In silico metabolic

reconstruction of Fe-S
cluster biogenesis in yeast

Rui Alves
Ciencies Mediques Basiques
Universitat de Lleida
 Introduction

 Novel pathways and gene circuits emerge from genome

and metagenome sequencing and from other non-
genomic research

 Creates a two-fold problem

 Identify proteins involved in pathways
 Assemble the pathway structure.

 Assembling these circuits/pathways is not simple

 Developing data integration methods that facilitate

solving both problems is not trivial

02/17/11 2
Objective of the research line

 Develop coherent framework where

different computational methods and
data sets are integrated to assemble
biological pathways & circuits and
generate hypothesis about how they
work
 Today: Example of integration and
application to the reconstruction of FeS
cluster biogenesis in yeast
 Using literature for network
reconstruction

• A huge amount of information is buried

in literature databases
• Literature co-occurrence of genes
indicates functional
relationship/interaction
• iHOP performs this type of analysis
automatically
02/17/11 4
Using phylogenetic profiles to predict
network interactions

Sequence
A 1
Homology
Protein A C 0.9
… …
analysis
Target Homologue in BHomologue in
0.11 …
Genome Genome 1? Genome 2?
…N …
A Y Calculate
B Proteins (ANand C) that are present
Y
…
and absent…in the same set of
C genomes are coincidence index
Y likely to be involved
N in the same …
process and therefore
… interact … … …

Similarly, if A is absent in all genomes in which protein B is present

Database of
A C
there is a likelihood that they perform the same function! phylogenetic
1
2
profiles for
j/number of genomes each protein in
02/17/11 an organism5
Study of docking interactions
in silico
…SSQIE… Sequence of known structure
…SSQEE… Homologue sequence for structure prediction
THREAD

D
If structures already exist, even easier
O as the
OPTIMIZE C
modeling part can be skipped
K

Ranking of scores for

docking between the
02/17/11 different proteins 6
Validating alternative network
connectivity
 Expert knowledge for curation and
selection of alternative connectivity for
the network
 Why? No good methods to automatically
create novel causal networks for
mathematical modeling
 Mathematical modeling crucial for
testing and comparing the dynamic
behavior of alternative networks to wet
lab experiments
02/17/11 7
 The modeling
A,…

… C E+F

d ( C)
≈ α ... − β Ah1 C h 2 ...
dt
g<0 inhibits flux
g=0 no influence on flux
02/17/11
g>0 activates flux 8
A reconstruction method

02/17/11 9
 Fe-S clusters

 Iron-Sulfur clusters are coordinated clusters of

ions that participate in electron transfer during
biological processes

e-
e-
S
Protein Protein
Fe Fe
Cysteine Cysteine
S
 What is known about FeSC
biogenesis

 About 15 different mitochondrial proteins

contribute to biogenesis and assembly in yeast

 The assembly process is ill-understood

 All 15 proteins have one thing in common

02/17/11 11
Phenotype of FeSC machinery
deletion mutants
FeSC Dependent
Fe Level Protein Activity

WT ∆ WT ∆
Fe Accumulates FeSC dependent protein
activity is impaired
FeSC biogenesis in a nutshell

Fe S
Scaffold Scaffold Scaffold Scaffold

Synthesis FeSC
(S)
FeSC
Transfer
Holo-P (T)
Apo-P
Damaged
FeSC
Repair
(R) Holo-P
The proteins and their function

 Understanding the role of seven proteins in

S. cerevisiae. FeS cluster biogenesis:
Grx5, Arh1-Yah1, Ssq1-Jac1-Mge1, Nfs1

02/17/11 14
 Grx5 is involved in FeSC
biogenesis in S. cerevisiae
Glutaredoxin
 Mediates glutathionylation state of Cys
residues
 May mediate protein-protein disulfide bridge

reduction (Belli et al. 2002, Tamarit et al.

