A Network-Based Approach on Elucidating the Multi-
Faceted Nature of Chronological Aging in S. cerevisiae
Esra Borklu Yucel*, Kutlu O. Ulgen
Department of Chemical Engineering, Bogazici University, Istanbul, Turkey
Background: Cellular mechanisms leading to aging and therefore increasing susceptibility to age-related diseases are a
central topic of research since aging is the ultimate, yet not understood mechanism of the fate of a cell. Studies with model
organisms have been conducted to ellucidate these mechanisms, and chronological aging of yeast has been extensively
used as a model for oxidative stress and aging of postmitotic tissues in higher eukaryotes.
Methodology/Principal Findings: The chronological aging network of yeast was reconstructed by integrating protein-
protein interaction data with gene ontology terms. The reconstructed network was then statistically ‘‘tuned’’ based on the
betweenness centrality values of the nodes to compensate for the computer automated method. Both the originally
reconstructed and tuned networks were subjected to topological and modular analyses. Finally, an ultimate ‘‘heart’’ network
was obtained via pooling the step specific key proteins, which resulted from the decomposition of the linear paths
depicting several signaling routes in the tuned network.
Conclusions/Significance: The reconstructed networks are of scale-free and hierarchical nature, following a power law
model with c = 1.49. The results of modular and topological analyses verified that the tuning method was successful. The
significantly enriched gene ontology terms of the modular analysis confirmed also that the multifactorial nature of
chronological aging was captured by the tuned network. The interplay between various signaling pathways such as TOR,
Akt/PKB and cAMP/Protein kinase A was summarized in the ‘‘heart’’ network originated from linear path analysis. The
deletion of four genes, TCB3, SNA3, PST2 and YGR130C, was found to increase the chronological life span of yeast. The
reconstructed networks can also give insight about the effect of other cellular machineries on chronological aging by
targeting different signaling pathways in the linear path analysis, along with unraveling of novel proteins playing part in
Citation: Borklu Yucel E, Ulgen KO (2011) A Network-Based Approach on Elucidating the Multi-Faceted Nature of Chronological Aging in S. cerevisiae. PLoS
ONE 6(12): e29284. doi:10.1371/journal.pone.0029284
Editor: Vladimir N. Uversky, University of South Florida College of Medicine, United States of America
Received July 26, 2011; Accepted November 23, 2011; Published December 21, 2011
Copyright: ? 2011 Borklu Yucel, Ulgen. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was financially supported by Bogazici Research Fund through project 5681 and also by TUBITAK through project 110M428. The funders had
no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: email@example.com
Aging is usually defined as the progressive loss of function
accompanied by decreasing fertility and increasing mortality with
advancing time, due to the accumulation of molecular, cellular
and organ damage. Although it is clear and evident that aging
‘‘occurs’’, the reasons, pathways and regulators responsible for the
mentioned accumulation of deleterious effects are still vaguely
described, rendering the mechanisms that contribute to aging and
age-associated diseases a central topic of interest. Recent works on
model organisms such as yeast, worms and flies have yielded
promising discoveries regarding these mechanisms [1,2] which
may be projected to higher eukaryotes. The yeast Saccharomyces
cerevisiae, an extensively used model organism, harbors two models
of aging: Replicative and Chronological Aging. Replicative aging
term is used for the aging of mitotically active yeast cells, involving
the capacity of daughter cell production of a mother cell, before
senescence . However, yeast chronological life span is the length
of time a population remains viable in the non-dividing, quiescent
state , which is thought to be a suitable model for aging of
post-mitotic tissues . Chronologically aged yeast cultures die
exhibiting typical markers of apoptosis, accumulate oxygen
radicals, and show caspase activation , i.e. processes crucial
for the cell fate of other higher eukaryotes. Several alterations in
signaling pathways such as TOR, Akt/PKB and cAMP/Protein
kinase A, which are also conserved between yeast and higher
eukaryotes such as Homo sapiens, have been demonstrated to affect
the damage accumulation previously mentioned [7–10]. In yeast,
these pathways may be represented by orthologous proteins like
Tor1p, Sch9p and Ras2p respectively. These points altogether,
render chronological aging machinery of yeast as a promising
candidate for gaining insight about aging and age-related diseases
Recently, research has been conducted to comprehend
connectivity between longevity and age-related diseases along
with the determination of genes regulating life span, using systems
biology approaches [11–18]. Almost all of the stated studies benefit
from published protein-protein interaction (PPI) data to construct
a biological network, which is then topologically analyzed. Studies
investigating aging and age-related diseases in humans employ
PLoS ONE | www.plosone.org1 December 2011 | Volume 6 | Issue 12 | e29284
different topological techniques, such as shortest path length [7,14]
and connectivity [7,11,12,14,16,18] analyses on the reconstructed
PPI networks. Also, the integration of intracellular PPI data with
extracellular ones is another approach in network reconstruction
employed by human aging studies . The networks of
individual signaling pathways affecting aging, such as TOR
pathway  and glucose repression pathway , are important
examples of network based approaches in elucidating aging
process. In S. cerevisiae, the two aging processes encountered have
also been subjected to network-based analysis. The application of
shortest path length analysis on a longevity network constructed
with PPI data of proteins related to replicative aging process ,
and topological analysis of a hybrid aging network, reconstructed
by integrating both replicative and chronological aging processes,
 gave information about novel genes and processes which
impact both types of aging in yeast. Moreover, examples of the
‘‘bottom-up’’ systems biology which involve the construction of an
in silico model with genes, proteins and processes as parameters
have also been encountered while investigating aging in yeast 
and in higher eukaryotes .
In the current study, Chronological Aging Network of S.
cerevisiae is reconstructed using Selective Permissibility Algorithm
(SPA) which integrates Gene Ontology (GO) annotation terms
with protein-protein interaction (PPI) data, in an automated
manner . False positives naturally occurring in PPI data and
insignificant PPI’s are eliminated from the reconstructed network
by statistical methods based on betweenness centrality values, and
the tuned network is then clustered and subjected to linear path
analysis. Via linear path analysis, routes starting with proteins
previously demonstrated to regulate life span such as Tor1p
(homologous to mammalian mTOR), Sch9p (homologous to
mammalian Akt/PKB) and Ras2p (homologous to mammalian
Ras proto-oncogenes) together with 3 other proteins (Gpa2p,
Pga3p and Ptk2p) and ending at Sir2p and Gts1p are investigated.
Simultaneous analysis of the linear path spectra of these input-
output pairs enable one to unravel intermediate players of the
signaling events that lead to chronological aging. Step-specific key
protein determination is the chosen method in the current study
for the mentioned in depth analysis, yielding a denser final
network of 92 nodes for the 6 input and 2 output proteins. This
dense ‘‘heart’’ network depicts the routes highly participating to
the information flow in the network by identifying fundamental
proteins for the proceeding of the signal transduction for studied
input-output pairs. Indeed, four proteins of this heart network,
Tcb3p, Sna3p, Pst2p and YGR130Cp, which have not been
reported to affect chronological aging and also have unknown GO
process terms, are demonstrated to be involved in life span
Reconstruction and dissection of the reconstructed network as
well as its topological analysis, helps us unravel and enlighten
the inner dynamics of chronological aging mechanism of
Saccharomyces cerevisiae. Only the members of the nutrient sensing
pathways (Tor1p, Gpa2p, Ras2p, and Sch9p) with some other
input proteins such as Pga3p, which is proved to regulate the life
span, and Ptk2p, which is involved in cellular ion homeostasis,
are investigated in the current study. Further analyses of linear
paths starting with other proteins taking part in different
signaling pathways will provide data to illuminate the possible
machineries by which the mentioned signaling pathways affect
the chronological life span as well as to decipher the important
proteins responsible for these effects. The proposed framework
can effectively be used as a tool to give insight about other
biological networks, regardless of the species of which they
Network Reconstruction and Reduction
In the present study, the ‘‘chronological aging network’’ (CAN)
in Saccharomyces cerevisiae is reconstructed via integration of protein–
protein interactions with Gene Ontology terms. To achieve this
goal, all proteins which share the ‘‘chronological aging’’ term
under biological process were selected as the core constituents of
the network to be reconstructed (Table S1). The network was then
expanded as described in the Methods section (Figure S1). The 18
core proteins led to an undirected graph composed of 2359 nodes
and 12314 edges as the final network (Table S2). The network
diameter and the mean path length are found to be 9 and 3.37
respectively. These distance measures are orders of magnitude
significantly smaller than the number of proteins, meaning that
despite the large size of the network, any two nodes in the network
can be connected by relatively short paths along existing links,
emphasizing the small world architecture of the reconstructed
network. Moreover, as in many biological networks, the
distribution of the nodes in the current reconstructed network
has a scale-free nature following nearly a power law model, P(k)
<k2c, having c = 1.49 with R2= 0.88 (Figure 1a). Other
topological parameters such as average degree, critical path
length, diameter and average clustering coefficient are also in close
vicinity with the values reported for other protein-protein
interaction networks in literature (Table 1). Further analysis of
average clustering coefficient values versus degree reveals that the
current system is actually a hierarchical network, resembling the
Barabasi-Albert model discussed in elsewhere [23,24]. The
distribution of average clustering coefficients with respect to
degree follows a power law model with C(k) <k2w, having w
<0.75 with R2= 0.69 (Figure 1a). These topological parameters,
c and w, imply that the network reconstructed is made of
numerous small, highly integrated modules, preserving both the
high degree of clustering and the scale-free property. In fact, when
the same topological analysis is carried out with the network
comprised of the whole PPI data in BioGrid (BioGrid network), it
is observed that both c and goodness of fit value, R2, decreases.
