simulation software, which performs
parameter estimation using fluorescence images.
We believe that a quantitative analysis of this pathway will lead to new
insights in this important pathway, since it is well established by now that
kinetic differences lead to very different cellular
behaviour.
[1] Vieira, A.V. et al (1996) Science
274, 2086 [2]Bastiaens, P. et al (2004)
J Biol Chem 279, 36972
[3] Bentele, M. et al (2004) J Cell Biol 166, 839 [4] Haj, F.G et al (2002), Science, 295
M-P23 Retroelement
insertion polymorphism in cell line identification.
Svetlana V. Ustyugova, Anna L. Amosova, Yuri B. Lebedev
and Eugene D. Sverdlov
Laboratory of
Structure and Functions of Human Genes, Shemyakin-Ovchinnikov Institute of
Bioorganic Chemistry, Miklukho-Maklaya 16/10, Moscow 117997, Russia,
Phone: + 7(095) 330 6329, FAX: + 7(095) 330 6538, e-mail: sveta@humgen.siobc.ras.ru
Cell
lines are an indispensable, renewable resource for numerous studies in various
fields of modern biology and medicine. It is assumed that the cell features in
a particular cell line are similar with that in the human tissue from which
this line originated. However, most human cell lines are prone to various sorts
of genetic rearrangements that affect biochemical, regulatory and other
phenotypic features of cells during their cultivation. Various chromosomal
abnormalities including aneuploidy, numerous rearrangements and loss of
chromosome regions are characteristic alterations of cell genomes, especially
if the cells are of tumor origin. Therefore, it is highly
desirable to permanently monitor the authenticity of the cell lines and/or to
have reliable cell line identification techniques to make sure that the cell
lines to be used in experiments are exactly what is expected. To this end, we
developed a set of informative markers based on insertion polymorphism of human
retroelements. The set includes 47 pairs of PCR primers corresponding to
introns of the human genes with dimorphic L1 and Alu insertions. Using locus
specific PCR assay, we have genotyped 10 human cell lines of various origin.
For each of these cell lines characteristic fingerprints were obtained. An
estimated probability that two different cell lines possess the same marker
genotype is about 10- 18. Therefore, the proposed set of markers
provides a reliable tool for cell line identification.
M-P24 RNAi screening for novel components of mammalian Hedgehog and Wnt pathways
Markku Varjosalo, Antti Oinas and Jussi Taipale
Biomedicum,
Molecular/Cancer Biology Research program, University of Helsinki,
Haartmaninkatu 8, Helsinki FI-00014, Finland, Phone: +358 9 191 25546,
FAX: +358 9 191 25554, e-mail: markku.varjosalo@helsinki.fi
RNA-mediated
interference (RNAi) is a powerful reverse genetic tool to silence gene expression in organisms ranging from plants
to humans. The initial finding that siRNAs and constructs encoding short
hairpin RNAs (shRNAs) could trigger gene silencing in mammals has been refined
and extended by the combination of shRNA with viral and episomal vectors that
allow transient or stable silencing in mammals (1). This has encouraged and
broadened interest in using mammalian cells for both forward and reverse
genetic screens.
In this study we tested the feasibility of two different vector-based
Gateway-compatible RNAi systems (siRNA and shRNA) by analyzing their knockdown
efficiency by measuring the ability of these vectors to specifically suppress
target gene –Renilla-luciferase fusion proteins. In our hands, the shRNA
vectors had much greater efficiency in this assay. We next validated the shRNA
system by testing the ability of co-transfected shRNAs targeting known
components of the Hedgehog or Wnt signaling pathway to inhibit luciferase-based
Hh or Wnt signaling reporters, respectively. The luminescence based signalling
assays allows us to analyze the effect of knockdown of different genes in
high-troughput manner in cultured cells. To identify novel effectors of these
pathways, we designed shRNAs targeting nearly all mouse kinases using
thermodynamic criteria (2), and genereted the corresponding arrayed shRNA-library.
Using this vector-based shRNA-library, we show that transfected as well as
endogenous genes can be efficiently inhibited. We are currently using this
library to identify new components of mammalian Hedgehog and Wnt signaling
pathways. These findings highlight the general utility of this vector-based
RNAi technology in suppressing gene expression in mammalian cells in
high-throughput and cost-efficient manner.
