Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Hindawi Publishing CorporationBioMed Research InternationalVolume 2013 Article ID 703849 15 pageshttpdxdoiorg1011552013703849
Research ArticleA Systemsrsquo Biology Approach to Study MicroRNA-MediatedGene Regulatory Networks
Xin Lai12 Animesh Bhattacharya3 Ulf Schmitz1 Manfred Kunz3
Julio Vera2 and Olaf Wolkenhauer14
1 Department of Systems Biology and Bioinformatics University of Rostock 18051 Rostock Germany2 Laboratory of Systems Tumor Immunology Department of Dermatology Faculty of Medicine University of Erlangen-NurembergUlmenweg 18 91054 Erlangen Germany
3Department of Dermatology Venereology and Allergology University of Leipzig 04155 Leipzig Germany4 Institute for Advanced Study (STIAS) Wallenberg Research Centre at Stellenbosch University Stellenbosch 7600 South Africa
Correspondence should be addressed to Julio Vera juliovera-gonzalezuk-erlangende and Olaf Wolkenhauerolafwolkenhaueruni-rostockde
Received 31 July 2013 Revised 12 September 2013 Accepted 17 September 2013
Academic Editor Tao Huang
Copyright copy 2013 Xin Lai et al This is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
MicroRNAs (miRNAs) are potent effectors in gene regulatory networks where aberrant miRNA expression can contribute tohuman diseases such as cancer For a better understanding of the regulatory role of miRNAs in coordinating gene expressionwe here present a systems biology approach combining data-driven modeling and model-driven experiments Such an approachis characterized by an iterative process including biological data acquisition and integration network construction mathematicalmodeling and experimental validation To demonstrate the application of this approach we adopt it to investigate mechanisms ofcollective repression on p21 by multiple miRNAs We first construct a p21 regulatory network based on data from the literatureand further expand it using algorithms that predict molecular interactions Based on the network structure a detailed mechanisticmodel is established and its parameter values are determined using data Finally the calibrated model is used to study the effect ofdifferent miRNA expression profiles and cooperative target regulation on p21 expression levels in different biological contexts
1 Introduction
Although microRNAs (miRNAs) are physically small theyhave been shown to play an important role in gene regulation[1] Currently an increasing number of studies are beingcarried out to deepen our understanding of miRNA reg-ulatory mechanisms and functions However experimentalapproaches have limitations when dealing with complexbiological systems composed of multiple layers of regu-lation such as the transcriptional and post-transcriptionalregulation by transcription factors (TFs) and miRNAs [2]Most experimental approaches focus on the identificationof miRNA targets and the investigation of physiologicalconsequences when perturbing miRNA expressions but areunsuited to provide a system-level interpretation for observedphenomena Therefore the introduction of a systematicapproach which can unravel the underlying mechanisms by
which miRNAs exert their functions becomes increasinglyappealing
The systems biology approach combining data-drivenmodeling andmodel-driven experiments provides a system-atic and comprehensive perspective on the regulatory rolesof miRNAs in gene regulatory networks [3ndash5] To investigatea gene regulatory network an iterative process of four stepsis needed (I) Biological network construction a map isconstructed that shows interactions amongmolecular entities(such as genes proteins and RNAs) using information fromliterature and databases (II) Model construction dependingon the biological problem investigated and experimental dataavailable the interactionmap can be translated into a detailedmechanistic model that can simulate the temporal evolutionof molecular entities The values of parameters in this modelcan be determind from literature databases or they aredirectly estimated from quantitative experimental data using
2 BioMed Research International
optimizationmethods (III)Computational experiments oncea model is established it can be simulated andor analyzedfor its general behavior (IV) Experimental validation modelpredictions together with biological explanations are inte-grated to guide the design of new experiments which inturn validate or falsify the model If model predictions arein agreement with the experiments the model justifies thebiological hypotheses behind it In turn these hypotheseswhich provide reasonable explanations for the biologicalphenomenon lead to an enhanced understanding of thegene regulatory network Otherwise the structure of themathematical model is modified to generate new hypothesesand suggest new experiments
The application of the systems biology approach to theanalysis of a gene regulatory network is demonstrated witha case study of the regulation of p21 by multiple miRNAs [4]The network combining putative targets of TF and miRNAregulation with experimentally proven molecular interac-tions was constructed and visualized Next the network wastranslated into a detailed mechanistic model which wascharacterized and validated with experimental data Finallythe integration of quantitative data and modeling helped usto generate and validate hypotheses about mechanisms ofcollective miRNA repression on p21
2 Results
21 The Systems Biology Approach to Study miRNA-MediatedGene Regulatory Networks
211 Mathematical Modeling The aim of our analysis is tounravel the complex mechanisms by which gene regulatorynetworks involving miRNAs are regulated We iterativelyintegrate data from literature experiments and biologicaldatabases into a detailed mechanistic model of a gene regula-tory network The model is then used to formulate and testhypotheses about mechanisms of miRNA target regulationand cellular process-related variability The methodologyincludes four steps which are briefly summarized in Table 1In the coming sections we present these steps in detail
(1) Data Retrieval To construct a gene regulatory networkcomposed by different levels of regulation we collect infor-mation from different resources which are briefly describedbelow and more resources for data retrieval are introducedin Table 1
(a) Transcriptional level regulators Experimentally veri-fied TFs for a gene can be extracted from literatureor databases such as TRED TRANSFAC or HTRIdb[6ndash8] the putative TFs which are associated withconserved TF-binding sites residing in the promoterregion of a gene can for example be extracted fromthe table of TFs with conserved binding sites in theUCSC genome browser or the TRANSFAC database[9] miRGen 20 is a database that provides bothpredicted and experimentally verified informationabout miRNA regulation by TFs [10]
(b) Post-transcriptional regulators Databases such asmiRecords Tarbase and miRTarBase are repositoriesof experimentally validatedmiRNAgene interactions[11ndash13] Predictions of miRNAgene interactions areaccumulated in databases like miRWalk [14]
(c) Protein-protein interactions Both the Human ProteinReference Database (HPRD) [15] and the STRINGdatabase [16] document experimentally verifiedprotein-protein interactions besides STRING alsoprovides putative protein-protein interactions rankedby confidence scores Further details about the exactmechanism of protein-protein interactions can befound in Reactome [17]
(2) Network Construction and Visualization Based on theinformation collected a gene regulatory network is con-structed and visualized for providing an overview For thispurpose we recommend CellDesigner which uses standard-ized symbols (Systems Biology Graphical NotationmdashSBGN)[18] for visualization and stores gene regulatory networksin the SBML format (Systems Biology Markup Language)[19] CellDesigner also provides the possibility to simulatetemporal dynamics of the gene regulatory network due tothe integration of the SBML ODE (ordinary differentialequation) solver Besides Cytoscape is another powerful toolfor integration of biological networks and gene expressiondata [20]
For assessing the reliability of interactions considered ingene regulatory networks confidence scores can be computedas being documented in our previous publication [4] Thefactors that are used to determine the confidence score formolecular interactions can be the number of publicationsreporting an interaction experimental methods used toidentify an interaction interaction types and computationalpredictionsThe computed confidence scores range from 0 to1 where values towards 1 indicate higher confidence whereasvalues towards 0 indicate lower confidence in a given inter-action For example the confidence score for a miRNAgeneinteraction can be calculated using the following equation
119878miRNAgene =119908119901sdot 119878119901+ 119908119898sdot 119878119898+ 119908bs sdot 119878bs
119908119901+ 119908119898+ 119908bs
(1)
where 119908⟨119901119898bs⟩ are weights that are assigned to the scores
which account for the number of publications (119878119901) detection
method (119878119898) and the number of predicted binding sites
(119878bs) The values of the weights can be assigned based onexpert knowledge and the higher the value of the weightis the bigger impact it has on the confidence score forthe interaction The values of 119878
119901and 119878bs can be calculated
using the logarithmic equation 119878⟨119901bs⟩(119899) = log
119898+1(119899)
where 119899 denotes the number of publications describing themiRNAgene interaction or the number of binding sites thatthe miRNA has in the 31015840 UTR (untranslated region) ofthe gene The value of 119898 is a cut-off that represents thenumber of publications or binding sites required for 119878
⟨119901bs⟩ toobtaining their maximum values Various methods such aswestern blots qRT-PCR and reporter assays can be appliedto support the miRNAgene interaction but these methods
BioMed Research International 3
Table 1 Overview of themethodology Key points in each step of themethodology and themain resources for constructingmiRNA-mediatedgene regulatory networks are given
Step 1 data retrievalRegulation types Resources
Transcriptional generegulation
TRED (httprulaicshleducgi-binTREDtredcgiprocess=home) a database that provides an integratedrepository for both cis- and transregulatory elements in mammalsTRANSFAC (httpwwwgene-regulationcompubdatabaseshtml) a database that collects eukaryotictranscriptional regulation comprising data on TFs their target genes and binding sitesThe UCSC table browser (httpgenomeucscedu) a popular web-based tool for querying the UCSC GenomeBrowser annotation tablesHTRIdb (httpwwwlbbcibbunespbrhtri) an open-access database for experimentally verified humantranscriptional regulation interactionsMIRNTN (httpmaiaunilumirontonphp) an integrative resource based on a metaregulation networkmodel including TFs miRNAs and genesPuTmiR (httpwwwisicalacinsimbioinfo miuTF-miRNAphp) a database of predicted TFs for humanmiRNAsTransmiR (http20238126151hmddmirnatf) a database of validated TF-miRNA interactionsmiRGen 20 (httpdianacslabecentuagrmirgen) a database of miRNA genomic information and regulation
Posttranscriptionalgene regulation
miRecords (httpmirecordsbioleadorg) a resource for animal miRNA-target interactionsTarbase (httpwwwmicrornagrtarbase) a database that stores detailed information for each miRNA-geneinteraction the experimental validation methodologies and their outcomesmiRTarBase (httpmirtarbasembcnctuedutw) a database that collects validated miRNA-target interactionsby manually surveying the pertinent literaturemiRWalk (httpwwwummuni-heidelbergdeappszmfmirwalk) a comprehensive database that providesinformation on miRNAs from human mouse and rat on their predicted as well as validated binding sites intarget genes
Protein-proteininteraction
HPRD (httpwwwhprdorg) a centralized platform to visually depict and integrate information pertaining todomain architecture posttranslational modifications interaction networks and disease association for eachprotein in the human proteomeSTRING (httpstring-dborg) a database of known and predicted protein interactions The interactionsinclude direct (physical) and indirect (functional) associationsMPPI (httpmipshelmholtz-muenchendeprojppi) a collection of manually curated high-quality PPI datacollected from the scientific literature by expert curatorsDIP (httpdipdoe-mbiuclaedudipMaincgi) a catalog of experimentally determined interactions betweenproteinsIntAct (httpwwwebiacukintactmainxhtml) a platform that provides a database system and analysis toolsfor molecular interaction dataReactome (httpwwwreactomeorg) an open-source open access manually curated and peer-reviewedpathway database
GO annotation
Amigo GO (httpamigogeneontologyorgcgi-binamigogocgi) the official GO browser and search enginemiR2Disease (httpwwwmir2diseaseorg) a manually curated database that aims at providing acomprehensive resource of miRNA deregulation in various human diseasesmiRCancer (httpmircancerecuedu) a miRNA-cancer association database constructed by text mining on theliteraturePhenomiR (httpmipshelmholtz-muenchendephenomir) a database that provides information aboutdifferentially expressed miRNAs in diseases and other biological processesmiRGator (httpmirgatorkobicrekr) a novel database and navigator tool for functional interpretation ofmiRNAsmiRo (httpferrolabdmiunictitmiro) a web-based knowledge base that provides users withmiRNA-phenotype associations in humans
Step 2 network construction and visualization(i) Visualize regulatory interactions in platforms such as CellDesigner and Cytoscape that support standardized data formats(ii) Calculate confidence scores for assessing reliability of interactions in gene regulatory networksStep 3 model construction and calibration(i) Formulate equations using rate equations(ii) Fix parameter values using available biological information(iii) Estimate the other unknown and immeasurable parameter values using optimization methods which can minimize the distancebetween model simulations and experimental data such as time course qRT-PCR and western blot dataStep 4 model validation and analysis(i) Design new experiments and generate new data to verify the calibrated model(ii) Study complex properties and behavior of the system
4 BioMed Research International
provide experimentalists with different levels of confidencethus differing confidences can be reflected using differentvalues for 119878
119898based on the experience of experimentalists
Of note although the confidence scores cannot be directlyconverted into a mathematical model with the help of thesescores we can discard non-reliable putative interactions togenerate the ultimate version of a gene regulatory networkThe final version of the network can be further analyzedto identify regulatory motifs like feedforward loops (FFLs)for example with the help of the Cytoscape plugin NetDS[21] Thereafter the complete network or parts of it canbe converted into a detailed mechanistic model which isdescribed in detail in the following section
(3) Model Construction and Calibration After the construc-tion visualization and refinement of a gene regulatory net-work it is converted into a detailedmechanistic model whichenables the investigation of unanswered biological questionsand validation of hypotheses Ordinary differential equations(ODEs) describe how processes of synthesis biochemicalmodification andor degradation affect the temporal con-centration profile of biochemical species like proteins RNAsand metabolites An ODE model can be constructed usingappropriate kinetic laws such as the law ofmass-action whichstates that the rate of a chemical reaction is proportional tothe probability that the reacting species collideThis collisionprobability is in turn proportional to the concentration ofthe reactants [22] A general representation of ODE-basedmodels using mass-action kinetics is given by the followingequation
119889119909119894
119889119905=
119898
sum
120583=1
119888119894120583sdot 119896120583sdot
119899
prod
119895=1
119909119895
119892120583119895 119894 isin 1 2 119899 (2)
where 119909119894represents state variables which denote the molar
concentration of the 119894th biochemical specie Every biochem-ical reaction 120583 is described as a product of a rate constant(119896120583) and biochemical species (119909
119895 119895 isin 1 2 119899) that are
involved in this reaction 119888119894120583 the so-called stoichiometric
coefficients relate the number of reactant molecules con-sumed to the number of product molecules generated inthe reaction 120583 119892
120583119895denotes kinetic orders which are equal
to the number of species of 119909119894involved in the biochemical
reaction 120583 The rate constants kinetic orders and the initialconditions of state variables are defined as model parametersBesides mass-action kinetics other kinetic rate laws such asMichaelis-Menten kinetics Hill equation and power-laws arealso frequently used in ODE models
After ODEs are formulated the model requires to becalibrated a process by which parameter values are adjustedin order to make model simulations match experimentalobservations as good as possible To do so there are twopossible means characterization of parameter values usingavailable biological information or estimation of parametervalues using optimization methods Some parameter valuescan be directly measured or obtained from literature ordatabases For example the half-life (119905
12) of some molecules
(eg protein) can be measured in vitro via western blot-ting This information can be used to characterize their
degradation rate constants through the equation 119896deg =
(ln 211990512) The database SABIO-RK provides a platform for
modelers of biochemical networks to assemble informationabout reactions and kinetic constants [27] However formost model parameters whose values cannot be measuredin laboratories or be accessed from literature or databasesparameter estimation is a necessary process to characterizetheir values Before running parameter estimation initialparameter values and boundaries should be set within phys-ically plausible ranges To do so the database BioNumbersprovides modelers with key numbers in molecular and cellbiology ranging from cell sizes to metabolite concentrationsfrom reaction rates to generation times from genome sizesto the number of mitochondria in a cell [28] After parameterestimation unknown parameter values are determined usingoptimization methods which can minimize a cost functionthat measures the goodness of fit of the model with respect togiven quantitative experimental data sets Parameter estima-tion using optimization algorithms is an open research fieldin which several methods have been developed accordingto the nature and numerical properties of biological dataanalyzed The discussion for choosing proper optimizationmethods for parameter estimation is beyond the scope ofthis paper but the interested reader is referred to the paperpublished by Chou and Voit [29]
