HAL Id: hal-01500393
https://hal.archives-ouvertes.fr/hal-01500393
Submitted on 3 Apr 2017
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires
Yolanda Sanchez-Dehesa, David P. Parsons, Jose Maria Pena, Guillaume Beslon
To cite this version:
Yolanda Sanchez-Dehesa, David P. Parsons, Jose Maria Pena, Guillaume Beslon. Modelling Evolution of Regulatory Networks in Artificial Bacteria. Mathematical Modelling of Natural Phenomena, EDP Sciences, 2008, 2, 3, pp.27-66. �10.1051/mmnp:2008054�. �hal-01500393�
Modelling Evolution of Regulatory Networks
in Artiial Bateria
Y. Sanhez-Dehesa
a,c
, D. Parsons
a
, J.M. Peña
b
, and G. Beslon 1,a,c
a
LIRIS CNRS UMR5205, INSA-Lyon, Université de Lyon, 69621 Villeurbanne,Frane
b
DATSI, UniversidadPoliténiade Madrid, 28660 Madrid,Spain
c
Institut Rhne-Alpindes Systèmes Complexes (IXXI), Lyon, Frane
Abstrat. Studying the evolutive and adaptative mehanisms of prokaryotes is a ompli-
atedtask. Asthesemehanismsannotbeeasilystudiedinvivo,itisneessarytoonsider
other methods. We have therefore developed the RAevol model, a modeldesigned to study
the evolutionof bateriaand theiradaptationtothe environment. Our modelsimulates the
evolution of apopulationof artiialbateriaina hangingenvironment, providingus with
aninsight intothe strategies that digitalorganismsdevelop toadapt tonew onditions.
In this paper we desribe the priniples and arhiteture of the model, fousing on the
mehanismsof the regulatorynetworksof artiialorganisms. Experiments were onduted
onpopulations of artiialbateriaunder onditions of stress. We study the ways in whih
organismsadapttoenvironmentalhanges andexamine thestrategies they adopt. Ananal-
ysisoftheseadaptationstrategiesispresented andabriefoverviewwasproposedonerning
the patterns and topologialharateristis of the evolved regulatory networks.
Key words: evolution,regulatory networks, modelling, motifs,adaptationmehanisms
AMS subjet lassiation: 9204, 92D10, 92D15
1 Introdution
Prokaryote organisms are very diverse, livingindierentenvironments and developingvari-
ous abilities. Bateria are found in every eosystem some being olonized only by miro-
organisms illustratingthe impressive adaptation apabilities of prokaryotes. They an be
1
Correspondingauthor. Email: guillaume.beslonliris.nrs.fr
Article available at http://www.mmnp-journal.org or http://dx.doi.org/10.1051/mmnp:2008054
ganisms(e.g., Buhnera aphidiola,whihlivesinsymbiosis withaphids,providingessential
amino aids for their host), or even in the human intestine where Esherihia oli favors
digestionand absorption of nutrients.
Bateria are goodexamplesof organismadaptation. They areable toreattovariations
in their environment at dierent levels: bateria strains an adapt to major environmental
hangesbyadarwinianevolutionaryproessandindividualbateriaanadapttoshort-term
hangesintheir environment. Toahievethis kindofadaptation atdierentlevels, bateria
havedeveloped a largerepertoireof strategies that may themselvesbeoptimized depending
onthe harateristisofthe environment: stability,periodiity,stohastiity,ompetition...
Although a lot of dierent strategies (e.g. evolution, regulation, bet-hedging, adaptive
mutation, gene ampliation, Baldwin eet) have been identied and are relatively well
haraterized individually, we only have a very partial insight into how they ombine with
one another: in an idealizedenvironment, one an identify the optimal strategy and math-
ematially nd the optimal parameters. However real environments are far from ideal and
there will generally be a wide range of viable adaptation strategies, ombining e.g., regu-
lation and evolution, evolution and bet-hedging, regulation and gene ampliation or any
ombination of these. For instane, if the environment hanges slowly, bateria may have
enoughtimetomutateanddarwinianevolutionan besuienttoadapttonewonditions.
