Reconstruction and analysis of genome-scale metabolic model of a photosynthetic bacterium.

Abstract

Synechocystis sp. PCC6803 is a cyanobacterium considered as a candidate photo-biological production platform--an attractive cell factory capable of using CO2 and light as carbon and energy source, respectively. In order to enable efficient use of metabolic potential of Synechocystis sp. PCC6803, it is of importance to develop tools for uncovering stoichiometric and regulatory principles in the Synechocystis metabolic network.
We report the most comprehensive metabolic model of Synechocystis sp. PCC6803 available, iSyn669, which includes 882 reactions, associated with 669 genes, and 790 metabolites. The model includes a detailed biomass equation which encompasses elementary building blocks that are needed for cell growth, as well as a detailed stoichiometric representation of photosynthesis. We demonstrate applicability of iSyn669 for stoichiometric analysis by simulating three physiologically relevant growth conditions of Synechocystis sp. PCC6803, and through in silico metabolic engineering simulations that allowed identification of a set of gene knock-out candidates towards enhanced succinate production. Gene essentiality and hydrogen production potential have also been assessed. Furthermore, iSyn669 was used as a transcriptomic data integration scaffold and thereby we found metabolic hot-spots around which gene regulation is dominant during light-shifting growth regimes.
iSyn669 provides a platform for facilitating the development of cyanobacteria as microbial cell factories.

Full text

Page 1

RESEARCH ARTICLEOpen Access
Reconstruction and analysis of genome-scale
metabolic model of a photosynthetic bacterium
Arnau Montagud1,3*, Emilio Navarro2, Pedro Fernández de Córdoba1, Javier F Urchueguía1, Kiran Raosaheb Patil3
Abstract
Background: Synechocystis sp. PCC6803 is a cyanobacterium considered as a candidate photo-biological
production platform - an attractive cell factory capable of using CO2and light as carbon and energy source,
respectively. In order to enable efficient use of metabolic potential of Synechocystis sp. PCC6803, it is of importance
to develop tools for uncovering stoichiometric and regulatory principles in the Synechocystis metabolic network.
Results: We report the most comprehensive metabolic model of Synechocystis sp. PCC6803 available, iSyn669,
which includes 882 reactions, associated with 669 genes, and 790 metabolites. The model includes a detailed
biomass equation which encompasses elementary building blocks that are needed for cell growth, as well as a
detailed stoichiometric representation of photosynthesis. We demonstrate applicability of iSyn669 for stoichiometric
analysis by simulating three physiologically relevant growth conditions of Synechocystis sp. PCC6803, and through
in silico metabolic engineering simulations that allowed identification of a set of gene knock-out candidates
towards enhanced succinate production. Gene essentiality and hydrogen production potential have also been
assessed. Furthermore, iSyn669 was used as a transcriptomic data integration scaffold and thereby we found
metabolic hot-spots around which gene regulation is dominant during light-shifting growth regimes.
Conclusions: iSyn669 provides a platform for facilitating the development of cyanobacteria as microbial cell
factories.
Background
Cyanobacteria, which have been model organisms since
the early 70s of the past century [1], are a widespread
group of photoautotrophic microorganisms, which origi-
nated, evolved, and diversified early in Earth’s history
[2]. It is commonly accepted that cyanobacteria played a
crucial role in the Precambrian phase by contributing
oxygen to the atmosphere [3]. All cyanobacteria com-
bine the ability to perform an oxygenic photosynthesis
(resembling that of chloroplasts) with typical prokaryotic
features, like performing anoxygenic photosynthesis by
using hydrogen sulfide (H2S) as the electron donor or
fixing atmospheric dinitrogen (N2) into ammonia (NH3).
Relevance of this phylum covers from evolutionary stu-
dies [4] to biotechnological applications, including bio-
fuel production [5]. Synechocystis sp. PCC6803 is a
cyanobacterium that is considered as a good candidate
for developing a photo-biological cell factory towards
production of a variety of molecules of socio-economic
interest, with CO2(and/or sugars) as carbon source and
light (and/or sugars) as energy source [6]. The diversity
of potential applications in this sense is broad. Works
have been published on heterologous production of
metabolites such as isoprene [6], poly-beta-hydroxybuty-
rate [7], biofuels [8] and bio-hydrogen [9,10] - an energy
vector of global interest [11].
Synechocystis sp. PCC6803 is capable of growing under
three different growth conditions as marked by the uti-
lized carbon source (/s) [12]. This causes that three dis-
tinct modes of operation are interweaved over the same
metabolic network, viz., i) photoautotrophy, where
energy comes from light and carbon from CO2; ii) het-
erotrophy, where energy and carbon source is a sacchar-
ide, for instance glucose; and, iii) mixotrophy, a
combination of the above two, where light is present as
well as a combination of two carbon sources: glucose
and CO2. Reconstruction of a genome-scale metabolic
model for this model photo-synthetic bacterium is one
* Correspondence: armontag@mat.upv.es
1Instituto Universitario de Matemática Pura y Aplicada, Universidad
Politécnica de Valencia, Camino de Vera 14, 46022 Valencia, Spain
Full list of author information is available at the end of the article
Montagud et al. BMC Systems Biology 2010, 4:156
http://www.biomedcentral.com/1752-0509/4/156
© 2010 Montagud et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative
Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.

Page 2

of the main goals of the current study. Genome-scale
metabolic network reconstruction is, in essence, a sys-
tematic assembly and organization of all the reactions
which build up the metabolism of a given organism; and
has been of great interest in the post-genomic era. The
variety of applications of such a metabolic model [13]
includes the possibility of assessing projects for the pro-
duction and optimization of an added value metabolite.
If a model is formulated properly, it is expected to allow
simulating environmental and genetic perturbations in
the metabolic network. Thus, together with appropriate
constraints, a metabolic model would partially represent
a virtual organism - an in silico model that allows prob-
ing possible flux distributions inside the cell under dif-
ferent environmental conditions and for a given genetic
make-up. Towards this end, a variety of tools/algorithms
are available [14], including flux balance analysis (FBA)
[15,16], minimization of metabolic adjustments
(MOMA) [17], regulatory on-off minimization (ROOM)
[18] and metabolic control analysis (MCA) [19,20].
Synechocystis sp. PCC6803 genome was sequenced,
annotated and made publicly available in 1996 [21,22]
and has been the target of some metabolic modeling
effort, especially for central carbon metabolic recon-
structions [23,24]. The work from Yang et al [23]
focused on a metabolic model of glycolysis, tricarboxylic
acid cycle and pentose phosphate pathway that was
simulated under heterotrophic and mixotrophic condi-
tions. Shastri and Morgan [24] studied a metabolic
model with the same pathways under autotrophic condi-
tions and compared their results to the ones from Yang
et al. These two works considered one lumped reaction
for the photosynthesis of the system. More recently, an
uncurated reaction list with a biomass composition
represented by central carbon metabolites has been pub-
lished [25]. This model, however, is not suitable for
simulations due to lack of proper biomass equation,
lumped nature of some key reactions and missing
reactions.
The large quantity of information featured in public
databases, like details about genomes [26], pathways
[27], enzymes [28] or proteins [29] can be used from
different databases to gather all published data for one
specific organism. However, the lack of quality must be
considered as a major drawback of some of the data-
bases: false positives, false negatives as well as wrongly
annotated objects may hinder efforts of collecting accu-
rate data [30]. Consequently, manual reconstruction by
detailed inspection of each and every reaction, biomass
equation based on metabolic building blocks (such as
amino acids and nucleotides), consistency and integrity
of the network is a pre-requisite for creating a high
quality and useful metabolic model [31]. The current
study presents such manually curated reconstruction for
Synechocystis sp. PCC6803 and demonstrates some of its
potential applications.
The present model features a detailed biomass equa-
tion which encompasses all the building blocks that are
needed for a flux distribution simulation that reflects
observed phenotype. No lumped reactions are present
and photosynthesis is described as a set of 19 reactions,
thus enabling the tracing of the corresponding fluxes.
Furthermore, different analyses are performed by using
this metabolic reconstruction, including reaction knock-
out simulations, flux variability analysis and identifica-
tion of transcriptional regulatory hotspots. Overall,
iSyn669 is a valuable tool towards the development of a
photo-biological production platform. The model will
also contribute to the existing set of genome-scale mod-
els with a virtue of being one of the first stoichiometric
models that account for photosynthesis.
Results and Discussion
Genome-scale metabolic network reconstruction
A complete literature examination, including databases,
biochemistry textbooks and the annotated genome
sequence, was needed in order to extract the current
state of the art on known metabolic reactions within the
metabolic network of Synechocystis sp. PCC6803. For a
thorough overview of the process of metabolic model
reconstruction, refer to very instructive work by Forster
et al [32] as well as review by Feist et al [31]. In detail,
the reconstruction started with the annotation and
genomic sequence files of Synechocystis sp. PCC6803
[21,22]. These files were used with Pathway Tools soft-
ware [33] in order to build a database of all the genes,
proteins and metabolites presents in the organism. The
list of reactions was then retrieved from Pathway Tools;
EC numbers and stoichiometry of the reactions were
checked and verified with the help of the Enzyme
nomenclature database [34] and KEGG pathway data-
base [27]. Reactions were elementally balanced except
for protons, so that chemical conversions were coherent.
In some of the reactions present in these databases,
metabolites were reported in a non-specific form (e.g.
‘an alcohol’). This is insufficient for metabolic model
simulation and, so, corresponding organism-specific
metabolites had to be identified [32]. Additionally, in a
large number of reactions cofactors were not completely
clarified: an enzyme being capable of using NADH or
NADPH or both. In the latter, two reactions were
included in the reconstructed metabolic network. Deter-
mination of reversibility of the reactions was assisted by
specific enzyme databases, like BRENDA [28]. If no con-
clusive evidence was reported, reactions were set to be
reversible.
In the reconstruction of the metabolic model, many
reactions (a total of 79 reactions, see Table 1) were
Montagud et al. BMC Systems Biology 2010, 4:156
http://www.biomedcentral.com/1752-0509/4/156
Page 2 of 16