(4) Model Validation and Analysis Usually the model sim-ulations are compared with the experimental data used forthe parameter estimation but a good agreement betweenboth is not enough to guarantee the predictive ability of themodel Therefore it is necessary to validate the model withdata sets that are not used during the parameter estimationThis process is called model validation and can ensure morereliable and accurate model predictions To do so the datagenerated in new experiments or extracted from literature arecompared with model simulations which are obtained afterconfiguring the model according to the new experimentalsettings Once a model is validated it can be used to per-form predictive simulations which are helpful to studythe dynamic properties of biochemical systems guide thedesign of new experiments in the laboratory and formulateadditional hypotheses In addition tools such as sensitivityand bifurcation analysis can be used to study complex proper-ties and behavior of the modeled system Sensitivity analysisis used to evaluate the influence of model parameters (eginitial concentrations of the state variables and rate constants)on model outputs such as the temporal behavior of networkcomponents [30] Bifurcation analysis is employed to detectcontrol parameters (also known as bifurcation parameters)whose variations are able to change drastically the dynamicalproperties of the biochemical system as well as the stabilityof its fixed points [31] The application of these tools tomathematical modeling is beyond the scope of this paper butthe interested reader is referred to the publication of Zhouet al [32] and Marino et al [33]
212 Experiment Methods As mentioned in the previoussections after a model is established it can be calibrated
BioMed Research International 5
and validated using temporal experimental data To do sothe data can either be derived from literature or generatedto calibrate the model by own experiments In case of ODEmodels the most suitable data for model calibration isquantitative time-series data obtained from perturbation orquantitative dose-response experimentsThe experiments inwhich time-series data are measured for different regulators(such as miRNAs and TFs) of a gene regulatory network canbe obtained using the techniques described in the subsectionsbelow
(1) qRT-PCR Quantitative real time polymerase chain reac-tion (qRT-PCR) has been used to identify mRNAs regulatedby overexpression or silencing of a specific miRNA [3435] miRNAs typically exhibit their regulatory effects byassociating with specific 31015840 UTR regions of the mRNAscalled miRNA seed regions [34] This association can leadeither to a temporary inhibition of translation or completedegradation of the mRNA in which case qRT-PCR mediateddetection is beneficial For this the cells are transfectedwith the miRNA and non-targeting control oligonucleotidesat an appropriate concentration using either a lipid basedtransfection reagent or nucleofection The transfected cellsare then incubated for the necessary time periods (eg 24 h48 h 72 h etc) after which lysates are prepared and total RNAis extracted 1000ndash2000 ng of the RNA is then converted intocDNA using a reverse transcription kit Taqman qRT-PCR isperformed using 10ndash20 ng of cDNA and primers labeled withfluorescence probes to detect the transcriptional levels of thetarget mRNA Housekeeping genes like GAPDH and HPRTare used as endogenous controls for data normalizationRelative expression at different time points is determinedby comparing the Ct values of the miRNA transfected cellswith Ct values of non-targeting control transfected cells andexpressed as relative expression [36] However qRT-PCR inspite of being highly sensitive is applicable specifically underconditions of complete or partial degradation of the targetmRNA miRNA-mediated inhibition in translation can bebetter demonstrated by techniques like immunoblotting
(2) Western BlotImmunblotting Western blotting or proteinimmunoblotting is a technique to detect the expression of agene at protein level This technique is particularly useful indetermining the regulatory effects of a miRNA on expressionof a target gene which is temporarily inhibited For thisthe cells are transfected with a miRNA or an antagomiRas mentioned above and cell lysates are prepared at appro-priate time points using protein lysis buffers (eg Radio-Immunoprecipitation Assay buffer or RIPA) containing lysisagents like Dithiothreitol (DTT) and protease inhibitors Theproteins from each sample are quantitated using Bradfordor BCA reagents and compared with bovine serum albumin(BSA) standards for accurate protein estimation 20ndash40 120583g ofprotein is then loaded and resolved on a sodium dodecyl sul-fate polyacrylamide gel (SDS-PAGE) along with a pre-stainedprotein marker The protein bands are then transferred ontoa nitrocellulose membrane followed by incubation with theappropriate primary and secondary antibodies linked tofluorescent dyes or horse radish peroxidase enzyme (HRP)
The protein expression is then analyzed using either fluores-cence or chemiluminiscenceHRP substrates in gel documen-tation systems (eg LI-COR Odyssey) Housekeeping geneslike 120573-actin or 120573-tubulin are used for protein normalizationThe time point for maximum target gene suppression gen-erally varies depending on the number of miRNA bindingsites at the 31015840 UTR of the target gene and the extent ofcomplementarity of the seed region [37] Immunoblotting isa widely used technique to provide confirmatory evidence forthe inhibitory effects ofmiRNA at the protein level but it failsto explain the underlying interaction mechanisms
(3) Reporter Gene Assay As each miRNA can inhibit theexpression of a large number of genes regulation of a par-ticular target gene may either be by direct interaction or bean indirect consequence of it In direct regulation a miRNAbinds to the complementary sequences at the 31015840 UTR of atarget gene and thereby suppresses its expression As a con-sequence of this the expression levels of a number of down-stream genes (indirect targets) are also dysregulated makingit crucial to differentiate between primary and secondarymiRNA targets To determine the interaction specificitya reporter construct (luciferase) with intact or mutated 31015840UTR of the target gene cloned at the 51015840 end is co-transfectedinto the cells along with the miRNA The regulatory effect ofthe miRNA on the target gene expression is then measuredusing the expression of a reporter gene In the absence ofthe appropriate binding sequences (mutated 31015840 UTR) themiRNA cannot suppress the reporter mRNA suggesting thatthe suppressive effect of the miRNA is mediated by a directinteraction The reporter activity can be analyzed at differenttime points such as 24 hr 48 hr and 72 hr to determine thetime dependent suppression of a target gene expression bya miRNA
22 Case Study The Regulation of p21 by Multiple andCooperative miRNAs
221 Construction and Visualization of p21 Regulatory Net-work By using the approach described above we investi-gated the regulation of p21 by its multiple targeting miRNAsp21 also known as cyclin-dependent kinase inhibitor 1(CDKN1A) is a transcriptional target of p53 It is requiredfor proper cell cycle progression and plays a role in cell deathDNA repair senescence and aging (reviewed in [38]) Inter-estingly p21 was the first experimentally validated miRNAtarget hub which is a gene that is simultaneously regulatedby many miRNAs This made it an ideal candidate for a casestudy of our approach [23] To do so we first constructeda p21 regulatory network using the following steps
(a) We extracted miRNA-target interactions from thepublication of Wu et al [23] where a list of predictedp21-regulating miRNAs was subjected to experimen-tal validation
(b) Experimentally verified TFs of p21 were extractedfrom literature and putative TFs having conservedbinding sites in the 5 kb upstream region of the p21open reading frame were extracted from UCSC table
6 BioMed Research International
browser A list of TFs controlling the expression of themiRNAswas constructed using information of exper-imentally proven TF-miRNA interactions extractedfrom TransmiR (release 10) [39] In addition wegenerated a list of putative TFs of miRNAs withbinding sites in the 10 kb upstream region of themiRNA genes using information from the databasesPuTmiR (release 10) [40] and MIRNTN (version121) [41] and from the table of TFs with conservedbinding sites in theUCSC genome browser (hg18) [9]
(c) Information about protein interactions was extractedfrom theHuman Protein ReferenceDatabase (HPRDrelease 90) [15] and the STRING database (release90) [16] Only the experimentally verified p21-proteininteractions were used to construct the network
(d) Additionally we associated the TFs in the network tonine biological processes based on the Gene Ontol-ogy (GO) [42] The corresponding GO terms werecell proliferation cell apoptosis immune responseinflammatory response cell cycle DNA damage cellsenescence DNA repair and cell migration
Next we visualized the network in CellDesigner andcomputed a confidence score for each interaction in thenetwork (Figure 1) The confidence scores provide us withthe reliability of the interactions considered in p21 regulatorynetwork With the help of these scores we discarded non-reliable interactions and constructed the mechanistic modelaccounting for p21 regulation by its targeting miRNAsBesides the interested experimentalists can further use thisinformation to choose reliable interacting candidates of p21for designing relevant experiments The scores for eachinteraction of p21 regulatory network are shown in Table 2
222 Mechanistic Modeling of p21 Regulation byIts Targeting miRNAs
(1) Model Construction After constructing the regulatorynetwork a detailed mechanistic model of ODEs whichdescribes the biochemical reactions underlying the regula-tion of p21 was established We chose the ODE modelingapproach as it is a simple formalism for describing temporaldynamics of biochemical systems and a wide range of toolsare available to explore their properties Precisely the modelconsidered themRNA(mp21 (3)) andprotein (p21 (6)) of themiRNA target hub p21 the p21-targeting miRNAs (miR
119894 119894 isin
(1 15) (5)) and the complexes formed by p21 mRNAand miRNA [mp21 | miR
119894] (4) Altogether the model is
constituted by 32 state variables and 64 parameters
119889mp21119889119905
= 119896mp21syn sdot 119891act (TFmp21)
minusmp21 sdot (119896mp21deg +sum
119894
119896complex
119894
ass sdotmiR119894)
(3)
119889 [mp21 | miR119894]
119889119905
= 119896complex
119894
ass sdotmp21 sdotmiR119894minus 119896
complex119894
deg sdot [mp21 | miR119894]
(4)
119889miR119894
119889119905= 119896
miR119894syn sdot 119891act (TFmiR119894)
minusmiR119894sdot (119896
miR119894deg + 119896
complex119894
ass sdotmp21) (5)
dp21119889119905
= 119896p21syn sdotmp21 + 119896p21deg sdot p21 (6)
mp21Total = mp21 +sum119894
[mp21 | miR119894] (7)
For mp21 processes considered in the model were (i) itssynthesis (119896mp21
syn ) mediated by TFs (119891act(TFmp21)) (ii) itsdegradation (119896mp21
deg ) and (iii) its association with a miRNA(119896complex
119894
ass ) For each miR119894 processes considered were (i)
the synthesis (119896miR119894syn ) mediated by TFs (119891act(TFmiR119894)) (ii) the
degradation (119896miR119894deg ) and (iii) the association with the p21
mRNA target (119896complex119894
ass ) For each [mp21 | miR119894] complex
processes considered were (i) the formation of the complexby a miR
119894and the p21 mRNA (119896complex
119894
ass ) and (ii) the complexdegradation (119896complex
119894
deg ) For p21 processes considered were(i) its synthesis (119896p21syn ) and (ii) its degradation (119896p21deg) Anadditional algebraic equation accounting for the total mea-surable amount of p21 mRNA (mp21Total) was also includedThe SBML file of the model is available for download athttpwwwsbiuni-rostockdeuploadstx templavoilap21TargetHub 03092013xml
(2) Model Calibration and Validation For model calibrationwe fixed some parameter values using published data andestimated the other unknown parameters using the time-series data published by Wu et al [23] in which the p21mRNA (northern plot) and protein levels (western plot)were measured 48 hr after transfection of individual p21-targeting miRNAs into human embryonic kidney 293 cellsThe unknown parameter values were estimated using aniterative method combining global (particle swarm patternsearch) [43] and local (downhill simplex method in multi-dimensions) [44] optimization algorithms For each miR
119894
considered in the model the method minimizes the distancebetween model simulations and experimental data using thefollowing cost function
119865miR119894cost =
[mp21miR119894sim (119905) minusmp21miR119894
exp (119905)]2
(120575mp21119894
)2
+
[p21miR119894sim (119905) minus p21miR119894
exp (119905)]2
(12057511990121
119894)2
119894 isin [1 15]
(8)
BioMed Research International 7
1 2 4 1 2 4 5 6 1 2 3 7
E2F TP53 RELA1 2 4 7 2 3 7 6
STAT5 SRF1 2 3 4 1 2
JUN STAT1
SP1 RUNX1 TFAP2A
Common TFs Modulation
CDKN1A
1 2 1 2 4 1 4
1 4 1 2 3 7
1 2 5 2 4 5 7 1
VDR
CUX1
E2A
RAR BRCA1
TP63 TP73 TBX2
STAT6 SP3 STAT3
STAT2 RUNX2
EHMT2 p21 TFs
1 2 4 5 8 9
CEBP120573NF120581B1
(1) Cell proliferation (6) Cell senescence
(2) Apoptosis (7) Inflammatory response
(3) Immune response (8) DNA repair
(4) Cell cycle control (9) Cell migration
(5) DNA damage
miRNAmiRNA
Translational repressionmiRNA
Translation
Transcription
Association
miR-132 miR-28-5p
miR-299-5p miR-298
miR-345 miR-125a-5p
miR-208a
miRNA
miRNADeadenylation followed by decay
Protein-protein interaction
miR-93 miR-654-3p
miR-657 miR-572
miR-423-3p miR-363
miR-639 miR-515-3p
p21
CCND2 CDK3 GMNN CCNE1 POLD2 CSNK2B C1orf123 CCDC85B
PIM1 CCNB2 KIAA0432 GADD45B HDAC11 ITGB1BP3 TTLL5 SLC25A11
GADD45G CDC7 BCCIP CDK1 GNB2LI RAB1A psmA3 TK1
CDC45 CCNB1 CCNA2 GADD45A MAPK8 TSG101 SET CIZ1
CDK14 CDK5 MDM2 CDC27 RNF144B ESR1 PRKCH RPA1
CDC20 SKP2 CDK2 CDK4 DHX9 PCNA CDC5L CCDC5
CCND3 CDK6 CCNE2 CCNA1 TEX11 MAP3K5 MCM10 Caspase-3
CDC6 CCND1 XRCC6 PARP1 DAPK3 NR4A1 CEBPA
Interacting proteins
AKT1
miRNA genes
5 8 1 2 1 2 4 3 1 2 3 3 2 2 3 3 1 4 1 3 1 4
1 4
NFATC2 YY1 CREB1 CEBPA ARNT
FOXCI1
ELK1 SOX9 FOS PAX5 MAX MEF2A RFX1 JUNB AKT1 GATA3 MYC NRSFEGR1
E47 PPARG POU3F2 FOXF2FOXL1 MZF1 NFIC SRY
FOXC1
HNFA4 NFYA MYF1 BACF1 ATF6 ZID RORA POU2F1
miRNA TFs
11 2 34
12 31 5 11
HNF1A
1 2
RB1
Cellular processes
Figure 1 p21 regulatory network The network contains several layers of regulators of p21 TFs (light blue and red boxes) miRNAs (darkblue and green boxes) and proteins (grey boxes) In each big box there are small boxes which represent individual components of this layerof regulation miRNAs are classified into two groups according to the mechanisms by which the expression of p21 is repressed One groupcauses p21 translation repression (green box) These miRNAs bind to p21 mRNA resulting in the repressed translation of p21 but unchangedmRNA expression level The other group of miRNAs (dark blue box) decreases the stability of p21 mRNA by modifying its structure leadingto mRNA decay and finally the downregulation of p21 TFs are classified into three groups p21 TFs (red) miRNA TFs (yellow) and theircommon TFs (light blue) The p21 interacting-proteins are framed in the grey boxes The purple boxes represent nine processes and the TFsassociated with these processes are indicated in the ellipses above them using corresponding figures This data is adapted from our previouspublication [4]
where mp21miR119894sim (119905) and mp21miR119894
exp (119905) represent the simulatedp21 mRNA and protein expression levels for each miR
119894at
time point 119905 p21miR119894sim (119905) and p21miR119894
exp (119905) represent the mea-sured value for each miR
119894at time point 119905 and their standard
deviations are 120575mp21119894
and 120575p21119894
Here 119905 is the time point (48 hr)after overexpression of the individual miRNAs in embryonickidney 293 cells at which the expression levels of the p21 andits mRNA were measured [23] The model calibration resultsare shown in Figure 2(a) and the obtained parameter valuesare listed in Table 3
Experimental results showed that a stronger repressionof the target gene can occur when two miRNA binding siteson the target mRNA are in close proximity [24 45] To testthe consequences of this hypothesis we predicted cooperativemiRNA pairs for p21 with seed site distances between 13ndash35 nt in the p21 31015840 UTR To substantiate the cooperativeeffect associated with pairs of miRNAs we introduced agroup of new state variables ([mp21 | miR
119894| miR
119894]) into
the original model These state variables account for the
ternary complexes composed of p21 mRNA and two puta-tively cooperating miRNAs (miR
119894and miR
119895) For these new
variables processes considered are (i) the association of p21mRNA with miR
119894and miR
119895into a complex (119896co-complex
119894119895
ass )and (ii) the degradation of the complex (119896co-complex
119894119895
deg ) Afterexpansion the corresponding modified and new ODEs arelisted below
dmp21119889119905
= 119896mp21syn sdot 119891act (TFmp21)
minusmp21 sdot (119896mp21deg +sum
119894
119896complex
119894
ass sdotmiR119894
+sum
119894119895
119896co-complex
119894119895
ass sdotmiR119894sdotmiR119895)
(9)
8 BioMed Research InternationalTa
ble2Th
econ
fidence
scores
forthe
interactions
inp21regulatorynetworkTh
enetworkconsistso
ffou
rtypes
ofinteractionsm
iRNA-
p21TF
-miRNAT
F-p21andp21-p
rotein
(a)miRNA-
p21interactio
nscores
miRNA
miR-125a-5p
miR-132
miR-208a
miR-28-5p
miR-298
miR-299-5p
miR-345
miR-363
Score
073
073
073
080
073
080
073
073
miRNA
miR-423-3p
miR-515-3p
miR-572
miR-639
miR-654-3p
miR-657
miR-93
Score
085
073
073
073
073
080
095
(b)TF
-miRNAinteractionscores
Score
Score
Score
Score
Score
Score
miR-208
miR-132
miR-657
miR-125a-5p
miR-28-5p
miR-93
MAX
027
EGR1
093
TFAP2
A034
EGR1
086
MZF
1028
TFAP2
A034
EGR1
029
CREB
1086
SP1
034
PAX5
024
E47
027
SOX9
024
ARN
T027
ARE
B6027
MZF
1034
NRS
F024
FOXC
1031
GAT
A3
027
RFX1
024
E47
024
MAX
029
ZID
024
FOXI1
024
MZF
1031
ATF6
024
ELK1
027
USF1
027
PPARG
027
SRY
027
SP1
034
YY1
027
EGR2
029
EGR1
034
GAT
A3
034
JUN
029
POU2F1
024
JUNB
024
FOS
024
RELA
029
MZF
1034
POU2F1
024
EGR1
029
POU2F1
024
NFIC
024
miR-654-3p
TP53
034
SRF
027
CEBP
A024
NFIC
027
RELA
027
YY1
027
NFIC
034
CEBP
A029
NFY
A024
miR-345
miR-423-3p
BACH
1024
miR-299-5p
RORA
024
RUNX1
024
SP1
034
BACH
1027
FOXL
1029
CREB
1024
miR-363
STAT
1024
RELA
027
STAT
1027
SRY
024
SRY
027
E2F1
062
E2F1
085
NFA
TC2
024
POU3F2
024
NFA
TC2
024
SRF
029
FOXC
1033
MYC
084
HNF4
A031
STAT
5A027
FOS
024
RUNX1
027
MZF
1034
miR-298
NFY
A024
SRF
027
POU2F1
027
miR-572
miR-639
JUNB
031
POU2F1
024
MEF
2A027
NF120581
B1024
FOXF
2024
NFY
A024
NFIC
031
POU2F1
024
HNF1A
024
(c)TF
-p21
interactionscores
TF(verified)
SP1
SP3
E2F1
Runx
1Ru
nx2
STAT
1E2
AST
AT5
TP53
TP63
TP73
Score
100
100
077
077
077
072
072
067
067
067
067
TF(verified)
STAT
3CU
X1TF
AP2
ABR
CA1
VDR
RARA
CEB
P120572CEB
P120573RB
1Tb
x2EH
MT2
Score
066
066
060
060
060
060
060
060
060
060
060
TF(putative)
NF120581
B1RE
LAST
AT2
STAT
6SR
FScore
044
044
044
044
043
(d)p21-p
rotein
interactionscores
Protein
TP53
PCNA
CASP
3CC
NA1
CCND1
SKP2
BCCI
PCC
NA2
CCND2
CCNE2
AKT
1C1
orf123
CCDC8
5BCC
NB1
CCNB2
CCND3
Score
100
092
087
087
087
087
081
081
081
081
072
072
072
072
072
072
ProteinCC
NE1
CDC4
5CD
C5L
CDC6
CDC7
CDK1
CDK14
CDK2
CDK3
CDK4
CDK6
CEBP
ACI
Z1CS
NK2
A1
CSNK2
BDAPK
3Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
ESR1
GADD45A
GADD45B
GADD45G
GMNN
GNB2
L1HDAC
11ITGB1BP
3MAP3
K5MAPK
8MCM
10PA
RP1
PIM1
POLD
2PS
MA3
RAB1A
Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
SET
SLC2
5A11
STAT
3TE
X11
TK1
TSG101
TTLL
5XR
CC6
CDC2
0CD
C27
CDK5
DHX9
MDM2
NR4
A1
PRKC
HRP
A1
Score
072
072
072
072
072
072
072
072
056
056
056
056
056
056
056
056
Verifi
edexp
erim
entally
verifi
edinteraction
putativ
epredictedinteraction
BioMed Research International 9
Con
trol
miR
-298
miR
-208
am
iR-1
32m
iR-2
8-5p
miR
-125
-5p
miR
-299
-5p
miR
-345
miR
-93
miR
-423
-3p
miR
-515
-3p
miR
-363
miR
-657
miR
-639
miR
-572
miR
-654
-3p
0
05
1
15m
RNA
expr
essio
n (a
u)
Data Model Data ModelC