2003, JBC)

 FeSC coordinate (mostly) with Cys residues

 Is Grx5 regulation of Cys reduction state in any

specific protein(s) involved in FeSC biogenesis
sufficient for phenotype?
Predicting partners for Grx5:
the protocol

•Combine literature analysis, phylogenetic

analysis of fully sequenced genomes and in
silico protein docking to predict the most
likely targets of Grx5

02/17/11 16
 Possible partners of Grx5 in
FeSC biogenesis

Isu1 Bibliography
Docking

Isa1 Phylogeny+
Docking

Isa2

Nfs1

Grx5

02/17/11 17
Possible roles of Grx5 in FeSC
biogenesis:
regulation of glutathionylation
S
HSGlutathione

P P S-SG

Grx5
02/17/11 18
Possible roles of Grx5 in FeSC
Biogenesis:
Recovery of dead-end complex

Scaffold SH HS Nfs1

Grx5

Scaffold S S Nfs1
 Studying the effect of Grx5:
the protocol

•Create models for alternative networks

•Normalize equations and scan parameters

•Compare simulations with known systemic

behavior to validate or invalidate alternatives
02/17/11 20
 gene deletion on protein
activity if Grx5 recovers Nfs1
activity
FeSC Dependent Protein Activity

1 1

1000s of
0.5 0.5 simulations

0.1 0.1

WT ∆ Not Recovering
Belli et al. MBC 13:1109 recovering Nfs1 and
Nfs1 and Scaffold
Scaffold
 gene deletion on protein
activity if Grx5 recovers Nfs1
activity
Fe Levels

1000s of
simulations

1 1

WT ∆ Not Recovering
Belli et al. MBC 13:1109 recovering Nfs1 and
Nfs1and Scaffold
Scaffold
 Conclusions:
Possible Modes of
action for Grx5

Reproducing
9 experimental No 6
phenotype?

Yes 3 Nfs1-Scaffold

Prediction:
Grx5 modulates Nfs1 and Scaffold activity/Interactions
Grx5 interacts with scaffold in
two-hybrid assay

Negative Controls Grx5 Positive

Scaffold Control

02/17/11 24
 Arh1-Yah1 proteins are
involved in FeSC biogenesis in
S. cerevisiae
 Arh1-Yah1
 Ferredoxin Reductase – Ferredoxin

 These proteins supply/drain electrons from other

processes

 What is their role in FeSC biogenesis?

 Possible partners of Arh1-
Yah1 in FeSC biogenesis
Docking +
Isu1
Yah1 Phylogeny
Isu2
Bibliography +
Isa1 Arh1 Docking +
Phylogeny

Isa2

Nfu1

02/17/11 26
Possible modes of Arh1-Yah1
action
Arh1/Yah1
Fe S
Scaffold Scaffold Scaffold Scaffold

Synthesis FeSC
(S)
FeSC Arh1/Yah1
Transfer
Holo-P
(T)
Apo-P
Damaged
FeSC
Repair
(R) Holo-P
Arh1/Yah1
 Model reproduces gene
deletion effect on FeSC
dependent activity if Arh1-
FeSC Dependent Protein Activity

Yah1 act on ST
1
1000s of
simulations

0.2

WT ∆ T R S
Li et al. JBC 276:1503 RS ST
Lange et al. PNAS 97:1050 RST
RT
 accumulation upon gene
deletion if Arh1-Yah1 act on
ST
Fe Levels

1000s of
simulations
1

WT ∆ S R S
Li et al. JBC 276:1503 T RT
Lange et al. PNAS 97:1050 ST
RS RST
 Conclusions for Arh1-Yah1:
Possible Modes of Arh1-
Yah1 action

Reproducing
7 experimental No 5
phenotype?

S
Yes 2
ST
 Arh1-Yah1 act on ST steps of FeSC biogenesis
 Consistent with results from
Li et al. JBC 276:1503
Lange et al. PNAS 97:1050
 Ssq1-Jac1-Mge1 proteins are
involved in FeSC biogenesis in
S. cerevisiae
 Ssq1-Jac1-Mge1
 Chaperone – Co-Chaperone – Nucleotide Release

Factor
 These proteins help fold-stabilize other proteins

 It has been suggested that they help stabilize the

FeSC in the scaffold for transfer.
 Is this role necessary to justify the ∆ phenotypes?
 Model reproduces gene
deletion effect on FeSC
dependent activity if
chaperones act on stability
FeSC Dependent Protein Activity

1
1000s of
simulations

0.2

WT ∆ Stability Fold

02/17/11 32
 accumulation upon gene
deletion if chaperones act on
stability
Fe Levels

1000s of
simulations
1

WT ∆ Stability Fold
 Conclusions:
Possible Modes of Ssq1-
Jac1-Mge1 action

Reproducing
3 experimental No 1
phenotype?