Moreover, the BioGrid network does not follow a hierarchical
nature, since w is found to be 0.615 with a considerably small R2
value of 0.46 (Figure S2).
It is well known that the aging process is a multifactorial
phenomenon; i.e. various parameters affect the lifespan of yeast as
well as of higher eukaryotes via different branches in cell
[12,25,18]. Hence it is not a surprising result that the size of the
reconstructed network is fairly large, integrating various mecha-
nisms attached to the process of chronological aging of yeast. The
current reconstruction of the network was carried out in an
automated fashion to eliminate possible biases which may arise
from manual curation. But unfortunately, that also renders the
network prone to the incorporation of incomplete and/or
erroneous data (e.g. false positives) which originate unavoidably
when high throughput experiments are carried out. To counter-
balance this side-effect of the method, a hypothesis testing based
on the betweenness centrality (BC) values of individual nodes was
carried out as described in the Methods section. Briefly, it was
assumed that if the BC value of a node does not change
significantly for both the reconstructed and randomized networks
(the average value of 100 networks in the randomized case), the
node was considered to be included in the network randomly and
therefore discarded, since its contribution to the putative
information flow in the random networks is the same as its
contribution to the real information flow in the original network.
Three different significance levels were employed for this
Chronological Aging Network of Yeast
PLoS ONE | www.plosone.org2December 2011 | Volume 6 | Issue 12 | e29284
Figure 1. Connectivity and average clustering coefficient distributions for a) CAN and b) tCAN.
Table 1. Topological properties of protein-protein interaction networks.
Model node # edge #
, ,k. .
, ,CC. .
CAN (S. cerevisiae) 2359 1231410.43.49 0.157 this study
tCAN (S. cerevisiae) 1736 84589.7 3.49 0.167 this study
Signaling (S. cerevisiae)
Ca2+signaling (S. cerevisiae)
1363 3649 5.4 6.89- 
182610078 11.043.5611 0.150
DIP (M. musculus)329 286-3.69 0.155
DIP (H. sapiens)10651369- 6.8 21 0.206
Wnt signaling (H. sapiens) 3489 10092- 4.415- 
EGFR signaling (H. sapiens)329 179510.91 4.711- 
Hedgehog Signaling (D. melanogaster) 568975- 4.8 14-
Wnt/b-catenin Signaling (D. melanogaster)656 1253- 4.813- 
‘‘,k.’’ denotes the average connectivity, ‘‘CPL’’ stands for critical path length, ‘‘d’’ is diameter and ‘‘,CC.’’ is the average clustering coefficient value.
Chronological Aging Network of Yeast
PLoS ONE | www.plosone.org3 December 2011 | Volume 6 | Issue 12 | e29284
hypothesis testing: 0.1, 0.01 and 0.001. Ultimately the significance
level value (a) of 0.001 was selected to be used as the threshold
since the resulting network with this significance value yielded the
highest goodness of fit score, R2= 0.86, compared to the other
two significance level values for the power law model (Figure S3).
Following this hypothesis testing, approximately 26% of the
original nodes were considered as statistically insignificant and
they were filtered out from the originally reconstructed network
along with their interactions (31% of the original interactions).
Ultimately, the ‘‘tuned’’ network obtained (tCAN) had 1734 nodes
and 8458 interactions (Table S3). Moreover, when tCAN was
topologically analyzed, it was observed that many of the statistical
parameters remained nearly unchanged, compared to those of
CAN (Figure 1b, Table 1).
Hubs and Clusters of CAN and tCAN
The first 20 of the highly connected nodes,
referred to as the ‘‘hubs’’ of the network, are the same proteins in
both CAN and tCAN, although their ranking differed slightly for
the two networks (Figure 2). 13 of these hubs are also among the
first 20 hubs of BioGrid network; however their specificity in CAN
differs considerably compared to that in BioGrid network. The
individual deletion of the 10 common hubs results in up to 6 and 5
fold increase in the number of connected components in CAN and
in tCAN respectively, compared to BioGrid network, except the
deletion of TPK1 (1.33 fold change in BioGrid Network compared
to CAN and tCAN), RPT5 (connected component number is the
same for all three networks) and RPN11 (2.5 fold change in CAN
while created connected component number does not change in
tCAN, compared to BioGrid Network) (Table S4). This result
demonstrates that in CAN, these hub proteins are more important
in terms of network stability and robustness: CAN is more prone to
‘‘attacks’’ targeting these hubs than the BioGrid network.
When these 20 hubs are analyzed thoroughly, it is observed that
they are indeed strongly related to the chronological aging and
quiescence processes in yeast. A very recent genome-wide study
demonstrated that the deletion mutants of the genes encoding 6 of
these 20 hub proteins had either shorter (WHI3, TPK1, RVS167) or
longer (PHO85, BRE5, SSA1) chronological life spans compared to
that of wild type strain . Although the effect of the essential
gene encoding the hub protein with the highest degree, RPN11, on
chronological aging of S. cerevisiae has not yet been investigated,
Tonoki and colleagues demonstrated that the loss of function of
RPN11 resulted in a shorter life span for D. melanogaster . Harris
et al. proved that reduced levels of Hsp82p activity in S. cerevisiae
resulted in a longer chronological life span, by increasing the stress
resistance of the cells .
The remaining 12 hub proteins have not been subject to
experimentation yet to determine their effects on the chronological
life span of yeast. However, when their roles in the cellular
machinery are concerned it is seen that they take part in crucial
cellular processes closely linked to aging such as maintenance of
genomic stability, actin dynamics, protein degradation and
regulation of pH. For instance, Cdc28p was required to generate
post senescence survivors at a normal rate in telomerase-negative
S. cerevisiae cells  together with a role in maintenance of
genomic stability . The dysfunction of telomeres induces
senescence  and has also been hardwired with chronological
aging in yeast [32,33]. Rpt5p, another hub protein of the network,
was identified to be among the proteins which affect the telomere
length in S. cerevisiae . The 3rdhub protein of CAN (and 4thof
tCAN), Dsn1p, is a member of the MIND kinetochore and is
responsible for the accurate segregation of chromosomes ;
therefore altered Dsn1p activity results in the disruption of
genomic stability, which in turn may trigger chronological aging
. The mRNA-binding protein Yra1p, another hub of the
networks under study, has a role in the DNA damage response of
the yeast cells via nucleotide excision repair (NER) system. This
repair system was proved to provide protection against both
cancer and aging [36,37]; hence Yra1p may affect the chrono-
Figure 2. The first 20 hub nodes in the two networks with their corresponding connectivity values.
Chronological Aging Network of Yeast
PLoS ONE | www.plosone.org4 December 2011 | Volume 6 | Issue 12 | e29284
logical aging mechanism in S. cerevisiae by altering the genomic
The three hub proteins Ptk2p, Yck1p and Yck2p are found to
be involved in the phosphorylation of Pma1p, a cell surface
protein which is the major regulator of cytoplasmic pH and
plasma membrane potential [38,39]. The limiting effect of
extracellular acidity on the chronological aging of S. cerevisiae is
well established , therefore these 3 hub proteins may affect the
response of the cell to acid toxicity via tampering Pma1p activity.
Sla1p, protein required for the assembly of the cortical actin
cytoskeleton, is another hub protein of the network. It was
suggested that the actin cytoskeleton interacts with the mitochon-
dria, and this interaction plays a significant role in programmed
cell death and aging . SSB1 encodes Ssb1p, a member of the
Hsp70 family together with another hub protein, Ssa1p, and is
considered to take part in the proper folding of newly synthesized
proteins as well as in the glucose sensing pathway in S. cerevisiae
[42,43]. Both protein homeostasis  and glucose signaling 
are suggested to be the key factors in aging process. In fact the two
hub proteins mentioned previously, Yck1p and Yck2p are also
taking parts in Snf3/Rgt2-mediated glucose signaling of yeast .
Protein homeostasis includes proper degradation as well as proper
folding machinery of proteins, since accumulation of misfolded or
damaged proteins leads to protein toxicity and aging. The hub
protein Srp1p was demonstrated to be a part of this degradation
machinery in S. cerevisiae through ubiquitin-proteasome system
, and may therefore be influential on chronological aging
mechanism via maintenance of protein homeostasis in the cell.