ontro
lm
iR-2
98m
iR-2
08a
miR
-132
miR
-28-
5pm
iR-1
25-5
pm
iR-2
99-5
pm
iR-3
45m
iR-9
3m
iR-4
23-3
pm
iR-5
15-3
pm
iR-3
63m
iR-6
57m
iR-6
39m
iR-5
72m
iR-6
54-3
p
0
02
04
06
08
1
Prot
ein
expr
essio
n (a
u)
(a)
DoxorubicinCellWell
Control miR-572 miR-93
Measure p21 expression at the time points of
miR-572+ miR-93
0 2 4 6 8 24hr
miR-572miR-93
Immunoblotting
(b)
0 4 8 12 16 20 240
4
8
12
16
20
Time (hr)
p53
Actin
Doxo (hrs)
fac
t(TF
mp2
1)
C 0 1 2 4 6 8 24
1 08 12 19 29 32 3 16
(c)
0 2 4 6 8 2402468
1012
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)
p21
expr
essio
n (a
u)
02468
1012
miR-572Control
02468
1012
miR-93
02468
1012
Model Data
0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24
1 23 46 107124 124 23 28 48 66 53 31 08 16 21 38 50 18 06 06 12 25 27 12
miR-572+ 93Doxo (hrs)
p21
Actin
p21
Actin
p21
Actin
p21
Actin
(d)
Figure 2 Model calibration and validation (a) Model calibration The figures show the relative change of the p21 mRNA and proteinexpression levels after overexpression of the indicated miRNAs (Model model simulation Data experimental data) These data werenormalized to the control group in which the p21 mRNA and protein expression levels were measured when the miRNAs were normallyexpressed (au arbitrary unit) (b) Experimental workflow In the experiments Sk-Mel-147 cells were seeded in six well plates Then maturemiRNA mimics were transfected individually at a concentration of 100 nM (miR-572 and miR-93) or in combination at 50 nM each (miR-572 + miR-93) After 48 hr transfection with miRNA mimics the cells were pulse treated with 250 nM doxorubicin for 1 hour after whichnormal growth medium was replenished The immunblotting were performed to measure p21 expression at 0 2 4 6 8 and 24 hr post-doxorubicin treatment (c) Temporal dynamics of p21 transcriptional function Afterdoxorubicin treatment the expression of p53 a TF ofp21 was measured using immunblotting and these data were used to characterize the transcriptional function of p21 using MATLAB linearinterpolation function (d) Model validation We measured the expression of p21 protein in response to genotoxic stress in the four scenariosas described in the main text The measured data (Data) were compared with the model simulations (Model) The figures (a) (c) and (d) areadapted from our previous publication [4]
119889 [mp21 | miR119894| miR
119895]
119889119905
= 119896co-complex
119894119895
ass sdotmp21 sdotmiR119894sdotmiR119895
minus 119896co-complex
119894119895
deg sdot [mp21 | miR119894| miR
119895]
(10)
To model stronger repression of the target gene bycooperating miRNAs we assumed a stronger association rateconstant for the complex [mp21 | miR
119894| miR
119895] which is
equal to the sum of their individual association rate constants
(119896co-complex119894119895
ass = 119896complex
119894
ass + 119896complex
119895
ass ) Similarly the degradationrates of the complexes [mp21 | miR
119894| miR
119895] were assumed
to be equal to the sum of degradation rate constants ofsingle miRNA binding complexes (119896co-complex
119894119895
deg = 119896complex
119894
deg +
119896complex
119895
deg ) However it has to be noted that these addedequations are an abstract description ofmiRNAcooperativitybecause the details of this mechanism are not yet known
To experimentally validate the capability of our model topredict the relative p21 concentrations regulated by coopera-tivemiRNAs we selectedmiR-572 andmiR-93 as a case study
10 BioMed Research International
Table 3 Initial concentrations of model variables and model parameter values Based on the experimental data the p21-targeting miRNAsverified byWu et al [23] were divided into two groups the translation repression group (marked with asterisk) and the mRNA deadenylationgroup AmiRNAwas classified into themRNA deadenylation group if its overexpression can result in 20 ormore downregulation of the p21mRNA level (ie p21mRNA levelle 08 the basal level is 1) otherwise it was classified into the translation repression group For the translationrepression group only 119896complex119894
ass was estimated and 119896complex119894deg was fixed For the other group both 119896complex119894
deg and 119896complex119894ass were estimatedThe initial
concentrations of p21 and mp21 were set to 1 and this value was used as their basal expression levels During the parameter estimation theinitial concentrations of p21-targeting miRNAs were set to 100 because in the publication [23] the expression levels of p21 and mp21 weremeasured after the individual introduction of the p21-targeting miRNAs with amount of 100 nM Due to the lack of biological information tocharacterize the transcriptional activation function (119891act) of p21 and its targeting miRNAs the corresponding functions were assumed to be1 for simplicity The data is adapted from our previous publication [4]
Initial concentration of variables and TF functionsVariable Description Initial concentration (au)p21 p21 protein 1
mp21 p21 mRNA 1
miR119894=115
p21-targeting miRNAs 100
[mp21 | miR119894=115
] Complexes formed by miR119894and mp21 0
119891act(TFmp21) p21rsquos transcriptional activation function 1119891act(TFmiR119894 (119894=115)) The transcriptional activation function of miR
1198941
Fixed parameter valuesParameter Description Value (hrminus1) Reference119896mp21syn Synthesis rate constant of mp21 01155 fixed119896mp21deg Degradation rate constant of mp21 01155 [24]119896miR119894syn (119894 = 1 15) Synthesis rate constant of miR
11989400289 fixed
119896miR119894deg (119894 = 1 15) Degradation rate constant of miR
11989400289 [25]
119896p21syn Synthesis rate constant of p21 13863 fixed119896p21deg Degradation rate constant of p21 13863 [26]
Estimated parameter values
miRNA (statevariable)
119896complex119894deg
(119894 = 1 15) (hrminus1)
119896complex119894ass
(119894 = 1 15)
(au minus1sdot hrminus1)
119865miR119894cost
(119894 = 1 15)Experimental data of p21 (protein mRNA plusmn SD)
miR-298 (miR1)lowast 01155 00254 34119890 minus 004 (016 1074 plusmn 0025)
miR-208a (miR2)lowast 01155 00041 20119890 minus 003 (051 1192 plusmn 0022)
miR-132 (miR3)lowast 01155 00275 24119890 minus 003 (015 121 plusmn 0147)
miR-28-5p (miR4)lowast 01155 00119 59119890 minus 003 (028 135 plusmn 006)
miR-125-5p (miR5)lowast 01155 00018 18119890 minus 003 (069 085 plusmn 0051)
miR-299-5p (miR6)lowast 01155 00080 18119890 minus 004 (036 095 plusmn 0038)
miR-345 (miR7)lowast 01155 00051 11119890 minus 004 (046 096 plusmn 0039)
miR-93 (miR8) 01564 00235 41119890 minus 014 (017 07776 plusmn 003)
miR-423-3p (miR9) 09118 00055 28119890 minus 009 (044 05102 plusmn 011)
miR-515-3p (miR10) 02098 00253 12119890 minus 013 (016 0616 plusmn 0037)
miR-363 (miR11) 02261 00399 22119890 minus 014 (011 056 plusmn 015)
mR-657 (miR12) 03465 00158 21119890 minus 014 (023 048 plusmn 012)
miR-639 (miR13) 04305 00327 18119890 minus 017 (013 036 plusmn 0084)
miR-572 (miR14) 03039 00360 94119890 minus 023 (012 045 plusmn 0044)
miR-654-3p (miR15) 97485 00024 30119890 minus 014 (064 063 plusmn 0053)
These two miRNAs were chosen because their predictedtarget sites in the p21 31015840 UTR are in close proximity to eachother and thereby they can induce cooperative repression onp21 as suggested in [24] The experiments were performed asfollows
(i) Melanoma cells (Sk-Mel-147) were transfected withthe mature miRNAmimics of the twomiRNAs eitherindividually (100 nM) or in combination (50 nMeach) whereas untreated cells were used as control(Figure 2(b))
BioMed Research International 11
p21
FOXF2
miR-572
Activation
Inhibition120572
Incoherent FFL + cascaded regulation
AF2
miR-93
(a)
miR-93
miR-572
TP53
MYC
p21
FOXF2
Coherent FFL + cascaded regulation
(b)
Saturation
Pulse
p21
(au
)
p21
(au
)Time (hr) Time (hr)
Saturation Pulse
Pulse
Saturation
Saturation
di
(au
)
102
101
102
100
101
10minus1
10010minus2
10minus1
102
101
100
10minus1
10minus2
FOXF2 (au) FOXF2 (au)
TP53
(au
)
1
09
08
07
06
05
04
03
02
01
10minus2 10210110010minus110minus20
AF2120572
(au
)
(c)
Figure 3 Different p21 dynamics in different network motifs We ran simulations to show the different dynamics of p21 for two differentnetwork motifs (a) and (b) Through simulations two dynamical patterns of p21 were identified saturation and pulse ((c) top) For eachnetwork motif the corresponding distributions of the two dynamical patterns were plotted ((c) bottom) For different combinations ofthe transcriptional strengths the normalized distance (119889
119894) between peaks (119901
119894) and steady states (ss
119894) of p21 is determined by the equation
119889119894= (119901119894minus ss119894)119901max 119901max = max(119901
1 119901
119899) If 119889
119894= 0 for the corresponding combination of transcriptional strengths the p21 dynamics is
saturation otherwise it is pulse The regions showing different dynamical patterns of p21 are separated using the white lines
(ii) Next the cells were treated with doxorubicina genotoxic-stress inducing agent The agent canupregulate the expression of p53 which is a knownTFof p21 and therefore it can result in the upregulationof p21 (Figure 2(c))
(iii) After doxorubicin treatment the expression levels ofp21 were measured by immunoblotting at differenttime points (0 2 4 6 8 24 hr) The p21 expressionvalues were normalized based on the p21 expression
level in the control group measured at time point 0 hr(Figure 2(d))
By doing so we obtained the p21 response after genotoxicstress in four scenarios (1) endogenous miRNA expression(Control) (2) overexpression of miR-572 (3) overexpressionof miR-93 and (4) both miRNAs moderately overexpressedThereafter we derived a model of seven ODEs based onthe original equations which was configured according tothe designed experiments making the simulation results
12 BioMed Research International
00001
0001
001
01
1Cell cycle Cell growth DNA damage Senescence Apoptosis
G1G2
MS
p21
expr
essio
n (a
u)
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
log10
wow
Figure 4 p21 expression regulated by cooperative miRNAs for different cellular processesThe associations of themiRNAs with these cellularprocesses were derived from GO terms of their TFs A miRNA was supposed to be expressed (in bold black font) in a cellular process onlyif its TF is related to the corresponding GO term of this process The p21 expression levels are computed for each process with (w)without(wo) considering the cooperative effect among the p21 targeting miRNAs
comparable with the experimental data As shown in theFigure 2(c) the simulations are in good agreement with theexperimental observations The individual overexpression ofmiR-572 or miR-93 led to the reduction of the upregulationof p21 response after genotoxic stress induction The twomiRNAs cause different degrees of repression due to theirdifferent repression efficiencies on p21 Interestingly the com-bined overexpression of both miRNAs induced the strongestdownregulation of p21 and therefore verifying the hypothesisof their cooperative regulation of p21 Above all the resultsnot only validated themodel but also demonstrated the abilityof our method to identify cooperative miRNA pairs for p21
(3) Predictive Simulations As there are abundant of networkmotifs such as FFLs in p21 regulatory network and thesenetwork motifs are important for determining p21 dynamicsit is interesting to investigate the dynamics of p21 in networkmodules where FFLs are involved To do so two networkmodules including both miR-93 and miR-572 and theirTFs were exemplified In Figure 3(a) the network modulecontains an incoherent FFL composed by AF2120572 miR-93 andp21 and a cascaded regulation in which p21 is repressed
by FOXF2 via miR-572 in Figure 3(b) the same cascadedregulation together with a coherent FFL composed of TP53MYC miR-93 and p21 forms another regulatory moduleof p21 By modulating the transcriptional strengths of thetwo miRNAs by their TFs two types of p21 dynamics wereidentified saturation and pulse (Figure 3(c) top) In theformer the p21 expression increases and reaches its steadystate at the highest level in contrast in the latter the p21expression increases to a peak and thereafter drops to asteady state at lower level For the two different networkmodules various combinations of transcriptional strengthsof the two miRNAs lead to different distributions of thetwo p21 dynamical patterns For the network module of anincoherent FFL plus the cascaded regulation (Figure 3(c)bottom left) the saturation pattern appears in two distinctregions one with weakened transcriptional strength of miR-93 and the other with enhanced transcriptional strength ofmiR-93 for the other module (Figure 3(c) bottom right)the saturation pattern only appears in the region in whichthe transcriptional strength of miR-93 is enhanced Takentogether the results showed that for the two different networkmotifs the dynamical pattern of p21 is changing according to
BioMed Research International 13
different combinations of its upstream regulators suggestingthe adaptation of p21 dynamics for different biological con-texts
Furthermore we performed a number of simulations toshow the influence of miRNA regulation on p21 expressionlevels in different cellular processes arranged in a consecutivemanner In this procedure a cell is first in the process ofcell cycle followed by proliferation (cell growth) then thecell responses to DNA damage and enters into the processof senescence and finally apoptosis is initiated As shown inFigure 4 during the cell cycle process the p21 expression islow due to the activation of most of its targeting miRNAswhen the cell starts proliferating the p21 expression declinesto an even lower level because of more activated miRNAsin this process after responding to DNA damage the p21expression soars to a high level which is caused by theactivation of its TFs like p53 and fewer expressed miRNAsalthough the p21 expression keeps at a similar level whilethe cell is undergoing senescence the expressed miRNAs aredifferent from the previous process finally the p21 expressiondecreases again to a low level due to the reemergence ofmost of its targeting miRNAs and the cell enters apoptosisInterestingly the model simulations are also consistent withexperimental observations under non-stressed condition thelow expression level of p21 is needed for cell proliferation theupregulation of p21 happens after response to DNA damagevia p53 and the increased p21 expression further results incell cycle arrest leading to senescence and apoptosis [25]Besides when considering the effect of cooperatingmiRNAsthe p21 expression levels were indistinguishable from theprevious simulation of DNA damage response However forthe other processes p21 expression levels were significantlylower compared to the simulations without considering thecooperative effect of the miRNAs Above all these resultsindicated that selective expression of cooperative miRNAscould be adopted by cells to ensure diverse expression levelsof p21 tomeet the requirements of different cellular processes
3 Conclusions and Discussion
In this paper we presented a systems biology approach com-bining data-drivenmodeling andmodel-driven experimentsto investigate the role of miRNA-mediated repression in generegulatory networksThis approach provides a systematicwayto gain a deeper understanding of the regulation of targetgenes by mutiple and cooperative miRNAs Using the regu-lation of p21 by multiple miRNAs as a case study we showedhow the ODE-basedmodel which is calibrated and validatedby means of experimental data is suitable for predicting thetemporal dynamics of molecular concentrations involved inbiochemical systems
Provided there are sufficiently rich quantitative data setsavaliable to characterize the model the use of the methodol-ogy here shown can be extended to more complex regulatorynetworks involving multiple targets cooperating TFs andmiRNAs and signaling pathways displaying cross-talk viapost-tranlational modifications In this case the criticalelement is the quality and quantitiy of the available data
Insufficiency and low quality of experimental data can causeerrors in the process of model construction and overfitting inparameter estimation can lead to uncertainties in the modelpredictions We believe that the quick development of quan-titative high throughput techniques such as transcriptomicsproteomics and miRNomics will facilitate the constructionand characterization of larger miRNA-mediated regulatorynetworks
Other modeling frameworks than ODE-based modelscan be used to describe biological systems such as prob-abilistic (eg Bayesian) or logical (eg Boolean) modelsImportantly different modeling frameworks have differentproperties and perform well regarding different perspectivesand levels of mechanistic details of biochemical systems [46]For example Bayesian models are helpful in the constructionof connections in signaling networks and can reveal the mostlikely underlying structure of the network in a probabilisticmanner Boolean models use binary values (0 and 1) andlogical gates (AND NOT and OR) to describe activities ofnetwork components and the information flow among themWe believe that in the coming future hybrid models whichconsist of modeling framework and experimental techniquespecific sub-modules will provide the necessary compromisebetween quantitativequalitative accuracy and scalability forthe investigation of large biochemical networks [47]
Disclosures
The authors declare that they have no competing financialinterests
Authorsrsquo Contribution
Julio Vera and Olaf Wolkenhauer are equal contributers
Acknowledgments
The authors would like to acknowledge the funding fromthe German Federal Ministry of Education and Research(BMBF) eBio-miRSys (0316175A to Xin Lai and Julio Vera)eBio-SysMet (0316171 to OlafWolkenhauer) and Gerontosys-ROSAge (0315892A to Olaf Wolkenhauer)
References
[1] A E Pasquinelli ldquoMicroRNAs and their targets recognitionregulation and an emerging reciprocal relationshiprdquo NatureReviews Genetics vol 13 no 4 pp 271ndash282 2012
[2] J Vera X Lai U Schmitz and O Wolkenhauer ldquoMicroRNA-regulated networks the perfect storm for classical molecularbiology the ideal scenario for systems biologyrdquo Advances inExperimental Medicine and Biology vol 774 pp 55ndash76 2013
[3] S Nikolov J Vera U Schmitz and O Wolkenhauer ldquoA model-based strategy to investigate the role of microRNA regulation incancer signalling networksrdquo Theory in Biosciences vol 130 no1 pp 55ndash69 2011
14 BioMed Research International
[4] X Lai U Schmitz S K Gupta et al ldquoComputational analysisof target hub gene repression regulated by multiple and cooper-ative miRNAsrdquoNucleic Acids Research vol 40 no 18 pp 8818ndash8834 2012
[5] X Lai O Wolkenhauer and J Vera ldquoModeling miRNA regu-lation in cancer signaling systems miR-34a regulation of thep53Sirt1 signaling modulerdquoMethods in Molecular Biology vol880 pp 87ndash108 2012
[6] C Jiang Z Xuan F Zhao and M Q Zhang ldquoTRED atranscriptional regulatory element database new entries andother developmentrdquoNucleic Acids Research vol 35 supplement1 pp D137ndashD140 2007
[7] L A Bovolenta M L Acencio and N Lemke ldquoHTRIdban open-access database for experimentally verified humantranscriptional regulation interactionsrdquo BMCGenomics vol 13article 405 2012
[8] VMatys O V Kel-Margoulis E Fricke et al ldquoTRANSFAC andits module TRANSCompel transcriptional gene regulation ineukaryotesrdquo Nucleic Acids Research vol 34 database issue ppD108ndashD110 2006
[9] D Karolchik R Baertsch M Diekhans et al ldquoThe UCSCgenome browser databaserdquo Nucleic Acids Research vol 31 no1 pp 51ndash54 2003