Page 3

found to be necessary for the production of the mono-
mers, precursors or building blocks, that are considered
in the biomass equation but which have no correspond-
ing enzyme coding gene assigned. In consequence, many
genes that were not annotated before should be consid-
ered, as they code for enzymes that should be present to
allow the formation of biomass. For instance, enzymes
malyl-CoA lyase and isocitrate lyase were not allocated
in the annotation of the genome albeit their activities
have been measured [35,36] and their presence is neces-
sary to complete the glyoxylate shunt; consequently,
they were included in the model.
The product of this reconstruction process was a set
of reactions that encompass all the known metabolite
conversions that take place in Synechocystis sp.
PCC6803. The resulting network, iSyn669, consists of
882 metabolic reactions and 790 metabolites (see Table
1 for more information). A total of 669 genes were
included, to which 639 reactions were assigned (see
Additional file 1 for details); the difference between the
number of genes and assigned reactions is due to the
presence of considerable number of protein complexes
(e.g. photosynthetic or respiratory activities) and isoen-
zymes. Reactions with no cognate genes are also present
in iSyn669, 20 passive transport reactions and 47 chemi-
cal conversions (not mediated by enzymes) were
included. Additionally, a total of 79 reactions were
included on the basis of biochemical evidence or physio-
logical considerations, but currently with no annotated
Open Reading Frame (ORF). iSyn669 genome-scale
metabolic model is available in Additional file 2 (in Opt-
Gene [37] format).
iSyn669 spans all the biologically relevant flux nodes
in the Synechocystis metabolism. Pyruvate, phosphoenol-
pyruvate (PEP), 3-phosphoglycerate, erythrose-4-phos-
phate and 2-oxoglutarate are main flux nodes for amino
acids biosynthesis. Acetyl-CoA is an important flux
node for fatty acids production, with high relevance for
metabolic engineering towards biofuel production.
Biosynthesis of nucleic acids comes from different meta-
bolites, namely, ribose-5-phosphate, 5-phospho-beta-D-
ribosyl-amine, L-histidine and L-glutamine. Moreover,
with the information publicly available on databases, we
can conclude that Synechocystis sp. PCC6803 bears an
incomplete tricarboxylic acid cycle (TCA cycle), as it
lacks 2-ketoglutarate dehydrogenase (EC 1.2.4.2). It has
been published that glyoxylate shunt completes this
cycle [35], permitting the recycling of TCA metabolites.
Alternatively, aspartate transaminase (reaction 2.6.1.1a
in iSyn669) can interconvert 2-ketoglutarate and oxaloa-
cetate, thus bridging the gap of 2-ketoglutarate dehydro-
genase, but short-circuiting TCA cycle.
From the network topology perspective, iSyn669 dis-
plays the connectivity distribution pattern similar to that
of the other microbial genome-scale networks, e.g. yeast
[32] and Escherichia coli [38] (Table 2). While most of
the metabolites have few connections, few metabolites
are involved in very many reactions and are often
referred to as metabolic hubs. Homeostasis of such
highly connected metabolites will affect globally the
metabolic phenotype (as reflected in metabolite levels
and fluxes) and therefore of interest for studying the
organization of regulatory mechanisms on the genome-
wide scale. Most connected metabolites include those
related to energy harvesting (e.g. ATP, NADP+, oxygen),
a key metabolite in the porphyrin and chlorophyll meta-
bolism (S-adenosyl methionine), a couple of amino acids
Table 1 Distribution of the model reactions as per
cognate genes
Number of reactions882
-With assigned genes669
·Protein-mediated transport78
-With no cognate gene221
·Chemical conversion47
·Transport reactions20
·EC reactions not annotated 79
·Needed for biomass simulation 75
Table 2 Most connected metabolites in the iSyn669
metabolic network
MetaboliteNeighborsNeighbors
in E. coli
Neighbors
in yeast
H2O 213697-
ATP144338 166
phosphate10881113
ADP103253131
diphosphate 97 28-
H+74 923188
CO2 72 5366
NADP+ 643961
NADPH 636657
NAD+467958
L-glutamate 455256
NADH 42 75 52
AMP 36 86 48
oxygen O2 36 40 31
ammonia 2822-
S-adenosyl-L-methionine2518 19
glutathione 251710
a malonyl-ACP 23 15 10
L-glutamine 22 1823
coenzyme A 21 7139
Montagud et al. BMC Systems Biology 2010, 4:156
http://www.biomedcentral.com/1752-0509/4/156
Page 3 of 16

Page 4

and its precursors (L-glutamate, L-glutamine and glu-
tathione) and a key metabolite in the lipid biosynthesis
pathway (malonyl-ACP). High connectivity of these
metabolites hints to their potential central role in the
re/adjustments of fluxes following environmental
changes/perturbations. In order to discover the corre-
sponding regulatory mechanisms, additional studies
should be done - e.g. putative regulatory sequence
motifs associated with the neighbors of these highly
connected metabolites [39]. Furthermore, most con-
nected metabolites with filtered cofactors can be found
in Additional file 3.
Simulations of the three metabolic modes
iSyn669, together with appropriate physiological con-
straints, was used as a stoichiometric simulation model
by using FBA algorithm [40]. The FBA model simulates
steady state behavior by enforcing mass balances con-
straints for the all metabolic intermediates (Methods).
Biomass synthesis, a theoretical abstraction for cellular
growth, is considered as a drain of some of these inter-
mediates, i.e. building blocks, into a general biomass
component. Different studies have reported that the
simulation results do not usually vary drastically when
using a common biomass equation for different growth
condition [15,24]. Nevertheless, experimental efforts
should be directed at the depiction of the best precur-
sors and composition that could characterize, at least,
the three main growth modes, viz., autotrophy, hetero-
trophy and mixotrophy, in the scope of recent results
[41]. Due to the lack of such data, the present work
uses one single biomass equation in the simulations of
all three metabolic states (Table 3). Presence of photo-
synthesis allows iSyn669 to “grow” under the all three
metabolic states (i.e., FBA with biomass formation as an
objective function results in a feasible solution): carbon
dioxide and light (autotrophic), sugars (heterotrophic),
carbon dioxide, light and sugars (mixotrophic).
Growth under pure heterotrophy, or dark heterotro-
phy (in the absence of light) is a subject under study
[42,43], being the regular experimental design to give a
short light pulse prior to the pure heterotrophic phase
(light-activated heterotrophy). Nevertheless, the theoreti-
cal flux distribution under heterotrophic conditions is
interesting by itself - especially in comparison with the
flux distribution in a light-fed energy metabolism. More-
over, fluxes in the heterotrophy mode may help in
obtaining insight into the variations under the mixo-
trophic condition, which is of high relevance for indus-
trial applications [9].
All FBA simulations were carried out under the
appropriate constraints so as to match an autotrophic
specific growth rate of 0.09 h-1. This growth rate corre-
sponds to a light input of 0.8 mE gDW-1h-1and to a net
carbon flux of 3.4 mmol gDW-1h-1into the cell, with
HCO3-and CO2as carbon sources. For the sake of
comparison across the different conditions, uptake rates
for the corresponding carbon sources were matched
based on normalization per number of carbon atoms
(this does not affect mono-carbon compounds like car-
bon dioxide and carbonic acid, but has importance in
glucose feeding). Results of the subsequent FBA simula-
tions for the three different growth conditions are pre-
sented in the following. Some of the reactions that are
physiologically relevant for each of the conditions are
summarized in Table 4 and Figure 1. Flux values for the
rest of the reactions, including the upper and lower
bounds are provided in Additional file 4.
Heterotrophy
Heterotrophy was simulated by considering glucose as
the sole carbon source with uptake rate of 0.567 mmol
gDW-1h-1, entering the system through glcP glucose
transporter (reaction TRANS-RXN59G-152 in iSyn669).
With the purpose of having a pure heterotrophic state,
Table 3 iSyn669 Biomass composition
Metabolitemmole/g
DCW
Metabolitemmole/g
DCW
Amino acids
[38]
Deoxyribonucleotides
[58]
Alanine 0.499149dATP0.0241506
Arginine0.28742dTTP0.0241506
Aspartate0.234232 dGTP0.02172983
Asparagine0.234232dCTP 0.02172983
Cysteine0.088988
Ribonucleotides [1]
Glutamine 0.255712AMP 0.14038929
Glutamate0.255712UMP0.14038929
Glycine0.595297GMP 0.12374585
Histidine 0.092056CMP 0.12374585
Isoleucine0.282306
Lipids [59]
Leucine0.437778 16C-lipid 0.20683718
Lysine 0.333448 (9Z)16C-lipid0.01573412
Methionine0.14933618C-lipid 0.00351776
Phenylalanine0.180021(9Z)18C-lipid0.03188596
Proline 0.214798(9Z,12Z)18C-lipid 0.03568367
Serine0.209684(9Z,12Z,15Z)18C-lipid0.01797109
Threonine 0.246506(6Z,9Z,12Z)18C-lipid0.05031906
Tryptophan0.055234(6Z,9Z,12Z,15Z)18C-lipid0.01448179
Tyrosine 0.133993
Antenna chromophores
[60]
Valine
Carbohydrates [61]
0.411184Chlorophyll a
Carotenoids
0.02728183
0.00820225
Glycogen0.01450617
Biomass composition description with references where the information was
retrieved from. All this building blocks with their respective stoichiometric
coefficient is converted into one gram of dry cell weight. Biomass equation is
reaction Biomass in Additional files 2 and 4.
Montagud et al. BMC Systems Biology 2010, 4:156
http://www.biomedcentral.com/1752-0509/4/156
Page 4 of 16