Fold
Yes 2 Stability/Fold

 Scaffolds need to act only on folding in FeSC biogenesis

 Nfs1 is involved in FeSC
biogenesis in S. cerevisiae
 Nfs1
 Cisteine Desulfurase

 This protein provides sulfur for the biogenesis

 It has been shown in vitro that it is able to repair

FeSC in situ and bypass the biogenesis pathway
 Is this role important for the phenotype?
 Model reproduces gene
deletion effect on FeSC
dependent activity if Nfs1 acts
on SR
FeSC Dependent Protein Activity

1
1000s of
simulations

0.2

WT ∆ R S
RS
 Model reproduces Fe
accumulation upon gene
deletion if Nfs1 acts on S
Fe Levels

1000s of
simulations
1

WT ∆ R S
RS
 Conclusions:
Possible Modes of Nfs1
action

Reproducing
3 experimental No 1
phenotype?

Yes 2 S
SR
 Nfs1 needs only to act on synthesis but can also act on
repair of FeSC
 Summary
 Arh1-Yah1 acts on Synthesis and transfer of FeSC

 Grx5 modulates Cysteine Desulfurase (Nfs1) activity and

maybe Scaffold activity

 Ssq1-Jac1-Yah1 act on folding FeSC proteins

 Nfs1 acts on in situ synthesis of clusters

 Yfh1 does not modulate Fe import into the mitochondria

Alves & Sorribas 2007 BMC Systems Biology

Alves et. al. 2004 Proteins 56:354
Vilella et. al. 2004 Comp. Func. Genomics 5:328
Alves et. al. 2004 Proteins 57:481
 Acknowledgments
 Albert Sorribas
 FCT
 Spanish Government
 Enric Herrero  NIH (Mike Savageau)
 Armindo Salvador

02/17/11 40
 Grx5 is predicted to dock
facing the Nfs1 active center
Active
center Grx5
Cys residue

Alternative
Grx5
Binding
Alternative solutions
Grx5
Binding Nfs1 dimer
solution

Active
center Nfs1
Cys residue
02/17/11 41
 Predictions for Grx5:

 Grx5 modulates Cysteine Desulfurase

(Nfs1) and scaffold activity and maybe the
interaction between both

02/17/11 42
Possible Modes of Chaperone
Action
Fe S
Scaffold Scaffold Scaffold Scaffold

Synthesis FeSC
(S)
FeSC Ssq1/Jac1/Mge1
Transfer
Holo-P
(T)
Apo-P
Damaged
FeSC
Repair
(R) Holo-P

02/17/11 43
Possible modes of Nfs1 action
Nfs1
Fe S
Scaffold Scaffold Scaffold Scaffold

Synthesis FeSC
(S)
FeSC
Transfer
Holo-P
(T)
Apo-P
Damaged
FeSC
Repair
(R) Holo-P

02/17/11
Nfs1 44
 Possible partners of Nfs1 in
FeSC biogenesis
Isu1
Isu2 Docking

Docking +
Isa1 Phylogeny
Bibliography +
Isa2 Docking +
Phylogeny
Nfu1 Nfs1

02/17/11 45
 Metabolic Reconstruction:
FeSC biogenesis
The view from here
 Test the predictions

 Extend the analysis to a variety of bacteria

(Preliminary results for E. coli and Buchnera)

 Create an interactive database/server to

implement the methodology and apply it to other
systems
 Process of interest

Determining Refining using Design new

Refining using
gene+proteins involved in phylogenetic and experiments to
automated
process “omics” data distinguish
literature analysis
between
alternatives

Protein Structure Two Hybrid Co-evolution Omics data

(PDB/ Models) Screens analysis analysis
Some models are
No model is validated by
validated by existing comparison with
data existing data
In silico Protein Predict
networks Predict Predict
Docking networks networks

Predict networks Analysis of model behavior

Baeysian network/human curation for

alternative network structures
Automated creation of mathematical
models
Reconstructing
Metabolism and
Investigating Design
Principles in Molecular
Biology II