Ksp1p, a ser/thr protein kinase required for growth in nutrient
limited conditions, is another hub protein of the network. It has
recently been demonstrated that Ksp1p is amongst the Tor1p-
regulated phosphoproteome; much specifically it is one of the
rapamycin-sensitive phosphoproteins of yeast coordinated by the
Sch9p branch of the TOR signaling [48,49]. Moreover Ksp1p
regulates the translocation of Bcy1p, the regulatory subunit of the
cAMP-dependent PKA . Both TOR and cAMP/Protein
kinase A signaling pathways are reported to be responsible for life
extension during calorie restriction in yeast, hence it is not
surprising to detect Ksp1p as a hub protein in CAN and tCAN
. The last one of the 20 hub proteins of CAN and tCAN is
Orc1p, the largest subunit of the origin recognition complex of
yeast. Orc1p is the paralog of Sir3p , the silencing protein
required for spreading of silenced chromatin. Although the
sequences of ORC1 and SIR3 have diverged significantly, Orc1p
is also involved in transcriptional silencing, rendering itself as a
suitable hub protein of CAN and tCAN since gene silencing leads
to cellular quiescence .
The reconstructed networks (CAN and
tCAN) are of hierarchical nature, whereas the BioGrid network is
not, which implies a reigning modular topology within CAN and
tCAN. The modules (the highly connected protein subgroups) are
expected to give insight about underlying cellular machineries
leading to chronological aging process in yeast (Tables S5 and S6).
Actually, when modularity analysis is performed to the whole PPI
data in BioGrid (Table S7), it is observed that only 5% of the
proteins encountered in the modules of CAN and tCAN are
shared by the protein spectrum of the BioGrid network.
Furthermore, the enrichment analyses of the modules of the
BioGrid network (having scores greater than 3) result in parent
GO terms predominantly,complicating
differential information from the analysis (Table S8). However,
the enriched GO process terms of modules of CAN and tCAN
reflect the cellular reprogramming necessary for quiescence which
is the hallmark of chronological aging, validating the fact that the
reconstructed networks indeed encompass different aspects
connected to chronological aging of yeast through this modular
topology. The filtered enriched terms for the clusters of CAN and
tCAN are summarized in Table S9 and Table S10 respectively.
Many terms related to cell cycle (S phase, interphase etc.) but
especially to its mitotic (M) phase (DNA replication initiation,
cytoskeleton organization, chromosome segregation and organi-
zation, cytokinesis, etc.) stand out among the enriched categories
following the modular analysis on the complete set of proteins of
CAN and tCAN (see Tables S9 and S10 for p-values). It was
surprising to have GO terms related to mitotic cell cycle, since
chronologically aged yeast cells are in a quiescent state (G0
phase), and do not divide. However, it has recently been
reported that along with the increase in genomic instability, a
breakdown in mitotic asymmetry is also encountered in chrono-
logically aged yeast cells . This renders the mitotic division
mechanism of importance for not solely replicative aging process,
but for both types of aging in yeast, coherent with the findings of
this study. Apart from cell cycle terms, the enriched GO categories
for both networks may be gathered into three main cellular
processes which are required to maintain viability in the quiescent
state: i) reorganization of metabolism, ii) redox homeostasis and iii)
Reorganization of metabolism.
B6 (pyridoxine) biosynthesis, which are the most significantly
enriched terms of the two clusters (15 & 16) in both networks (p-
values,10E-5), are examples of the metabolic rearrangement
encountered in quiescent cells. SNZ1, the product of which is
required for pyridoxine biosynthesis, is identified to be expressed
after entry into quiescence , and then a role of vitamin B6 as a
cofactor in stationary-phase specific processes or as an antioxidant
has been suggested . Similarly, trehalose was proposed to
protect proteins against oxidative damage in quiescent cells 
and moreover, cells metabolize trehalose also for fuel upon exit
from the quiescent state . The GO process terms belonging to
glycolysis and gluconeogenesis along with respiration and
fermentation stood out in the enrichment results (p-values,2E-
2). Apart from the tight regulation of these processes in quiescent
cells, alternative energy production routes are well activated in
chronologically aging cells. The terms ‘‘fatty acid metabolism’’ and
‘‘triacylglycerol mobilization’’ are among the enriched categories
in this study, indicating that the networks reconstructed contained
triacylglycerol and/or fatty acid oxidation . The possibility
of deriving energy from fatty acid catabolism implies an important
role for peroxisome and mitochondria in the maintenance of a
quiescent cell. In fact many terms related to the mitochondrial as
well as peroxisomal functions appear in the enrichment results of
tCAN (and similarly of CAN), in accordance with the literature
[61,62]. The terms ‘‘cell wall organization and biogenesis’’, ‘‘cell
wall chitin metabolic process’’ and ‘‘ergosterol biosynthetic
process’’ give insight about another aspect of the metabolic
reorganization encountered in quiescence, the remodeling of the
cell wall. Actually, quiescent cells develop thickened cell walls 
and that cell wall organization in yeast cells is highly dependent on
PKA signaling pathway which may negatively regulate longevity
in yeast . Moreover, ergosterol, which is an essential lipid for
the membrane, is taking role in the response to oxidative stress and
is shown to be a part of the longevity network of yeast [65,17].
metabolism in quiescent cells is closely linked to the redox
homeostasis machinery. As mentioned above, the storage of
carbohydrates such as trehalose and glycogen together with
hydrolysis of lipid stores in quiescent cells probably leads to the
The trehalose and vitamin
Chronological Aging Network of Yeast
PLoS ONE | www.plosone.org5December 2011 | Volume 6 | Issue 12 | e29284
accumulation of free fatty acids. Peroxisomes come onto stage at
this point, and oxidize these fatty acids to acetyl-CoA, which is
subsequently oxidized in mitochondria to generate ATP in
quiescent cells . The terms involving the electron transport
chain along with ‘‘TCA cycle’’ and ‘‘Acetyl-CoA catabolism’’
terms in the enrichment results also support this hypothesis.
However, this reorganization in the cell generates considerable
quantities of mitochondrial ROS and oxidative stress, pointed out
by the enriched ‘‘superoxide metabolic process’’, ‘‘oxygen and
reactive oxygen species metabolic process’’, and ‘‘age-dependent
response to reactive oxygen species during chronological cell
aging’’ terms in GO enrichment analysis. In fact, ROS
homeostasis must be tightly regulated in quiescent cells, since it
is debated that ROS play a dual role in determining the fate of a
cell : ROS are speculated to increase the chronological
lifespan of the whole culture, below a certain threshold, probably
by triggering autophagy [67,68] and/or apoptosis [69,6] but they
may very well decrease it, via exactly the same mechanisms, likely
when their level exceeds the threshold [70–72]. Coherent with the
mentioned points above, the apoptosis and autophagy related
terms are also present in the enrichment results of both networks.
The most significant GO process term for
both networks is ‘‘ubiquitin-dependent protein catabolic process’’,
having the highest corrected p-value. Ubiquitin/proteasome
system is one of the protein turnover mechanisms in yeast and
was demonstrated to contribute to chronological aging in yeast
[19,73]. Whether the ubiquitin dependent protein degradation is a
mechanism necessary for oxidative stress resistance of yeast cells is
a highly debated issue. Recently, however, it was demonstrated
that ubiquitin, as well as proteasome, is necessary for oxidative
stress resistance . This information relates the above
mentioned ROS homeostasis and protein turnover machineries
in quiescent cells: proper degradation of oxidized proteins
maintaining viability in quiescent state. Moreover, autophagy
and ubiquitin/proteasome system have been proved to be cross-
linked , hence this mechanism may also be effective in the
amino acid recycling of stationary phase yeast cells along with
autophagy, which has been proved to affect the chronological life
span of S. cerevisiae. ‘‘Protein amino acid deacetylation’’ is another
term that comes across the results of this study concerning protein
modification, implying its importance for quiescent cells. Indeed,
studies adopting spermidine and resveratrol, both of which are
known to extend the chronological life span of yeast by inducing
autophagy, have hinted that these agents activate the autophagic
cascade in the cell via deacetylation reactions [76,77]. The
majority of the remaining enriched GO terms are related to
intracellular (especially vesicle-mediated) transport involving
endosomes, membrane invagination, endocytosis, exocytosis, and
actin cytoskeleton organization. These processes are also closely
linked to the autophagy machinery which promotes the survival of
quiescent cells [78–80].
Considering the similarities in the topological properties as well
as in the enrichment results of CAN and tCAN, it was decided to
adopt tCAN in the forthcoming analyses.
Linear Path and Key Protein Analysis
Linear Path Analysis.
between particular proteins in tCAN it is aimed to gain insight
about the information flow of the intracellular machineries leading
to chronological aging in yeast. Six input (Pga3p, Tor1p, Gpa2p,
Ptk2p, Ras2p and Sch9p) and two output (Sir2p and Gts1p)
proteins were selected for linear path analysis. The role of Tor1p
and Ras2p, the two membrane proteins, is well established in
By investigating the linear paths
chronological aging [4,81]. Pga3p is a putative cytochrome b5
reductase on the plasma membrane, and one of the core proteins
of the reconstructed network . GPA2 encodes a subunit of the
heterotrimeric G protein that interacts with the receptor Gpr1p,
which has a signaling role in response to glucose and its deletion
mutant has been demonstrated to have a longer chronological life
span compared to wild type strain . Ptk2p regulates the ion
transport across plasma membrane and enhances spermidine
uptake . Moreover it is one of the first 20 hub proteins of the
network. The involvement of Sch9p in the regulation of life span
in yeast is well documented , but it is not localized to the
plasma membrane. The two end proteins, Sir2p and Gts1p, are
among the core proteins of the network, and both have a
transcription regulator activity. By decomposing tCAN into linear
paths, depicting the signaling routes between mentioned inputs
and outputs, the intermediate proteins taking part in the
embedded information flow can be unraveled.