Page 5

Table 4 Comparison of selected fluxes across different growth conditions
Reaction
name
Autotrophy Minimum
flux
Maximum
flux
Mixotrophy Minimum
flux
Maximum
flux
Dark
Heterotrophy
Minimum
flux
Maximum
flux
Light
Heterotrophy
Minimum
flux
Maximum
flux
Reaction description
2.7.1.2a
0000.5670.566 0.5670.567 0.5660.567 0.5670.5660.567 beta-D-glucose + ATP ®
beta-D-glucose-6-
phosphate + ADP
4.2.1.2
12.6712.667+∞
14.67 14.657+∞
0.905 0.884+∞
2.1481.836+∞
malate ↔ fumarate + H2O
D-ribose-5-phosphate ↔
D-ribulose-5-phosphate
PSII* + UQ + 2 H+ ® PSII
+ UQH2
NADH + UQ + 7 H+ ®
NAD+ + UQH2 + 4 H
+_peribac
5.3.1.6
1.2011.2+∞
1.2691.269+∞
-0.054-0.051 -0.0550.0660.067+∞
_UQ
0.800.80.800.80000.800.8
_1.6.5.3
00+∞
00+∞
2.1340+∞
00+∞
_3.6.3.14
38.348 15.7+∞
21.727 21.7+∞
4.984.95+∞
6.2926.281+∞
3 H+_peribac + phosphate
O4P + ADP ↔ 3 H+ +
H2O + ATP
6.2.1.1
0.008-∞
+∞
-30.017-∞
+∞
-2.124 -∞
+∞
-4.635-∞
+∞
coenzyme A + acetate +
ATP ↔ acetyl-CoA +
diphosphate + AMP
Units in mmol gDW-1h-1. 2.7.1.2a, glucokinase, is the reaction that phosphorylates beta-D-glucose upon entrance in the cell, marking the start of the glycolysis. The flux direction changes can be seen in reaction
4.2.1.2, fumarate hydratase, from TCA cycle and 5.3.1.6, ribose-5-phosphate isomerase, from the pentose phosphate pathway. _UQ and _1.6.5.3 are reactions that reduce UQH2 from photosystem II or NADH
oxidation, respectively, causing a pumping of protons to the thylakoid. _3.6.3.14 is the ATP synthase that forms ATP shuttling protons from the thylakoid to the cytosol. 6.2.1.1, acetate-CoA ligase, is the reaction that
generates acetyl-CoA from acetate and coenzyme A, that would be a major flux hub in an ethanol-producing strain, standing as the first step of fermentation.
Montagud et al. BMC Systems Biology 2010, 4:156
http://www.biomedcentral.com/1752-0509/4/156
Page 5 of 16

Page 6

photon uptake rate was constrained to 0; this caused
photosynthesis fluxes to be shut down. In this case, glu-
cose will be the source for the formation of carbon
backbones for the building blocks of the cell, depicted
in the biomass equation. The glycolytic and the oxida-
tive mode of the pentose phosphate pathway were found
to be active. Oxidative pentose phosphate pathway is the
major pathway for glucose catabolism as was reported in
reference [44]. PEP carboxylase (reaction 4.1.1.31 in
iSyn669) is the main anaplerotic flux to the TCA cycle.
Carbon fixation efficiency is around 60%, the rest being
released in the form of CO2, as reported in our previous
work [9].
In contrast to dark heterotrophy, if a light-activated
heterotrophy simulation is run, light enters the system
and RuBisCO enzyme is active (reaction 4.1.1.39), fixing
all the CO2that was released in dark heterotrophy,
boosting carbon efficiency to a theoretical 100%. In this
case, global flux distribution as well as flux ranges
resemble that of autotrophy more than that of the dark
heterotrophy. Carbon skeletons are still produced
through glycolysis and NAD(P)H is reduced along the
glycolysis, pyruvate metabolism and TCA cycle. On the
other hand, pentose phosphate pathway has shifted to
the reductive mode due to RuBisCO activation and the
corresponding flux is increased in magnitude. Carbon
fixation happens at the RuBisCO level, thereby assimilat-
ing the CO2produced by the glucose metabolism, and
the production of ATP and NADPH through photo-
synthesis relieves the oxidative phosphorylation from
draining NADPH to generate ATP.
Autotrophy
Photoautotrophy was initially simulated considering an
illumination of 0.15 mE m-2s-1. Assuming that the mass
of a typical Synechocystis sp. PCC6803 cell is 0.5 pg [45]
and its radius is 1.75 μm [46], we estimated that the
theoretical maximum illumination is 41563.26 mE gDW-1
h-1. An additional optimization step was performed in
order to estimate physiologically meaningful photon
uptake values that are closer to the experimental mea-
surements [24]. First, carbon uptake rate was found that
resulted in a specific growth rate of 0.09 h-1, while the
light intake was unconstrained. Next, the growth rate
was constrained to this value and the second optimiza-
tion problem was solved where light uptake was mini-
mized. This minimization resulted in photon uptake for
photosystem I (reaction _lightI) and photosystem II
(reaction _lightII) being 0.8 mE gDW-1h-1. Carbon
sources used in simulating photoautotrophy conditions
were carbon dioxide and carbonic acid, and its entrance
to the system was mediated by RuBisCO (reaction
4.1.1.39 in iSyn669) and carbonic anhydrase (reaction
4.2.1.1b) respectively. As iSyn669 biomass equation
encompasses all essential metabolite precursors, these
will be the sinks of our network, while photons, carbon
dioxide and/or carbonic acid will be the sources. Thus
4.2.1.2
_3.6.3.14
_1.6.5.3
amino acid
lipid
cofactor
nucleic acid
metabolite
6.2.1.1
acetyl-CoA +
diphosphate + AMP
acetate +
coenzyme A + ATP
fumarate +
H2O
malate
ADP + 3 H+_thylac +
phosphate
ATP + 3 H+ + H2O
NADH + UQ + 7 H+
NAD+ + UQH2
+ 4 H+_thylac
beta-D-glucose + ATP
beta-D-glucose-6-
phosphate + ADP
2.7.1.2a
D-ribose-5-
phosphate
D-ribulose-5-
phosphate
5.3.1.6
12.67
14.67
2.148
0.905
autotrophy
mixotrophy
light heterot
dark heterot
0
0.567
0.567
0.567
PSIIPSI
0.008
-30.02
-4.635
-2.124
38.35
21.73
6.292
4.98
1.201
1.269
0.066
-0.054
0
0
0
2.134
Figure 1 Selected reactions in iSyn669 network that display flux changes across the four studied growth modes. Flux values (in mmol
gDW-1h-1) for selected reactions in the Synechocystis sp. PCC6803 metabolism. These reactions mark changes across four growth modes, viz.,
autotrophy, mixotrophy and light and dark heterotrophy. Corresponding flux ranges can be found in Table 4 and in Additional file 4 for all the
reactions in iSyn669.
Montagud et al. BMC Systems Biology 2010, 4:156
http://www.biomedcentral.com/1752-0509/4/156
Page 6 of 16