Rui Alves
Ciencies Mediques Basiques
Universitat de Lleida
 Outline
 Metabolic Network Reconstruction
 Iron-Sulfur Cluster Biogenesis Pathway in S.
cerevisiae.
 Pathway Evolution
 Amino
acid biosynthetic pathways protein
composition
 Design Principles
 Regulatory Design in Networks
 Two Component Systems
• Mono-functionality vs. Bi-functionality of Sensor Proteins
Studying an organism

…
ACTG…

S
tres >Dna
s MAACTG…

Measure >DNA Pol

Response MTC…

Find signatures for

physiological
dynamics in
 Stress Regulation
 Some genes are expressed specifically
in response to a stimulus.

 For such genes, the amino acid

composition of the proteins can be
biased to facilitate their own synthesis
and the physiological response.

 This creates a signature of the

response in the protein composition
 Cognate Bias
 Cognate Bias

 Carbon & Sulfur fixing enzymes have a lower

content of amino acids with sidechains
containing a lot of carbon or sulfur atoms
(Baudouin-Cornu et al. Science 293:297)

 Amino acid biosynthetic enzymes have a lower

content of their cognate amino acid (Alves &
Savageau 2005 Mol. Microbiol. 56:1017-34)
Regulation of amino acid
biosynthetic pathways
gene expression
aa Switch to AA
levels poor medium
protei
n No cognate aa in
levels biosynthetic
proteins
Lots of cognate
aa in
biosynthetic
proteins
aa rich aa
mediu t
poor
m mediu
Effect of relative content of cognate
amino acid on protein synthesis
Low cognate aa content in biosynthetic
pathway
aa Protein
levels Synthesis

t
High cognate aa content in biosynthetic pathway
aa Protein
levels Synthesis

t
 The Low Cognate Bias
Hypothesis
 Thus, we hypothesize that evolution would lead to the
selection of amino-acid biosynthetic enzymes that have a
relatively low content of their cognate amino acid. We
call this the “cognate-bias hypothesis”.
 Test Cases:
 E. coli
 S. typhimurium
 B. subtilis
 Calculating the bias
aa- Calculate aa
biosynthesis composition
proteins
KEGG
Rank w/respect
Metacyc Protein to proteome
Sequences
P(aa)
Literature
Control
High
GeneBank Groups:
1

Proteome Av.
Specific 0.5
genome Growing cells
aa composition
… 0
Low
There is low cognate bias in aa
biosynthesis
Positive correlation between
relative cognate aa composition
and specific activity of proteins

× =
Specific activity Protein copies aa synthesis

× =
Specific activity Protein copies aa synthesis

P<0.003
Negative correlation between
relative cognate aa composition
and number of proteins in
pathway
Average aa
content
+

++
+++

++++
1 2 3 4

P<0.003 (except Bs)

Higher bias is due to functional
requirements

Activ
e
Cente
r

02/17/11 60
 Exceptions to the
cognate bias hypothesis

 Functional Reasons
 ActiveCenters (Phe, Tyr)
 Dimerization Domains (Asp)

02/17/11 61
Low cognate bias is present in
organisms with the full
complement of pathways
 Environmental effects
in cognate bias
 The relative abundance of amino acid should influence the
pressure to keep a low cognate amino acid content in
biosynthetic pathways

 E. coli (or S. typhimurium) and B. subtilis live in different

environments

 Amino acid availability is different

 Is there a signal for these differences?

02/17/11 63
Differences in amino acid
composition of distinct habitats

Soil Gut
Low amounts Arg, Glu, Lys, Low amounts Ala, Asp, Gly,
Trp Tyr Ser, Thr
Intermediate Ala, Asp, Thr High amounts Arg, Glu, Lys,
amounts Trp, Tyr
High amounts Gly, Ser

02/17/11 Savageau 1983 Am. Naturalist. 122:732 64

Positive correlation between
environmental levels of aa
and cognate bias
Rank of
Environmen
tal aa

Rank of Cognate
Bias

Presence of aa in environment decreases pressure

for low cognate bias P<0.004
02/17/11 65
 Environmental
availability influences
cognate bias

 Thus the effect of amino acid availability in

the environment and the consequent
regulation and physiological dynamics of
gene expression does leaves a signature in
the cognate bias of the different pathways.