Using NetSearch algorithm , different path lengths starting
from 3 to 8 were tested to reach a transcriptional regulator from
an input protein and it was seen that as the linear path length
increases, the number of paths to be analyzed increased in an
exponential fashion (unpublished data). Furthermore, when the
number of steps is greater than 7, the data analysis becomes
computationally tedious, since the final number of linear paths for
a given couple of proteins is in the order of 106for this number of
steps. So, the path length is chosen to be 6 for this study, with an
acceptable core and network protein coverage value of 65% and
55% respectively. Alternatively, the same analysis was carried out
for Tor1p-Sir2p pair in BioGrid network and it was observed that
the number of linear paths increased almost 35 fold (729960 linear
paths) while the number of proteins involved in these linear paths
increased about 5 fold (3479 proteins) when compared to tCAN.
Actually, the total number of proteins taking part in the interaction
data of BioGrid release 3.1.73 is 5626, and 61% of these proteins
appeared in the results of the linear path analysis for this branch
only. Moreover, the protein set obtained by the analysis performed
only on Tor1p-Sir2p pair in BioGrid network has a strong
similarity to complete tCAN, having 1200 proteins in common.
Therefore, it is not surprising to have an increased core protein
coverage and percentage of common network proteins, 86% and
69% respectively, for a path length of 6 in BioGrid network for
Tor1p-Sir2p branch (unpublished data).
Linear path analysis supplies analytical measures to distinguish
the relative activity of the signaling routes under study, aside from
pointing out candidate proteins belonging to a signaling cascade.
As a general result, it can be deduced that the Sir2p branch of the
pathway is more active compared to Gts1p branch, since an
approximately 2-fold increase is observed in the number of linear
paths for all input proteins (Table 2). When the input proteins are
ranked according to the abundance of linear paths, Ptk2p and
Tor1p are the most active ones, followed by Gpa2p. Sch9p and
Ras2p, clustered as the third active input proteins, pursuit Gpa2p
and finally the least active input protein is Pga3p for both outputs
with less than 100 linear paths. The activity of the input proteins
may hint to the robustness and therefore to the complexity of the
cellular machineries in which these proteins take part. For
instance, the higher path numbers belonging to Ptk2p and Tor1p
indicate that the signal flowing through these nodes has many
alternative routes, implying that these proteins are probably
involved in regulating multitudinous complex processes which
have an impact on chronological aging. In case a perturbation
occurs in one route, there are several alternative routes that may
maintain an intact signal transduction if the start proteins are
Ptk2p and/or Tor1p . According to this point of view, the
Chronological Aging Network of Yeast
PLoS ONE | www.plosone.org6December 2011 | Volume 6 | Issue 12 | e29284
signaling cascades involving Pga3p are expected to be very
sensitive to any perturbation and should affect the chronological
aging in yeast in a relatively direct manner. Indeed, Pga3p is the
only membrane protein which is also one of the core proteins of
tCAN, and the relatively high percentage of core protein coverage
of linear paths starting with Pga3p despite the extremely low total
protein coverage implies a relatively direct involvement of this
protein in aging . Moreover, all of the linear paths of the
Pga3p branch directed to both Sir2p and Gts1p, contain at least
one protein which is encoded by a gene whose deletion is lethal to
the cell (SGD phenotype data, release 11328), emphasizing the
sensitivity of the transduction routes towards perturbations .
When the proteins involved in linear path spectra are
considered, it was noticed that approximately 84% of the proteins
involved in Sch9p branch are common with Tor1p branch for
Sir2p and this percentage mounts to 95% for the output protein
Gts1p. Also, Tor1p branch surpasses Sch9p branch (for all output
proteins) in both the unique protein number and linear path
number. Hence linear path analysis points to the Sch9p branch to
be a sub-cluster of Tor1p branch, and the fact that Sch9p is a
substrate of Tor1p supports this finding . The other input
protein which has a similar activity (in terms of linear path
abundance) with Sch9p is Ras2p. Ras2p and Gpa2p are involved
in the transcriptional response to glucose and are members of G-
proteins [90,91]. However, Ras2p branch has a higher unique
protein number compared to that of Gpa2p despite a considerably
lower linear path number. This fact, when combined with a
common protein percentage being lower than 70% (for both Sir2p
and Gts1p) between the two branches, indicates that Ras2p
branch cannot be considered as a sub-cluster of Gpa2p branch
(Table 2). In fact, although both Gpa2p and Ras2p were
demonstrated to function similarly in the cell, e.g. to induce
cAMP signaling and to mediate the transcriptional response to
glucose in yeast , they act in redundant pathways, rather than
in sequential steps in the same pathway [93,94].
Reconstruction ofthe heart
quantitative analysis, the linear paths of each input-output
protein pair were qualitatively investigated to have deeper
information about the participations of individual proteins in the
information flow. A classical approach for determining the
important proteins for an input-output pair is the global
investigation of their participation percentage values [95,96].
Participation percentage value of a protein is the percent ratio of
the number of linear paths, in which the mentioned protein is
involved, over the total number of linear paths, disregarding the
step at which it is contributing to the information flow. Although
the important proteins (ranging from 3 to 12 for all pairs) are
determined via this method for an input-output pair successfully,
they are not always interacting with each other since they do not
necessarily emerge from successive steps of a path. But the
successive structure of a signaling network, which is crucial in the
information flow of biological networks, is enclosed in the
currently developed decomposition method (step-specific key
protein determination), yielding a more complete spectrum for
important proteins, compared to the global percentage analysis.
By this decomposition method, four successive groups of most
active proteins having a role in the information flow were
extracted, yielding a key protein subset of approximately 20–30
proteins for each input-output pair. The assembly of these subsets
resulted in a smaller but denser ‘‘heart’’ network having 92 nodes
and 477 interactions, depicting the most frequent information flow
routes for each input-output pair (more than 50% of the linear
paths were covered for each pair in the final ‘‘heart’’ network),
generated with Cerebral plugin of Cytoscape  (Figure 3). The
same decomposition method was also applied to the linear path
spectrum of Tor1p-Sir2p pair in BioGrid network, and 107 step-
specific key proteins emerged for this pair solely, 74 of them absent
in the current ‘‘heart’’ network. Moreover, these key proteins did
not include almost one fourth of the key proteins determined for
Tor1p-Sir2p pair in tCAN (Table S11). Although the number of
step-specific key proteins increases approximately five fold in
BioGrid network for Tor1p-Sir2p pair compared to tCAN,
analysis on BioGrid network fails to detect all of the proteins
determined in tCAN. For example, proteins such as Sac6p, Slt2p
and Pma1p, which are involved in chronological aging [79,98,99]
process of S. cerevisiae and fission yeast, are not captured in the
results of BioGrid Network, but of tCAN. This result implies that
the reconstructed network provides a smaller yet more distinctive
subset of proteins for chronological aging in S. cerevisiae.
The ‘‘heart’’ network (comprised of the key proteins) also
includes 19 of the 20 hub proteins given in Figure 2, except Yra1p.
This result is actually expected, since hub proteins, having higher
connectivity values compared to other proteins, are indeed proved
to be active in the information flow in a network . Apart from
the 19 hub proteins, several key proteins of the heart network have
been reported either to be involved in regulation of chronological
life span or to be associated with quiescence in yeast. BCK1,
FMP48, SNF1, TOR1, CKA1 and RIM15 are among the signal
transduction genes whose expression values have been demon-
strated to be significantly higher in quiescent cells . Similarly,
the deletion mutants of BCK1, CKA2, UFD2, DFM1, SSD1 and
OSH6 have recently been reported to be among the outgrowing
Table 2. The quantitative results of the linear path analysis.
To Sir2p To Gts1p
# of paths# of proteins# of UP’s* CPC** (%)OPC*** (%)# of paths# of proteins# of UP’s*CPC** (%) OPC*** (%)
Ptk2p33117 75015 55.6 43.316580684 106 44.439.4
Tor1p21109689116 44.4 39.7 10200 65185 38.937.5
Gpa2p934451028 33.329.44443 485 3327.827.9
Ras2p559158183 38.933.5 2500 4465738.925.7
Sch9p 6075 51835 38.9 29.9 24003463 27.8 19.9
Pga3p73647 22.23.7 4046511.1 2.7
*UP’s: Unique proteins, designate proteins solely present in the linear path spectra starting with the mentioned input protein.
**CPC: Core protein coverage, is the percent ratio of the core proteins present in the linear path spectra of the specific input-output pair over those of tCAN.
***OPC: Overall protein coverage, is the percent ratio of the proteins present in the linear path spectra of the specific input-output pair over those of tCAN.