Page 7

autotrophic fluxes will flow in the gluconeogenic direc-
tion and through the Calvin cycle, which is the reductive
mode of the pentose phosphate pathway. PEP carboxy-
lase is the main anaplerotic flux to the TCA cycle and
glyoxylate shunt is inactive.
Mixotrophy
Photons, carbon dioxide and glucose are independent
feed fluxes in this simulation. These fluxes entered the
system through the same reactions as described for the
previous growth modes. Carbon source presents, in
this case, one more degree of freedom than in the rest
of the conditions. In order to keep a comparative cri-
terion across conditions, we normalized CO2and glu-
cose inputs to the same carbon uptake flux as in the
case of the autotrophy and the heterotrophy. Photon
uptake rates were also normalized in a similar manner
to match the autotrophic state. Having the same meta-
bolic sinks as the two previous modes and the sources
from the both of them, it is logical to think that the
resulting flux distribution will be a mixture of the
autotrophic and heterotrophic simulations. Indeed, we
observed that the mixotrophic flux distribution lies in-
between the previous two states, being a bit closer to
the heterotrophy. Glycolysis is present and glyoxylate
is shut down; an active photosynthesis is present, oxi-
dative phosphorylation is less stressed than in hetero-
trophy as the energy can be produced from the photon
uptake; and Calvin cycle is active, as carbon sources
are CO2and glucose.
Flux variability analysis
Flux balance analysis presented above guarantees to
find the optimal objective function value (biomass for-
mation rate). However, the predicted intra-cellular
flux distribution is not necessarily unique due to the
presence of multiple pathways that are equivalent in
terms of their overall stoichiometry. Thus, often the
system exhibits multiple optimal solutions and further
elucidation requires additional constraints based on
experimental evidences (e.g. carbon labeling data).
Alternatively, physiological insight can be still
obtained by studying the variability at each flux node
given the objective function value - a procedure
referred to as flux variability analysis. In order to gain
insight into the flux changes underlying the changes
in the Synechocystis metabolism due to (un)availability
of light, we have compared the autotrophic growth
with the other two by using flux variability analysis
(Figure 2). Interestingly, autotrophy permits an overall
broader flux landscape than heterotrophy (let it be
dark or light-activated). On the other hand auto-
trophic flux ranges are in general narrower than the
mixotrophic ranges. Figure 1 and Table 4 depict some
of the physiologically relevant reactions for which the
feasible flux range differs across conditions. These
include glucokinase from glycolysis, fumarate hydra-
tase from TCA cycle, ribose-5-phosphate isomerase
from pentose phosphate pathway, NADH dehydrogen-
ase from oxidative phosphorylation or photosystem II
oxidation. These reactions mark the key nodes in the
metabolism network that must be appropriately regu-
lated in order to adapt in response to the available
energy/carbon source. Mechanisms underlying such
changes will be of particular interest not only for bio-
technological applications but also from the biological
point of view. As a glimpse of the detailed flux (re-)
distributions in each of the studied growth conditions,
Additional file 5 describes fluxes in the pyruvate
metabolism.
166
98
26
186
28
151
58
60
38
197
405
94
0
2
3
0 100200 300400500
same
broad
narrow
up
down
autotrophy - light het
autotrophy - dark het
autotrophy - mixotrophy
Figure 2 Overview of the flux adjustments between different
growth conditions. Comparison of flux variability between
autotrophy vs. mixotrophy, autotrophy vs. dark heterotrophy and
autotrophy vs. light-activated heterotrophy. Minimum and maximum
flux ranges were compared for each reaction, 378 reactions were
found blocked in all the studied conditions.
no growthconstrained growthwild type growth
35%
5%
60%
Synechocystis
15%
5%
80%
E. coli
11%
4%
85%
S. cerevisiae
34%
2%
64%
15%
5%
80%
10%
2%
88%
FBA
MOMA
Figure 3 Essential genes in Synechocystis sp. PCC6803.
Distribution of gene knock-out results for three model organisms,
simulated by using FBA and MOMA algorithm, classified as wild-
type growth, constrained growth and no growth.
Montagud et al. BMC Systems Biology 2010, 4:156
http://www.biomedcentral.com/1752-0509/4/156
Page 7 of 16

Page 8

Gene/Reaction knock-out analysis
The comprehensive set of reconstructed biochemical
equations of iSyn669 and FBA simulations enabled us to
further analyze the characteristics and potential of the
Synechocystis metabolic network. This can be oriented
towards the study of the reactions (and thereby the cor-
responding genes) that are necessary for the growth, or
to in silico metabolic engineering for identification of
targets for maximization of a given metabolite of socio-
economic interest.
Essential Genes
iSyn669 network consists of 790 metabolites and 882
reactions. Among these, 350 genes (36% of the total,
Figure 3) were found to be necessary for the formation
of the biomass under the mixotrophic growth conditions
by using FBA and MOMA algorithms. This set of genes
can be divided in to two categories: i) essential genes,
deletion of which completely inhibits biomass growth
(304 genes, 34% of the total, with FBA): and ii) genes
deletion of which causes a reduced growth rate
(46 genes, 2% of the total, with FBA). The set of 304
essential genes can be understood as the core of the
metabolism, as deleting them would produce an unvi-
able organism. The results based on MOMA algorithm
essentially tally these numbers: 311 essential genes, 35%
of the total, and 45 that cause a reduced growth rate,
5% of the total, (Additional file 6).
Interestingly, if we compare the proportion of the
essential genes under FBA simulation in the metabolic
networks of E. coli (187 genes, 15% of the total) [38]
and Saccharomyces cerevisiae (148, 10% of the total)
[32] with iSyn669, we find that Synechocystis has a sig-
nificantly larger fraction of metabolic genes whose dele-
tion obliterates biomass formation (304 genes, 34% of
the total). One possible explanation for the difference in
the relative proportion of essential genes in these three
organisms would be an incomplete/incorrect annotation
of the genome of Synechocystis sp. PCC6803. For exam-
ple, if only one of the isoenzymes corresponding to a
reaction is annotated, the corresponding in silico knock-
out will result in a false negative prediction. It is impor-
tant to note that the computational predictions of gene
essentiality based on FBA are highly dependent on the
growth medium used for the simulations. Thus, the
comparison across different species may not be straight-
forward. Moreover, it is also possible that the natural
growth conditions of Synechocystis may have dictated
selection for a relatively high proportion of essential
genes. Such hypotheses need careful consideration of
several factors and are beyond the scope of this work.
Production of value-added compounds
Synechocystis sp. PCC6803 is considered as a candidate
photobiological production platform - it can potentially
produce molecules of interest by using CO2and light
[6]. To this end, iSyn669 can be used to perform simu-
lations, not only for assessing the feasibility of producing
a given compound, but also to identify potential meta-
bolic engineering targets towards improved productivity.
For example, FBA simulations can help in estimating
maximum theoretical yields for the products/intermedi-
ates of interest. A product of obvious interest is hydro-
gen. In our previous work [9], we have estimated
maximum theoretical hydrogen production values that
are far from the current state of experimental reports.
In silico studies can direct the efforts and counsel the
scientists towards a hydrogen producing cyanobacteria
that could be of impact. iSyn669 predicts, in autotrophic
conditions, a theoretical H2evolution rate of 0.17 mmol
gDW-1h-1obliterating biomass growth. Else, the stoi-
chiometry permits the evolution of 0.156 mmol gDW-1h-
1of hydrogen with a biomass growth of 10% of the wild
type (0.007 mmol gDW-1h-1).
Succinate is an important metabolite for its biotechno-
logical applications as well as for being a metabolite that
bridges the TCA cycle with the electron transfer chain.
As an example of the usefulness of the present meta-
bolic model we have designed an in silico metabolic
engineering strategy to improve the production of succi-
nate. The underlying idea is to design a succinate over-
producing metabolic network (through reaction knock-
out simulations), whereas the intracellular fluxes are dis-
tributed so as to maximize the biological objective func-
tion (e.g. growth) [47]. To this end, OptGene algorithm
[37] was used together with Minimization Of Metabolic
Adjustment (MOMA) [17] as a biological objective func-
tion. MOMA has been reported to provide better
description of flux distributions in mutants or under
un-natural growth conditions as opposed to FBA. A
design objective function which copes with the metabo-
lite of interest, succinate, has been determined maintain-
ing the biological objective function as the biomass
formation.
OptGene simulations for single, double and triple
knock-out strategies were performed to obtain solutions
with improved succinate production, but without drasti-
cally diminishing the biomass production. We used mix-
otrophic conditions, for which wild type optimal growth
rate was 0.17909 mmol gDW-1h-1. The best single
knock-out was found to be the mutant of pyruvate
kinase (reaction 2.7.1.40c in iSyn669 and genes sll0587
and sll1275) that has a succinate evolution of 0.5695
mmol gDW-1h-1with a growth rate of 0.0714 mmol
gDW-1h-1. Blocking this reaction, preventing pyruvate
and phosphoenolpyruvate from using GTP and GDP
would drive a high increase in succinate production.
The flux between pyruvate and phosphoenolpyruvate
can still be accomplished with reactions 2.7.1.40a and
2.7.9.2, but using ATP and ADP as cofactors. Double
Montagud et al. BMC Systems Biology 2010, 4:156
http://www.biomedcentral.com/1752-0509/4/156
Page 8 of 16