But,they maynot beabletoonserve omplexregulationstrategiessine mutationsquikly
degrade regulationmehanismswhen theseare inative[14℄. Now, if the environmentvaries
alittlefaster, evolution an beless eient than regulation,provided that bateriaare able
tosensetheirenvironmentatanaeptableost andthat environmentalhangesshowsome
regularities (e.g., swithes between two dierent energy soures as in the well-known la
operon). Onthe ontrary, rarebut unpreditable eventsput organismsunderstress and are
known to promote spei adaptive strategies suh as the development of mutator strains
[44℄. All these dierent strategies imply plastiity at dierent levels: geneti, metaboli,
physiologi, phenotypi,all of whih are involved inomplex interations.
Theseadaptationmehanismshelpbateriatoadapttohangingenvironments. However
eah has its own tempo, ranging fromslow (i.e., darwinian strategy) to fast (i.e. stohasti
perturbationsleading tophenotypivariability). In themiddle, genetiregulationenablesa
fast dynamiadaptation, enablingells toreat tohemialsignals. Regulationis the main
mehanismtoprovideadaptivebehavioratametabolilevel. However, regulationneverats
alone,itisobviouslyombinedwithevolution: genetivariations,genedupliation,geneloss
orhromosomalalterations[19℄onstituteavast repertoireofvariationsthatanbeusedby
abaterialstraintoadapttoitsenvironment,butthat analsoprovidebateriaindividuals
withtoolstodevelopmoreomplexadaptationmehanisms. Inspei onditionsevolution
givesrisetoregulatorysystemsthatenablefastadaptationtorapidlyhangingenvironments.
Intheaseofthelaoperon,regulationenablestheorganismtosaveenergywhenseveralfood
soures are available. It is supposed that regulation is a result of adaptation to hanging
environments. Yet, it an be shown that suh a system an be very sensitive to hanges
in the environment onditions: Dekel [14℄ has shown that only a few hundred generations
are neessary for E. oli to drastially hange its la operon behavior when plaed in new
onditions. At the other end of the time sale, the laoperon is known to have a stohasti
behavior [11, 17℄ and it an be shown that stohastiity of transription interats with the
regulatory ativity of the operon, delaying the operon swith [23℄. Thus, while regulation
ativity has long been supposed to be independent of slow evolutionary hanges or fast
stohasti variations, it is beoming more and more lear that the interations of all these
adaptation strategies must bestudied tofully understand their behavior [22℄.
Itisstillamatterofdebateinwhatkindofsituation/environmentevolutionpromotesthe
emergeneofregulatoryproessesandhowregulationinteratswiththeevolutionaryproess
itself. Hypotheses annot be easily studied on real living systems. Although experimental
evolution is possible with miro-organisms [16℄, traking hanges in genomes, regulatory
networks and even phenotypes is almost impossible in in vivo tests. An alternative is to
usedigitalorganismstostudythe genetibasesofadaptationinsilio [2℄. Insuhartiial
models,organisms(i.e.,omputationaldatastrutures)areplaedinasynthetienvironment
that provides them with resoures. In this environment the organisms reprodue, mutate
and ompete for the resoures, thus resulting in darwinian evolution. Sine the organisms
aswellasthe environmentare artiiallydened they anbothbeperfetlyand ompletely
desribed [38℄. Suh models have already shown their usefulness in studying evolution of
robustness[47℄ orinidentifyingindiretmutationalpressurethatregulates genomesize [29℄.
Yet, sine most of these models fous on mutational adaptation, they annot be used to
study omplex interations between the dierent adaptationmehanisms.
The denition of a suitablemodelto desribe this biologialproess would be useful to
takle many open questions in the literature of this domain: How do organisms adapt to
environmentalhanges? What isthe originof regulatorynetworks? Why doregulatorynet-
works appear during evolution? Howdo networks evolve over time? Studyingthe inlusion
of new nodes in already existing regulatory networks and studying the development of new
regulatory networks ould help to answer some of these questions and provide us with a
better understanding of network evolution.