[10] P Alexiou T Vergoulis M Gleditzsch et al ldquomiRGen 20 adatabase of microRNA genomic information and regulationrdquoNucleic Acids Research vol 38 supplement 1 pp D137ndashD1412010
[11] F Xiao Z Zuo G Cai S Kang X Gao and T Li ldquomiRecordsan integrated resource for microRNA-target interactionsrdquoNucleic Acids Research vol 37 supplement 1 pp D105ndashD1102009
[12] S-D Hsu F-M Lin W-Y Wu et al ldquomiRTarBase a databasecurates experimentally validated microRNA-target interac-tionsrdquo Nucleic Acids Research vol 39 supplement 1 pp D163ndashD169 2011
[13] P Sethupathy B Corda and A G Hatzigeorgiou ldquoTarBase acomprehensive database of experimentally supported animalmicroRNA targetsrdquo RNA vol 12 no 2 pp 192ndash197 2006
[14] H Dweep C Sticht P Pandey and N Gretz ldquomiRWalkmdashdatabase prediction of possible miRNA binding sites byldquowalkingrdquo the genes of three genomesrdquo Journal of BiomedicalInformatics vol 44 no 5 pp 839ndash847 2011
[15] T S Keshava Prasad RGoel K Kandasamy et al ldquoHumanpro-tein reference databasemdash2009 updaterdquo Nucleic Acids Researchvol 37 supplement 1 pp D767ndashD772 2009
[16] D Szklarczyk A Franceschini M Kuhn et al ldquoThe STRINGdatabase in 2011 functional interaction networks of proteinsglobally integrated and scoredrdquo Nucleic Acids Research vol 39supplement 1 pp D561ndashD568 2011
[17] D Croft G OrsquoKelly G Wu et al ldquoReactome a database ofreactions pathways and biological processesrdquo Nucleic AcidsResearch vol 39 supplement 1 pp D691ndashD697 2011
[18] N le Novere M Hucka H Mi et al ldquoThe systems biologygraphical notationrdquoNature Biotechnology vol 27 no 8 pp 735ndash741 2009
[19] A Funahashi Y Matsuoka A Jouraku M Morohashi NKikuchi and H Kitano ldquoCellDesigner 35 a versatile modelingtool for biochemical networksrdquo Proceedings of the IEEE vol 96no 8 pp 1254ndash1265 2008
[20] M S Cline M Smoot E Cerami et al ldquoIntegration ofbiological networks and gene expression data using CytoscaperdquoNature Protocols vol 2 no 10 pp 2366ndash2382 2007
[21] D-H Le andY-KKwon ldquoNetDS aCytoscape plugin to analyzethe robustness of dynamics and feedforwardfeedback loopstructures of biological networksrdquo Bioinformatics vol 27 no 19pp 2767ndash2768 2011
[22] E O Voit Computational Analysis of Biochemical SystemsA Practical Guide for Biochemists and Molecular BiologistsCambridge University Press Cambridge UK 2000
[23] S Wu S Huang J Ding et al ldquoMultiple microRNAs modulatep21Cip1Waf1 expression by directly targeting its 31015840 untranslatedregionrdquo Oncogene vol 29 no 15 pp 2302ndash2308 2010
[24] P Saeligtrom B S E Heale O Snoslashve Jr L Aagaard J Alluin andJ J Rossi ldquoDistance constraints between microRNA target sitesdictate efficacy and cooperativityrdquo Nucleic Acids Research vol35 no 7 pp 2333ndash2342 2007
[25] Y-S Jung YQian andXChen ldquoExamination of the expandingpathways for the regulation of p21 expression and activityrdquoCellular Signalling vol 22 no 7 pp 1003ndash1012 2010
[26] W Wang H Furneaux H Cheng et al ldquoHuR regulatesp21 mRNA stabilization by UV lightrdquo Molecular and CellularBiology vol 20 no 3 pp 760ndash769 2000
[27] U Wittig R Kania M Golebiewski et al ldquoSABIO-RKmdashdatabase for biochemical reaction kineticsrdquo Nucleic AcidsResearch vol 40 database issue pp D790ndashD796 2012
[28] R Milo P Jorgensen U Moran G Weber and M SpringerldquoBioNumbersmdashthe database of key numbers in molecular andcell biologyrdquo Nucleic Acids Research vol 38 supplement 1 ppD750ndashD753 2010
[29] I-C Chou and E O Voit ldquoRecent developments in parameterestimation and structure identification of biochemical andgenomic systemsrdquoMathematical Biosciences vol 219 no 2 pp57ndash83 2009
[30] A Saltelli K Chan and E M Scott Sensitivity Analysis JohnWiley amp Sons New York NY USA 1st edition 2000
[31] S Strogatz Nonlinear Dynamics and Chaos Applications toPhysics Biology Chemistry and Engineering With Applicationsto Physics Biology Chemistry and Engineering Westview PressBoulder Colo USA 2000
[32] P Zhou S Cai Z Liu and R Wang ldquoMechanisms generatingbistability and oscillations in microRNA-mediated motifsrdquoPhysical Review E vol 85 no 4 apart 1 Article ID 041916 9pages 2012
[33] S Marino I B Hogue C J Ray and D E Kirschner ldquoAmethodology for performing global uncertainty and sensitivityanalysis in systems biologyrdquo Journal of Theoretical Biology vol254 no 1 pp 178ndash196 2008
[34] L P LimNC Lau PGarrett-Engele et al ldquoMicroarray analysisshows that some microRNAs downregulate large numbers of-target mRNAsrdquo Nature vol 433 no 7027 pp 769ndash773 2005
[35] J Krutzfeldt N Rajewsky R Braich et al ldquoSilencing ofmicroRNAs in vivowith lsquoantagomirsrsquordquoNature vol 438 no 7068pp 685ndash689 2005
[36] MWPfafflGWHorgan andLDempfle ldquoRelative expressionsoftware tool (REST) for group-wise comparison and statisticalanalysis of relative expression results in real-time PCRrdquoNucleicAcids Research vol 30 no 9 p e36 2002
[37] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[38] T Abbas and A Dutta ldquop21 in cancer intricate networks andmultiple activitiesrdquo Nature Reviews Cancer vol 9 no 6 pp400ndash414 2009
[39] J Wang M Lu C Qiu and Q Cui ldquoTransmiR a transcriptionfactor microRNA regulation databaserdquo Nucleic Acids Researchvol 38 supplement 1 pp D119ndashD122 2010
BioMed Research International 15
[40] S Bandyopadhyay andM Bhattacharyya ldquoPuTmiR a databasefor extracting neighboring transcription factors of humanmicroRNAsrdquo BMC Bioinformatics vol 11 article 190 2010
[41] A Le Bechec E Portales-Casamar G Vetter et alldquoMIRNTN a framework integrating transcription factorsmicroRNAs and their targets to identify sub-network motifs ina meta-regulation network modelrdquo BMC Bioinformatics vol12 article 67 2011
[42] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyTheGene Ontology ConsortiumrdquoNature Genetics vol 25 no 1 pp 25ndash29 2000
[43] A I F Vaz and L N Vicente ldquoA particle swarm pattern searchmethod for bound constrained global optimizationrdquo Journal ofGlobal Optimization vol 39 no 2 pp 197ndash219 2007
[44] W H PressNumerical Recipes The Art of Scientific ComputingCambridge University Press Cambridge UK 2007
[45] J G Doench and P A Sharp ldquoSpecificity of microRNA targetselection in translational repressionrdquo Genes and Developmentvol 18 no 5 pp 504ndash511 2004
[46] B Kholodenko M B Yaffe and W Kolch ldquoComputationalapproaches for analyzing information flow in biological net-worksrdquo Science Signaling vol 5 no 220 2012
[47] F M Khan U Schmitz S Nikolov et al ldquoHybrid modeling ofthe crosstalk between signaling and transcriptional networksusing ordinary differential equations and multi-valued logicrdquoBiochimica et Biophysica Acta 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
2 BioMed Research International
optimizationmethods (III)Computational experiments oncea model is established it can be simulated andor analyzedfor its general behavior (IV) Experimental validation modelpredictions together with biological explanations are inte-grated to guide the design of new experiments which inturn validate or falsify the model If model predictions arein agreement with the experiments the model justifies thebiological hypotheses behind it In turn these hypotheseswhich provide reasonable explanations for the biologicalphenomenon lead to an enhanced understanding of thegene regulatory network Otherwise the structure of themathematical model is modified to generate new hypothesesand suggest new experiments
The application of the systems biology approach to theanalysis of a gene regulatory network is demonstrated witha case study of the regulation of p21 by multiple miRNAs [4]The network combining putative targets of TF and miRNAregulation with experimentally proven molecular interac-tions was constructed and visualized Next the network wastranslated into a detailed mechanistic model which wascharacterized and validated with experimental data Finallythe integration of quantitative data and modeling helped usto generate and validate hypotheses about mechanisms ofcollective miRNA repression on p21
2 Results
21 The Systems Biology Approach to Study miRNA-MediatedGene Regulatory Networks
211 Mathematical Modeling The aim of our analysis is tounravel the complex mechanisms by which gene regulatorynetworks involving miRNAs are regulated We iterativelyintegrate data from literature experiments and biologicaldatabases into a detailed mechanistic model of a gene regula-tory network The model is then used to formulate and testhypotheses about mechanisms of miRNA target regulationand cellular process-related variability The methodologyincludes four steps which are briefly summarized in Table 1In the coming sections we present these steps in detail
(1) Data Retrieval To construct a gene regulatory networkcomposed by different levels of regulation we collect infor-mation from different resources which are briefly describedbelow and more resources for data retrieval are introducedin Table 1
(a) Transcriptional level regulators Experimentally veri-fied TFs for a gene can be extracted from literatureor databases such as TRED TRANSFAC or HTRIdb[6ndash8] the putative TFs which are associated withconserved TF-binding sites residing in the promoterregion of a gene can for example be extracted fromthe table of TFs with conserved binding sites in theUCSC genome browser or the TRANSFAC database[9] miRGen 20 is a database that provides bothpredicted and experimentally verified informationabout miRNA regulation by TFs [10]
(b) Post-transcriptional regulators Databases such asmiRecords Tarbase and miRTarBase are repositoriesof experimentally validatedmiRNAgene interactions[11ndash13] Predictions of miRNAgene interactions areaccumulated in databases like miRWalk [14]
(c) Protein-protein interactions Both the Human ProteinReference Database (HPRD) [15] and the STRINGdatabase [16] document experimentally verifiedprotein-protein interactions besides STRING alsoprovides putative protein-protein interactions rankedby confidence scores Further details about the exactmechanism of protein-protein interactions can befound in Reactome [17]
(2) Network Construction and Visualization Based on theinformation collected a gene regulatory network is con-structed and visualized for providing an overview For thispurpose we recommend CellDesigner which uses standard-ized symbols (Systems Biology Graphical NotationmdashSBGN)[18] for visualization and stores gene regulatory networksin the SBML format (Systems Biology Markup Language)[19] CellDesigner also provides the possibility to simulatetemporal dynamics of the gene regulatory network due tothe integration of the SBML ODE (ordinary differentialequation) solver Besides Cytoscape is another powerful toolfor integration of biological networks and gene expressiondata [20]
For assessing the reliability of interactions considered ingene regulatory networks confidence scores can be computedas being documented in our previous publication [4] Thefactors that are used to determine the confidence score formolecular interactions can be the number of publicationsreporting an interaction experimental methods used toidentify an interaction interaction types and computationalpredictionsThe computed confidence scores range from 0 to1 where values towards 1 indicate higher confidence whereasvalues towards 0 indicate lower confidence in a given inter-action For example the confidence score for a miRNAgeneinteraction can be calculated using the following equation
119878miRNAgene =119908119901sdot 119878119901+ 119908119898sdot 119878119898+ 119908bs sdot 119878bs
119908119901+ 119908119898+ 119908bs
(1)
where 119908⟨119901119898bs⟩ are weights that are assigned to the scores
which account for the number of publications (119878119901) detection
method (119878119898) and the number of predicted binding sites
(119878bs) The values of the weights can be assigned based onexpert knowledge and the higher the value of the weightis the bigger impact it has on the confidence score forthe interaction The values of 119878
119901and 119878bs can be calculated
using the logarithmic equation 119878⟨119901bs⟩(119899) = log
119898+1(119899)
where 119899 denotes the number of publications describing themiRNAgene interaction or the number of binding sites thatthe miRNA has in the 31015840 UTR (untranslated region) ofthe gene The value of 119898 is a cut-off that represents thenumber of publications or binding sites required for 119878
⟨119901bs⟩ toobtaining their maximum values Various methods such aswestern blots qRT-PCR and reporter assays can be appliedto support the miRNAgene interaction but these methods
BioMed Research International 3
Table 1 Overview of themethodology Key points in each step of themethodology and themain resources for constructingmiRNA-mediatedgene regulatory networks are given
Step 1 data retrievalRegulation types Resources
Transcriptional generegulation
TRED (httprulaicshleducgi-binTREDtredcgiprocess=home) a database that provides an integratedrepository for both cis- and transregulatory elements in mammalsTRANSFAC (httpwwwgene-regulationcompubdatabaseshtml) a database that collects eukaryotictranscriptional regulation comprising data on TFs their target genes and binding sitesThe UCSC table browser (httpgenomeucscedu) a popular web-based tool for querying the UCSC GenomeBrowser annotation tablesHTRIdb (httpwwwlbbcibbunespbrhtri) an open-access database for experimentally verified humantranscriptional regulation interactionsMIRNTN (httpmaiaunilumirontonphp) an integrative resource based on a metaregulation networkmodel including TFs miRNAs and genesPuTmiR (httpwwwisicalacinsimbioinfo miuTF-miRNAphp) a database of predicted TFs for humanmiRNAsTransmiR (http20238126151hmddmirnatf) a database of validated TF-miRNA interactionsmiRGen 20 (httpdianacslabecentuagrmirgen) a database of miRNA genomic information and regulation
Posttranscriptionalgene regulation
miRecords (httpmirecordsbioleadorg) a resource for animal miRNA-target interactionsTarbase (httpwwwmicrornagrtarbase) a database that stores detailed information for each miRNA-geneinteraction the experimental validation methodologies and their outcomesmiRTarBase (httpmirtarbasembcnctuedutw) a database that collects validated miRNA-target interactionsby manually surveying the pertinent literaturemiRWalk (httpwwwummuni-heidelbergdeappszmfmirwalk) a comprehensive database that providesinformation on miRNAs from human mouse and rat on their predicted as well as validated binding sites intarget genes
Protein-proteininteraction
HPRD (httpwwwhprdorg) a centralized platform to visually depict and integrate information pertaining todomain architecture posttranslational modifications interaction networks and disease association for eachprotein in the human proteomeSTRING (httpstring-dborg) a database of known and predicted protein interactions The interactionsinclude direct (physical) and indirect (functional) associationsMPPI (httpmipshelmholtz-muenchendeprojppi) a collection of manually curated high-quality PPI datacollected from the scientific literature by expert curatorsDIP (httpdipdoe-mbiuclaedudipMaincgi) a catalog of experimentally determined interactions betweenproteinsIntAct (httpwwwebiacukintactmainxhtml) a platform that provides a database system and analysis toolsfor molecular interaction dataReactome (httpwwwreactomeorg) an open-source open access manually curated and peer-reviewedpathway database
GO annotation
Amigo GO (httpamigogeneontologyorgcgi-binamigogocgi) the official GO browser and search enginemiR2Disease (httpwwwmir2diseaseorg) a manually curated database that aims at providing acomprehensive resource of miRNA deregulation in various human diseasesmiRCancer (httpmircancerecuedu) a miRNA-cancer association database constructed by text mining on theliteraturePhenomiR (httpmipshelmholtz-muenchendephenomir) a database that provides information aboutdifferentially expressed miRNAs in diseases and other biological processesmiRGator (httpmirgatorkobicrekr) a novel database and navigator tool for functional interpretation ofmiRNAsmiRo (httpferrolabdmiunictitmiro) a web-based knowledge base that provides users withmiRNA-phenotype associations in humans
Step 2 network construction and visualization(i) Visualize regulatory interactions in platforms such as CellDesigner and Cytoscape that support standardized data formats(ii) Calculate confidence scores for assessing reliability of interactions in gene regulatory networksStep 3 model construction and calibration(i) Formulate equations using rate equations(ii) Fix parameter values using available biological information(iii) Estimate the other unknown and immeasurable parameter values using optimization methods which can minimize the distancebetween model simulations and experimental data such as time course qRT-PCR and western blot dataStep 4 model validation and analysis(i) Design new experiments and generate new data to verify the calibrated model(ii) Study complex properties and behavior of the system
4 BioMed Research International
provide experimentalists with different levels of confidencethus differing confidences can be reflected using differentvalues for 119878
119898based on the experience of experimentalists
Of note although the confidence scores cannot be directlyconverted into a mathematical model with the help of thesescores we can discard non-reliable putative interactions togenerate the ultimate version of a gene regulatory networkThe final version of the network can be further analyzedto identify regulatory motifs like feedforward loops (FFLs)for example with the help of the Cytoscape plugin NetDS[21] Thereafter the complete network or parts of it canbe converted into a detailed mechanistic model which isdescribed in detail in the following section
(3) Model Construction and Calibration After the construc-tion visualization and refinement of a gene regulatory net-work it is converted into a detailedmechanistic model whichenables the investigation of unanswered biological questionsand validation of hypotheses Ordinary differential equations(ODEs) describe how processes of synthesis biochemicalmodification andor degradation affect the temporal con-centration profile of biochemical species like proteins RNAsand metabolites An ODE model can be constructed usingappropriate kinetic laws such as the law ofmass-action whichstates that the rate of a chemical reaction is proportional tothe probability that the reacting species collideThis collisionprobability is in turn proportional to the concentration ofthe reactants [22] A general representation of ODE-basedmodels using mass-action kinetics is given by the followingequation
119889119909119894
119889119905=
119898
sum
120583=1
119888119894120583sdot 119896120583sdot
119899
prod
119895=1
119909119895
119892120583119895 119894 isin 1 2 119899 (2)
where 119909119894represents state variables which denote the molar
concentration of the 119894th biochemical specie Every biochem-ical reaction 120583 is described as a product of a rate constant(119896120583) and biochemical species (119909
119895 119895 isin 1 2 119899) that are
involved in this reaction 119888119894120583 the so-called stoichiometric
coefficients relate the number of reactant molecules con-sumed to the number of product molecules generated inthe reaction 120583 119892
120583119895denotes kinetic orders which are equal
to the number of species of 119909119894involved in the biochemical
reaction 120583 The rate constants kinetic orders and the initialconditions of state variables are defined as model parametersBesides mass-action kinetics other kinetic rate laws such asMichaelis-Menten kinetics Hill equation and power-laws arealso frequently used in ODE models