Chronological Aging Network of Yeast
PLoS ONE | www.plosone.org7December 2011 | Volume 6 | Issue 12 | e29284
strains in a competition experiment on life span regulation .
Overexpression of ADH1, which encodes the alcohol dehydroge-
nase, another key protein in the heart network, resulted in the
extension of chronological life span . More importantly,
when GO process terms of these 92 key proteins are investigated
globally, it is noticed that they overlap with the GO enrichment
results attributed to modules of tCAN (Table S12), indicating that
the sub network they formed is successful in reflecting the
characteristics of tCAN. Moreover, the step-specific key proteins
comprising the ‘‘heart’’ network also include the ‘‘important’’
proteins obtained via global participation percentage method.
Determining New Proteins Regulating the Chronological
Life Span (CLS) of Yeast
The proteins in one of the clusters of the heart network (cluster
19 in Figure 3) are of particular interest for further analysis since
these four proteins, Tcb3p, Sna3p, Pst2p and YGR130Cp, do not
have a known biological process term. Therefore, these key
proteins are taken as good candidates to check the validity of the
assumption that the members of the heart network are actively
contributing to the chronological aging process in yeast, and
chronological life span determination experiments were carried
out as described in Materials and Methods section with deletion
mutants of genes encoding these proteins along with Dras2 strain.
The deletion of RAS2 is known to extend the chronological life
span ; hence this strain is used as a reference to compare the
extent of gene deletion effect on life span.
When the survival curves belonging to these strains are
observed, it is noticed that chronological life span was increased
for all mutants compared to wild type strain (Figure 4), although
the increase was not as pronounced as it is in the case of RAS2
deletion. This difference is in fact more clearly observed in terms
of mean (the day on which survival reaches 50%) and maximum
(the day on which the survival reaches 10%) life spans (Table 3).
Parallel to our findings, Pst2p and YGR130Cp have recently
been reported to be a part of a longevity network of yeast .
Pst2p is a flavodoxin-like protein which plays a role in the stress
response of yeast, and rapamycin treatment induces a serious
growth defect in the homozygous deletion mutant of PST2 .
As for YGR130C, whose expression is approximately doubled
following a rapamycin treatment in a study investigating nitrogen
assimilation in yeast , Pst2p is found to be involved in
endocytotic machinery of the cell and speculated to control protein
turnover [106,107]. The other membrane protein affecting CLS,
Sna3p, is a multivesicular body cargo protein  and is a part of
the endosomal network. Moreover, overexpression of SNA3 results
in a more stabilized Tat2p, the high-affinity tryptophan permease
which is normally degraded upon nutrient starvation or rapamycin
treatment . Finally, Tcb3p which is one of the three yeast
tricalbins, is another membrane protein involved in membrane
trafficking  and deletion of TCB3 renders cells more resistant
to rapamycin treatment .
In summary, all these 4 proteins share a role in the intracellular
transport, mainly endocytotic pathway, and are responsive to
rapamycin treatment in yeast cells. Observing the similar survival
profiles of the deletion strains of the genes encoding these proteins,
one may speculate that the four key proteins Tcb3p, Sna3p, Pst2p
and YGR130Cp, affect chronological life span by altering the
Figure 3. The ‘‘heart’’ network depicting the key proteins of the 12 branches. The node colors reflect the degree of the proteins in tCAN.
The GO enrichment results of the numbered clusters are given in Table S4. The circular and triangular nodes depict intracellular and nuclear proteins
respectively, diamond nodes are the membrane proteins and square nodes are the proteins with the GO compartment term ‘‘unknown’’.
Chronological Aging Network of Yeast
PLoS ONE | www.plosone.org8 December 2011 | Volume 6 | Issue 12 | e29284
endosomal network, since endocytosis and vacuolar protein sorting
processes are already known to be involved in the regulation of
chronological life span in yeast [78,80].
In this study, the chronological aging network of S. cerevisiae is
reconstructed in an automated manner, by integrating protein-
protein interaction data available in literature with gene ontology
terms. It is observed that the resulting network reflects qualities
similar with other biological networks such as a scale-free nature
and small world architecture, but the current network also
possesses a hierarchical nature, implying a regulatory organization
within itself. This result is particularly interesting because it stresses
out the presence of an information flow embedded in the network
from particular nodes to final ones, in a structured and organized
manner, as it is the case with signaling networks. In fact, many
signaling pathways have proven to alter the chronological life span
of yeast, and the current study integrates most of these pathways
into a large and hierarchical network.
Network refining is a necessary measure taken to counteract the
automated method adopted, to remove insignificant (to the current
context of chronological aging) as well as erroneous protein-
protein interaction data. In other words, the tuning enabled the
Figure 4. Survival curves of the deletion mutants and wild type strains. The percentages are the average values of 4 experiments (two
biological along with their two technical replicates) and error bars denote the standard deviation of the indicated sample.
Table 3. Percent increase and p–values of maximum and
mean life spans of deletion mutants compared to wild type
Dsna342 271.52E-05 2.84E-02
Dras258 82 1.11E-04 1.24E-05
Chronological Aging Network of Yeast
PLoS ONE | www.plosone.org9December 2011 | Volume 6 | Issue 12 | e29284
investigation of the chronological aging process of yeast with a
<30% smaller network, refined from erroneous data. The strong
overlapping of the results obtained from both topological and
modular analyses of the two networks, CAN and tCAN,
demonstrated that indeed tCAN is a more refined and to the
point version of the initial CAN. Thus betweenness centrality of
the nodes is an accurate measure to decide whether the particular
node is important when investigating signaling networks, where
the participation of nodes to the information flow is crucial.
Several gene ontology process terms involving cell cycle, protein
and ROS homeostasis together with reorganization of metabolism,
proven to be related to chronological aging in S. cerevisiae emerged
from enrichment analysis of the modules of CAN and tCAN,
fortifying the hypothesis that the reconstructed networks encom-
pass the multi-faceted mechanisms which impact chronological
aging in yeast. Protein ubiquitination/degradation and processes
pertinent to mitochondria and peroxisomes especially stand out
among the enrichment results, stressing out the strong interplay
between protein and ROS homeostasis in the cell when
chronological aging is concerned. Processes involved in autophagy
and apoptosis, especially endocytotic machinery, are also among
the key effectors of chronological aging according to the
The in-depth investigation of tCAN is achieved via linear path
analysis between selected input-output protein pairs, followed by
the identification of ‘‘step-specific key proteins’’ of the linear
spectra. The quantitative evaluation of the linear path analysis
provided preliminary results about the activity and robustness of
the branch starting with a specific input protein in the
chronological aging network. These characteristics give hints
about the complexity of the machinery by which the mentioned
input protein affects the life span; increased complexity may imply
that the input protein in question alter aging via a broad range of
cellular processes. The identification of step-specific key proteins
constitutes the qualitative aspect of the linear path analysis, aiming
the detection of proteins whose impact on aging is not obviously
In fact, a majority of the members of the ‘‘heart’’ network
(which is composed of step-specific key proteins) is actually proved
to be involved in life span alteration in studies conducted by
different groups, hinting that the computational approach adopted
is successful in capturing the global picture. Moreover, 4 genes
(TCB3, SNA3, PST2 and YGR130C) which have not been
previously reported to be related to the chronological life span
of S. cerevisiae, encode the key proteins that are not involved in any
known process in the cell. The experimental chronological life
span assays reveal that the deletion of the genes resulted in an
increase in the life span compared to wild type, albeit the rise is not
as significant as in the case of the deletion of RAS2. The fact that
the four proteins are all responsive to rapamycin treatment and
related to the endocytotic machinery of the cell implies that they
may affect the chronological life span by modifying the endosomal
network of yeast. Further experimental work, such as a whole
genome analysis may be conducted to comprehend more
thoroughly the underlying reorganization of the cellular machin-
ery of these deletion mutant strains leading to an increased life
Due to its multifactorial nature, aging remains one of the most
complicated and therefore intriguing phenomena of the cell to
investigate. The involvement of various signaling pathways is
responsible for its multi-faceted nature, as proved here by modular
analysis. The disruption of the collaboration between these
pathways is essential for aging and also increases the susceptibility
of an organism towards age-related diseases. Therefore, the
reconstructed network may not only shed light to chronological
aging process itself, but also to the fundamental bottleneck points
responsible for the degeneration of the cooperation between these
signaling branches, which determine the fate of a cell.
Materials and Methods
Network Reconstruction by Selective Permissibility
To reconstruct the signaling network of chronological aging in
yeast, the Selective Permissibility Algorithm (SPA) was adopted
. The inputs of the algorithm were the core proteins of the
network and the Annotation Collection Table which is used to
expand the network from the mentioned core proteins. To
determine the core proteins, gene products which share the
‘‘chronological cell aging’’ GO biological process term were
extracted from the manually curated literature data of Saccharo-
myces Genome Database (SGD), released on 29.01.2011 (Table
S1). Next, the Annotation Collection table was created by pooling
the process, function and component GO annotations of the
determined core proteins only. As the third step, all physical
interactions of the core protein(s) were extracted from BioGRID
database  of protein and genetic interactions, release 3.1.73.