Page 9

deletion did not improve the results from the single
knock-out strain, evolving the same succinate produc-
tion with the same growth rate. The best triple knock-
out was found to be the combination of pyruvate kinase
(reaction 2.7.1.40c in iSyn669 and genes sll0018 and
sll0587), fructose-bisphosphate aldolase (reaction
4.1.2.13b in iSyn669 and genes slr0943 and sll1275) and
succinate dehydrogenase (reaction _1.3.99.1 in iSyn669
and genes sll0823, sll1625 and slr1233). This simulated
strain has a succinate evolution of 0.6999 mmol gDW-1
h-1with a growth rate of 0.0688 mmol gDW-1h-1. This
design combines the blocking of the oxidation of succi-
nate on the electron chain transfer through succinate
dehydrogenase with the prevention of using GTP
between pyruvate and phosphoenolpyruvate and the
lack of an aldolase needed in the reductive mode of the
pentose phosphate pathway. This leads to a situation
where flux is directed to TCA cycle in order to meet
with an overproduction of succinate.
These studies on knock-outs are reaction centered,
even though the in vivo knock-out building will ulti-
mately be through gene manipulations. This is the rea-
son underlying the fact that we found 2.7.1.40c knock-
out as the best result. This design would hint at the idea
of selection of a mutated pyruvate kinase protein speci-
fic for ATP cofactor. This may be difficult to achieve on
the bench, but has high biotechnological expectations.
iSyn669 as a data integration scaffold
Apart from the flux simulations, another important pro-
blem in the field of metabolic systems biology that can
be addressed by using reconstructed genome-scale mod-
els is the integration of the different genome-wide bio-
molecular abundance datasets, i.e. omics datasets, such
as transcriptome and metabolome. An example of algo-
rithms for carrying out such an integrative analysis
through the use of genome-scale metabolic networks is
Reporter Features [48,49]. Reporter algorithm allows
integration of omics data with bio-molecular interaction
networks, thereby allowing identification of cellular reg-
ulatory focal points (i.e. reporter features), for instance
reporter metabolites as regulatory hubs in the metabolic
network.
In this work, Reporter Features software was used to
integrate transcriptional information over the recon-
structed Synechocystis sp. PCC6803 network allowing us
to infer regulatory principles underlying metabolic flux
changes following shifts in growth mode. In particular,
we analyzed the data from a work [50] that reports the
transcriptional changes caused in Synechocystis sp.
PCC6803 by shifts from darkness to illumination condi-
tions and back. As it can be understood from the ratio-
nale beneath the metabolic capabilities of this
cyanobacterium, the presence or absence of light drives
big changes in the flux distribution through the net-
work, as discussed in the previous sections. We have
focused our study on the relationship between the tran-
scription of Synechocystis sp. PCC6803 genes and the
reactions of the metabolic network. Associations
between genes and reactions were identified, listing all
the genes that performed or were involved in a specific
reaction. With this information and the metabolic
model, Reporter Features analysis was carried out. In
brief, the analysis helped to identify metabolites around
which the transcriptional changes are significantly con-
centrated. These metabolites are termed reporter meta-
bolites as they represent key regulatory nodes in the
network.
Gill et al [50] designed the experiment so that Syne-
chocystis was grown to mid-exponential phase (A730=
0.6 to 0.8). Then, the lights were extinguished and RNA
samples were taken after 24 h in the dark (full dark).
Illumination was then turned back on for 100 min (tran-
sient light), followed immediately by an additional 100
min in the dark (transient dark).
We were interested in two aspects of this study: i) to
identify metabolites around which regulation is centered
during the light regime transitions; and ii) to find the
metabolic genes that were collectively significantly co-
regulated across these transitions [49].The analysis was
divided in three parts: an analysis of the data arrays
from the whole experimental profile ("all time points”),
an analysis of the shift from darkness to a light environ-
ment ("dark to light”) and from light back to dark ("light
to dark”). For a study of the overall genome and its light
regulation, refer to Gill et al [50]. In this study, as the
relationship between the metabolism and this regulation
was investigated, genes with no direct relationship to a
metabolic reaction were not considered. Distributions of
the genes across KEGG Orthologies related to the meta-
bolism altered with the light shift are depicted in Table
5.
All time points
When all seven arrays were used, reporter metabolites
were found to be quite scattered across the metabolism
spanning several metabolic pathways, and thus offering
a global view of the transcriptional response in the
metabolic network (see Figure 4a and Table 6a). Pre-
sence of some amino acids (L-tyrosine, L-isoleucine),
nucleic acids and its precursors (GTP, dihydroorotate),
carbon metabolism metabolites (D-ribulose-5-phosphate,
succinyl-CoA), lipids precursors (myo-inositol, D-myo-
inositol 3-monophosphate), cofactors (thioredoxin,
p-aminobenzoate) and photosynthesis metabolites (plas-
tocyanin) pictures a scenario of a global regulation
throughout the different metabolic pathways.
By using the metabolic sub-network search algorithm,
we found 212 genes that have their expression changed
Montagud et al. BMC Systems Biology 2010, 4:156
http://www.biomedcentral.com/1752-0509/4/156
Page 9 of 16

Page 10

across the arrays and that have a relationship with the
metabolites of iSyn669 network. Furthermore, 50 genes
were identified that are strongly co-regulated all along
the profile of the experiment (Additional File 7, section
a). This set of genes is characterized in two groups. The
first set consists of the genes from photosynthesis
(93.85%) and oxidative phosphorylation (6.15%). The
second set is representative of a variety of genes from
different pathways such as amino acid metabolism
(39%), carbohydrate metabolism (22%), nucleotide meta-
bolism (13%), nitrogen metabolism (13%) and metabo-
lism of cofactors (9%) that globally regulates the entire
metabolic network (see Table 5 for further details).
It can be expected that an experimental design like the
one we have based our work on, which combines a shift
from dark to light with a shift back to darkness, will
encompass an important part of the regulatory changes
the cell is undergoing in its natural habitat. In a glu-
cose-deficient environment, the presence or absence of
light is the main condition around which the Synecho-
cystis metabolism gravitates [9]. Indeed, one of the
co-regulated sets consists of the genes coding for the
proteins that work on, and around, the thylakoid mem-
brane, let it be photosynthesis or oxidative phosphoryla-
tion genes.
Dark to light
Next, we considered the arrays that represent the shift
from darkness to light, the first three arrays (from “24
hours of darkness” array to “60 minutes of light” array).
Reporter metabolites were found to be largely within the
nucleotide and amino acid metabolism (Table 6b). Some
cofactors were also identified as regulation hubs like tet-
rahydrofolate, thioredoxin and adenosylcobinamide.
Sub-network search yielded set of 247 genes that have
their expression changed across the first three arrays
and that are related with iSyn669 reactions. Further-
more, 84 genes were identified that are strongly co-
regulated across the three arrays (Additional File 7, sec-
tion b). This set of genes cover photosynthesis (25%),
oxidative phosphorylation (24%), amino acid metabolism
(11%), carbohydrate metabolism (11%), nucleotide meta-
bolism (10%) and metabolism of cofactors (10%).
This set of data arrays are indeed a good example of
a cell’s metabolic machinery starting up. After a 24
hour period in darkness where cell density did not
Table 5 KEGG orthology groups for the metabolic genes altered with the light shift.
All time pointsDark to LightLight to Dark
Number
of genes
%Number
of genes
%Number
of genes
%
Energy Metabolism12860.3812851.8212761.65
Amino Acid Metabolism2511.793112.552411.65
Carbohydrate Metabolism2411.322811.332311.16
Metabolism of Cofactors and Vitamins136.132610.53125.83
Nucleotide Metabolism125.66239.32125.83
Lipid Metabolism73.352.0262.91
Membrane Transport31.4241.6320.97
Biosynthesis of Secondary
Metabolites
0010.400
Biosynthesis of Polyketides
and Nonribosomal Peptides
0010.400
Total212100247100206100
a)
50 hours of light
cultivation
24 hours of darkness100 min of
light
100 min of
darkness
: whole-genome array
L-isoleucine
L-tyrosine
dihydroorotate
D-ribulose-5-
phosphate
GTP
plastocyanin
PSIIPSI
b)
amino acid
lipid
cofactor
nucleic acid
metabolite
reporter
metabolite
Figure 4 Reporter metabolites under light/dark regime. a)
Reporter metabolites for all time points set of arrays depicted on the
iSyn669 network. b) Light/dark-shift profiles and localization of the
genome arrays for the work from Gill et al. [47].
Montagud et al. BMC Systems Biology 2010, 4:156
http://www.biomedcentral.com/1752-0509/4/156
Page 10 of 16