After ODEs are formulated the model requires to becalibrated a process by which parameter values are adjustedin order to make model simulations match experimentalobservations as good as possible To do so there are twopossible means characterization of parameter values usingavailable biological information or estimation of parametervalues using optimization methods Some parameter valuescan be directly measured or obtained from literature ordatabases For example the half-life (119905
12) of some molecules
(eg protein) can be measured in vitro via western blot-ting This information can be used to characterize their
degradation rate constants through the equation 119896deg =
(ln 211990512) The database SABIO-RK provides a platform for
modelers of biochemical networks to assemble informationabout reactions and kinetic constants [27] However formost model parameters whose values cannot be measuredin laboratories or be accessed from literature or databasesparameter estimation is a necessary process to characterizetheir values Before running parameter estimation initialparameter values and boundaries should be set within phys-ically plausible ranges To do so the database BioNumbersprovides modelers with key numbers in molecular and cellbiology ranging from cell sizes to metabolite concentrationsfrom reaction rates to generation times from genome sizesto the number of mitochondria in a cell [28] After parameterestimation unknown parameter values are determined usingoptimization methods which can minimize a cost functionthat measures the goodness of fit of the model with respect togiven quantitative experimental data sets Parameter estima-tion using optimization algorithms is an open research fieldin which several methods have been developed accordingto the nature and numerical properties of biological dataanalyzed The discussion for choosing proper optimizationmethods for parameter estimation is beyond the scope ofthis paper but the interested reader is referred to the paperpublished by Chou and Voit [29]
(4) Model Validation and Analysis Usually the model sim-ulations are compared with the experimental data used forthe parameter estimation but a good agreement betweenboth is not enough to guarantee the predictive ability of themodel Therefore it is necessary to validate the model withdata sets that are not used during the parameter estimationThis process is called model validation and can ensure morereliable and accurate model predictions To do so the datagenerated in new experiments or extracted from literature arecompared with model simulations which are obtained afterconfiguring the model according to the new experimentalsettings Once a model is validated it can be used to per-form predictive simulations which are helpful to studythe dynamic properties of biochemical systems guide thedesign of new experiments in the laboratory and formulateadditional hypotheses In addition tools such as sensitivityand bifurcation analysis can be used to study complex proper-ties and behavior of the modeled system Sensitivity analysisis used to evaluate the influence of model parameters (eginitial concentrations of the state variables and rate constants)on model outputs such as the temporal behavior of networkcomponents [30] Bifurcation analysis is employed to detectcontrol parameters (also known as bifurcation parameters)whose variations are able to change drastically the dynamicalproperties of the biochemical system as well as the stabilityof its fixed points [31] The application of these tools tomathematical modeling is beyond the scope of this paper butthe interested reader is referred to the publication of Zhouet al [32] and Marino et al [33]
212 Experiment Methods As mentioned in the previoussections after a model is established it can be calibrated
BioMed Research International 5
and validated using temporal experimental data To do sothe data can either be derived from literature or generatedto calibrate the model by own experiments In case of ODEmodels the most suitable data for model calibration isquantitative time-series data obtained from perturbation orquantitative dose-response experimentsThe experiments inwhich time-series data are measured for different regulators(such as miRNAs and TFs) of a gene regulatory network canbe obtained using the techniques described in the subsectionsbelow
(1) qRT-PCR Quantitative real time polymerase chain reac-tion (qRT-PCR) has been used to identify mRNAs regulatedby overexpression or silencing of a specific miRNA [3435] miRNAs typically exhibit their regulatory effects byassociating with specific 31015840 UTR regions of the mRNAscalled miRNA seed regions [34] This association can leadeither to a temporary inhibition of translation or completedegradation of the mRNA in which case qRT-PCR mediateddetection is beneficial For this the cells are transfectedwith the miRNA and non-targeting control oligonucleotidesat an appropriate concentration using either a lipid basedtransfection reagent or nucleofection The transfected cellsare then incubated for the necessary time periods (eg 24 h48 h 72 h etc) after which lysates are prepared and total RNAis extracted 1000ndash2000 ng of the RNA is then converted intocDNA using a reverse transcription kit Taqman qRT-PCR isperformed using 10ndash20 ng of cDNA and primers labeled withfluorescence probes to detect the transcriptional levels of thetarget mRNA Housekeeping genes like GAPDH and HPRTare used as endogenous controls for data normalizationRelative expression at different time points is determinedby comparing the Ct values of the miRNA transfected cellswith Ct values of non-targeting control transfected cells andexpressed as relative expression [36] However qRT-PCR inspite of being highly sensitive is applicable specifically underconditions of complete or partial degradation of the targetmRNA miRNA-mediated inhibition in translation can bebetter demonstrated by techniques like immunoblotting
(2) Western BlotImmunblotting Western blotting or proteinimmunoblotting is a technique to detect the expression of agene at protein level This technique is particularly useful indetermining the regulatory effects of a miRNA on expressionof a target gene which is temporarily inhibited For thisthe cells are transfected with a miRNA or an antagomiRas mentioned above and cell lysates are prepared at appro-priate time points using protein lysis buffers (eg Radio-Immunoprecipitation Assay buffer or RIPA) containing lysisagents like Dithiothreitol (DTT) and protease inhibitors Theproteins from each sample are quantitated using Bradfordor BCA reagents and compared with bovine serum albumin(BSA) standards for accurate protein estimation 20ndash40 120583g ofprotein is then loaded and resolved on a sodium dodecyl sul-fate polyacrylamide gel (SDS-PAGE) along with a pre-stainedprotein marker The protein bands are then transferred ontoa nitrocellulose membrane followed by incubation with theappropriate primary and secondary antibodies linked tofluorescent dyes or horse radish peroxidase enzyme (HRP)
The protein expression is then analyzed using either fluores-cence or chemiluminiscenceHRP substrates in gel documen-tation systems (eg LI-COR Odyssey) Housekeeping geneslike 120573-actin or 120573-tubulin are used for protein normalizationThe time point for maximum target gene suppression gen-erally varies depending on the number of miRNA bindingsites at the 31015840 UTR of the target gene and the extent ofcomplementarity of the seed region [37] Immunoblotting isa widely used technique to provide confirmatory evidence forthe inhibitory effects ofmiRNA at the protein level but it failsto explain the underlying interaction mechanisms
(3) Reporter Gene Assay As each miRNA can inhibit theexpression of a large number of genes regulation of a par-ticular target gene may either be by direct interaction or bean indirect consequence of it In direct regulation a miRNAbinds to the complementary sequences at the 31015840 UTR of atarget gene and thereby suppresses its expression As a con-sequence of this the expression levels of a number of down-stream genes (indirect targets) are also dysregulated makingit crucial to differentiate between primary and secondarymiRNA targets To determine the interaction specificitya reporter construct (luciferase) with intact or mutated 31015840UTR of the target gene cloned at the 51015840 end is co-transfectedinto the cells along with the miRNA The regulatory effect ofthe miRNA on the target gene expression is then measuredusing the expression of a reporter gene In the absence ofthe appropriate binding sequences (mutated 31015840 UTR) themiRNA cannot suppress the reporter mRNA suggesting thatthe suppressive effect of the miRNA is mediated by a directinteraction The reporter activity can be analyzed at differenttime points such as 24 hr 48 hr and 72 hr to determine thetime dependent suppression of a target gene expression bya miRNA
22 Case Study The Regulation of p21 by Multiple andCooperative miRNAs
221 Construction and Visualization of p21 Regulatory Net-work By using the approach described above we investi-gated the regulation of p21 by its multiple targeting miRNAsp21 also known as cyclin-dependent kinase inhibitor 1(CDKN1A) is a transcriptional target of p53 It is requiredfor proper cell cycle progression and plays a role in cell deathDNA repair senescence and aging (reviewed in [38]) Inter-estingly p21 was the first experimentally validated miRNAtarget hub which is a gene that is simultaneously regulatedby many miRNAs This made it an ideal candidate for a casestudy of our approach [23] To do so we first constructeda p21 regulatory network using the following steps
(a) We extracted miRNA-target interactions from thepublication of Wu et al [23] where a list of predictedp21-regulating miRNAs was subjected to experimen-tal validation
(b) Experimentally verified TFs of p21 were extractedfrom literature and putative TFs having conservedbinding sites in the 5 kb upstream region of the p21open reading frame were extracted from UCSC table
6 BioMed Research International
browser A list of TFs controlling the expression of themiRNAswas constructed using information of exper-imentally proven TF-miRNA interactions extractedfrom TransmiR (release 10) [39] In addition wegenerated a list of putative TFs of miRNAs withbinding sites in the 10 kb upstream region of themiRNA genes using information from the databasesPuTmiR (release 10) [40] and MIRNTN (version121) [41] and from the table of TFs with conservedbinding sites in theUCSC genome browser (hg18) [9]
(c) Information about protein interactions was extractedfrom theHuman Protein ReferenceDatabase (HPRDrelease 90) [15] and the STRING database (release90) [16] Only the experimentally verified p21-proteininteractions were used to construct the network
(d) Additionally we associated the TFs in the network tonine biological processes based on the Gene Ontol-ogy (GO) [42] The corresponding GO terms werecell proliferation cell apoptosis immune responseinflammatory response cell cycle DNA damage cellsenescence DNA repair and cell migration
Next we visualized the network in CellDesigner andcomputed a confidence score for each interaction in thenetwork (Figure 1) The confidence scores provide us withthe reliability of the interactions considered in p21 regulatorynetwork With the help of these scores we discarded non-reliable interactions and constructed the mechanistic modelaccounting for p21 regulation by its targeting miRNAsBesides the interested experimentalists can further use thisinformation to choose reliable interacting candidates of p21for designing relevant experiments The scores for eachinteraction of p21 regulatory network are shown in Table 2
222 Mechanistic Modeling of p21 Regulation byIts Targeting miRNAs
(1) Model Construction After constructing the regulatorynetwork a detailed mechanistic model of ODEs whichdescribes the biochemical reactions underlying the regula-tion of p21 was established We chose the ODE modelingapproach as it is a simple formalism for describing temporaldynamics of biochemical systems and a wide range of toolsare available to explore their properties Precisely the modelconsidered themRNA(mp21 (3)) andprotein (p21 (6)) of themiRNA target hub p21 the p21-targeting miRNAs (miR
119894 119894 isin
(1 15) (5)) and the complexes formed by p21 mRNAand miRNA [mp21 | miR
119894] (4) Altogether the model is
constituted by 32 state variables and 64 parameters
119889mp21119889119905
= 119896mp21syn sdot 119891act (TFmp21)
minusmp21 sdot (119896mp21deg +sum
119894
119896complex
119894
ass sdotmiR119894)
(3)
119889 [mp21 | miR119894]
119889119905
= 119896complex
119894
ass sdotmp21 sdotmiR119894minus 119896
complex119894
deg sdot [mp21 | miR119894]
(4)
119889miR119894
119889119905= 119896
miR119894syn sdot 119891act (TFmiR119894)
minusmiR119894sdot (119896
miR119894deg + 119896
complex119894
ass sdotmp21) (5)
dp21119889119905
= 119896p21syn sdotmp21 + 119896p21deg sdot p21 (6)
mp21Total = mp21 +sum119894
[mp21 | miR119894] (7)
For mp21 processes considered in the model were (i) itssynthesis (119896mp21
syn ) mediated by TFs (119891act(TFmp21)) (ii) itsdegradation (119896mp21
deg ) and (iii) its association with a miRNA(119896complex
119894
ass ) For each miR119894 processes considered were (i)
the synthesis (119896miR119894syn ) mediated by TFs (119891act(TFmiR119894)) (ii) the
degradation (119896miR119894deg ) and (iii) the association with the p21
mRNA target (119896complex119894
ass ) For each [mp21 | miR119894] complex
processes considered were (i) the formation of the complexby a miR
119894and the p21 mRNA (119896complex
119894
ass ) and (ii) the complexdegradation (119896complex
119894
deg ) For p21 processes considered were(i) its synthesis (119896p21syn ) and (ii) its degradation (119896p21deg) Anadditional algebraic equation accounting for the total mea-surable amount of p21 mRNA (mp21Total) was also includedThe SBML file of the model is available for download athttpwwwsbiuni-rostockdeuploadstx templavoilap21TargetHub 03092013xml
(2) Model Calibration and Validation For model calibrationwe fixed some parameter values using published data andestimated the other unknown parameters using the time-series data published by Wu et al [23] in which the p21mRNA (northern plot) and protein levels (western plot)were measured 48 hr after transfection of individual p21-targeting miRNAs into human embryonic kidney 293 cellsThe unknown parameter values were estimated using aniterative method combining global (particle swarm patternsearch) [43] and local (downhill simplex method in multi-dimensions) [44] optimization algorithms For each miR
119894
considered in the model the method minimizes the distancebetween model simulations and experimental data using thefollowing cost function
119865miR119894cost =
[mp21miR119894sim (119905) minusmp21miR119894
exp (119905)]2
(120575mp21119894
)2
+
[p21miR119894sim (119905) minus p21miR119894
exp (119905)]2
(12057511990121
119894)2
119894 isin [1 15]
(8)
BioMed Research International 7
1 2 4 1 2 4 5 6 1 2 3 7
E2F TP53 RELA1 2 4 7 2 3 7 6
STAT5 SRF1 2 3 4 1 2
JUN STAT1
SP1 RUNX1 TFAP2A
Common TFs Modulation
CDKN1A
1 2 1 2 4 1 4
1 4 1 2 3 7
1 2 5 2 4 5 7 1
VDR
CUX1
E2A
RAR BRCA1
TP63 TP73 TBX2
STAT6 SP3 STAT3
STAT2 RUNX2
EHMT2 p21 TFs
1 2 4 5 8 9
CEBP120573NF120581B1
(1) Cell proliferation (6) Cell senescence
(2) Apoptosis (7) Inflammatory response
(3) Immune response (8) DNA repair
(4) Cell cycle control (9) Cell migration
(5) DNA damage
miRNAmiRNA
Translational repressionmiRNA
Translation
Transcription
Association
miR-132 miR-28-5p
miR-299-5p miR-298
miR-345 miR-125a-5p
miR-208a
miRNA
miRNADeadenylation followed by decay
Protein-protein interaction
miR-93 miR-654-3p
miR-657 miR-572
miR-423-3p miR-363
miR-639 miR-515-3p
p21
CCND2 CDK3 GMNN CCNE1 POLD2 CSNK2B C1orf123 CCDC85B
PIM1 CCNB2 KIAA0432 GADD45B HDAC11 ITGB1BP3 TTLL5 SLC25A11
GADD45G CDC7 BCCIP CDK1 GNB2LI RAB1A psmA3 TK1
CDC45 CCNB1 CCNA2 GADD45A MAPK8 TSG101 SET CIZ1
CDK14 CDK5 MDM2 CDC27 RNF144B ESR1 PRKCH RPA1
CDC20 SKP2 CDK2 CDK4 DHX9 PCNA CDC5L CCDC5
CCND3 CDK6 CCNE2 CCNA1 TEX11 MAP3K5 MCM10 Caspase-3
CDC6 CCND1 XRCC6 PARP1 DAPK3 NR4A1 CEBPA
Interacting proteins
AKT1
miRNA genes
5 8 1 2 1 2 4 3 1 2 3 3 2 2 3 3 1 4 1 3 1 4
1 4
NFATC2 YY1 CREB1 CEBPA ARNT
FOXCI1
ELK1 SOX9 FOS PAX5 MAX MEF2A RFX1 JUNB AKT1 GATA3 MYC NRSFEGR1
E47 PPARG POU3F2 FOXF2FOXL1 MZF1 NFIC SRY
FOXC1
HNFA4 NFYA MYF1 BACF1 ATF6 ZID RORA POU2F1
miRNA TFs
11 2 34
12 31 5 11
HNF1A
1 2
RB1
Cellular processes
Figure 1 p21 regulatory network The network contains several layers of regulators of p21 TFs (light blue and red boxes) miRNAs (darkblue and green boxes) and proteins (grey boxes) In each big box there are small boxes which represent individual components of this layerof regulation miRNAs are classified into two groups according to the mechanisms by which the expression of p21 is repressed One groupcauses p21 translation repression (green box) These miRNAs bind to p21 mRNA resulting in the repressed translation of p21 but unchangedmRNA expression level The other group of miRNAs (dark blue box) decreases the stability of p21 mRNA by modifying its structure leadingto mRNA decay and finally the downregulation of p21 TFs are classified into three groups p21 TFs (red) miRNA TFs (yellow) and theircommon TFs (light blue) The p21 interacting-proteins are framed in the grey boxes The purple boxes represent nine processes and the TFsassociated with these processes are indicated in the ellipses above them using corresponding figures This data is adapted from our previouspublication [4]
where mp21miR119894sim (119905) and mp21miR119894
exp (119905) represent the simulatedp21 mRNA and protein expression levels for each miR
119894at
time point 119905 p21miR119894sim (119905) and p21miR119894
exp (119905) represent the mea-sured value for each miR
119894at time point 119905 and their standard
deviations are 120575mp21119894
and 120575p21119894
Here 119905 is the time point (48 hr)after overexpression of the individual miRNAs in embryonickidney 293 cells at which the expression levels of the p21 andits mRNA were measured [23] The model calibration resultsare shown in Figure 2(a) and the obtained parameter valuesare listed in Table 3
Experimental results showed that a stronger repressionof the target gene can occur when two miRNA binding siteson the target mRNA are in close proximity [24 45] To testthe consequences of this hypothesis we predicted cooperativemiRNA pairs for p21 with seed site distances between 13ndash35 nt in the p21 31015840 UTR To substantiate the cooperativeeffect associated with pairs of miRNAs we introduced agroup of new state variables ([mp21 | miR
119894| miR
119894]) into
the original model These state variables account for the
ternary complexes composed of p21 mRNA and two puta-tively cooperating miRNAs (miR
119894and miR
119895) For these new
variables processes considered are (i) the association of p21mRNA with miR
119894and miR
119895into a complex (119896co-complex
119894119895
ass )and (ii) the degradation of the complex (119896co-complex
119894119895
deg ) Afterexpansion the corresponding modified and new ODEs arelisted below
dmp21119889119905
= 119896mp21syn sdot 119891act (TFmp21)
minusmp21 sdot (119896mp21deg +sum
119894
119896complex
119894
ass sdotmiR119894
+sum
119894119895
119896co-complex
119894119895
ass sdotmiR119894sdotmiR119895)
(9)
8 BioMed Research InternationalTa
ble2Th
econ
fidence
scores
forthe
interactions
inp21regulatorynetworkTh
enetworkconsistso
ffou