By integrating GO annotation terms with the interactome data of
yeast, the chronological aging network of S. cerevisiae is recon-
structed. Briefly, a candidate protein was included into the
network if all of the three GO annotations (component/function/
process) of the protein are present in the Annotation Collection
and if it physically interacts with the core proteins, as a first
neighbor. Proteins included via this procedure become the ‘‘new’’
input proteins and the algorithm expands the network in this cyclic
way until no new interacting proteins are added to the network
In order to prevent exclusion of proteins solely due to the lack of
available literature data, along with the GO annotations of the
core proteins, the ‘‘biological_process’’, ‘‘cellular_component’’ and
‘‘molecular_function’’ terms, namely the ‘‘unknown’’ terms, were
also included into the Annotation Collection Table. The
Annotation Collection created by this approach covers 178
annotations extracted out of a total of 4,208 annotations (about
4%) (Table S13).
The reconstructed network was statistically ‘‘tuned’’ using the
betweenness centrality (BC, the number of shortest paths passing
through a node (or an edge) given a shortest path algorithm) of a
node, which is a measure of a node’s importance to the network. It
is assumed that the nodes of the reconstructed network should
differ significantly in their participation to the information flow,
compared to their role in random networks. 100 random networks
were generated by shuffling the edges of the original graph, i.e. by
preserving the degree of each node. In other words, in the null
distribution composed of 100 random networks, all nodes had the
same ‘‘degree’’ value as they had in CAN, whereas their
interaction partners were randomly selected, opposed to CAN.
The original BC value distribution of CAN tended to follow a
skewed distribution, hence this trait observed in the distribution
was used as a basis in the control chart utilized to check the
suitability of the computed null distribution (Figure S4). The
randomization procedure and computation of BC values of all
nodes (for both the original and randomized networks) are
implemented in MATLAB 7.0 (MathWorks, Inc.,Natick, MA)
using the MatlabBGL package (written by David Gleich). For
randomized networks, average values along with the estimated
Chronological Aging Network of Yeast
PLoS ONE | www.plosone.org 10 December 2011 | Volume 6 | Issue 12 | e29284
variances of BC values corresponding to each node are computed
and a hypothesis testing is carried out with the following
The protein in question is included in the
network randomly; its contribution to the putative information
flow in the random networks is the same as its contribution to the
real information flow of the original network (BC value in CAN is
equal to the average BC value of 100 random networks)
question to the information flow in the original network differs
significantly from that in the random networks (BC value in CAN
is not equal to the average BC value of 100 random networks)
The contribution of the protein in
H1: BCoriginal=BCave, rand
This hypothesis testing is then carried out for all nodes using a
dependent, two-tailed t-test for paired samples with a confidence
level of 99.9%.
Determination of Network Topology
Topological properties of both the tuned (reduced) and initial
networks, such as degrees, betweenness centralities, diameter,
average shortest path length and clustering coefficients were
calculated using the algorithm implemented in MATLAB 7.0
(MathWorks, Inc.,Natick, MA) with the MatlabBGL package.
Cluster Identification and Functional Enrichment
Highly densely connected proteins of the reconstructed
networks were identified with MCODE  plugin of Cytoscape.
In MCODE, loops were included while scoring the networks and
the degree cutoff (the minimum degree necessary in order for a
node to be scored) value was set to 2. The set cutoff value (the
threshold score that determines the inclusion of a node to a cluster,
depending on the seed node’s score) was 0.2 for cluster expansion;
the fluff parameter (the threshold score that determines the
inclusion of the neighbors of a node to a cluster, depending on the
node’s neighborhood density) was turned ‘‘off’’ while the haircut
option was ‘‘on’’ (all singly-connected nodes were removed from
clusters). Finally, the K-Core value (the minimum degree of a
maximally inter-connected sub-cluster within a cluster) was set to
2, and the maximum depth (the distance from the seed node while
searching for cluster members) was set to 100. The overrepre-
sented categories within the clusters with an MCODE score
greater than 1 or with a node number greater than 3, were further
investigated with BINGO plugin of Cytoscape . The
enrichment was assessed with the hypergeometric test for the
cluster under study versus the whole annotation. The significance
level was chosen to be 0.05 and the false discovery rate was
controlled with Benjamini and Hochberg correction.
Identification of Linear Paths and Step-specific Key
To gain insight on the signal flow in the chronological aging
network of S. cerevisiae, the linear paths of length 6, between a
starting (usually a membrane) protein and a target protein (a
transcriptional regulator) were evaluated using NetSearch algo-
Figure 5. Schematic representation of the network reconstruction algorithm, SPA. The black filled circles represent core proteins, empty
circles are possible candidates for first neighbors, and grey filled circles show eliminated candidates via annotation collection table. Process continues
on with determination of new candidate proteins as next neighbors based on published PPI data and validation of these candidates by annotation
collection table until no new protein is added to the network.
Chronological Aging Network of Yeast
PLoS ONE | www.plosone.org11December 2011 | Volume 6 | Issue 12 | e29284
rithm . Step-specific key proteins of the linear paths of an
input-output protein pair were then determined via decomposing
the linear paths into their steps (Figure 6). Briefly, for the first one
of the successive 5 steps constituting a linear path of length 6, the
percent participation values (ratio of the number of linear paths in
which the mentioned protein is present over the total number of
linear paths) of involved proteins were calculated. Then a
cumulative histogram representing the frequency distribution of
these percentage values was created. mi, the histogram function,
counts the number of proteins that fall into each of the disjoint
categories (known as ‘‘bins’’) of participation frequencies. The
cumulative histogram Mi, counts the cumulative number of
proteins in all of the frequency bins up to the specified bin,
represented by eqn. 1:
where i is from 1 to k (the square root of the number of proteins
participating to the histogram, the bin number). The fold change
in the percent difference of the cumulative frequency for each bin
is denoted by eqn. 2:
from i=1 to k-1, with M0=0 and Mk=100. The ithfrequency bin
in which the largest percent fold change occurred in frequency was
chosen as the threshold bin, if the number of proteins having
frequency values larger than that threshold did not exceed 10% of
the total number of proteins under study. Otherwise, the
frequency bin having the next greater fold change is determined
as the threshold. Ultimately, proteins which have frequency values
larger than that of the threshold frequency bin are determined as
the key proteins specific to the first step. For the successive step,
these key proteins were used as ‘‘baits’’: linear paths involving
these proteins as the second proteins were selected and the
procedure was repeated to this reduced subset of linear paths to
yield ‘‘hits’’, the key proteins of the third step. Via this
decomposition analysis, 4 sets of key proteins (specific to the 1st,
2nd, 3rdand 4thsteps) were determined for each input-output pair
in the study. When these sets were pooled and united with the
input and output proteins, a relatively smaller subset of proteins
describing the whole linear spectra of the 12 branches was
obtained (Figure 3).
Figure 6. Schematic representation of key protein determination of the linear paths for an input-output pair. The putative input and
output proteins are A and K respectively. The procedure is depicted for a path length of 5. For details of the Frequency Histogram Analysis, see
Materials and Methods section.
Chronological Aging Network of Yeast
PLoS ONE | www.plosone.org 12 December 2011 | Volume 6 | Issue 12 | e29284
Chronological Life Span Assay
All of the experiments were carried out in liquid synthetic
dextrose complete (SDC) medium  with 2% glucose. The
deletion strains used in the study (Table 4) were derived from
BY4742. Overnight cultures grown in SDC, were diluted to an
OD600value of 0.1, and inoculated into 15 ml centrifuge tubes
with a 3 ml fresh SDC medium, maintaining a volume ratio of 1:5.
The cultures were then incubated at 30uC and 180 rpm. Serial
dilutions of the culture were spread onto four YPD plates (two
biological and two technical replicates) for each strain and time
point. Colony formation was monitored after two days. All
cultures were presumed to be 100% viable at day 3, with
subsequent colony forming unit (CFU) measurements normalized
to CFUs of day 3 to obtain survival data.
The methodology followed in the present work is summarized in
Table 4. Yeast strains used in the study.
BY4742Mat a; his3D1; leu2D0; lys2D0; ura3D0 EUROSCARF
DSNA3 Mat a; his3D1; leu2D0; lys2D0; ura3D0;
DTCB3 Mat a; his3D1; leu2D0; lys2D0; ura3D0;
DYGR130C Mat a; his3D1; leu2D0; lys2D0; ura3D0;
DPST2Mat a; his3D1; leu2D0; lys2D0; ura3D0;
Figure 7. Algorithm of the methodology followed in this work.
Chronological Aging Network of Yeast
PLoS ONE | www.plosone.org13 December 2011 | Volume 6 | Issue 12 | e29284
at each step during expansion.
Number of proteins included in the network
Connectivity and b) average clustering coefficient dis-
Topological analysis of BioGrid Network: a)
tCAN for a) a=0.1 and b) a=0.01.
Connectivity distributions of the resulting
computed. The chart is based on the skewness property of BC
value distribution of the original CAN. CL, UCL (CL + 3s) and
LCL (CL-3s) correspond to center line, upper and lower control
Control chart for the 100 random networks
Core proteins of the network.