1. Paddison PJ & Hannon GJ. (2003)
Curr Opin Mol Ther. 5(3):217-24
2. Schwarz, D. S. et al. (2003) Cell 115, 199-208
M-P25 Modeling
emergent networks by dynamic reconstruction in silico
Hao Zhu and
Pawan Dhar
Systems Biology,
Bioinformatics Institute, Biopolis Street, Singapore 138671, Singapore,
Phone: +65-64788303, FAX: +65-64789048, e-mail: zhuhao@bii-sg.org
Molecular
interactions in cells of metazoans are context dependent, with rich semantics,
evolve with cell fate specification, reprogrammable, and create emergent
networks. To unveil the dynamic process of how signaling and mis-signaling
among molecules leads to patterning and mis-patterning, functioning and
mis-functioning of cells, modeling and simulation that couples intra- and
inter-cellular, molecular and cellular events are required to systematically
probe the evolution of networks and the order and timing of signaling. We
advocate a modeling paradigm and introduce a modeling platform, which are based
on the combining of cellular automata parallelism and object-oriented
programming. The proposed method features: (1) object-oriented extensions to a
language-based cellular automaton to create a two-tier parallelism; (2)
event-driven signaling; (3) flexible qualitative and quantitative computation;
(4) dynamic capture and display of signaling events and molecular attributes;
(5) dynamic reconstruction of signaling networks. The cell-object(molecule)
two-tier structure enables the modeling of parallel molecular networks in
cells; and the separation of discrete signaling simulated as message passing
among objects, from continuous computation in the form of ODE and so on, allows
a realistic description of large scale systems. The dynamic captured signaling
explains both the evolution of networks and the order and timing of signaling
in cells. With an exemplary 2D model on mouse somite segmentation, we exhibit
that networks show dynamics and evolve with cell fate specification. Complex,
including seemingly chaotic, global behaviors can be well elucidated with the
captured signaling events. In addition to the partnership of molecular
interactions which determines network topology, we find the order and timing of
signaling, termed as network dynamics here, are also crucial. Signaling
at wrong time and in wrong order may explain a large group of anomalies, which
may not necessarily have a wrong molecular interaction topology. An ectopic or
unduly signal may fundamentally change the signaling inside
a cell, and thus the fate of the cell, as widely observed in many cell fate
transformation events. These factors often mingle together context dependently,
making signaling emergent and elusive and demanding effective computational
simulation.
M-PoP1 Niels Aarsaether
M-PoP2 SYMBIONIC: A European initiative on the Systems
Biology of the neuronal cell
Ivan Arisi
PKO-Bioinformatics,
Lay Line Genomics, via di Castel Romano, 100, Roma 00128, Italy,
Phone: +39/06/80319041, FAX: +39/06/80319065, e-mail: i.arisi@laylinegenomics.com,
Web: www.laylinegenomics.com
No abstract submitted.
M-PoP3 In vitro systems for modelling of signal transduction in hepatocytes
Patricio Godoy, Katja Breitkopf, Loredana Ciuclan,
Eliza Wiercinska, et al. [i.e.,
BMBF-Network Systems Biology], and Steven Dooley
Molecular Alcohol
Research in Gastroenterology, II. Medizinische Klinik, Universitätsklinikum Mannheim, Theodor-Kutzer-Ufer
1-3, Mannheim 68167, Germany, Phone: 0049-621-383-3768,
FAX: 0049-621-383-1467, e-mail: steven.dooley@med.ma.uni-heidelberg.de
The aim
of our consortium is to model specific signal transduction pathways as well as
detoxification processes in hepatocytes. For this purpose, standardized in
vitro systems with hepatocytes are needed. We established standardized
hepatocyte systems including (i) cultured primary human and mouse hepatocytes,
(ii) stem cell derived hepatocyte like cells, including hepatopancreatic
precursor cells that have been conditionally immortalizeed by Bmi-1 and hTERT,
(iii) constitutively as well as conditionally immortalized adult and fetal
hepatocytes.
TGF-b participates in several aspects of liver damage, including hepatocyte apoptosis. We
established tools to monitor TGF-b signalling in murine and human hepatocytes.
Functional ectopic protein expression of YFP and CFP fusion
proteins for TGF-b type I (ALK5) and type II receptors was confirmed
biochemically and with confocal microscopy. The constructs will be used with
the Opera System (Evotec) to study endocytotic trafficking of the receptor
complex during signalling (TGF-b treatment) and degradation processes
(Smad7/Smurf1). TGF-b signalling was documented by phosphorylation of Smad2,
activation of a Smad binding element containing reporter construct,
(CAGA)9-MLP-Luc, and by triggering apoptosis, presented as annexin staining or
DNA laddering assay. To interfere with specific
steps of TGF-b signalling in hepatocytes, we used dominant negative TGF-b
receptors, Alk5 inhibitor SB 431542, Smad7, Ski and SnoN.
For modelling purposes, TGF-b receptor complex cargo will be monitored
optically with the Opera based on time course experiments after ligand binding
and/or after the above mentioned treatments.