Genetinetworksappeartobehighlyorganized: theyare modular[21℄,sale-free[7℄and
some motifs are overrepresented [4℄. Yet, the preise origin of these strutures is not fully
understood. In partiular, it is quite diult to distinguish between seletive origin (the
struture of the network is seleted beause it ensures a orret funtion in the organism's
environment), mutational origin (the mutational proess tends tofavor some strutures, as
inthepreferentialattahmentmodel[7℄)and indiretseletiveorigin(thenetwork struture
is seleted beause it is robust to mutation or, on the opposite, highly adaptable). It has
been shown that in some spei onditions, modular strutures an be seleted inevolved
networks [20, 25℄. Here again, modelling isan essentialtoolto takle suh questions.
Struture and dynamis of regulatorynetworks are atthe heartof systems biology. The
rapid development of this eld has been followed by the development of a very ative mod-
ellingativity ofsuhnetworks. Asfar asevolutionof regulatorynetworksisonerned, the
workhasbeenfousedonthequestionoftopologyevolution[25,26,49℄,evolutionofnetwork
robustness [3,12,42℄and evolutionof artiialfuntions[5,6,18, 32℄. Mostofthese papers
deal with diret evolution of geneti networks (i.e., in the model the network struture is
diretlymodiedbythegenetioperatorsmutations,rossing-overandrearrangements)or
seletionoftheindividualsonthe basisof thenetworkproperties(e.g.,seletionofaspei
topologyor seletionof a spei regulationdynami).
Additionally, many studies have been onduted to understand evolution of regulatory
networks from a bioinformati perspetive. Phylogeneti studies and sequene omparison
provide a quite preise view of the fores that shape bateria genomes and inuenes the
evolution of their regulatory networks [35℄. Thanks to these studies, it is now learer that
largegenomieventssuhasgenomirearrangement,horizontalgenetransfer(HGT)[19,31℄
or gene dupliation play a key role in the evolution of networks [45℄ and that the topology
of the network is fora large part indiretly shaped by the mutational dynami[13℄.
All these approahes fous on a spei fore that shape the network topology (e.g.,
mutational dynami, seletion for funtion, seletion for robustness - either mutational or
funtional robustness, ...). However, in a real biologialregulation network, allthese fores
are atwork simultaneouslyand thenetworktopologyresultsfromaompromisebetweenall
the onstraintsa networkand anorganism must fae. These onstraintsthemselves depend
ontheenvironmentalproperties: inastati environment,seletionforfuntionalrobustness
isimportantwhileinarandomly(butslowly)evolvingenvironment,themutationaldynami
and/or evolvability property may be ruial for the organism. Thus, to better understand
howthe environmentmodulatesthe emergene of spei network properties,anintegrated
modelisneededinwhihtheappearaneofdierentnetworktopologiesduringtheevolution
depends on the dynamial properties of the environment. Moreover, this model should
respet the main lines of organisms' evolution. Organisms should own a geneti sequene
that allows alarge variety of mutationalevents, aomplexgenotype-to-phenotype mapping
that inludes a proteome level and enables the evolution of a geneti network inside the
organism. Thus, it should be stratied from a genomi level (the sequene being diretly
modiedbymutationaleventswhileallotherorganizationlevelsareonlyindiretlymodied
depending on the eet of the random mutations) to a phenotype level (the phenotype
level being the only one subjet to seletion while the other organization levels are only
indiretly seleted depending on their inuene on the phenotype). The proteome level
must respet the ore properties of regulatory networks' evolution: the regulation network
isneitherdiretlymutatednor diretlyseleted. The nodes ofthe networks are theproteins
of the organism but the links result from a omplex interation between the organisms
proteinsand itsgenomisequene: eahprotein mayor may not interatwith the sequene
atspei loations,modifyingthetransriptionalativity ofapromoter and,onsequently,
the transription rate of one or many genes. Eah gene is then transribed at a spei
rate that depends on the intrinsi properties of its promoter and on the inuene of the
regulation network (inluding ativation, inhibition and self-regulation - see below). The
proteinonentrationis thengoverned bythe transriptionrate and by adegradationterm.
Moreover, the whole transription/translation proess is highly stohasti and it is now
reognized that stohastiity inuenes the fateof organisms[17℄.