02/17/11 66
 Conclusions:
 The low cognate bias hypothesis is supported by:
 Analysis of protein composition
 Analysis of protein specific activity
 Analysis of pathway length

 Some factors that overcome selective pressure

for low cognate bias
 Functional requirements for specific cognate residues
 Environmental factors

Alves & Savageau 2005 Mol. Microbiol. in press

Amino acid composition:
the view from here
 Extend the work for other bacteria and attempt
to create organism/environment biosignatures

 Analyze ribosomal proteins amino acid bias,

amino acid transport proteins bias and catabolic
protein bias

 Analyze influence of oxydizability on selection of

surface amino acids in proteins
 Outline
 Metabolic Network Reconstruction
 Iron-Sulfur Cluster Biogenesis Pathway in S.
cerevisiae.
 Pathway Evolution
 Amino
acid biosynthetic pathways protein
composition
 Design Principles
 Regulatory Design in Networks
 Two Component Systems
• Mono-functionality vs. Bi-functionality of Sensor Proteins
 Alternative sensor
design in Two
Component Systems
S R* S R*

S* R S* R

Q1 Q2 Q1 Q2

Monofunctional Sensor Bifunctional Sensor

 Studying physiological
differences of
... alternative designs
α β
...
X& = '3 X 4g '34 − X 2h 32 X 3h 33 X 6h 36
X&3 = α 3 X 1g 31 X 4g 34 − β 3 X 2h 32 X 3h 33 X 6h 36 3 3

... ...

A A

Q Q

Q
 Physiological
Predictions
 Bifunctional design lowers Q2 signal
amplification
 prefered when cross-talk is undesirable

 Monofunctional design elevates Q2 signal

amplification
 prefered when cross-talk is desirable.
Predicting Monofunctionality
from structure
Differences
in ATP lid

Bifunctional Monofunctional
Sensor Sensor

~1000 sequences from genomic data of dozens of bacteria

100s predicted structures by modeling

25 monofunctional sensors
Alves & Savageau 2003 Mol. Microbiol. 48:25
 Two Component
Systems – the view from
here
 Independent phosphatase activity (directly
dependent on HK vs. Not directly dependent
on HK)
 Analysis of Phosphorelays
 TCS vs. Eukaryotic signal transduction
 Alternative designs of TCS (Nar system)
 Metabolic Reconstruction
 Summary
 Development and integration of computational
tools to address quantitative biological problems

 Biological results form the application of these

methods
 Reconstruction of Metabolism
 Testing evolutionary hypothesis using large scale
genome data
 Understand the design principles of biological
networks.
 Acknowledgments
 Mike Savageau
 PGDBM
 JNICT
 Mike Sternberg  FCT
 Albert Sorribas  Spanish Government
 Enric Herrero  Portuguese Government
 Armindo Salvador  NIH (Mike Savageau)
 DOD (ONR) (Mike Savageau)
Metabolic
Reconstruction: the view
from here (I)
 The “Putting it
together” Section
Continue Use Know how to
FeSC work reconstruct TCS
network in M.
xanthus

Analyze aa
biosyntesis Analyze more
and enzyme TCS designs
networks
evolution
Grx5 interacts with Scaffold in
Two-Hybrid assay

Negative Controls Grx5 Positive

Scaffold Control

Two-hybrid analysis of the interaction between Grx5 and Scaffold. Numbers over bars indicate the beta-
galactosidase activity (Miller units) in cultures of S. cerevisiae cells co-transformed with plasmids pGBT9
and pACT2 vectors alone, or derivatives expressing the respective Gal4 fusion proteins with Grx5, Scaffold,
and two proteins known to interact. Results are the mean of three independent experiments.
ATP binding domain important in
functionality of sensor
…SSQIE… Sensor with known structure
…SSQ-E… Sensor sequence for structure prediction

100s of structure predictions

Differences
in ATP lid

Predicted to be Predicted to be
Bifunctional Monofunctional
Sensor Sensor

Alves & Savageau 2003 Mol. Microbiol. 48:25