M-PoP4 Module
dynamics of the GnRH signal transduction network
Karen Page 1 and David Krakauer 2
1 Department of
Computer Science, University College London, Gower Street, London WC1E 6BT, UK,
Phone: +44 20 7679 3683, FAX: +44 20 7387 1397, e-mail: k.page@cs.ucl.ac.uk, Web: www.cs.ucl.ac.uk/staff/K.Page
2 Santa Fe Institute, Santa
Fe, NM 87501, USA
We
analyze computational modules of a frequency decoding signal transduction network. The
gonadotropin releasing hormone (GnRH) signal transduction network mediates the
biosynthesis and release of the gonadotropins, luteinizing hormone (LH) and
follicle stimulating hormone (FSH). The pulsatile pattern of GnRH production by the hypothalamus has a
critical influence on the release and synthesis of gonadotropins in the
pituitary. In humans, slower pulses lead to the expression of the beta subunit
of the LH protein and cause anovulation and
amenorrhea. Higher frequency pulses lead to expression of the alpha subunit and
a hypogonadal state. The frequency sensitivity is a consequence of the
structure of the GnRH signal transduction network. We analyze individual
components of this network, organized into three network architectures, and
describe the frequency-decoding capabilities of each of these modules. We find
that these modules are comparable to simple circuit elements, some of which
integrate and others which perform as frequency sensitive filters. We propose
that the cell computes by exploiting variation in the time scales of gene activation (phosphorylation) and gene
expression.
M-PoP5 Experimental
design for model discrimination in cellular signal transduction
Clemens Kreutz 1,
Jörg Stelling 2, Thomas Maiwald 1 and Jens Timmer 1
1 Centre for Data Analysis and Modelling, University of Freiburg,
Eckerstr. 1, Freiburg 79104, Germany, Phone: +49 761 203 5829,
FAX: +49 761 203 5967, e-mail: jeti@fdm.uni-freiburg.de, Web: http://www.fdm.uni-freiburg.de/~jeti/
2 Max-Planck-Institute for
Dynamics of Complex Technical Systems, Sandtorstr.1, 39106 Magdeburg, Germany
Dynamic modeling of signal transduction pathways helps to understand cellular processes. This
systems biology approach requires quantitative, time-resolved measurements of
the participating molecules, which are usually expensive and time consuming.
Hence, experiments should be designed in advance being capable to (a) specify
the correct model and (b) determine the kinetic parameters with a minimal effort.
Several methods are proposed specifying the optimal experimental design in
order to estimate the kinetic parameters, given a mechanistic model. However,
often no explicit model exists for a biochemical system or serveral models are
suggested and the question arises, which model describes the reality most
accurately.
We propose a new approach to determine the optimal experimental design
addressing model selection. We defined a model sensitivity, a measure
for the dependency of the dynamics of a protein concentration on the model structure. The model sensitivity can be
used for ranking the most important system players which should be measured to
discriminate models and identifies a minimal combination of measured proteins
for a significant model discrimination.
Given this combination of proteins, the sampling time of the measurement is
optimized to improve furthermore the expected significance of the model
selection procedure.
As an example, two different models of the MAP-kinase signalling cascade are investigated. We identified the minimum set
of necessary measurements to distinguish both models.
M-PoP6 Integration of genomics and proteomics with metabolic/signaling pathways for
generating/improving novel anti-cancer drug targets
He Yang
Systembiology,
Bioinformatics Institute, 30 Biopolis Street, Singapore 138671, Singapore,
Phone: +65/64788268, FAX: +65/64789047, e-mail: henryy@bii.a-star.edu.sg
cDNA microarray has become a
powerful tool for profiling differential gene expression of cancer cells versus normal cells. With extraction of marker genes based
on cDNA microarray data, we can better understand the regulatory genes, leading
to improvement of drug targeting. A direct suggestion in drug targeting can come from
differentially expressed proteins obtained from comparative proteomic data. A
complete metabolic/signaling pathway depicts the interaction mechanisms among
metabolites (or proteins) and proteins (or genes) and thus can also be a
powerful tool for designing anti-cancer drug targets. It seems that any of
these methods (genomics, proteomics and metabolic/signaling pathways) could be
individually used for finding the drug targets for curing cancers. However, all
these methods are currently incomplete. Genomics and proteomics data will
produce a false discovery rate of marker genes and proteins, while
metabolic/signaling pathways are most likely incomplete due to the lack of
thorough understanding of cancer development. In this project, gene and protein expression data are used to check the correctness of a pathway
from a component A to another combination B. Furthermore, these data are
employed to determine whether there is an additional hidden component C between
the components A and B. The likelihood of the existence of such a component
will also be estimated. Such integration of genomic and proteomic data into the
pathway leads to construction of a better pathway model and thus improved drug
targeting.