Page 11

change (see Figure 1 in Gill et al [50]), light enters the
system and the cell starts to synthesize new bio-mole-
cules, mostly nucleotides so it can copy its genetic
material and amino acids to build up proteins.
Light to dark
Finally, we considered the arrays that represent the shift
from light to dark, data from “90 minutes of light” array
to “60 minutes of dark” array. Similar to the previous
case study, reporter metabolites were found to be
focused on the nucleotide and amino acid metabolism
(Table 6c). Additionally, the presence of metabolite a
1,4-alpha-D-glucan_n and its cognate a 1,4-alpha-D-glu-
can_n1 also stands out as they are involved in carbon
reserves catabolism and anabolism.
With the help of the sub-network search, 133 genes
were identified as being significantly co-regulated across
those three arrays (Additional File 7, section c). This set
comprises of the genes from photosynthesis (34%), oxi-
dative phosphorylation (26%), amino acid metabolism
(12%), carbohydrate metabolism (12%), nucleotide meta-
bolism (7.5%) and metabolism of cofactors (4.5%).
This last set of data array is a scenario where metabo-
lism is being shut down, as a consequence of the dark-
ness and lack of carbohydrate source. Without light,
photosynthesis is blocked and carbon fixation is nearly
obliterated. Cells strive to build up carbon reserves
(hence the presence of a 1,4-alpha-D-glucan_n as a
reporter metabolite) and oxidative phosphorylation is the
Table 6 Reporter metabolites for the light shift experiment.
a)b)c)
MetaboliteNumber of
neighbors
MetaboliteNumber of
neighbors
MetaboliteNumber of
neighbors
All time points Dark to LightLight to Dark
L-tyrosine4N-carbamoyl-L-aspartate35-phosphoribosyl-N-formylglycineamidine3
N-carbamoyl-L-
aspartate
3dihydroorotate3diphosphate76
dTDP45-phosphoribosyl 1-
pirophosphate
9a 1,4-alpha-D-glucan_n2
L-isoleucine3L-valine3a 1,4-alpha-D-glucan_n12
D-ribulose-5-
phosphate
45-phospho-ribosyl-
glycineamide
3 UDP-N-acetylmuramoyl-L-alanyl-D-glutamyl-meso-2,6-
diaminoheptanedioate
2
D-myo-inositol (3)-
monophosphate
2 O-phospho-L-
homoserine
2 pyridoxine-5’-phosphate2
myo-inositol2peptidylproline (omega
= 180)
4(E, E)-farnesyl diphosphate3
L-valine3peptidylproline (omega
= 0)
4 GMP6
succinyl-CoA3indole-3-glycerol-
phosphate
2phosphoribosylformiminoAICAR-phosphate2
adenosine25-aminoimidazole
ribonucleotide
3L-aspartyl-4-phosphate2
GTP13tetrahydrofolate
cofactors
8 pantothenate2
thioredoxin11 GTP13undecaprenyl-diphospho-N-acetylmuramoyl-L-alanyl-D-
glutamyl-meso-2,6-diaminopimeloyl-D-alanyl-D-alanine
2
thioredoxin
disulfide
11L-glutamate gamma-
semialdehyde
2MurAc(oyl-L-Ala-D-gamma-Glu-L-Lys-D-Ala-D-Ala)-
diphospho-undecaprenol
2
p-aminobenzoate2inosine-5’-phosphate5undecaprenyl-diphospho-N-acetylmuramoyl-L-alanyl-D-
glutamyl-L-lysyl-D-alanyl-D-alanine
2
acetylphosphate2pantetheine 4’-
phosphate
2 L-aspartate-semialdehyde2
glycine7UDP-N-acetylmuramoyl-
L-alanyl-D-glutamate
25-phospho-ribosyl-glycineamide3
succinate7phytoene25’-phosphoribosyl-N-formylglycineamide4
dihydroorotate 3thioredoxin11sulfur2
PC12thioredoxin disulfide 11glycine7
Reporter metabolites for each set of arrays analysed with Reporter Features software.
Montagud et al. BMC Systems Biology 2010, 4:156
http://www.biomedcentral.com/1752-0509/4/156
Page 11 of 16

Page 12

main energy pathway that remains present. Regulation is
centered on the energy metabolism shift (60% of the
total co-regulated sub-network), withholding amino
acids and nucleotide precursors and keeping the cofac-
tors available in a low-profile metabolism.
Conclusions
We have successfully reconstructed a genome-scale
metabolic network for Synechocystis sp. PCC6803, called
iSyn669, which allows simulating production of all the
metabolic precursors of the organism. The metabolic
reconstruction represents an up-to-date database that
encompasses all knowledge available in public databases,
scientific publications and textbooks on the metabolism
of this cyanobacterium.
From the annotation publicly available, our metabolic
network includes 882 metabolic reactions and 790 meta-
bolites, as well as the information from 669 genes that
have some relationship with the metabolic reactions.
This model is the most complete and comprehensive
work for Synechocystis sp. PCC6803 to date, which has
its potential as the photosynthetic model organism.
Interestingly, the reconstruction identified 79 reactions
that should be present in the metabolism but with no
cognate gene discovered yet; this should direct experi-
mental work at the discovery of these genes. Topological
characteristics of the network resemble those of other
reconstructed microbial metabolic networks and thus
provide an additional input for the analysis of their
structural and organizational properties from evolution-
ary perspective.
Applicability of iSyn669 metabolic model was demon-
strated by using a variety of computational analyses.
Flux balance analysis was applied in order to simulate
the three physiologically important growth conditions of
cyanobacteria, viz., heterotrophic, mixotrophic and auto-
trophic. Our metabolic model was capable of simulating
the production of the monomers or building blocks that
build up the cells, in the range that is in agreement with
the reported growth experiments. Our photosynthetic
metabolic model includes all of the central metabolic
pathways that previous works [23-25] considered.
Regarding the parts from our model that overlap with
the previous works (part of the central carbon metabo-
lism), the predictions for the flux directionality changes
following light shift match between those models and
iSyn669. In fact, iSyn669 expands the flux study to all
the pathways described in the Synechocystis sp.
PCC6803 genome annotation. Further work should be
directed at the definition of a detailed and descriptive
biomass cell composition, so as to have a better repre-
sentation of the biomass equation for simulation
purposes.
Single reaction/gene knock-out simulations revealed
311 genes that are essential for the survival. Bearing in
mind the distance from the efforts taken in the annota-
tion of the genome of the bacteria and yeast models to
that of the cyanobacterium, our study shows that Syne-
chocystis sp. PCC6803 has a larger fraction of genes that
are essential for producing biomass, as opposed to
Escherichia coli and Saccharomyces cerevisiae. Further
investigation of the causes for this difference will be of
definite interest in understanding the genome annota-
tion and/or the evolution of the metabolic network of
Synechocystis.
Evaluation of the theoretical potential of this organism
to produce hydrogen was assessed, in support of the
efforts directed to this direction from several groups and
scientific council initiatives. Present hydrogen produc-
tion projects are far from the theoretical potential, but
efforts in this field can trigger a very significant increase
of the present hydrogen evolution rates in Synechocystis
sp. PCC6803 or other photobiological production plat-
forms candidates, e.g. Chlamydomonas reinhardtii, Nos-
toc punctiforme and Synechococcus species.
Suitability of the presented model for performing in
silico metabolic engineering analysis was demonstrated
by using OptGene software framework. Furthermore, we
also show that iSyn669 can be used as a scaffold to inte-
grate network-wide omics data. As a case study, we
identified key reporter metabolites around which regula-
tion during light shifts is organized, as well as gene sub-
networks that were co-regulated across the light
conditions.
Altogether, the genome-scale metabolic network of
Synechocystis sp. PCC6803 (iSyn669) will be a valuable
tool for the applied and fundamental research of Syne-
chocystis sp. PCC6803, as well as for the broad field of
metabolic systems biology. iSyn669 represents an impor-
tant step for the integration of tools and knowledge
from different disciplines towards development of
photo-biological cell factories.
Methods
Metabolic network reconstruction
Pathway Tools software [33] was used to construct a
Synechocystis-specific database of genes, proteins,
enzymes and metabolites. Synechocystis sp. PCC6803
genome and annotation files were downloaded from
NCBI Entrez Genome repository as of date 10 of Sep-
tember of 2008 [51]. Pathway tools retrieved a first ver-
sion of the network, which had to be checked with
different kinds of databases depending on the informa-
tion they bear. Databases used towards this purpose
included Enzyme nomenclature database [34], KEGG
pathway database [27], BioCyc genome database [26],
Montagud et al. BMC Systems Biology 2010, 4:156
http://www.biomedcentral.com/1752-0509/4/156
Page 12 of 16

Page 13

BRENDA Enzyme database [28] and UniProt protein
database [29].
Parts that characterize Synechocystis network, like the
incomplete TCA cycle [52,53], the presence of the
glyoxylate shunt [35], the interconnected photosynthesis
and oxidative phosphorylation [54] or the cyclic and
non-cyclic electron transport related to these latter pro-
cesses [55-57], were accounted for in detail.
At the end of the reconstruction process, four kinds of
relationships were present in the database: reaction with
cognate genes, reactions that needed to be included in
the model in order to have metabolic precursors in the
network (with no assigned genes), non-enzymatic reac-
tions that have no related gene, and genes described in
the annotations but with no assigned function. For an
overview of the underlying process, please refer to Fort-
ser et al [32] work on the reconstruction of Saccharo-
myces cerevisiae metabolic network.
Linear programming for Flux Balance Analysis
The set of biochemical reactions of the genome-scale
metabolic model were formulated as a steady state stoi-
chiometric model:
S v ⋅= 0
The details are described elsewhere, for example in Ste-
phanopoulos et al [40]. This model describes cellular
behavior under pseudo steady-state conditions, where S is
stoichiometric matrix that contains the stoichiometric
coefficients corresponding to all internal (balanced) meta-
bolites. v is flux vector that corresponds to the columns of
S. Given a set of experimentally-driven constraints, former
equation was solved by using linear programming, the
approach known as flux balance analysis, or FBA [16].
Since the number of reactions is typically larger than
the number of metabolites, the system becomes under-
determined. In order to obtain a feasible solution for the
intracellular fluxes, an optimization criterion on meta-
bolic balances has to be imposed. This can be formu-
lated by maximizing one of the biochemical reactions, e.
g. biomass equation, subject to the mass balance and
the capacity constraints.
For instance,
Maxsubject to N
R
R
R




ij
j irr
,
j rev
,
j const
,
Sj
()
= ∀ ∈
∈
∈
<
<
∈
∈
+
·0
, ,
,
,
,,
vv
vv
min j const