Protein-protein interactions of CAN.
Protein-protein interactions of tCAN.
first 20 hubs of CAN, in CAN and BioGrid Network.
Connected Component (CC) Analysis for the
MCODE results of CAN.
MCODE results of tCAN.
MCODE results of BioGrid Network.
BINGO results for the first 6 clusters of
BINGO results for clusters of CAN.
BINGO results for clusters of tCAN.
of Tor1p-Sir2p pair for tCAN and BioGrid Network.
Comparison of the step-specific key proteins
GO process terms of the clusters of the
Annotation Collection Table of the core
We thank Prof. Stephen G. Oliver and Dr. Pınar Pir for providing the yeast
strains. We thank Assist. Prof. Tunahan C ¸akır for implementing the
calculation of topological properties and network randomization processes
in MATLAB. We also thank Dr. Serpil Eraslan for her help with the
verification of the absence of the deleted genes for mutant strains.
Conceived and designed the experiments: EBY KOU. Performed the
experiments: EBY. Analyzed the data: EBY. Contributed reagents/
materials/analysis tools: EBY KOU. Wrote the paper: EBY KOU.
1. Fontana L, Partridge L, Longo VD (2010) Extending healthy life span--from
yeast to humans. Science (New York, N.Y.) 328: 321–6. Available: http://
www.ncbi.nlm.nih.gov/pubmed/20395504. Accessed 19 Jul 2011.
2. Kaeberlein M (2010) Lessons on longevity from budding yeast. Nature 464:
513–9. Available: http://www.ncbi.nlm.nih.gov/pubmed/20336133. Accessed
10 Jun 2011.
3. Mortimer RK, Johnston JR (1959) Life Span of Individual Yeast Cells. Nature
183: 1751–1752. Available: http://dx.doi.org/10.1038/1831751a0. Accessed
19 Jul 2011.
4. Fabrizio P, Longo VD (2003) The chronological life span of Saccharomyces
cerevisiae. Aging cell 2: 73–81. Available: http://www.ncbi.nlm.nih.gov/
5. Fabrizio P, Longo VD (2008) Chronological aging-induced apoptosis in yeast.
Accessed 20 Jun 2011.
6. Herker E, Jungwirth H, Lehmann K a, Maldener C, Fro ¨hlich K-U, et al. (2004)
Chronological aging leads to apoptosis in yeast. The Journal of cell biology 164:
501–7. Available: http://www.pubmedcentral.nih.gov/articlerender.fcgi?
artid=2171996&tool=pmcentrez&rendertype=abstract. Accessed 23 Jun 2011.
7. Bell R, Hubbard A, Chettier R, Chen D, Miller JP, et al. (2009) A human protein
interaction network shows conservation of aging processes between human and
invertebrate species. PLoS genetics 5: e1000414. Available: http://www.
rendertype=abstract. Accessed 19 Jul 2011.
8. Evans DS, Kapahi P, Hsueh W-C, Kockel L (2011) TOR signaling never gets old:
aging, longevity and TORC1 activity. Ageing research reviews 10: 225–37.
&tool=pmcentrez&rendertype=abstract. Accessed 19 Jul 2011.
9. Hlavata ´ L, Nachin L, Jezek P, Nystro ¨m T (2008) Elevated Ras/protein kinase
A activity in Saccharomyces cerevisiae reduces proliferation rate and lifespan
by two different reactive oxygen species-dependent routes. Aging cell 7:
148–57. Available: http://www.ncbi.nlm.nih.gov/pubmed/18081742.
Accessed 19 Jul 2011.
10. Nie K, Yu J-C, Fu Y, Cheng H-Y, Chen F-Y, et al. (2009) Age-related decrease
in constructive activation of Akt/PKB in SAMP10 hippocampus. Biochemical
and biophysical research communications 378: 103–7. Available: http://www.
ncbi.nlm.nih.gov/pubmed/19013131. Accessed 19 Jul 2011.
11. Ahmed SS, Ahameethunisa AR, Santosh W, Chakravarthy S, Kumar S (2011)
Systems biological approach on neurological disorders: a novel molecular
connectivity to aging and psychiatric diseases. BMC systems biology 5: 6.
&tool=pmcentrez&rendertype=abstract. Accessed 19 Jul 2011.
12. Budovsky A, Abramovich A, Cohen R, Chalifa-Caspi V, Fraifeld V (2007)
Longevity network: construction and implications. Mechanisms of ageing and
development 128: 117–24. Available: http://www.ncbi.nlm.nih.gov/pubmed/
17116322. Accessed 19 Jul 2011.
13. Chautard E, Thierry-Mieg N, Ricard-Blum S (2010) Interaction networks as a
tool to investigate the mechanisms of aging. Biogerontology 11: 463–73.
Available: http://www.ncbi.nlm.nih.gov/pubmed/20213321. Accessed 16 Jun
14. Dezso Z, Nikolsky Y, Nikolskaya T, Miller J, Cherba D, et al. (2009) Identifying
disease-specific genes based on their topological significance in protein networks.
BMC systems biology 3: 36. Available: http://www.pubmedcentral.nih.gov/
19 Jul 2011.
15. Lorenz DR, Cantor CR, Collins JJ (2009) A network biology approach to aging
in yeast. Proceedings of the National Academy of Sciences of the United States
of America 106: 1145–50. Available: http://www.pubmedcentral.nih.gov/
16. Simko ´ GI, Gyurko ´ D, Veres DV, Na ´na ´si T, Csermely P (2009) Network
strategies to understand the aging process and help age-related drug design.
Genome medicine 1: 90. Available: http://www.pubmedcentral.nih.gov/
17. Tan Y, Bush JM, Liu W, Tang F (2009) Identification of longevity genes with
systems biology approaches. Advances and Applications in Bioinformatics and
Chemistry 2: 49–56.
18. Wolfson M, Budovsky A, Tacutu R, Fraifeld V (2009) The signaling hubs at the
crossroad of longevity and age-related disease networks. The international
Chronological Aging Network of Yeast
PLoS ONE | www.plosone.org 14December 2011 | Volume 6 | Issue 12 | e29284
journal of biochemistry & cell biology 41: 516–20. Available: http://www.ncbi. Download full-text
nlm.nih.gov/pubmed/18793745. Accessed 19 Jul 2011.
19. Managbanag JR, Witten TM, Bonchev D, Fox LA, Tsuchiya M, et al. (2008)
Shortest-path network analysis is a useful approach toward identifying genetic
determinants of longevity. PloS one 3: e3802. Available: http://www.
rendertype=abstract. Accessed 19 Jul 2011.
20. Barea F, Bonatto D (2009) Aging defined by a chronologic-replicative protein
network in Saccharomyces cerevisiae: an interactome analysis. Mechanisms of
ageing and development 130: 444–60. Available: http://www.ncbi.nlm.nih.
gov/pubmed/19433103. Accessed 19 Jul 2011.
21. Kriete A, Bosl WJ, Booker G (2010) Rule-based cell systems model of aging using
feedback loop motifs mediated by stress responses. PLoS computational biology
6: e1000820. Available: http://www.pubmedcentral.nih.gov/articlerender.
fcgi?artid=2887462&tool=pmcentrez&rendertype=abstract. Accessed 17 Jun 2011.
22. Arga KY, Onsan ZI, Kirdar B, Ulgen KO, Nielsen J (2007) Understanding
signaling in yeast: Insights from network analysis. Biotechnology and
bioengineering 97: 1246–1258. Available: http://onlinelibrary.wiley.com/
doi/10.1002/bit.21317/abstract. Accessed 19 Jul 2011.
23. Albert R, Baraba ´si AL (2002) Statistical mechanics of complex networks.
Reviews of modern physics 74: 47. Available: http://rmp.aps.org/abstract/
RMP/v74/i1/p47_1. Accessed 19 Jul 2011.
24. Ravasz E, Baraba ´si A-L (2003) Hierarchical organization in complex networks.
Physical Review E 67: 1–7. Available: http://link.aps.org/doi/10.1103/
PhysRevE.67.026112. Accessed 18 Jul 2011.
25. Magalha ˜esJPde,BudovskyA,LehmannG,CostaJ,LiY,etal.(2009)TheHuman
cell 8: 65–72. Available: http://www.pubmedcentral.nih.gov/articlerender.fcgi?
artid=2635494&tool=pmcentrez&rendertype=abstract. Accessed 19 Jul 2011.
26. Burtner CR, Murakami CJ, Olsen B, Kennedy BK, Kaeberlein M (2011) A
genomic analysis of chronological longevity factors in budding yeast. Cell Cycle
10: 1385–1396. Available: http://www.landesbioscience.com/journals/cc/
article/15464/. Accessed 19 Jul 2011.
27. Tonoki A, Kuranaga E, Tomioka T, Hamazaki J, Murata S, et al. (2009)
Genetic evidence linking age-dependent attenuation of the 26S proteasome with
the aging process. Molecular and cellular biology 29: 1095–106. Available:
pmcentrez&rendertype=abstract. Accessed 20 Jun 2011.