Additional Teacher Abstracts
Looking for new targets in cancer therapy from a Systems Biology Approach
Marta Cascante 1, Pedro Vizan 1,
Antonio Ramos-Montoya 1, Begoña Comin-Anduix 1,
Pedro de Arauri 1, Laszlo G. Boros 2,
Sybille Mazurek 2, Erich Eigenbrodt 1,
Wilma Frederiks 1, W.-N. Paul Lee 2, Josep
J Centelles 1 and Joan Boren 1
1 Departament of Biochemistry and Molecular Biology, University of
Darcelone, Marti i Franques 1, Barcelona 08028, Spain, EU,
Phone: +34934021217, FAX: +34934021219, e-mail: martacascante@ub.edu,
Web: http://www.bq.ub.es/bioqint/recerca.html
2 Harbor-UCLA Research and
Education Institute, UCLA School of Medicine, 1124 West Carson St. RB 1,
Torrance, CA 90502, USA
In the last few years, science has
concentrated their efforts in characterizing the molecular elements that
conform cells. Several techniques as DNA sequencing, expression arrays, and proteomic and metabolomic
experiments have provided us a large amount of new information that cannot be
easily interpreted. Moreover, the integration of all these information in in
vivo models is likely to be the most interesting tool to understand and to
complete an overview picture of the cellular processes. We must take into
account, that metabolic profile is in most cases the end point of the
signalling events, where changes caused by diseases like cancer may be reflected. Therefore, using integrative bioinformatics
tools we are able to identify the main steps that control a metabolic pathway after the integration of all the experimental
data. These control points may be used as new therapeutical targets. In this
way, we are working in new antitumoral treatments based on the inhibition of
the synthesis of DNA and RNA because of their importance for cell proliferation. We have
focused on ribose-5-phosphate synthesis, which is a component of nucleotides.
First of all, we characterized the metabolic pathways (utilizing gas
chromatography coupled to mass spectrometry) implied in glucose metabolism and ribose synthesis. This was followed by the
integration of the obtained data in mathematical models which led us to
identify the main enzymes controlling ribose-5-P synthesis: transketolase and
glucose-6-phosphate dehydrogenase. Finally, we validated the obtained targets using
specific inhibitors and then, we studied the effects produced in cell
proliferation as well as in the proteomic profile. Consequences of the
utilization of these inhibitors on protein-protein interactions and on
supramolecular organization of the tumor metabolism have also been studied.
Mathematical modeling of mammalian
nucleotide excision repair based on in
vivo
Martijn Mone 1,
Martijn Luijsterburg 1, Antonio Politi 2,
Reinhart Heinrich 2, Adriaan Houtsmuller 3, Wim Vermeulen 4
and Roel Van Driel 1
1 University of
Amsterdam, BioCentrum Amsterdam, Kruislaan 318,
Amsteradm NL-1098SM, Netherlands, Phone: +31/20/525/5150,
FAX: +31/20/525/7935, e-mail: van.driel@science.uva.nl
2 Department of Theoretical
Biophysics, Institute of Biology, Humboldt University, Berlin, Germany
3 Dept. of Pathology,
Erasmus Medical Centre, Rotterdam, the Netherlands
4 Dept. of Cell Biology and
Genetics, Erasmus Medical Centre, Rotterdam, the Netherlands
Nucleotide excision repair in
mammals requires the concerted action of many different proteins. The repair
machinery assembles itself at damaged-DNA sites in a strict sequential fashion. We have developed methods to
visualize and quantitatively analyse this process in the nucleus of the living
cell (Moné et al, 2001, 2004). Based on our measurements we constructed a
mathematical model that delineates hallmarks and general characteristics for
this repair pathway. These data allowed us to scrutinize the dynamic behavior
of the nucleotide excision repair process in detail. The strict sequential
assembly mechanism appears to be remarkably advantageous in terms of repair
efficiency. Our findings show that alternative mechanisms, for instance the
forming of repair protein complexes in the nucleoplasm before binding to damaged DNA, or
random sequence of protein assembly, can readily become kinetically
unfavorable. Our model provides a kinetic framework for nucleotide excision repair and rationalizes why many
multiprotein processes within the cell nucleus may opt for a strategy of
sequential assembly. The model constitutes a firm basis for further wet experiments.
Moné, M.J. et al. (2001) EMBO Reports.
2:1013
Moné, M.J. et al. (2004) Proc Natl Acad Sci U S A. 101:15933
Principles
of Systems Biology: the chapter on signal transduction
Frank Bruggeman, Nathan Brady,
Jorrit Hornberg and Hans V. Westerhoff
Molecular Cell
Physiology and Mathematical Biochemistry, BioCentrum Amsterdam, De Boelelaan 1085, Amsterdam NL-1081 HV, The
Netherlands, EU, Phone: +31 20 5987230, FAX: +31 20 5987229,
e-mail: hw@bio.vu.nl,
Web: http://www.bio.vu.nl/hwconf
Systems Biology is not only about
the analysis of massive datasets, or the modelling of complex networks. Life
follows certain principles. These derive from constraints such as steady state,
as well as a number of fundamental properties of living systems. In this
presentation we shall refer to principles such as that every process in a
living cell is catalyzed by a protein, and that every protein is encoded
by a (set of) gene(s). These principles are wrong if
applied strictly to certain pathways, close to reality for others but may serve
as a source of inspiration for all. They induce one to ask about the extent to
which biological functions are controlled by the process activities.