rtypes
ofinteractionsm
iRNA-
p21TF
-miRNAT
F-p21andp21-p
rotein
(a)miRNA-
p21interactio
nscores
miRNA
miR-125a-5p
miR-132
miR-208a
miR-28-5p
miR-298
miR-299-5p
miR-345
miR-363
Score
073
073
073
080
073
080
073
073
miRNA
miR-423-3p
miR-515-3p
miR-572
miR-639
miR-654-3p
miR-657
miR-93
Score
085
073
073
073
073
080
095
(b)TF
-miRNAinteractionscores
Score
Score
Score
Score
Score
Score
miR-208
miR-132
miR-657
miR-125a-5p
miR-28-5p
miR-93
MAX
027
EGR1
093
TFAP2
A034
EGR1
086
MZF
1028
TFAP2
A034
EGR1
029
CREB
1086
SP1
034
PAX5
024
E47
027
SOX9
024
ARN
T027
ARE
B6027
MZF
1034
NRS
F024
FOXC
1031
GAT
A3
027
RFX1
024
E47
024
MAX
029
ZID
024
FOXI1
024
MZF
1031
ATF6
024
ELK1
027
USF1
027
PPARG
027
SRY
027
SP1
034
YY1
027
EGR2
029
EGR1
034
GAT
A3
034
JUN
029
POU2F1
024
JUNB
024
FOS
024
RELA
029
MZF
1034
POU2F1
024
EGR1
029
POU2F1
024
NFIC
024
miR-654-3p
TP53
034
SRF
027
CEBP
A024
NFIC
027
RELA
027
YY1
027
NFIC
034
CEBP
A029
NFY
A024
miR-345
miR-423-3p
BACH
1024
miR-299-5p
RORA
024
RUNX1
024
SP1
034
BACH
1027
FOXL
1029
CREB
1024
miR-363
STAT
1024
RELA
027
STAT
1027
SRY
024
SRY
027
E2F1
062
E2F1
085
NFA
TC2
024
POU3F2
024
NFA
TC2
024
SRF
029
FOXC
1033
MYC
084
HNF4
A031
STAT
5A027
FOS
024
RUNX1
027
MZF
1034
miR-298
NFY
A024
SRF
027
POU2F1
027
miR-572
miR-639
JUNB
031
POU2F1
024
MEF
2A027
NF120581
B1024
FOXF
2024
NFY
A024
NFIC
031
POU2F1
024
HNF1A
024
(c)TF
-p21
interactionscores
TF(verified)
SP1
SP3
E2F1
Runx
1Ru
nx2
STAT
1E2
AST
AT5
TP53
TP63
TP73
Score
100
100
077
077
077
072
072
067
067
067
067
TF(verified)
STAT
3CU
X1TF
AP2
ABR
CA1
VDR
RARA
CEB
P120572CEB
P120573RB
1Tb
x2EH
MT2
Score
066
066
060
060
060
060
060
060
060
060
060
TF(putative)
NF120581
B1RE
LAST
AT2
STAT
6SR
FScore
044
044
044
044
043
(d)p21-p
rotein
interactionscores
Protein
TP53
PCNA
CASP
3CC
NA1
CCND1
SKP2
BCCI
PCC
NA2
CCND2
CCNE2
AKT
1C1
orf123
CCDC8
5BCC
NB1
CCNB2
CCND3
Score
100
092
087
087
087
087
081
081
081
081
072
072
072
072
072
072
ProteinCC
NE1
CDC4
5CD
C5L
CDC6
CDC7
CDK1
CDK14
CDK2
CDK3
CDK4
CDK6
CEBP
ACI
Z1CS
NK2
A1
CSNK2
BDAPK
3Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
ESR1
GADD45A
GADD45B
GADD45G
GMNN
GNB2
L1HDAC
11ITGB1BP
3MAP3
K5MAPK
8MCM
10PA
RP1
PIM1
POLD
2PS
MA3
RAB1A
Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
SET
SLC2
5A11
STAT
3TE
X11
TK1
TSG101
TTLL
5XR
CC6
CDC2
0CD
C27
CDK5
DHX9
MDM2
NR4
A1
PRKC
HRP
A1
Score
072
072
072
072
072
072
072
072
056
056
056
056
056
056
056
056
Verifi
edexp
erim
entally
verifi
edinteraction
putativ
epredictedinteraction
BioMed Research International 9
Con
trol
miR
-298
miR
-208
am
iR-1
32m
iR-2
8-5p
miR
-125
-5p
miR
-299
-5p
miR
-345
miR
-93
miR
-423
-3p
miR
-515
-3p
miR
-363
miR
-657
miR
-639
miR
-572
miR
-654
-3p
0
05
1
15m
RNA
expr
essio
n (a
u)
Data Model Data ModelC
ontro
lm
iR-2
98m
iR-2
08a
miR
-132
miR
-28-
5pm
iR-1
25-5
pm
iR-2
99-5
pm
iR-3
45m
iR-9
3m
iR-4
23-3
pm
iR-5
15-3
pm
iR-3
63m
iR-6
57m
iR-6
39m
iR-5
72m
iR-6
54-3
p
0
02
04
06
08
1
Prot
ein
expr
essio
n (a
u)
(a)
DoxorubicinCellWell
Control miR-572 miR-93
Measure p21 expression at the time points of
miR-572+ miR-93
0 2 4 6 8 24hr
miR-572miR-93
Immunoblotting
(b)
0 4 8 12 16 20 240
4
8
12
16
20
Time (hr)
p53
Actin
Doxo (hrs)
fac
t(TF
mp2
1)
C 0 1 2 4 6 8 24
1 08 12 19 29 32 3 16
(c)
0 2 4 6 8 2402468
1012
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)
p21
expr
essio
n (a
u)
02468
1012
miR-572Control
02468
1012
miR-93
02468
1012
Model Data
0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24
1 23 46 107124 124 23 28 48 66 53 31 08 16 21 38 50 18 06 06 12 25 27 12
miR-572+ 93Doxo (hrs)
p21
Actin
p21
Actin
p21
Actin
p21
Actin
(d)
Figure 2 Model calibration and validation (a) Model calibration The figures show the relative change of the p21 mRNA and proteinexpression levels after overexpression of the indicated miRNAs (Model model simulation Data experimental data) These data werenormalized to the control group in which the p21 mRNA and protein expression levels were measured when the miRNAs were normallyexpressed (au arbitrary unit) (b) Experimental workflow In the experiments Sk-Mel-147 cells were seeded in six well plates Then maturemiRNA mimics were transfected individually at a concentration of 100 nM (miR-572 and miR-93) or in combination at 50 nM each (miR-572 + miR-93) After 48 hr transfection with miRNA mimics the cells were pulse treated with 250 nM doxorubicin for 1 hour after whichnormal growth medium was replenished The immunblotting were performed to measure p21 expression at 0 2 4 6 8 and 24 hr post-doxorubicin treatment (c) Temporal dynamics of p21 transcriptional function Afterdoxorubicin treatment the expression of p53 a TF ofp21 was measured using immunblotting and these data were used to characterize the transcriptional function of p21 using MATLAB linearinterpolation function (d) Model validation We measured the expression of p21 protein in response to genotoxic stress in the four scenariosas described in the main text The measured data (Data) were compared with the model simulations (Model) The figures (a) (c) and (d) areadapted from our previous publication [4]
119889 [mp21 | miR119894| miR
119895]
119889119905
= 119896co-complex
119894119895
ass sdotmp21 sdotmiR119894sdotmiR119895
minus 119896co-complex
119894119895
deg sdot [mp21 | miR119894| miR
119895]
(10)
To model stronger repression of the target gene bycooperating miRNAs we assumed a stronger association rateconstant for the complex [mp21 | miR
119894| miR
119895] which is
equal to the sum of their individual association rate constants
(119896co-complex119894119895
ass = 119896complex
119894
ass + 119896complex
119895
ass ) Similarly the degradationrates of the complexes [mp21 | miR
119894| miR
119895] were assumed
to be equal to the sum of degradation rate constants ofsingle miRNA binding complexes (119896co-complex
119894119895
deg = 119896complex
119894
deg +
119896complex
119895
deg ) However it has to be noted that these addedequations are an abstract description ofmiRNAcooperativitybecause the details of this mechanism are not yet known
To experimentally validate the capability of our model topredict the relative p21 concentrations regulated by coopera-tivemiRNAs we selectedmiR-572 andmiR-93 as a case study
10 BioMed Research International
Table 3 Initial concentrations of model variables and model parameter values Based on the experimental data the p21-targeting miRNAsverified byWu et al [23] were divided into two groups the translation repression group (marked with asterisk) and the mRNA deadenylationgroup AmiRNAwas classified into themRNA deadenylation group if its overexpression can result in 20 ormore downregulation of the p21mRNA level (ie p21mRNA levelle 08 the basal level is 1) otherwise it was classified into the translation repression group For the translationrepression group only 119896complex119894
ass was estimated and 119896complex119894deg was fixed For the other group both 119896complex119894
deg and 119896complex119894ass were estimatedThe initial
concentrations of p21 and mp21 were set to 1 and this value was used as their basal expression levels During the parameter estimation theinitial concentrations of p21-targeting miRNAs were set to 100 because in the publication [23] the expression levels of p21 and mp21 weremeasured after the individual introduction of the p21-targeting miRNAs with amount of 100 nM Due to the lack of biological information tocharacterize the transcriptional activation function (119891act) of p21 and its targeting miRNAs the corresponding functions were assumed to be1 for simplicity The data is adapted from our previous publication [4]
Initial concentration of variables and TF functionsVariable Description Initial concentration (au)p21 p21 protein 1
mp21 p21 mRNA 1
miR119894=115
p21-targeting miRNAs 100
[mp21 | miR119894=115
] Complexes formed by miR119894and mp21 0
119891act(TFmp21) p21rsquos transcriptional activation function 1119891act(TFmiR119894 (119894=115)) The transcriptional activation function of miR
1198941
Fixed parameter valuesParameter Description Value (hrminus1) Reference119896mp21syn Synthesis rate constant of mp21 01155 fixed119896mp21deg Degradation rate constant of mp21 01155 [24]119896miR119894syn (119894 = 1 15) Synthesis rate constant of miR
11989400289 fixed
119896miR119894deg (119894 = 1 15) Degradation rate constant of miR
11989400289 [25]
119896p21syn Synthesis rate constant of p21 13863 fixed119896p21deg Degradation rate constant of p21 13863 [26]
Estimated parameter values
miRNA (statevariable)
119896complex119894deg
(119894 = 1 15) (hrminus1)
119896complex119894ass
(119894 = 1 15)
(au minus1sdot hrminus1)
119865miR119894cost
(119894 = 1 15)Experimental data of p21 (protein mRNA plusmn SD)
miR-298 (miR1)lowast 01155 00254 34119890 minus 004 (016 1074 plusmn 0025)
miR-208a (miR2)lowast 01155 00041 20119890 minus 003 (051 1192 plusmn 0022)
miR-132 (miR3)lowast 01155 00275 24119890 minus 003 (015 121 plusmn 0147)
miR-28-5p (miR4)lowast 01155 00119 59119890 minus 003 (028 135 plusmn 006)
miR-125-5p (miR5)lowast 01155 00018 18119890 minus 003 (069 085 plusmn 0051)
miR-299-5p (miR6)lowast 01155 00080 18119890 minus 004 (036 095 plusmn 0038)
miR-345 (miR7)lowast 01155 00051 11119890 minus 004 (046 096 plusmn 0039)
miR-93 (miR8) 01564 00235 41119890 minus 014 (017 07776 plusmn 003)
miR-423-3p (miR9) 09118 00055 28119890 minus 009 (044 05102 plusmn 011)
miR-515-3p (miR10) 02098 00253 12119890 minus 013 (016 0616 plusmn 0037)
miR-363 (miR11) 02261 00399 22119890 minus 014 (011 056 plusmn 015)
mR-657 (miR12) 03465 00158 21119890 minus 014 (023 048 plusmn 012)
miR-639 (miR13) 04305 00327 18119890 minus 017 (013 036 plusmn 0084)
miR-572 (miR14) 03039 00360 94119890 minus 023 (012 045 plusmn 0044)
miR-654-3p (miR15) 97485 00024 30119890 minus 014 (064 063 plusmn 0053)
These two miRNAs were chosen because their predictedtarget sites in the p21 31015840 UTR are in close proximity to eachother and thereby they can induce cooperative repression onp21 as suggested in [24] The experiments were performed asfollows
(i) Melanoma cells (Sk-Mel-147) were transfected withthe mature miRNAmimics of the twomiRNAs eitherindividually (100 nM) or in combination (50 nMeach) whereas untreated cells were used as control(Figure 2(b))
BioMed Research International 11
p21
FOXF2
miR-572
Activation
Inhibition120572
Incoherent FFL + cascaded regulation
AF2
miR-93
(a)
miR-93
miR-572
TP53
MYC
p21
FOXF2
Coherent FFL + cascaded regulation
(b)
Saturation
Pulse
p21
(au
)
p21
(au
)Time (hr) Time (hr)
Saturation Pulse
Pulse
Saturation
Saturation
di
(au
)
102
101
102
100
101
10minus1
10010minus2
10minus1
102
101
100
10minus1
10minus2
FOXF2 (au) FOXF2 (au)
TP53
(au
)
1
09
08
07
06
05
04
03
02
01
10minus2 10210110010minus110minus20
AF2120572
(au
)
(c)
Figure 3 Different p21 dynamics in different network motifs We ran simulations to show the different dynamics of p21 for two differentnetwork motifs (a) and (b) Through simulations two dynamical patterns of p21 were identified saturation and pulse ((c) top) For eachnetwork motif the corresponding distributions of the two dynamical patterns were plotted ((c) bottom) For different combinations ofthe transcriptional strengths the normalized distance (119889
119894) between peaks (119901
119894) and steady states (ss
119894) of p21 is determined by the equation
119889119894= (119901119894minus ss119894)119901max 119901max = max(119901
1 119901
119899) If 119889
119894= 0 for the corresponding combination of transcriptional strengths the p21 dynamics is
saturation otherwise it is pulse The regions showing different dynamical patterns of p21 are separated using the white lines
(ii) Next the cells were treated with doxorubicina genotoxic-stress inducing agent The agent canupregulate the expression of p53 which is a knownTFof p21 and therefore it can result in the upregulationof p21 (Figure 2(c))
(iii) After doxorubicin treatment the expression levels ofp21 were measured by immunoblotting at differenttime points (0 2 4 6 8 24 hr) The p21 expressionvalues were normalized based on the p21 expression
level in the control group measured at time point 0 hr(Figure 2(d))
By doing so we obtained the p21 response after genotoxicstress in four scenarios (1) endogenous miRNA expression(Control) (2) overexpression of miR-572 (3) overexpressionof miR-93 and (4) both miRNAs moderately overexpressedThereafter we derived a model of seven ODEs based onthe original equations which was configured according tothe designed experiments making the simulation results
12 BioMed Research International
00001
0001
001
01
1Cell cycle Cell growth DNA damage Senescence Apoptosis
G1G2
MS
p21
expr
essio
n (a
u)
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
log10
wow
Figure 4 p21 expression regulated by cooperative miRNAs for different cellular processesThe associations of themiRNAs with these cellularprocesses were derived from GO terms of their TFs A miRNA was supposed to be expressed (in bold black font) in a cellular process onlyif its TF is related to the corresponding GO term of this process The p21 expression levels are computed for each process with (w)without(wo) considering the cooperative effect among the p21 targeting miRNAs
comparable with the experimental data As shown in theFigure 2(c) the simulations are in good agreement with theexperimental observations The individual overexpression ofmiR-572 or miR-93 led to the reduction of the upregulationof p21 response after genotoxic stress induction The twomiRNAs cause different degrees of repression due to theirdifferent repression efficiencies on p21 Interestingly the com-bined overexpression of both miRNAs induced the strongestdownregulation of p21 and therefore verifying the hypothesisof their cooperative regulation of p21 Above all the resultsnot only validated themodel but also demonstrated the abilityof our method to identify cooperative miRNA pairs for p21
(3) Predictive Simulations As there are abundant of networkmotifs such as FFLs in p21 regulatory network and thesenetwork motifs are important for determining p21 dynamicsit is interesting to investigate the dynamics of p21 in networkmodules where FFLs are involved To do so two networkmodules including both miR-93 and miR-572 and theirTFs were exemplified In Figure 3(a) the network modulecontains an incoherent FFL composed by AF2120572 miR-93 andp21 and a cascaded regulation in which p21 is repressed
by FOXF2 via miR-572 in Figure 3(b) the same cascadedregulation together with a coherent FFL composed of TP53MYC miR-93 and p21 forms another regulatory moduleof p21 By modulating the transcriptional strengths of thetwo miRNAs by their TFs two types of p21 dynamics wereidentified saturation and pulse (Figure 3(c) top) In theformer the p21 expression increases and reaches its steadystate at the highest level in contrast in the latter the p21expression increases to a peak and thereafter drops to asteady state at lower level For the two different networkmodules various combinations of transcriptional strengthsof the two miRNAs lead to different distributions of thetwo p21 dynamical patterns For the network module of anincoherent FFL plus the cascaded regulation (Figure 3(c)bottom left) the saturation pattern appears in two distinctregions one with weakened transcriptional strength of miR-93 and the other with enhanced transcriptional strength ofmiR-93 for the other module (Figure 3(c) bottom right)the saturation pattern only appears in the region in whichthe transcriptional strength of miR-93 is enhanced Takentogether the results showed that for the two different networkmotifs the dynamical pattern of p21 is changing according to
BioMed Research International 13
different combinations of its upstream regulators suggestingthe adaptation of p21 dynamics for different biological con-texts
Furthermore we performed a number of simulations toshow the influence of miRNA regulation on p21 expressionlevels in different cellular processes arranged in a consecutivemanner In this procedure a cell is first in the process ofcell cycle followed by proliferation (cell growth) then thecell responses to DNA damage and enters into the processof senescence and finally apoptosis is initiated As shown inFigure 4 during the cell cycle process the p21 expression islow due to the activation of most of its targeting miRNAswhen the cell starts proliferating the p21 expression declinesto an even lower level because of more activated miRNAsin this process after responding to DNA damage the p21expression soars to a high level which is caused by theactivation of its TFs like p53 and fewer expressed miRNAsalthough the p21 expression keeps at a similar level whilethe cell is undergoing senescence the expressed miRNAs aredifferent from the previous process finally the p21 expressiondecreases again to a low level due to the reemergence ofmost of its targeting miRNAs and the cell enters apoptosisInterestingly the model simulations are also consistent withexperimental observations under non-stressed condition thelow expression level of p21 is needed for cell proliferation theupregulation of p21 happens after response to DNA damagevia p53 and the increased p21 expression further results incell cycle arrest leading to senescence and apoptosis [25]Besides when considering the effect of cooperatingmiRNAsthe p21 expression levels were indistinguishable from theprevious simulation of DNA damage response However forthe other processes p21 expression levels were significantlylower compared to the simulations without considering thecooperative effect of the miRNAs Above all these resultsindicated that selective expression of cooperative miRNAscould be adopted by cells to ensure diverse expression levelsof p21 tomeet the requirements of different cellular processes
3 Conclusions and Discussion
In this paper we presented a systems biology approach com-bining data-drivenmodeling andmodel-driven experimentsto investigate the role of miRNA-mediated repression in generegulatory networksThis approach provides a systematicwayto gain a deeper understanding of the regulation of targetgenes by mutiple and cooperative miRNAs Using the regu-lation of p21 by multiple miRNAs as a case study we showedhow the ODE-basedmodel which is calibrated and validatedby means of experimental data is suitable for predicting thetemporal dynamics of molecular concentrations involved inbiochemical systems
Provided there are sufficiently rich quantitative data setsavaliable to characterize the model the use of the methodol-ogy here shown can be extended to more complex regulatorynetworks involving multiple targets cooperating TFs andmiRNAs and signaling pathways displaying cross-talk viapost-tranlational modifications In this case the criticalelement is the quality and quantitiy of the available data
Insufficiency and low quality of experimental data can causeerrors in the process of model construction and overfitting inparameter estimation can lead to uncertainties in the modelpredictions We believe that the quick development of quan-titative high throughput techniques such as transcriptomicsproteomics and miRNomics will facilitate the constructionand characterization of larger miRNA-mediated regulatorynetworks
Other modeling frameworks than ODE-based modelscan be used to describe biological systems such as prob-abilistic (eg Bayesian) or logical (eg Boolean) modelsImportantly different modeling frameworks have differentproperties and perform well regarding different perspectivesand levels of mechanistic details of biochemical systems [46]For example Bayesian models are helpful in the constructionof connections in signaling networks and can reveal the mostlikely underlying structure of the network in a probabilisticmanner Boolean models use binary values (0 and 1) andlogical gates (AND NOT and OR) to describe activities ofnetwork components and the information flow among themWe believe that in the coming future hybrid models whichconsist of modeling framework and experimental techniquespecific sub-modules will provide the necessary compromisebetween quantitativequalitative accuracy and scalability forthe investigation of large biochemical networks [47]