Followingthesepriniples,wehavedevelopedtheRegulatoryArtiialEvolution model
(RAevol). In thismodel, artiialdigital bateriaevolveina variableenvironment. Along
theirevolution,thesebateriaaquiregenesandevolveaomplexgenome,aomplexregula-
tionnetworkandanadapted phenotype. Onanevolutionarytimesale, thebestindividuals
are those whih evolve the best mehanisms to fae environmental variations. We are then
able tounderstand whih ofthese mehanismsare eient dependingonthe environmental
onditions. Inthis paper, we rst desribethe generalprinipleof regulationin prokaryotes
and we expose the mehanisms that onstitute the ore of our model(Setion 2). Then we
preisely desribe the RAevol model (Setion 3), fousing on the regulation properties. Fi-
nallywe presenta simpleartiialevolution experimentthat illustratesthe mainproperties
of the model(Setion 4)and disuss evolutionary senariithat may be testedwith RAevol.
2 Priniples of Geneti Regulation in Prokaryotes
Thepriniplesoftransriptionregulationweredesribedinthe60'sbyJaobandMonod[24℄.
Experimenting with Esherihia oli, they showed that the transription rate of a spei
genetisequene depends onatleastthree fators: itspromoter,whihisthe initialbinding
sequene of the RNA polymerase, regulation sites (either ativators or inhibitors) where
somespeiproteinsanbind,thereafterinueningthetransriptionproess,andexternal
fators suh asthe onentration of RNA polymerase inthe ell. Note that these priniples
annot be onsidered universal: in eukaryoti organisms, the regulation of transription
ativity depends onmany dierent mehanisms,inludinghromatindynamis.
Contrary to eukaryotes, in whih promoters are generally inative in the absene of
transriptionfators (initiationomplexes are neessary forthe transription tostart and a
naked promoter willbe essentially inative), prokaryoti promoters and RNA polymerase
an diretly interat with one another. In the absene of regulatory elements, a promoter
will have an inherent ativity that mainly depends on its quality. When a promoter has a
primary sequene very similar tothe onsensus sequene, RNA-polymerasean easily bind
toit. The initiationoftransriptionwillthenregularly ourand the intrinsitransription
levelwillbehigh (possibly atamaximum levelif thepromoter has avery goodanity with
thepolymerase). Inthisase,thetransriptionratewillonlydependonextrinsifatorssuh
as the RNA polymerase onentrationand quality orthe transription elongation speed).
If the promoter anity to the RNA polymerase is weak, transription will only rarely
be initiated. The quality of the promoter thus determines the transriptional ground tran-
sription level
β
(or basal transription level, gure 1(a)) [43℄. Thus, in the absene of spei regulatory sequenes, genes are transribed at a rate that mainly depends on theirpromoterstrength, maximumtransriptionratebeingbounded byglobalfators suhasthe
polymerase properties and onentration.
The transription level an be modiedby the ation of regulatoryproteins. These pro-
teinsmodify thetransriptionlevels, enhaningorinhibitinggenetransription. Inprokary-
otes, this proess ismainlyused toontrolenergy onsumptioninorder tomaintain agood
balanebetween food availabilityand energy, and to adaptto environmentalhanges.
In prokaryotes,inhibitionorrepression oftransriptionours whenaregulatoryprotein
inhibits the initiation of transription or the elongation of the transript (i.e., repressor
proteins). Ativationoftransriptionourswhenaproteinpromotestransriptioninitiation
[48℄. Whenapromoterisativated,itsativity anonlyriseup toamaximumtransription
level (meaning that intrinsially eient promoters an onlybemarginally enhaned).
Transription fators (ativation and repression proteins) at by binding to spei re-
gionsoftheDNAthatarenearthepromoteroftheproteinthey regulate. Repressorproteins
bindtoaregionalledoperator(alsoalledinhibitoryregion)generallysituateddownstream
from the promoter region. When bound there, a repressor may prevent RNA polymerase
frombindingorblok itsdisplaementalongthe DNAthusdisturbing RNAelongation (g-
ure 1(b)). Ativator proteins target ativator-binding sites are usually loated upstream
of the promoter region. They promote RNA-polymerase binding, thus enhaning protein
prodution(gure 1()).
enhancer p r o m o t e r o p e r a t o r
DNA START STOP
t e r m i n a t o r
(a) Whenno proteinsbind theregulatory regionsthe RNA tran-
sriptionisdoneatgroundlevel.