It is this inspiration also that induced us to ask what the control of the total genome, i.e. of all activities combined,
is on biological functions such as steady state flux or the duration of a signal. We adress this question by
summing the control exercised by the individual processes, over all processes.
Importantly then, well defined answers are obtained for certain functions. We
here discuss the functional characteristics of signal transduction from this
perspective.
Aardema, 11, 23, 107, 177
Aarsaether, 11, 24, 173, 177
adaptation, 40, 116, 151
Aebersold, 25, 54, 67
Affourtit, 11, 24, 160, 177
Alberghina, 2, 7, 10, 11, 22, 177
Alon, 2, 7, 9, 11, 22, 25, 29, 48, 76, 80, 89,
106, 177
ALW-NWO, 3, 6, 156
Ambur, 11, 24, 144, 177
Arisi, 11, 27, 173, 177
AstraZeneca, 3, 7, 21, 35, 64, 124, 181
ATP, 24, 38, 64, 86, 113, 124, 127, 130, 143, 145, 151, 160, 163
Austria, 1, 2, 133, 148, 162, 178, 179, 181,
182
Bachmann, 11, 27, 144, 177
Bacillus subtilis, 31, 134, 146, 148, 152
Beirer, 11, 27, 161, 177
Beslon, 11, 29, 82, 177
Bezler, 11, 31, 161, 177
BioCentrum
Amsterdam, 1, 2, 3, 4, 7, 46, 70, 106, 107, 120, 151, 152,
155, 158, 175, 176, 177, 178, 181, 183, 185, 186
bioinformatics, 100, 159
BioSim NoE, 3, 7
blackboard, 9
Blank, 11, 23, 31, 76, 130, 145, 177
Blüthgen, 11, 30, 32, 100, 177
BMBF, 3, 6, 119, 145, 173
Boolean, 168
Boone, 7, 11, 23, 25, 56, 58, 67, 135, 177
Borodina, 11, 23, 126, 177
Breit, 11, 23, 24, 133, 162, 178
Brown, 11, 26, 35, 108, 178
Bruggeman, 7, 9, 11, 22, 23, 25, 46, 106, 176,
178
calcium, 94, 129, 160
Caldara, 11, 24, 145, 178
Camacho, 11, 30, 108, 178
cAMP, 23, 24, 40, 72, 110, 150
cancer, 26, 31, 32, 38, 54, 64, 70, 72, 94, 124, 134, 161, 163, 168, 171,
174, 175
Cascante, 2, 7, 10, 11, 23, 32, 33, 118, 163,
175, 178
cell cycle, 28, 31, 72, 74, 84, 110, 120, 126,
132, 142, 150, 163
chemostat, 26, 114, 138, 155
chromatin, 158
Combe, 11, 26, 126, 178
connectivity, 50, 100, 107, 133, 140, 159
control, 24, 26, 30, 35, 38, 54, 70, 74, 88, 92, 94, 98, 100, 127, 128, 143,
150, 151, 155, 157, 158, 159, 160, 161, 164, 165, 167, 169, 175, 176
Control analysis, 9, 22, 25
Control Analysis, 38
Conzelmann, 11, 27, 162, 178
Csikasz-Nagy, 11, 29, 30, 84, 142, 178
Curtis, 11, 23, 109, 178
Dach, 11, 30, 127, 178
Danø, 11, 14, 24, 25, 26, 30, 31, 62, 86, 111,
124, 141, 142, 178, 184
Davey, 11, 35, 146, 178
De Fabritiis, 11, 26, 109, 178
de la Fuente, 11, 30, 110, 178
De Monte, 11, 26, 29, 30, 86, 111, 142, 178
Derow, 11, 31, 163, 178
diabetes, 168, 169, 184
Diaz-Cuervo, 11, 23, 110, 178
diffusion, 64, 86, 92, 112, 114, 125, 136, 139, 142
DKFZ, 3, 7, 54, 70, 108, 178, 182
DNA, 24, 26, 48, 82, 88, 108, 114, 117, 118, 123, 126, 131, 143, 144, 147,
150, 153, 154, 158, 161, 173, 175
Donnet, 11, 24, 163, 178
Dooley,
11, 31, 173, 178
d'Ovidio, 11, 26, 30, 86, 111, 142, 178
Doyle, 7, 11, 22, 38, 42, 51, 179
Drosophila, 23, 40, 64, 112, 125
drug, 7, 25, 26, 31, 32, 56, 64, 92, 104, 112,