max
j uptake

min j uptakemax
<
<


R
where vjis the rate of the jthreaction. The elements of
the flux vector v were constrained for the definition of
reversible and irreversible reactions, vj, revand vj, irr,
respectively. Additionally, two set of equations were estab-
lished, νj, const, constrained metabolic reactions, and νj,
uptake, uptake reactions, which were bound by experimen-
tally determined values from the literature. Biomass synth-
esis was considered as a drain of precursors or building
blocks into a hypothetical biomass component. Flux
through biomass synthesis reaction, being the biomass for-
mation rate, is directly related to growth of the modeled
organism [40]. Table 3 shows the biomass composition
that was considered in the iSyn669 metabolic model.
Simulations were performed with the OptGene soft-
ware [37]. Some capacity constraints had to be added in
order to have a feasible solution for the linear program-
ming problem. As an example, maximum uptake rates
were determined as follows: maximum glucose uptake
rate under heterotrophic conditions was found to be 0.85
mmol glucose gDW-1h-1[23]. Maximum CO2uptake rate
was found to be 3.7 mmol CO2gDW-1h-1[24]. Addition-
ally, we fixed the maintenance requirement for the het-
erotrophic case to be 1.67 ATP moles per mole of
glucose consumed as was determined by ref [24], and
was maintained for autotrophic and mixotrophic growth.
MOMA algorithm
Segre et al [17] introduced the method of minimization
of metabolic adjustment (MOMA) to better understand
the flux states of mutants. MOMA is based on the same
stoichiometric constraints as FBA, but relaxes the
assumption of optimal growth flux for the mutants, test-
ing the hypothesis that the corresponding flux distribu-
tion is better approximated by the flux minimal
response to the perturbation than by the optimal one.
MOMA algorithm searches for a point in the feasible
space of the solutions space of the knock-out (Fj) that has
minimal distance from a given flux vector w. The goal is
to find the vector x ÎFjsuch that the Euclidean distance
D w x
( , )
wx
ii
i
N
()
=−
=∑
2
1
is minimized. For details, please address to Segre et al
[17].
Reporter Features algorithm
Reporter Features software [48] works on three kinds of
information - network, omics data and association
between genes and the nodes in the network. We have
used Reporter Features for a transcriptomic analysis, so
our three files were p-values file, resulting from a Stu-
dent t-test run on transcriptomic data, interaction file,
where reactions are connected to the corresponding
substrates and products, and association file, where gene
Montagud et al. BMC Systems Biology 2010, 4:156
http://www.biomedcentral.com/1752-0509/4/156
Page 13 of 16

Page 14

are associated to reactions they are involved in, either by
coding for the enzyme or by regulating the gene that
codes for the enzyme.
In brief, Reporter algorithm converts the p-value for a
given node to a z-score by using the inverse normal
cumulative distribution function (cdf-1).
z cdf1p
gene i
1
gene i
=
()
−
–
After scoring each non-feature node in this fashion,
we need to calculate the score of each feature j, zfeature j.
We used the scoring method based on distribution of
the means, which is a test for the null hypothesis “genes
adjacent to feature j display their normalized average
response by chance”. In particular, the score of each fea-
ture j is defined as the average of the scores of its neigh-
bour Njnodes (genes), i.e.:
z
N
z
feature j
j
genek
k
Nj
=∑
=
1
1
To evaluate the significance of each zfeature j, this value
should be corrected for the background distribution of z
scores in the data, by subtracting the mean (mN) and
dividing by the standard deviation (sN) of random aggre-
gates of size N.
z
zm
s
feature j
corrected
feature jN
N
=
−
()
Additional material
Additional file 1: iSyn669 reactions to gene connections. Excel file
with the list of iSyn669 reactions and its cognate list of genes.
Additional file 2: iSyn669 genome-scale metabolic model in
OptGene format. Text file with the stoichiometric model, in OptGene
[37] format, with all the constraints needed for its simulation with FBA
algorithm.
Additional file 3: Most connected metabolites with filtered
cofactors. Supplementary table with most connected metabolites once
the cofactors have been filtered.
Additional file 4: iSyn669 metabolic fluxes simulated under four
conditions. Excel file with all the reactions simulations and resulting flux
ranges from the model simulated under four growth conditions:
autotrophy, dark o pure heterotrophy, light-activated heterotrophy and
mixotrophy.
Additional file 5: Fluxes of reactions around pyruvate. Flux values (in
mmol/g DCW/h) for reactions that produce or drain pyruvate in
Synechocystis sp. PCC6803 metabolism. Negative sign in bidirectional
reactions means pyruvate consumption. Reactions names can be traced
in reaction list in Additional files 2 and fluxes can be found in Additional
file 4.
Additional file 6: FBA and MOMA simulation values for biomass
growth in Synechocystis sp. PCC6803, Escherichia coli and
Saccharomyces cerevisiae genome-scale metabolic models. Excel file
with the growth values under MOMA simulation for Synechocystis sp.
PCC6803, Escherichia coli and Saccharomyces cerevisiae. Data for
Synechocystis is original from present work, data for Escherichia coli has
been obtained from metabolic model from reference 18 and data for
Saccharomyces cerevisiae is from reference 30.
Additional file 7: iSyn669 groups of correlated genes in the three
sets of arrays of light shift experiments. Word file with the list of
iSyn669 correlated genes in “All time points”, “Dark to light” and “Light to
dark” analyses.
Abbreviations
BM: biomass; DCW: dry cell weight; FBA: flux balance analysis; MCA:
metabolic control analysis; MOMA: minimization of metabolic adjustments;
ORF: Open Reading Frame; PEP: phosphoenolpyruvate; ROOM: regulatory on-
off minimization of metabolic fluxes; RuBisCO: Ribulose-1,5-bisphosphate
carboxylase oxygenase; TCA cycle: tricarboxylic acid cycle
Acknowledgements
This work was financially supported by MICINN TIN2009-12359 project
ArtBioCom, EU FP7-KBBE-2007 project TarPol (contract n°212894) and EU
FP6-NEST-2005 project BioModularH2 (contract n° 043340). AM thanks to
Generalitat Valenciana grant BFPI/2007/283 and EN to Ministerio de
Educación y Ciencia de España through the program Juan de la Cierva.
Author details
1Instituto Universitario de Matemática Pura y Aplicada, Universidad
Politécnica de Valencia, Camino de Vera 14, 46022 Valencia, Spain.
2Departamento de Lenguajes y Ciencias de la Computación, Campus de
Teatrinos, Universidad de Málaga, 29071 Málaga, Spain.3Structural and
Computational Biology Unit, European Molecular Biology Laboratory,
Meyerhofstrasse 1, D-69117 Heidelberg, Germany.
Authors’ contributions
AM and EN conducted the reconstruction and the different analyses. PF and
JFU conceived of the study and participated in its design. AM and KRP
designed the study and wrote the manuscript. All authors contributed to,
read and approved the final manuscript.
Received: 5 February 2010 Accepted: 17 November 2010
Published: 17 November 2010
References
1. Allen MM, Smith AJ: Nitrogen chlorosis in blue-green algae. Arch Mikrobiol
1969, 69:114-120.
2.Tamagnini P, Axelsson R, Lindberg P, Oxelfelt F, Wunschiers R, Lindblad P:
Hydrogenases and hydrogen metabolism of cyanobacteria. Microbiol Mol
Biol Rev 2002, 66:1-20, table of contents.
3.Schopf J: The Fossil Record: Tracing the Roots of the Cyanobacterial
Lineage. In The ecology of cyanobacteria. Edited by: Whitton B, Potts M.
Dordrecht: Kluwer Academic Publishers; 2000:13-35.
4.Shi T, Falkowski PG: Genome evolution in cyanobacteria: the stable core
and the variable shell. Proc Natl Acad Sci USA 2008, 105:2510-2515.
5.Tamagnini P, Leitao E, Oliveira P, Ferreira D, Pinto F, Harris DJ, Heidorn T,
Lindblad P: Cyanobacterial hydrogenases: diversity, regulation and
applications. FEMS Microbiol Rev 2007, 31:692-720.
6.Lindberg P, Park S, Melis A: Engineering a platform for photosynthetic
isoprene production in cyanobacteria, using Synechocystis as the model
organism. Metab Eng 2010, 12:70-79.
7.Wu GF, Wu QY, Shen ZY: Accumulation of poly-beta-hydroxybutyrate in
cyanobacterium Synechocystis sp. PCC6803. Bioresour Technol 2001,
76:85-90.
8.Liu X, Curtiss R: Nickel-inducible lysis system in Synechocystis sp. PCC
6803. Proc Natl Acad Sci USA 2009, 106:21550-21554.
9.Navarro E, Montagud A, Fernández de Córdoba P, Urchueguía JF:
Metabolic flux analysis of the hydrogen production potential in
Synechocystis sp. PCC6803. Int J Hydrogen Energy 2009, 34:8828-8838.
10.McHugh K: Hydrogen production methods Alexandria, Virginia: MPR
Associates, Inc; 2005.
Montagud et al. BMC Systems Biology 2010, 4:156
http://www.biomedcentral.com/1752-0509/4/156
Page 14 of 16