RNA Polymerase
enhancer p r o m o t e r o p e r a t o r
DNA START STOP
t e r m i n a t o r transcription
translation
(b) Aregulatory proteinhas targetedthe operator. It bloks the
polymerase displaement along DNA and prevents it from tran-
sribingthegene. Thusthistransriptionfatorrepressesthepro-
dutionof theproteinassoiatedwiththisgene.
RNA Polymerase
enhancer p r o m o t e r o p e r a t o r
DNA START STOP
t e r m i n a t o r transcription
translation
() A protein binds the enhaner region, favoring the RNA-
polymerase(toparrow)bindingandtransriptioninitiation. Sine
no inhibitoryproteinbindtheoperator,theRNA-Polymerasean
transribe the gene more eiently, thus enhaning the protein
produtionlevel.
Figure1: Transriptional states inprokaryotes.
Inprokaryotes,multiplegenesoftenshareasinglepromoter,itsoperatoranditsativator
bindingsites. Thesegenesareo-transribedand thereforeo-regulated. Suhasequene in
whihseveral genesshare their promoter and regulatoryregions isalledan operon beause
allgenes are underthe ontrol of asingle operator (gure2).
enhancer o p e r a t o r
START STOP START STOP
p r o m o t e r t e r m i n a t o r
transcription translation RNA Polymerase
DNA
Figure2: Overview of an operon struture
The best known regulation system is probably the Latose (la) Operon whih ontrols
the latose-gluose metabolism in Esherihia oli. When Monod experimented with the
eets ofombiningsugarsasarbonsouresforE. oli,hefound thatif gluoseandlatose
are provided tothe baterium, itrst metabolizesgluose and the olony grows fast. When
gluose is depleted, the bateria stop growing. After a short period (lag-phase), bateria
start onsuming latose and the olony grows again. Jaob and Monod later showed that
this adaptivebehavior omes froma gene regulation mehanism.
InE.oli,thelatosemetabolismisontrolledbyanenzyme,the
β
-galatosidaseprotein, that breaks down latose into two simple sugars (galatoseand gluose)and by a permeaseprotein that transports latose from the environment to the ell. The former protein also
onverts part of the latose intoallolatose.
The
β
-galatosidaseproteinis enoded by the LaZ gene and the permease by the LaY gene. Both genes are grouped on an operon struture, the la operon, and are under theinueneofthesamepromoterandthesameoperator. Infatthelaoperonontainsathird
gene, LaA,that enodes for a
β
-galatosidasetransaetylase. A fourth gene, LaI, that is not on the same operon, ompletes the system by oding for a repressor of the la operon.The repressor proteinisable tobindtothe la operator,preventing the transriptionof the
operon(gure3). However, whenlatoseispresentintheell, itinteratswiththe repressor
protein,and hanges itsonformation, preventing itfrombinding tothe la operon. When,
the operon is nolonger repressed LaYand LaZ an betransribed. Due tothe permease,
latoseonentrationthusinreases,while
β
-galatosidaseisproduedanddegradeslatose.TheLaIontrolisanexampleofnegativeontrol. However,itisnotsuienttoexplain
the whole behavior of the la operon. In partiular, negative ontrol annot explain why,
in presene of both gluose and latose, the operon is not transribed. Indeed, the operon
is alsoontrolledby a positive loop: the onentration of gluose is sensedby the ellvia a
signalingmoleule, AMP;themoregluoseintheenvironment,the lowertheonentration
of AMP. AMP binds to an induer of the operon, the CAP protein, that itself binds on
the DNA upstream from the la promoter. Then, the la operon is transribed if and only
if latose is present in the environment and gluose is not (or no longer) present in the
environment 2
.
2
A lots of seondary mehanisms have been disovered. They slightly modify the behavior of the la
operon but thetwomain regulationloops are thenegativeloop due to LaI and the positive loop due to
AMPbindingonCAP (gure 3).