124, 135, 157, 163, 174
DSM, 4, 7
E. coli, 22, 23, 24, 25, 27, 31, 40, 46, 48,
50, 62, 76, 78, 82, 106, 107, 116, 118, 127, 134, 143, 145, 146, 151, 153, 170
Ebenhöh, 11, 30, 111, 113, 179
Ederer, 11, 23, 127, 179
Eichelbaum, 7, 12, 32, 92, 96, 103, 179
Eigel, 12, 31, 164, 179
Eils, 1, 8, 12, 25, 28, 30, 54, 58, 67, 70,
108, 179
Eisenacher, 12, 26, 128, 179
Eißing, 12, 27, 164, 179
EMBL, 3, 94, 100, 120, 159, 182, 184, 185
energy, 28, 72, 78, 82, 100, 110, 116, 130,
138, 148, 150, 151, 159, 168, 171
ESF, 3, 6, 10
FEBS, 1, 2, 3, 4, 5, 6, 7, 8, 10, 16, 17, 118
FEBS Journal, 4, 7, 17
Feeney, 12, 24, 165, 179
Fernandez-Lloris, 12, 27, 102, 160, 165, 179
Figueiredo, 12, 29, 30, 86, 143, 179
flux, 9, 22, 23, 25, 27, 29, 30, 40, 46, 66, 76, 78, 86, 106, 115, 116,
118, 120, 121, 122, 125, 129, 130, 132, 143, 145, 146, 151, 152, 155, 157, 163,
176
Fritsch, 12, 30, 128, 179
Fuhrer, 12, 31, 146, 179
Funahashi, 12, 23, 30, 114, 117, 129, 179
Gaubert, 12, 31, 166, 179
gene, 23, 24, 26, 30, 31, 35, 40, 46, 48, 50, 56, 62, 64, 66, 76, 78, 82,
84, 92, 94, 98, 102, 106, 107, 108, 109, 110, 112, 113, 119, 120, 123, 125,
126, 127, 128, 129, 131, 133, 134, 135, 136, 138, 139, 145, 146, 147, 148, 151,
152, 153, 154, 155, 156, 157, 158, 160, 161, 166, 167, 168, 169, 170, 172, 173,
174, 176
genome, 6, 30, 40, 54, 56, 62, 66, 82, 84,
108, 109, 110, 112, 113, 116, 117, 121, 125, 126, 128, 134, 139, 144, 145, 146,
147, 150, 152, 153, 156, 157, 158, 161, 165, 167, 176, 182, 186
Ghosh, 12, 21, 24, 139, 179
Gilboa-Geffen, 12, 24, 166, 179
glucose, 26, 27, 31, 48, 76, 86, 117, 122,
123, 130, 132, 136, 138, 143, 146, 148, 149, 150, 151, 152, 155, 156, 160, 169,
175
glycogen, 24, 169
Godoy, 11, 31, 173, 178
Goldbeter, 7, 12, 22, 24, 40, 42, 51, 145, 179
Gonze, 12, 27, 167, 179
Goryanin, 2, 7, 10, 12, 25, 179
Grabe, 12, 26, 129, 179
Grabherr, 12, 179
Grosu, 12, 24, 121, 179
Gursky, 12, 23, 64, 112, 125, 180
Hall, 12, 17, 18, 30, 130, 133, 162, 178, 180, 182
Hallén, 12, 26, 112, 180
Handorf, 12, 30, 111, 113, 180
Hartner, 12, 23, 24, 130, 148, 180
Hatakeyama, 12, 32, 98, 180
He, 12, 14, 23, 31, 113, 174, 180, 186
Hebben-Serrano, 12, 27, 147, 180
Heinrich, 7, 9, 12, 22, 23, 25, 30, 38, 42, 51, 111, 113, 136, 175, 180
Heßeler, 12, 26, 114, 180
Hiroi, 12, 30, 114, 180
Höfer, 12, 26, 27, 32, 98, 118, 161, 180
Hohmann, 2, 7, 10, 12, 28, 29, 33, 180
Ihekwaba, 12, 31, 35, 167, 180
Inácio, 12, 31, 148, 180
inhibitor, 64, 114, 124, 163, 167, 168, 173
insulin, 24, 160, 168
integration, 29, 30, 31, 32, 72, 86, 98, 143,
174
intercellular, 40, 136
Ivakhno, 12, 26, 131, 181
James, 12, 25, 26, 30, 64, 102, 124, 146, 160, 181
Jonson, 12, 30, 131, 181
Jouhten, 12, 23, 132, 181
Jules, 12, 26, 132, 181
Juvan, 12, 30, 133, 181
Kadirkamanathan, 12, 27, 121, 181
Kell, 7, 12, 21, 26, 31, 35, 108, 137, 146, 157, 167, 181
Kern, 12, 23, 24, 130, 148, 181
Kholodenko, 7, 12, 27, 32, 92, 96, 103, 162, 181
kinase, 32, 38, 48, 72, 78, 82, 92, 94, 100,