Page 15

11.Turner J, Sverdrup G, Mann M, Maness P, Kroposki B, Ghirardi M, Evans R,
Blake D: Renewable hydrogen production. International Journal Energy
Research 2008, 32:379-407.
Herrero A, Flores E: The cyanobacteria: molecular biology, genomics, and
evolution Norfolk, UK: Caister Academic Press; 2008.
Oberhardt MA, Palsson BO, Papin JA: Applications of genome-scale
metabolic reconstructions. Mol Syst Biol 2009, 5:320.
Patil KR, Akesson M, Nielsen J: Use of genome-scale microbial models for
metabolic engineering. Curr Opin Biotechnol 2004, 15:64-69.
Varma A, Palsson BO: Metabolic capabilities of Escherichia coli: II. Optimal
growth patterns. J Theor Biol 1993, 165:503-522.
Edwards J, Ramakrishna R, Schilling C, Palsson B: Metabolic flux balance
analysis. In Metabolic engineering. Edited by: Lee S, Papoutsakis E. New York:
Marcel Dekker Inc; 1999.
Segre D, Vitkup D, Church GM: Analysis of optimality in natural and
perturbed metabolic networks. Proc Natl Acad Sci USA 2002,
99:15112-15117.
Shlomi T, Berkman O, Ruppin E: Regulatory on/off minimization of
metabolic flux changes after genetic perturbations. Proc Natl Acad Sci
USA 2005, 102:7695-7700.
Rapoport TA, Heinrich R, Jacobasch G, Rapoport S: A linear steady-state
treatment of enzymatic chains. A mathematical model of glycolysis of
human erythrocytes. Eur J Biochem 1974, 42:107-120.
Kacser H, Burns JA: The control of flux. Symp Soc Exp Biol 1973, 27:65-104.
Kaneko T, Sato S, Kotani H, Tanaka A, Asamizu E, Nakamura Y, Miyajima N,
Hirosawa M, Sugiura M, Sasamoto S, et al: Sequence analysis of the
genome of the unicellular cyanobacterium Synechocystis sp. strain
PCC6803. II. Sequence determination of the entire genome and
assignment of potential protein-coding regions (supplement). DNA Res
1996, 3:185-209.
Kaneko T, Nakamura Y, Sasamoto S, Watanabe A, Kohara M, Matsumoto M,
Shimpo S, Yamada M, Tabata S: Structural analysis of four large plasmids
harboring in a unicellular cyanobacterium, Synechocystis sp. PCC 6803.
DNA Res 2003, 10:221-228.
Yang C, Hua Q, Shimizu K: Metabolic flux analysis in Synechocystis using
isotope distribution from 13C-labeled glucose. Metab Eng 2002, 4:202-216.
Shastri AA, Morgan JA: Flux balance analysis of photoautotrophic
metabolism. Biotechnol Prog 2005, 21:1617-1626.
Fu P: Genome-scale modeling of Synechocystis sp. PCC6803 and
prediction of pathway insertion. Journal of Chemical Technology &
Biotechnology 2009, 84:473-483.
Karp PD, Ouzounis CA, Moore-Kochlacs C, Goldovsky L, Kaipa P, Ahren D,
Tsoka S, Darzentas N, Kunin V, Lopez-Bigas N: Expansion of the BioCyc
collection of pathway/genome databases to 160 genomes. Nucleic Acids
Res 2005, 33:6083-6089.
Kanehisa M, Araki M, Goto S, Hattori M, Hirakawa M, Itoh M, Katayama T,
Kawashima S, Okuda S, Tokimatsu T, Yamanishi Y: KEGG for linking
genomes to life and the environment. Nucleic Acids Res 2008, 36:
D480-484.
Chang A, Scheer M, Grote A, Schomburg I, Schomburg D: BRENDA,
AMENDA and FRENDA the enzyme information system: new content
and tools in 2009. Nucleic Acids Res 2009, 37:D588-592.
The universal protein resource (UniProt). Nucleic Acids Res 2008, 36:
D190-195.
Weise S, Grosse I, Klukas C, Koschutzki D, Scholz U, Schreiber F, Junker BH:
Meta-All: a system for managing metabolic pathway information. BMC
Bioinformatics 2006, 7:465.
Feist AM, Herrgard MJ, Thiele I, Reed JL, Palsson BO: Reconstruction of
biochemical networks in microorganisms. Nat Rev Microbiol 2009,
7:129-143.
Forster J, Famili I, Fu P, Palsson BO, Nielsen J: Genome-scale reconstruction
of the Saccharomyces cerevisiae metabolic network. Genome Res 2003,
13:244-253.
Karp PD, Paley S, Romero P: The Pathway Tools software. Bioinformatics
2002, 18(Suppl 1):S225-232.
Bairoch A: The ENZYME database in 2000. Nucleic Acids Res 2000,
28:304-305.
Yang C, Hua Q, Shimizu K: Quantitative analysis of intracellular metabolic
fluxes using GC-MS and two-dimensional NMR spectroscopy. J Biosci
Bioeng 2002, 93:78-87.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.Pearce J, Carr NG: The metabolism of acetate by the blue-green algae,
Anabaena variabilis and Anacystis nidulans. J Gen Microbiol 1967,
49:301-313.
Patil KR, Rocha I, Forster J, Nielsen J: Evolutionary programming as a
platform for in silico metabolic engineering. BMC Bioinformatics 2005,
6:308.
Feist AM, Henry CS, Reed JL, Krummenacker M, Joyce AR, Karp PD,
Broadbelt LJ, Hatzimanikatis V, Palsson BO: A genome-scale metabolic
reconstruction for Escherichia coli K-12 MG1655 that accounts for 1260
ORFs and thermodynamic information. Mol Syst Biol 2007, 3:121.
Zelezniak A, Pers TH, Soares S, Patti ME, Patil KR: Metabolic network
topology reveals transcriptional regulatory signatures of type 2 diabetes.
PLoS Comput Biol 2010, 6:e1000729.
Stephanopoulos G, Aristidou AA, Nielsen JH: Metabolic engineering: principles
and methodologies San Diego: Academic Press; 1998.
Schuetz R, Kuepfer L, Sauer U: Systematic evaluation of objective
functions for predicting intracellular fluxes in Escherichia coli. Mol Syst
Biol 2007, 3:119.
Anderson SL, McIntosh L: Light-activated heterotrophic growth of the
cyanobacterium Synechocystis sp. strain PCC 6803: a blue-light-requiring
process. J Bacteriol 1991, 173:2761-2767.
Carr NG, Whitton BA: The Biology of cyanobacteria Berkeley: University of
California Press; 1982.
Pelroy RA, Rippka R, Stanier RY: Metabolism of glucose by unicellular
blue-green algae. Arch Mikrobiol 1972, 87:303-322.
Loferer-Krossbacher M, Klima J, Psenner R: Determination of bacterial cell
dry mass by transmission electron microscopy and densitometric image
analysis. Appl Environ Microbiol 1998, 64:688-694.
Lawrence BA, Suarez C, DePina A, Click E, Kolodny NH, Allen MM: Two
internal pools of soluble polyphosphate in the cyanobacterium
Synechocystis sp. strain PCC 6308: an in vivo 31P NMR spectroscopic
study. Arch Microbiol 1998, 169:195-200.
Stephanopoulos G, Alper H, Moxley J: Exploiting biological complexity for
strain improvement through systems biology. Nat Biotechnol 2004,
22:1261-1267.
Oliveira AP, Patil KR, Nielsen J: Architecture of transcriptional regulatory
circuits is knitted over the topology of bio-molecular interaction
networks. BMC Syst Biol 2008, 2:17.
Patil KR, Nielsen J: Uncovering transcriptional regulation of metabolism
by using metabolic network topology. Proc Natl Acad Sci USA 2005,
102:2685-2689.
Gill RT, Katsoulakis E, Schmitt W, Taroncher-Oldenburg G, Misra J,
Stephanopoulos G: Genome-wide dynamic transcriptional profiling of the
light-to-dark transition in Synechocystis sp. strain PCC 6803. J Bacteriol
2002, 184:3671-3681.
NCBI Entrez Genome for Synechocystis sp. PCC6803. [http://www.ncbi.
nlm.nih.gov/sites/entrez?Db=genome&Cmd=ShowDetailView&TermToSearch
= 112].
Pearce J, Leach CK, Carr NG: The incomplete tricarboxylic acid cycle in
the blue-green alga Anabaena variabilis. J Gen Microbiol 1969, 55:371-378.
Vazquez-Bermudez MF, Herrero A, Flores E: Uptake of 2-oxoglutarate in
Synechococcus strains transformed with the Escherichia coli kgtP gene.
J Bacteriol 2000, 182:211-215.
Peschek GA, Löffelhardt W, Schmetterer G: The phototrophic prokaryotes
New York: Kluwer Academic/Plenum; 1999.
Rubio FC, Camacho FG, Sevilla JM, Chisti Y, Grima EM: A mechanistic
model of photosynthesis in microalgae. Biotechnol Bioeng 2003,
81:459-473.
Albertsson P: A quantitative model of the domain structure of the
photosynthetic membrane. Trends Plant Sci 2001, 6:349-358.
Allen J: Photosynthesis of ATP-electrons, proton pumps, rotors, and
poise. Cell 2002, 110:273-276.
Herdman M, Janvier M, Waterbury J, Rippka R, Stanier R: Deoxyribonucleic
Acid Base Composition of Cyanobacteria. Journal of General Microbiology
1979, 111:63-71.
Tasaka Y, Gombos Z, Nishiyama Y, Mohanty P, Ohba T, Ohki K, Murata N:
Targeted mutagenesis of acyl-lipid desaturases in Synechocystis:
evidence for the important roles of polyunsaturated membrane lipids in
growth, respiration and photosynthesis. EMBO J 1996, 15:391-396.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
Montagud et al. BMC Systems Biology 2010, 4:156
http://www.biomedcentral.com/1752-0509/4/156
Page 15 of 16