Disclosures
The authors declare that they have no competing financialinterests
Authorsrsquo Contribution
Julio Vera and Olaf Wolkenhauer are equal contributers
Acknowledgments
The authors would like to acknowledge the funding fromthe German Federal Ministry of Education and Research(BMBF) eBio-miRSys (0316175A to Xin Lai and Julio Vera)eBio-SysMet (0316171 to OlafWolkenhauer) and Gerontosys-ROSAge (0315892A to Olaf Wolkenhauer)
References
[1] A E Pasquinelli ldquoMicroRNAs and their targets recognitionregulation and an emerging reciprocal relationshiprdquo NatureReviews Genetics vol 13 no 4 pp 271ndash282 2012
[2] J Vera X Lai U Schmitz and O Wolkenhauer ldquoMicroRNA-regulated networks the perfect storm for classical molecularbiology the ideal scenario for systems biologyrdquo Advances inExperimental Medicine and Biology vol 774 pp 55ndash76 2013
[3] S Nikolov J Vera U Schmitz and O Wolkenhauer ldquoA model-based strategy to investigate the role of microRNA regulation incancer signalling networksrdquo Theory in Biosciences vol 130 no1 pp 55ndash69 2011
14 BioMed Research International
[4] X Lai U Schmitz S K Gupta et al ldquoComputational analysisof target hub gene repression regulated by multiple and cooper-ative miRNAsrdquoNucleic Acids Research vol 40 no 18 pp 8818ndash8834 2012
[5] X Lai O Wolkenhauer and J Vera ldquoModeling miRNA regu-lation in cancer signaling systems miR-34a regulation of thep53Sirt1 signaling modulerdquoMethods in Molecular Biology vol880 pp 87ndash108 2012
[6] C Jiang Z Xuan F Zhao and M Q Zhang ldquoTRED atranscriptional regulatory element database new entries andother developmentrdquoNucleic Acids Research vol 35 supplement1 pp D137ndashD140 2007
[7] L A Bovolenta M L Acencio and N Lemke ldquoHTRIdban open-access database for experimentally verified humantranscriptional regulation interactionsrdquo BMCGenomics vol 13article 405 2012
[8] VMatys O V Kel-Margoulis E Fricke et al ldquoTRANSFAC andits module TRANSCompel transcriptional gene regulation ineukaryotesrdquo Nucleic Acids Research vol 34 database issue ppD108ndashD110 2006
[9] D Karolchik R Baertsch M Diekhans et al ldquoThe UCSCgenome browser databaserdquo Nucleic Acids Research vol 31 no1 pp 51ndash54 2003
[10] P Alexiou T Vergoulis M Gleditzsch et al ldquomiRGen 20 adatabase of microRNA genomic information and regulationrdquoNucleic Acids Research vol 38 supplement 1 pp D137ndashD1412010
[11] F Xiao Z Zuo G Cai S Kang X Gao and T Li ldquomiRecordsan integrated resource for microRNA-target interactionsrdquoNucleic Acids Research vol 37 supplement 1 pp D105ndashD1102009
[12] S-D Hsu F-M Lin W-Y Wu et al ldquomiRTarBase a databasecurates experimentally validated microRNA-target interac-tionsrdquo Nucleic Acids Research vol 39 supplement 1 pp D163ndashD169 2011
[13] P Sethupathy B Corda and A G Hatzigeorgiou ldquoTarBase acomprehensive database of experimentally supported animalmicroRNA targetsrdquo RNA vol 12 no 2 pp 192ndash197 2006
[14] H Dweep C Sticht P Pandey and N Gretz ldquomiRWalkmdashdatabase prediction of possible miRNA binding sites byldquowalkingrdquo the genes of three genomesrdquo Journal of BiomedicalInformatics vol 44 no 5 pp 839ndash847 2011
[15] T S Keshava Prasad RGoel K Kandasamy et al ldquoHumanpro-tein reference databasemdash2009 updaterdquo Nucleic Acids Researchvol 37 supplement 1 pp D767ndashD772 2009
[16] D Szklarczyk A Franceschini M Kuhn et al ldquoThe STRINGdatabase in 2011 functional interaction networks of proteinsglobally integrated and scoredrdquo Nucleic Acids Research vol 39supplement 1 pp D561ndashD568 2011
[17] D Croft G OrsquoKelly G Wu et al ldquoReactome a database ofreactions pathways and biological processesrdquo Nucleic AcidsResearch vol 39 supplement 1 pp D691ndashD697 2011
[18] N le Novere M Hucka H Mi et al ldquoThe systems biologygraphical notationrdquoNature Biotechnology vol 27 no 8 pp 735ndash741 2009
[19] A Funahashi Y Matsuoka A Jouraku M Morohashi NKikuchi and H Kitano ldquoCellDesigner 35 a versatile modelingtool for biochemical networksrdquo Proceedings of the IEEE vol 96no 8 pp 1254ndash1265 2008
[20] M S Cline M Smoot E Cerami et al ldquoIntegration ofbiological networks and gene expression data using CytoscaperdquoNature Protocols vol 2 no 10 pp 2366ndash2382 2007
[21] D-H Le andY-KKwon ldquoNetDS aCytoscape plugin to analyzethe robustness of dynamics and feedforwardfeedback loopstructures of biological networksrdquo Bioinformatics vol 27 no 19pp 2767ndash2768 2011
[22] E O Voit Computational Analysis of Biochemical SystemsA Practical Guide for Biochemists and Molecular BiologistsCambridge University Press Cambridge UK 2000
[23] S Wu S Huang J Ding et al ldquoMultiple microRNAs modulatep21Cip1Waf1 expression by directly targeting its 31015840 untranslatedregionrdquo Oncogene vol 29 no 15 pp 2302ndash2308 2010
[24] P Saeligtrom B S E Heale O Snoslashve Jr L Aagaard J Alluin andJ J Rossi ldquoDistance constraints between microRNA target sitesdictate efficacy and cooperativityrdquo Nucleic Acids Research vol35 no 7 pp 2333ndash2342 2007
[25] Y-S Jung YQian andXChen ldquoExamination of the expandingpathways for the regulation of p21 expression and activityrdquoCellular Signalling vol 22 no 7 pp 1003ndash1012 2010
[26] W Wang H Furneaux H Cheng et al ldquoHuR regulatesp21 mRNA stabilization by UV lightrdquo Molecular and CellularBiology vol 20 no 3 pp 760ndash769 2000
[27] U Wittig R Kania M Golebiewski et al ldquoSABIO-RKmdashdatabase for biochemical reaction kineticsrdquo Nucleic AcidsResearch vol 40 database issue pp D790ndashD796 2012
[28] R Milo P Jorgensen U Moran G Weber and M SpringerldquoBioNumbersmdashthe database of key numbers in molecular andcell biologyrdquo Nucleic Acids Research vol 38 supplement 1 ppD750ndashD753 2010
[29] I-C Chou and E O Voit ldquoRecent developments in parameterestimation and structure identification of biochemical andgenomic systemsrdquoMathematical Biosciences vol 219 no 2 pp57ndash83 2009
[30] A Saltelli K Chan and E M Scott Sensitivity Analysis JohnWiley amp Sons New York NY USA 1st edition 2000
[31] S Strogatz Nonlinear Dynamics and Chaos Applications toPhysics Biology Chemistry and Engineering With Applicationsto Physics Biology Chemistry and Engineering Westview PressBoulder Colo USA 2000
[32] P Zhou S Cai Z Liu and R Wang ldquoMechanisms generatingbistability and oscillations in microRNA-mediated motifsrdquoPhysical Review E vol 85 no 4 apart 1 Article ID 041916 9pages 2012
[33] S Marino I B Hogue C J Ray and D E Kirschner ldquoAmethodology for performing global uncertainty and sensitivityanalysis in systems biologyrdquo Journal of Theoretical Biology vol254 no 1 pp 178ndash196 2008
[34] L P LimNC Lau PGarrett-Engele et al ldquoMicroarray analysisshows that some microRNAs downregulate large numbers of-target mRNAsrdquo Nature vol 433 no 7027 pp 769ndash773 2005
[35] J Krutzfeldt N Rajewsky R Braich et al ldquoSilencing ofmicroRNAs in vivowith lsquoantagomirsrsquordquoNature vol 438 no 7068pp 685ndash689 2005
[36] MWPfafflGWHorgan andLDempfle ldquoRelative expressionsoftware tool (REST) for group-wise comparison and statisticalanalysis of relative expression results in real-time PCRrdquoNucleicAcids Research vol 30 no 9 p e36 2002
[37] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[38] T Abbas and A Dutta ldquop21 in cancer intricate networks andmultiple activitiesrdquo Nature Reviews Cancer vol 9 no 6 pp400ndash414 2009
[39] J Wang M Lu C Qiu and Q Cui ldquoTransmiR a transcriptionfactor microRNA regulation databaserdquo Nucleic Acids Researchvol 38 supplement 1 pp D119ndashD122 2010
BioMed Research International 15
[40] S Bandyopadhyay andM Bhattacharyya ldquoPuTmiR a databasefor extracting neighboring transcription factors of humanmicroRNAsrdquo BMC Bioinformatics vol 11 article 190 2010
[41] A Le Bechec E Portales-Casamar G Vetter et alldquoMIRNTN a framework integrating transcription factorsmicroRNAs and their targets to identify sub-network motifs ina meta-regulation network modelrdquo BMC Bioinformatics vol12 article 67 2011
[42] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyTheGene Ontology ConsortiumrdquoNature Genetics vol 25 no 1 pp 25ndash29 2000
[43] A I F Vaz and L N Vicente ldquoA particle swarm pattern searchmethod for bound constrained global optimizationrdquo Journal ofGlobal Optimization vol 39 no 2 pp 197ndash219 2007
[44] W H PressNumerical Recipes The Art of Scientific ComputingCambridge University Press Cambridge UK 2007
[45] J G Doench and P A Sharp ldquoSpecificity of microRNA targetselection in translational repressionrdquo Genes and Developmentvol 18 no 5 pp 504ndash511 2004
[46] B Kholodenko M B Yaffe and W Kolch ldquoComputationalapproaches for analyzing information flow in biological net-worksrdquo Science Signaling vol 5 no 220 2012
[47] F M Khan U Schmitz S Nikolov et al ldquoHybrid modeling ofthe crosstalk between signaling and transcriptional networksusing ordinary differential equations and multi-valued logicrdquoBiochimica et Biophysica Acta 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
BioMed Research International 3
Table 1 Overview of themethodology Key points in each step of themethodology and themain resources for constructingmiRNA-mediatedgene regulatory networks are given
Step 1 data retrievalRegulation types Resources
Transcriptional generegulation
TRED (httprulaicshleducgi-binTREDtredcgiprocess=home) a database that provides an integratedrepository for both cis- and transregulatory elements in mammalsTRANSFAC (httpwwwgene-regulationcompubdatabaseshtml) a database that collects eukaryotictranscriptional regulation comprising data on TFs their target genes and binding sitesThe UCSC table browser (httpgenomeucscedu) a popular web-based tool for querying the UCSC GenomeBrowser annotation tablesHTRIdb (httpwwwlbbcibbunespbrhtri) an open-access database for experimentally verified humantranscriptional regulation interactionsMIRNTN (httpmaiaunilumirontonphp) an integrative resource based on a metaregulation networkmodel including TFs miRNAs and genesPuTmiR (httpwwwisicalacinsimbioinfo miuTF-miRNAphp) a database of predicted TFs for humanmiRNAsTransmiR (http20238126151hmddmirnatf) a database of validated TF-miRNA interactionsmiRGen 20 (httpdianacslabecentuagrmirgen) a database of miRNA genomic information and regulation
Posttranscriptionalgene regulation
miRecords (httpmirecordsbioleadorg) a resource for animal miRNA-target interactionsTarbase (httpwwwmicrornagrtarbase) a database that stores detailed information for each miRNA-geneinteraction the experimental validation methodologies and their outcomesmiRTarBase (httpmirtarbasembcnctuedutw) a database that collects validated miRNA-target interactionsby manually surveying the pertinent literaturemiRWalk (httpwwwummuni-heidelbergdeappszmfmirwalk) a comprehensive database that providesinformation on miRNAs from human mouse and rat on their predicted as well as validated binding sites intarget genes
Protein-proteininteraction
HPRD (httpwwwhprdorg) a centralized platform to visually depict and integrate information pertaining todomain architecture posttranslational modifications interaction networks and disease association for eachprotein in the human proteomeSTRING (httpstring-dborg) a database of known and predicted protein interactions The interactionsinclude direct (physical) and indirect (functional) associationsMPPI (httpmipshelmholtz-muenchendeprojppi) a collection of manually curated high-quality PPI datacollected from the scientific literature by expert curatorsDIP (httpdipdoe-mbiuclaedudipMaincgi) a catalog of experimentally determined interactions betweenproteinsIntAct (httpwwwebiacukintactmainxhtml) a platform that provides a database system and analysis toolsfor molecular interaction dataReactome (httpwwwreactomeorg) an open-source open access manually curated and peer-reviewedpathway database
GO annotation
Amigo GO (httpamigogeneontologyorgcgi-binamigogocgi) the official GO browser and search enginemiR2Disease (httpwwwmir2diseaseorg) a manually curated database that aims at providing acomprehensive resource of miRNA deregulation in various human diseasesmiRCancer (httpmircancerecuedu) a miRNA-cancer association database constructed by text mining on theliteraturePhenomiR (httpmipshelmholtz-muenchendephenomir) a database that provides information aboutdifferentially expressed miRNAs in diseases and other biological processesmiRGator (httpmirgatorkobicrekr) a novel database and navigator tool for functional interpretation ofmiRNAsmiRo (httpferrolabdmiunictitmiro) a web-based knowledge base that provides users withmiRNA-phenotype associations in humans
Step 2 network construction and visualization(i) Visualize regulatory interactions in platforms such as CellDesigner and Cytoscape that support standardized data formats(ii) Calculate confidence scores for assessing reliability of interactions in gene regulatory networksStep 3 model construction and calibration(i) Formulate equations using rate equations(ii) Fix parameter values using available biological information(iii) Estimate the other unknown and immeasurable parameter values using optimization methods which can minimize the distancebetween model simulations and experimental data such as time course qRT-PCR and western blot dataStep 4 model validation and analysis(i) Design new experiments and generate new data to verify the calibrated model(ii) Study complex properties and behavior of the system
4 BioMed Research International
provide experimentalists with different levels of confidencethus differing confidences can be reflected using differentvalues for 119878
119898based on the experience of experimentalists
Of note although the confidence scores cannot be directlyconverted into a mathematical model with the help of thesescores we can discard non-reliable putative interactions togenerate the ultimate version of a gene regulatory networkThe final version of the network can be further analyzedto identify regulatory motifs like feedforward loops (FFLs)for example with the help of the Cytoscape plugin NetDS[21] Thereafter the complete network or parts of it canbe converted into a detailed mechanistic model which isdescribed in detail in the following section
(3) Model Construction and Calibration After the construc-tion visualization and refinement of a gene regulatory net-work it is converted into a detailedmechanistic model whichenables the investigation of unanswered biological questionsand validation of hypotheses Ordinary differential equations(ODEs) describe how processes of synthesis biochemicalmodification andor degradation affect the temporal con-centration profile of biochemical species like proteins RNAsand metabolites An ODE model can be constructed usingappropriate kinetic laws such as the law ofmass-action whichstates that the rate of a chemical reaction is proportional tothe probability that the reacting species collideThis collisionprobability is in turn proportional to the concentration ofthe reactants [22] A general representation of ODE-basedmodels using mass-action kinetics is given by the followingequation
119889119909119894
119889119905=
119898
sum
120583=1
119888119894120583sdot 119896120583sdot
119899
prod
119895=1
119909119895
119892120583119895 119894 isin 1 2 119899 (2)
where 119909119894represents state variables which denote the molar
concentration of the 119894th biochemical specie Every biochem-ical reaction 120583 is described as a product of a rate constant(119896120583) and biochemical species (119909
119895 119895 isin 1 2 119899) that are
involved in this reaction 119888119894120583 the so-called stoichiometric
coefficients relate the number of reactant molecules con-sumed to the number of product molecules generated inthe reaction 120583 119892
120583119895denotes kinetic orders which are equal
to the number of species of 119909119894involved in the biochemical
reaction 120583 The rate constants kinetic orders and the initialconditions of state variables are defined as model parametersBesides mass-action kinetics other kinetic rate laws such asMichaelis-Menten kinetics Hill equation and power-laws arealso frequently used in ODE models
After ODEs are formulated the model requires to becalibrated a process by which parameter values are adjustedin order to make model simulations match experimentalobservations as good as possible To do so there are twopossible means characterization of parameter values usingavailable biological information or estimation of parametervalues using optimization methods Some parameter valuescan be directly measured or obtained from literature ordatabases For example the half-life (119905
12) of some molecules
(eg protein) can be measured in vitro via western blot-ting This information can be used to characterize their
degradation rate constants through the equation 119896deg =
(ln 211990512) The database SABIO-RK provides a platform for
modelers of biochemical networks to assemble informationabout reactions and kinetic constants [27] However formost model parameters whose values cannot be measuredin laboratories or be accessed from literature or databasesparameter estimation is a necessary process to characterizetheir values Before running parameter estimation initialparameter values and boundaries should be set within phys-ically plausible ranges To do so the database BioNumbersprovides modelers with key numbers in molecular and cellbiology ranging from cell sizes to metabolite concentrationsfrom reaction rates to generation times from genome sizesto the number of mitochondria in a cell [28] After parameterestimation unknown parameter values are determined usingoptimization methods which can minimize a cost functionthat measures the goodness of fit of the model with respect togiven quantitative experimental data sets Parameter estima-tion using optimization algorithms is an open research fieldin which several methods have been developed accordingto the nature and numerical properties of biological dataanalyzed The discussion for choosing proper optimizationmethods for parameter estimation is beyond the scope ofthis paper but the interested reader is referred to the paperpublished by Chou and Voit [29]
(4) Model Validation and Analysis Usually the model sim-ulations are compared with the experimental data used forthe parameter estimation but a good agreement betweenboth is not enough to guarantee the predictive ability of themodel Therefore it is necessary to validate the model withdata sets that are not used during the parameter estimationThis process is called model validation and can ensure morereliable and accurate model predictions To do so the datagenerated in new experiments or extracted from literature arecompared with model simulations which are obtained afterconfiguring the model according to the new experimentalsettings Once a model is validated it can be used to per-form predictive simulations which are helpful to studythe dynamic properties of biochemical systems guide thedesign of new experiments in the laboratory and formulateadditional hypotheses In addition tools such as sensitivityand bifurcation analysis can be used to study complex proper-ties and behavior of the modeled system Sensitivity analysisis used to evaluate the influence of model parameters (eginitial concentrations of the state variables and rate constants)on model outputs such as the temporal behavior of networkcomponents [30] Bifurcation analysis is employed to detectcontrol parameters (also known as bifurcation parameters)whose variations are able to change drastically the dynamicalproperties of the biochemical system as well as the stabilityof its fixed points [31] The application of these tools tomathematical modeling is beyond the scope of this paper butthe interested reader is referred to the publication of Zhouet al [32] and Marino et al [33]