102, 106, 110, 117, 118, 119, 150, 159, 160, 161, 163, 167, 168, 174
kinetics, 24, 28, 46, 64, 70, 74, 92, 94, 98, 106, 115, 118, 119, 120, 122,
124, 125, 127, 129, 136, 149, 156, 164, 168, 171, 174, 175
Kitano, 2, 7, 10, 12, 23, 26, 30, 32, 88, 114,
117, 129, 143, 179, 180, 181
Klingmüller, 7, 13, 32, 94, 96, 103, 181
Klipp, 7, 13, 28, 30, 72, 80, 89, 119, 181
Kobayashi, 13, 23, 115, 181
Kollmann, 13, 22, 23, 48, 106, 181
Koster, 1, 2, 13, 23, 107, 181
Kozlov, 13, 23, 25, 26, 64, 112, 125, 181
Kresnowati, 13, 26, 115, 181
Kriegner, 13, 182
Kuchler, 1, 8, 10, 13, 21, 25, 28, 33, 182
Kummer, 2, 7, 13, 25, 28, 31, 58, 67, 70, 122,
182
Küpfer, 13, 27, 31, 76, 145, 149, 182
lactic acid bacteria, 31, 84, 144, 147, 149, 152, 156
Larsen, 13, 31, 149, 182
LCD projector, 17, 18
Le Novere, 13, 32, 94, 96, 103, 182
Lebiedz, 13, 31, 122, 182
lecture, 15, 17, 18, 35, 38, 104
Liao, 13, 27, 140, 182
lipid, 139
Liu, 13, 24, 30, 110, 122, 147, 182
liver, 30, 111, 113, 169, 173, 179
Ma, 13, 30, 62, 116, 182
Mahmood, 13, 24, 168, 182
MAPK, 30, 32, 92, 100, 159, 170
mass spectrometry, 23, 35, 76, 78, 107, 108, 122, 131, 162, 175
Meechai, 13, 24, 157, 182
membrane, 76, 92, 102, 116, 117, 129, 130, 136, 137, 139, 140, 148, 158, 160,
161, 163, 167, 168, 170
Merck A.G., 4
Millat, 13, 23, 116, 182
Minasbekyan, 13, 27, 158, 182
Modre-Osprian, 13, 23, 24, 133, 162, 182
modularity, 92
Moriya, 13, 26, 117, 182
Mowbray, 6
Mramor, 13, 26, 134, 182
, 13, 182
Müller, 13, 23, 24, 110, 150, 182
Murray, 13, 30, 88, 143, 182
Neumann, 13, 182
Neves, 13, 27, 150, 183
Nicolas, 13, 14, 24, 31, 32, 56, 94, 96, 103, 151, 154, 182, 183, 185
nitrogen, 126, 138, 147
NMR, 23, 132, 150
Notebaart, 7, 13, 30, 134, 183
NovoNordisk, 3, 7, 33, 104
Ohta, 13, 31, 140, 183
Orij, 13, 24, 27, 151, 152, 183
oscillations, 23, 26, 29, 30, 35, 38, 40, 62,
86, 92, 100, 111, 116, 119, 124, 132, 136, 142, 159, 167
Page, 13, 24, 173, 183
Papp, 13, 26, 66, 125, 183
Parsons, 13, 23, 135, 183
Patel, 13, 26, 135, 183
pattern, 78, 109, 112, 139, 165, 169, 173
Paveliev, 13, 27, 168, 183
Peddinti, 13, 30, 117, 183
phosphatase, 38, 78, 100, 118, 149, 159, 161,
171
phosphorylation, 78
Pitkänen, 13, 23, 50, 107, 183
poster, 7, 8, 9, 10, 11, 15, 16, 17, 18, 23, 26, 30
Postmus, 13, 24, 27, 151, 152, 183
powerposter, 5, 8, 9, 10, 11, 15, 18, 134, 139
, 13, 31, 136, 183
protein, 26, 27, 30, 32, 56, 62, 72, 76, 78, 82, 86, 92, 94, 98, 100, 102,
107, 108, 110, 111, 117, 118, 119, 120, 130, 131, 133, 138, 140, 141, 143, 147,
150, 151, 152, 153, 154, 155, 157, 158, 159, 160, 162, 163, 164, 166, 169, 170,
173, 174, 175, 176
Purac, 4, 7
Quistorff, 13, 31, 183
RAS, 30, 32, 100, 159
Redon, 13, 31, 152, 183
Reuss, 7, 13, 23, 24, 28, 29, 72, 80, 89, 110, 150, 184
Rischatsch, 13, 24, 153, 184
RNA, 62, 113, 133, 146, 154, 158, 161, 166, 172, 175
RNAi, 31, 100, 159, 172
robustness, 38, 50, 86, 88, 107, 115, 132, 143
Rocha, 13, 27, 153, 184