28. Harris N, MacLean M, Hatzianthis K, Panaretou B, Piper PW (2001) Increasing
Saccharomyces cerevisiae stress resistance, through the overactivation of the heat
shock response resulting from defects in the Hsp90 chaperone, does not extend
replicativelife spanbut can beassociated withslower chronological ageingofnon.
Molecular Genetics and Genomics 265: 258–263. Available: http://www.
Accessed 19 Jul 2011.
29. Grandin N, Charbonneau M (2003) Mitotic Cyclins Regulate Telomeric
Recombination in Telomerase-Deficient Yeast Cells. Molecular and Cellular
Biology 23: 9162–9177. doi:10.1128/MCB.23.24.9162.
30. Enserink JM, Hombauer H, Huang M-E, Kolodner RD (2009) Cdc28/Cdk1
positively and negatively affects genome stability in S. cerevisiae. The Journal of
cell biology 185: 423–37. Available: http://www.pubmedcentral.nih.gov/
Accessed 19 Jul 2011.
31. Ungar L, Yosef N, Sela Y, Sharan R, Ruppin E, et al. (2009) A genome-wide
screen for essential yeast genes that affect telomere length maintenance. Nucleic
acids research 37: 3840–9. Available: http://www.pubmedcentral.nih.gov/
Accessed 19 Jul 2011.
32. QiH,Chen Y,FuX,LinC-P, ZhengXFS,etal.(2008)TOR regulatescelldeath
induced by telomere dysfunction in budding yeast. PloS one 3: e3520. Available:
pmcentrez&rendertype=abstract. Accessed 19 Jul 2011.
metabolism and Ku70/80 function. The Journal of biological chemistry 283:
27514–24. Available: http://www.pubmedcentral.nih.gov/articlerender.fcgi?
artid=2562072&tool=pmcentrez&rendertype=abstract. Accessed 14 Jul 2011.
34. De Wulf P, McAinsh AD, Sorger PK (2003) Hierarchical assembly of the budding
yeast kinetochore from multiple subcomplexes. Genes & development 17:
2902–21. Available: http://www.pubmedcentral.nih.gov/articlerender.fcgi?
artid=289150&tool=pmcentrez&rendertype=abstract. Accessed 19 Jul 2011.
35. Charville GW, Rando T a (2011) Stem cell ageing and non-random
chromosome segregation. Philosophical transactions of the Royal Society of
London. Series B, Biological sciences 366: 85–93. Available: http://www.
rendertype=abstract.Accessed 28 Jun 2011.
36. Kamenisch Y, Fousteri M, Knoch J, Thaler A-K von, Fehrenbacher B, et al.
(2010) Proteins of nucleotide and base excision repair pathways interact in
mitochondria to protect from loss of subcutaneous fat, a hallmark of aging. The
Journal of experimental medicine 207: 379–90. Available: http://www.
&rendertype=abstract. Accessed 19 Jul 2011.
37. Mitchell JR, Hoeijmakers JH, Niedernhofer LJ (2003) Divide and conquer:
nucleotide excision repair battles cancer and ageing. Current Opinion in Cell
Biology 15: 232–240. Available: http://linkinghub.elsevier.com/retrieve/pii/
S0955067403000188. Accessed 19 Jul 2011.
38. Eraso P, Mazo ´n MJ, Portillo F (2006) Yeast protein kinase Ptk2 localizes at the
plasma membrane and phosphorylates in vitro the C-terminal peptide of the
H+-ATPase. Biochimica et biophysica acta 1758: 164–70. Available: http://
www.ncbi.nlm.nih.gov/pubmed/16510118. Accessed 19 Jul 2011.
39. Estrada E, Agostinis P, Vandenheede JR, Goris J, Merlevede W, et al. (1996)
Phosphorylation of yeast plasma membrane H+-ATPase by casein kinase I.
The Journal of biological chemistry 271: 32064–72. Available: http://www.
40. Burtner CR, Murakami CJ, Kennedy BK, Kaeberlein M (2009) A molecular
mechanism of chronological aging in yeast. Cell cycle (Georgetown, Tex.) 8:
1256–70. Available: http://www.pubmedcentral.nih.gov/articlerender.fcgi?
41. Gourlay CW, Ayscough KR (2005) The actin cytoskeleton in ageing and
apoptosis. FEMS yeast research 5: 1193–8. Available: http://www.ncbi.nlm.
nih.gov/pubmed/16144774. Accessed 19 Jul 2011.
42. Peisker K, Chiabudini M, Rospert S (2010) The ribosome-bound Hsp70
homolog Ssb of Saccharomyces cerevisiae. Biochimica et biophysica acta 1803:
662-72. pp 662–72. Available: http://www.ncbi.nlm.nih.gov/pubmed/
20226819. Accessed 19 Jul 2011.
43. Plehwe U von, Berndt U, Conz C, Chiabudini M, Fitzke E, et al. (2009) The
Hsp70 homolog Ssb is essential for glucose sensing via the SNF1 kinase
network. Genes & development 23: 2102–15. Available: http://www.
rendertype=abstract. Accessed 19 Jul 2011.
44. Kaeberlein M, Kennedy BK (2007) Protein translation, 2007. Aging cell 6:
731–4. Available: http://www.ncbi.nlm.nih.gov/pubmed/17941970. Accessed
19 Jul 2011.
45. Ruckenstuhl C, Carmona-Gutierrez D, Madeo F (2010) The sweet taste of
death: glucose triggers apoptosis during yeast chronological aging. Aging 2:
643–9. Available: http://www.pubmedcentral.nih.gov/articlerender.fcgi?
46. Santangelo GM (2006) Glucose Signaling in Saccharomyces cerevisiae.
Microbiology and Molecular Biology Reviews 70: 253–282. doi:10.1128/
47. Tabb MM, Tongaonkar P, Vu L, Nomura M (2000) Evidence for separable
functions of Srp1p, the yeast homolog of importin alpha (Karyopherin alpha):
role for Srp1p and Sts1p in protein degradation. Molecular and cellular biology
20: 6062–73. Available: http://www.pubmedcentral.nih.gov/articlerender.
48. Huber A, Bodenmiller B, Uotila A, Stahl M, Wanka S, et al. (2009)
Characterization of the rapamycin-sensitive phosphoproteome reveals that
Sch9 is a central coordinator of protein synthesis. Genes & development 23:
1929–43. Available: http://www.pubmedcentral.nih.gov/articlerender.fcgi?
artid=2725941&tool=pmcentrez&rendertype=abstract. Accessed 19 Jul
49. Soulard A, Cremonesi A, Moes S, Schu ¨tz F, Jeno ¨ P, et al. (2010) The
Rapamycin-sensitive Phosphoproteome Reveals That TOR Controls Protein
Kinase A Toward Some But Not All Substrates. Molecular Biology of the Cell
21: 3475–3486. doi:10.1091/mbc.E10.
50. Bharucha N, Ma J, Craig JD, Lawson SK, Yang Z, et al. (2008) Analysis of the
Yeast Kinome Reveals a Network of Regulated Protein Localization during
Filamentous Growth. Molecular Biology of the Cell 19: 2708 -2717: doi:10.1091/
51. Goldberg AA, Richard VR, Kyryakov P, Bourque SD, Beach A, et al. (2010)
Chemical genetic screen identifies lithocholic acid as an anti-aging compound that
extends yeast chronological life span in a TOR-independent manner, by
modulating housekeeping longevity assurance processes. Aging 2: 393–414.
52. Hickman MA, Rusche LN (2010) Transcriptional silencing functions of the
yeast protein Orc1/Sir3 subfunctionalized after gene duplication. Proceedings
of the National Academy of Sciences of the United States of America 107:
19384–9. Available: http://www.pubmedcentral.nih.gov/articlerender.fcgi?
artid=2984160&tool=pmcentrez&rendertype=abstract. Accessed 19 Jul
53. Camblong J, Iglesias N, Fickentscher C, Dieppois G, Stutz F (2007) Antisense
RNA Stabilization Induces Transcriptional Gene Silencing via Histone
Deacetylation in S. cerevisiae. Cell 131: 706–717. Available: http://
linkinghub.elsevier.com/retrieve/pii/S0092867407012019. Accessed 12 Jul
54. Gray JV, Petsko GA, Johnston GC, Ringe D, Singer RA, et al. (2004)
‘‘ Sleeping Beauty ’’: Quiescence in Saccharomyces cerevisiae. Microbiology
and Molecular Biology reviews 68: 187–206. doi:10.1128/MMBR.68.2.187.
55. Qin H, Lu M, Goldfarb DS (2008) Genomic instability is associated with natural
life span variation in Saccharomyces cerevisiae. PloS one 3: e2670. Available:
pmcentrez&rendertype=abstract. Accessed 19 Jul 2011.
56. BraunEL, FugeEK,Padilla PA,Werner-Washburne M (1996) A stationary-phase
gene in Saccharomyces cerevisiae is a member of a novel, highly conserved gene
Chronological Aging Network of Yeast
PLoS ONE | www.plosone.org 15December 2011 | Volume 6 | Issue 12 | e29284