Rodriguez, 13, 14, 23, 27, 118, 162, 170, 184
RodrÌguez-Caso, 14, 31, 169, 184
Ros, 14, 24, 169, 184
Saez-Rodriguez, 14, 27, 162, 170, 184
Salazar, 14, 26, 118, 184
Salzburg, 16, 18, 28, 34
Santos, 14, 27, 30, 32, 100, 150, 159, 184
Sauer, 2, 7, 14, 23, 27, 29, 31, 76, 80, 89, 130, 145, 146, 149, 184
Sauter, 14, 23, 27, 30, 32, 102, 127, 160, 162, 184
SBML, 70, 86, 129, 135, 143
Scandurra, 14, 31, 170, 184
Schaber, 14, 30, 119, 184
Schmidt-Glenewinkel, 14, 24, 171, 184
Schuster, 7, 9, 14, 22, 25, 40, 42, 51, 184
Schütze, 14, 23, 136, 184
signal transduction, 22, 23, 24, 25, 26, 27,
28, 31, 32, 38, 48, 54, 62, 64, 70, 72, 82, 92, 94, 98, 100, 102, 106, 108,
110, 115, 118, 123, 124, 127, 133, 138, 140, 150, 159, 160, 161, 162, 164, 167,
168, 170, 171, 172, 173, 174, 176
Silicon cell, 9
ski, 9, 18, 129
Snoep, 7, 9, 14, 22, 25, 28, 58, 67, 70, 184
Sourjik, 14, 29, 82, 184
Spasic, 14, 26, 35, 137, 184
spatial, 40, 64, 92, 94, 98, 100, 110, 114, 117, 122, 125, 129, 134, 136, 149,
150, 158, 159, 162, 164
Spieth, 14, 23, 119, 184
Staines, 14, 24, 141, 184
Stelling, 7, 9, 14, 22, 25, 27, 28, 40, 74, 80, 89, 149, 174, 185
stochastic, 22, 27, 40, 48, 76, 82, 88, 106,
113, 121, 139, 143
Streptomyces, 23, 126
symposium, 9, 10, 15, 18
Tamrazyan, 14, 31, 154, 185
Tans, 14, 31, 158, 185
Teranode, 3, 7
Teusink, 14, 24, 29, 30, 70, 84, 134, 156, 185
Thul, 14, 31, 137, 185
Timmer, 14, 23, 27, 48, 106, 174, 185
Tomita, 2, 7, 14, 29, 78, 80, 89, 185
TOR, 27, 147, 149
Tourasse, 14, 24, 154, 185
transport, 31, 92, 98, 126, 130, 138, 139, 140, 150, 161, 163, 164, 171
tuberculosis, 157
Tuzun, 14, 27, 155, 185
Unilever, 4
Usaite, 14, 23, 138, 185
USB, 7, 15, 16, 17
Ustyugova, 14, 27, 171, 185
Vainshtein, 14, 26, 120, 185
van den Brink, 14, 31, 155, 185
Van Driel, 2, 7, 10, 14, 33, 175, 185
van Enckevort, 14, 24, 30, 84, 134, 156, 185
van Eunen, 7, 14, 30, 120, 185
Varjosalo, 14, 31, 172, 185
Velagapudi, 14, 27, 156, 185
Vienna, 1, 7, 179, 182
Vitkup, 14, 22, 46, 185
Vujasinovic, 14, 27, 48, 123, 185
Wanner, 2, 7, 14, 29, 78, 80, 89, 185
Westerhoff, 1, 8, 9, 10, 14, 21, 22, 23, 25,
30, 33, 70, 106, 120, 176, 186
Wirtz, 6
Wnt, 31, 38, 172
Wodak, 2, 7, 14, 25, 56, 58, 67, 186
Wu, 13, 14, 30, 31, 115, 116, 157, 182, 186
Yang, 14, 27, 31, 121, 174, 186
yeast, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
35, 46, 48, 50, 56, 62, 66, 70, 72, 74, 84, 88, 107, 108, 110, 111, 112, 113,
117, 119, 120, 123, 124, 125, 128, 130, 131, 132, 135, 136, 137, 138, 142, 143,
145, 146, 147, 148, 149, 150, 151, 152, 153, 155, 156, 157, 158
Zampera, 14, 22, 27, 48, 123, 186
Zeng, 14, 23, 25, 30, 62, 113, 116, 186
Zhang, 14, 27, 31, 138, 140, 157, 186
Zhu, 14, 24, 172, 186
Zimmermann, 14, 31, 136, 139, 186
(westerhoffgosau3corr998.doc)
rinciples
of Systems Biology
rinciples
of Systems Biology
ools
and methods (part 1)
ools
and methods
ools
& Methods (part 2)
nicellular
Organisms (part 1)
nicellular
Organisms
ulticellular
Organisms