212 Experiment Methods As mentioned in the previoussections after a model is established it can be calibrated
BioMed Research International 5
and validated using temporal experimental data To do sothe data can either be derived from literature or generatedto calibrate the model by own experiments In case of ODEmodels the most suitable data for model calibration isquantitative time-series data obtained from perturbation orquantitative dose-response experimentsThe experiments inwhich time-series data are measured for different regulators(such as miRNAs and TFs) of a gene regulatory network canbe obtained using the techniques described in the subsectionsbelow
(1) qRT-PCR Quantitative real time polymerase chain reac-tion (qRT-PCR) has been used to identify mRNAs regulatedby overexpression or silencing of a specific miRNA [3435] miRNAs typically exhibit their regulatory effects byassociating with specific 31015840 UTR regions of the mRNAscalled miRNA seed regions [34] This association can leadeither to a temporary inhibition of translation or completedegradation of the mRNA in which case qRT-PCR mediateddetection is beneficial For this the cells are transfectedwith the miRNA and non-targeting control oligonucleotidesat an appropriate concentration using either a lipid basedtransfection reagent or nucleofection The transfected cellsare then incubated for the necessary time periods (eg 24 h48 h 72 h etc) after which lysates are prepared and total RNAis extracted 1000ndash2000 ng of the RNA is then converted intocDNA using a reverse transcription kit Taqman qRT-PCR isperformed using 10ndash20 ng of cDNA and primers labeled withfluorescence probes to detect the transcriptional levels of thetarget mRNA Housekeeping genes like GAPDH and HPRTare used as endogenous controls for data normalizationRelative expression at different time points is determinedby comparing the Ct values of the miRNA transfected cellswith Ct values of non-targeting control transfected cells andexpressed as relative expression [36] However qRT-PCR inspite of being highly sensitive is applicable specifically underconditions of complete or partial degradation of the targetmRNA miRNA-mediated inhibition in translation can bebetter demonstrated by techniques like immunoblotting
(2) Western BlotImmunblotting Western blotting or proteinimmunoblotting is a technique to detect the expression of agene at protein level This technique is particularly useful indetermining the regulatory effects of a miRNA on expressionof a target gene which is temporarily inhibited For thisthe cells are transfected with a miRNA or an antagomiRas mentioned above and cell lysates are prepared at appro-priate time points using protein lysis buffers (eg Radio-Immunoprecipitation Assay buffer or RIPA) containing lysisagents like Dithiothreitol (DTT) and protease inhibitors Theproteins from each sample are quantitated using Bradfordor BCA reagents and compared with bovine serum albumin(BSA) standards for accurate protein estimation 20ndash40 120583g ofprotein is then loaded and resolved on a sodium dodecyl sul-fate polyacrylamide gel (SDS-PAGE) along with a pre-stainedprotein marker The protein bands are then transferred ontoa nitrocellulose membrane followed by incubation with theappropriate primary and secondary antibodies linked tofluorescent dyes or horse radish peroxidase enzyme (HRP)
The protein expression is then analyzed using either fluores-cence or chemiluminiscenceHRP substrates in gel documen-tation systems (eg LI-COR Odyssey) Housekeeping geneslike 120573-actin or 120573-tubulin are used for protein normalizationThe time point for maximum target gene suppression gen-erally varies depending on the number of miRNA bindingsites at the 31015840 UTR of the target gene and the extent ofcomplementarity of the seed region [37] Immunoblotting isa widely used technique to provide confirmatory evidence forthe inhibitory effects ofmiRNA at the protein level but it failsto explain the underlying interaction mechanisms
(3) Reporter Gene Assay As each miRNA can inhibit theexpression of a large number of genes regulation of a par-ticular target gene may either be by direct interaction or bean indirect consequence of it In direct regulation a miRNAbinds to the complementary sequences at the 31015840 UTR of atarget gene and thereby suppresses its expression As a con-sequence of this the expression levels of a number of down-stream genes (indirect targets) are also dysregulated makingit crucial to differentiate between primary and secondarymiRNA targets To determine the interaction specificitya reporter construct (luciferase) with intact or mutated 31015840UTR of the target gene cloned at the 51015840 end is co-transfectedinto the cells along with the miRNA The regulatory effect ofthe miRNA on the target gene expression is then measuredusing the expression of a reporter gene In the absence ofthe appropriate binding sequences (mutated 31015840 UTR) themiRNA cannot suppress the reporter mRNA suggesting thatthe suppressive effect of the miRNA is mediated by a directinteraction The reporter activity can be analyzed at differenttime points such as 24 hr 48 hr and 72 hr to determine thetime dependent suppression of a target gene expression bya miRNA
22 Case Study The Regulation of p21 by Multiple andCooperative miRNAs
221 Construction and Visualization of p21 Regulatory Net-work By using the approach described above we investi-gated the regulation of p21 by its multiple targeting miRNAsp21 also known as cyclin-dependent kinase inhibitor 1(CDKN1A) is a transcriptional target of p53 It is requiredfor proper cell cycle progression and plays a role in cell deathDNA repair senescence and aging (reviewed in [38]) Inter-estingly p21 was the first experimentally validated miRNAtarget hub which is a gene that is simultaneously regulatedby many miRNAs This made it an ideal candidate for a casestudy of our approach [23] To do so we first constructeda p21 regulatory network using the following steps
(a) We extracted miRNA-target interactions from thepublication of Wu et al [23] where a list of predictedp21-regulating miRNAs was subjected to experimen-tal validation
(b) Experimentally verified TFs of p21 were extractedfrom literature and putative TFs having conservedbinding sites in the 5 kb upstream region of the p21open reading frame were extracted from UCSC table
6 BioMed Research International
browser A list of TFs controlling the expression of themiRNAswas constructed using information of exper-imentally proven TF-miRNA interactions extractedfrom TransmiR (release 10) [39] In addition wegenerated a list of putative TFs of miRNAs withbinding sites in the 10 kb upstream region of themiRNA genes using information from the databasesPuTmiR (release 10) [40] and MIRNTN (version121) [41] and from the table of TFs with conservedbinding sites in theUCSC genome browser (hg18) [9]
(c) Information about protein interactions was extractedfrom theHuman Protein ReferenceDatabase (HPRDrelease 90) [15] and the STRING database (release90) [16] Only the experimentally verified p21-proteininteractions were used to construct the network
(d) Additionally we associated the TFs in the network tonine biological processes based on the Gene Ontol-ogy (GO) [42] The corresponding GO terms werecell proliferation cell apoptosis immune responseinflammatory response cell cycle DNA damage cellsenescence DNA repair and cell migration
Next we visualized the network in CellDesigner andcomputed a confidence score for each interaction in thenetwork (Figure 1) The confidence scores provide us withthe reliability of the interactions considered in p21 regulatorynetwork With the help of these scores we discarded non-reliable interactions and constructed the mechanistic modelaccounting for p21 regulation by its targeting miRNAsBesides the interested experimentalists can further use thisinformation to choose reliable interacting candidates of p21for designing relevant experiments The scores for eachinteraction of p21 regulatory network are shown in Table 2
222 Mechanistic Modeling of p21 Regulation byIts Targeting miRNAs
(1) Model Construction After constructing the regulatorynetwork a detailed mechanistic model of ODEs whichdescribes the biochemical reactions underlying the regula-tion of p21 was established We chose the ODE modelingapproach as it is a simple formalism for describing temporaldynamics of biochemical systems and a wide range of toolsare available to explore their properties Precisely the modelconsidered themRNA(mp21 (3)) andprotein (p21 (6)) of themiRNA target hub p21 the p21-targeting miRNAs (miR
119894 119894 isin
(1 15) (5)) and the complexes formed by p21 mRNAand miRNA [mp21 | miR
119894] (4) Altogether the model is
constituted by 32 state variables and 64 parameters
119889mp21119889119905
= 119896mp21syn sdot 119891act (TFmp21)
minusmp21 sdot (119896mp21deg +sum
119894
119896complex
119894
ass sdotmiR119894)
(3)
119889 [mp21 | miR119894]
119889119905
= 119896complex
119894
ass sdotmp21 sdotmiR119894minus 119896
complex119894
deg sdot [mp21 | miR119894]
(4)
119889miR119894
119889119905= 119896
miR119894syn sdot 119891act (TFmiR119894)
minusmiR119894sdot (119896
miR119894deg + 119896
complex119894
ass sdotmp21) (5)
dp21119889119905
= 119896p21syn sdotmp21 + 119896p21deg sdot p21 (6)
mp21Total = mp21 +sum119894
[mp21 | miR119894] (7)
For mp21 processes considered in the model were (i) itssynthesis (119896mp21
syn ) mediated by TFs (119891act(TFmp21)) (ii) itsdegradation (119896mp21
deg ) and (iii) its association with a miRNA(119896complex
119894
ass ) For each miR119894 processes considered were (i)
the synthesis (119896miR119894syn ) mediated by TFs (119891act(TFmiR119894)) (ii) the
degradation (119896miR119894deg ) and (iii) the association with the p21
mRNA target (119896complex119894
ass ) For each [mp21 | miR119894] complex
processes considered were (i) the formation of the complexby a miR
119894and the p21 mRNA (119896complex
119894
ass ) and (ii) the complexdegradation (119896complex
119894
deg ) For p21 processes considered were(i) its synthesis (119896p21syn ) and (ii) its degradation (119896p21deg) Anadditional algebraic equation accounting for the total mea-surable amount of p21 mRNA (mp21Total) was also includedThe SBML file of the model is available for download athttpwwwsbiuni-rostockdeuploadstx templavoilap21TargetHub 03092013xml
(2) Model Calibration and Validation For model calibrationwe fixed some parameter values using published data andestimated the other unknown parameters using the time-series data published by Wu et al [23] in which the p21mRNA (northern plot) and protein levels (western plot)were measured 48 hr after transfection of individual p21-targeting miRNAs into human embryonic kidney 293 cellsThe unknown parameter values were estimated using aniterative method combining global (particle swarm patternsearch) [43] and local (downhill simplex method in multi-dimensions) [44] optimization algorithms For each miR
119894
considered in the model the method minimizes the distancebetween model simulations and experimental data using thefollowing cost function
119865miR119894cost =
[mp21miR119894sim (119905) minusmp21miR119894
exp (119905)]2
(120575mp21119894
)2
+
[p21miR119894sim (119905) minus p21miR119894
exp (119905)]2
(12057511990121
119894)2
119894 isin [1 15]
(8)
BioMed Research International 7
1 2 4 1 2 4 5 6 1 2 3 7
E2F TP53 RELA1 2 4 7 2 3 7 6
STAT5 SRF1 2 3 4 1 2
JUN STAT1
SP1 RUNX1 TFAP2A
Common TFs Modulation
CDKN1A
1 2 1 2 4 1 4
1 4 1 2 3 7
1 2 5 2 4 5 7 1
VDR
CUX1
E2A
RAR BRCA1
TP63 TP73 TBX2
STAT6 SP3 STAT3
STAT2 RUNX2
EHMT2 p21 TFs
1 2 4 5 8 9
CEBP120573NF120581B1
(1) Cell proliferation (6) Cell senescence
(2) Apoptosis (7) Inflammatory response
(3) Immune response (8) DNA repair
(4) Cell cycle control (9) Cell migration
(5) DNA damage
miRNAmiRNA
Translational repressionmiRNA
Translation
Transcription
Association
miR-132 miR-28-5p
miR-299-5p miR-298
miR-345 miR-125a-5p
miR-208a
miRNA
miRNADeadenylation followed by decay
Protein-protein interaction
miR-93 miR-654-3p
miR-657 miR-572
miR-423-3p miR-363
miR-639 miR-515-3p
p21
CCND2 CDK3 GMNN CCNE1 POLD2 CSNK2B C1orf123 CCDC85B
PIM1 CCNB2 KIAA0432 GADD45B HDAC11 ITGB1BP3 TTLL5 SLC25A11
GADD45G CDC7 BCCIP CDK1 GNB2LI RAB1A psmA3 TK1
CDC45 CCNB1 CCNA2 GADD45A MAPK8 TSG101 SET CIZ1
CDK14 CDK5 MDM2 CDC27 RNF144B ESR1 PRKCH RPA1
CDC20 SKP2 CDK2 CDK4 DHX9 PCNA CDC5L CCDC5
CCND3 CDK6 CCNE2 CCNA1 TEX11 MAP3K5 MCM10 Caspase-3
CDC6 CCND1 XRCC6 PARP1 DAPK3 NR4A1 CEBPA
Interacting proteins
AKT1
miRNA genes
5 8 1 2 1 2 4 3 1 2 3 3 2 2 3 3 1 4 1 3 1 4
1 4
NFATC2 YY1 CREB1 CEBPA ARNT
FOXCI1
ELK1 SOX9 FOS PAX5 MAX MEF2A RFX1 JUNB AKT1 GATA3 MYC NRSFEGR1
E47 PPARG POU3F2 FOXF2FOXL1 MZF1 NFIC SRY
FOXC1
HNFA4 NFYA MYF1 BACF1 ATF6 ZID RORA POU2F1
miRNA TFs
11 2 34
12 31 5 11
HNF1A
1 2
RB1
Cellular processes
Figure 1 p21 regulatory network The network contains several layers of regulators of p21 TFs (light blue and red boxes) miRNAs (darkblue and green boxes) and proteins (grey boxes) In each big box there are small boxes which represent individual components of this layerof regulation miRNAs are classified into two groups according to the mechanisms by which the expression of p21 is repressed One groupcauses p21 translation repression (green box) These miRNAs bind to p21 mRNA resulting in the repressed translation of p21 but unchangedmRNA expression level The other group of miRNAs (dark blue box) decreases the stability of p21 mRNA by modifying its structure leadingto mRNA decay and finally the downregulation of p21 TFs are classified into three groups p21 TFs (red) miRNA TFs (yellow) and theircommon TFs (light blue) The p21 interacting-proteins are framed in the grey boxes The purple boxes represent nine processes and the TFsassociated with these processes are indicated in the ellipses above them using corresponding figures This data is adapted from our previouspublication [4]
where mp21miR119894sim (119905) and mp21miR119894
exp (119905) represent the simulatedp21 mRNA and protein expression levels for each miR
119894at
time point 119905 p21miR119894sim (119905) and p21miR119894
exp (119905) represent the mea-sured value for each miR
119894at time point 119905 and their standard
deviations are 120575mp21119894
and 120575p21119894
Here 119905 is the time point (48 hr)after overexpression of the individual miRNAs in embryonickidney 293 cells at which the expression levels of the p21 andits mRNA were measured [23] The model calibration resultsare shown in Figure 2(a) and the obtained parameter valuesare listed in Table 3
Experimental results showed that a stronger repressionof the target gene can occur when two miRNA binding siteson the target mRNA are in close proximity [24 45] To testthe consequences of this hypothesis we predicted cooperativemiRNA pairs for p21 with seed site distances between 13ndash35 nt in the p21 31015840 UTR To substantiate the cooperativeeffect associated with pairs of miRNAs we introduced agroup of new state variables ([mp21 | miR
119894| miR
119894]) into
the original model These state variables account for the
ternary complexes composed of p21 mRNA and two puta-tively cooperating miRNAs (miR
119894and miR
119895) For these new
variables processes considered are (i) the association of p21mRNA with miR
119894and miR
119895into a complex (119896co-complex
119894119895
ass )and (ii) the degradation of the complex (119896co-complex
119894119895
deg ) Afterexpansion the corresponding modified and new ODEs arelisted below
dmp21119889119905
= 119896mp21syn sdot 119891act (TFmp21)
minusmp21 sdot (119896mp21deg +sum
119894
119896complex
119894
ass sdotmiR119894
+sum
119894119895
119896co-complex
119894119895
ass sdotmiR119894sdotmiR119895)
(9)
8 BioMed Research InternationalTa
ble2Th
econ
fidence
scores
forthe
interactions
inp21regulatorynetworkTh
enetworkconsistso
ffou
rtypes
ofinteractionsm
iRNA-
p21TF
-miRNAT
F-p21andp21-p
rotein
(a)miRNA-
p21interactio
nscores
miRNA
miR-125a-5p
miR-132
miR-208a
miR-28-5p
miR-298
miR-299-5p
miR-345
miR-363
Score
073
073
073
080
073
080
073
073
miRNA
miR-423-3p
miR-515-3p
miR-572
miR-639
miR-654-3p
miR-657
miR-93
Score
085
073
073
073
073
080
095
(b)TF
-miRNAinteractionscores
Score
Score
Score
Score
Score
Score
miR-208
miR-132
miR-657
miR-125a-5p
miR-28-5p
miR-93
MAX
027
EGR1
093
TFAP2
A034
EGR1
086
MZF
1028
TFAP2
A034
EGR1
029
CREB
1086
SP1
034
PAX5
024
E47
027
SOX9
024
ARN
T027
ARE
B6027
MZF
1034
NRS
F024
FOXC
1031
GAT
A3
027
RFX1
024
E47
024
MAX
029
ZID
024
FOXI1
024
MZF
1031
ATF6
024
ELK1
027
USF1
027
PPARG
027
SRY
027
SP1
034
YY1
027
EGR2
029
EGR1
034
GAT
A3
034
JUN
029
POU2F1
024
JUNB
024
FOS
024
RELA
029
MZF
1034
POU2F1
024
EGR1
029
POU2F1
024
NFIC
024
miR-654-3p
TP53
034
SRF
027
CEBP
A024
NFIC
027
RELA
027
YY1
027
NFIC
034
CEBP
A029
NFY
A024
miR-345
miR-423-3p
BACH
1024
miR-299-5p
RORA
024
RUNX1
024
SP1
034
BACH
1027
FOXL
1029
CREB
1024
miR-363
STAT
1024
RELA
027
STAT
1027
SRY
024
SRY
027
E2F1
062
E2F1
085
NFA
TC2
024
POU3F2
024
NFA
TC2
024
SRF
029
FOXC
1033
MYC
084
HNF4
A031
STAT
5A027
FOS
024
RUNX1
027
MZF
1034
miR-298
NFY
A024
SRF
027
POU2F1
027
miR-572
miR-639
JUNB
031
POU2F1
024
MEF
2A027
NF120581
B1024
FOXF
2024
NFY
A024
NFIC
031
POU2F1
024
HNF1A
024
(c)TF
-p21
interactionscores
TF(verified)
SP1
SP3
E2F1
Runx
1Ru
nx2
STAT
1E2
AST
AT5
TP53
TP63
TP73
Score
100
100
077
077
077
072
072
067
067
067
067
TF(verified)
STAT
3CU
X1TF
AP2
ABR
CA1
VDR
RARA
CEB
P120572CEB
P120573RB
1Tb
x2EH
MT2
Score
066
066
060
060
060
060
060
060
060
060
060
TF(putative)
NF120581
B1RE
LAST
AT2
STAT
6SR
FScore
044
044
044
044
043
(d)p21-p
rotein
interactionscores
Protein
TP53
PCNA
CASP
3CC
NA1
CCND1
SKP2
BCCI
PCC
NA2
CCND2
CCNE2
AKT
1C1
orf123
CCDC8
5BCC
NB1
CCNB2
CCND3
Score
100
092
087
087
087
087
081
081
081
081
072
072
072
072
072
072
ProteinCC
NE1
CDC4
5CD
C5L
CDC6
CDC7
CDK1
CDK14
CDK2
CDK3
CDK4
CDK6
CEBP
ACI
Z1CS
NK2
A1
CSNK2
BDAPK
3Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
ESR1
GADD45A
GADD45B
GADD45G
GMNN
GNB2
L1HDAC
11ITGB1BP
3MAP3
K5MAPK
8MCM
10PA
RP1
PIM1
POLD
2PS
MA3
RAB1A
Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
SET
SLC2
5A11
STAT
3TE
X11
TK1
TSG101
TTLL
5XR
CC6
CDC2
0CD
C27
CDK5
DHX9
MDM2
NR4
A1
PRKC
HRP
A1
Score
072
072
072
072
072
072
072
072
056
056
056
056
056
056
056
056
Verifi
edexp
erim
entally
verifi
edinteraction
putativ
epredictedinteraction
BioMed Research International 9
Con
trol
miR
-298
miR
-208
am
iR-1
32m
iR-2
8-5p
miR
-125
-5p
miR
-299
-5p
miR
-345
miR
-93
miR
-423
-3p
miR
-515
-3p
miR
-363
miR
-657
miR
-639
miR
-572
miR
-654
-3p
0
05
1
15m
RNA
expr
essio
n (a
u)
Data Model Data ModelC
ontro
lm
iR-2
98m
iR-2
08a
miR
-132
miR
-28-
5pm
iR-1
25-5
pm
iR-2
99-5
pm
iR-3
45m
iR-9
3m
iR-4
23-3
pm
iR-5
15-3
pm
iR-3
63m
iR-6
57m
iR-6
39m
iR-5
72m
iR-6
54-3
p
0
02
04
06
08
1
Prot
ein
expr
essio
n (a
u)
(a)
DoxorubicinCellWell
Control miR-572 miR-93
Measure p21 expression at the time points of
miR-572+ miR-93
0 2 4 6 8 24hr
miR-572miR-93
Immunoblotting
(b)
0 4 8 12 16 20 240
4
8
12
16
20
Time (hr)
p53
Actin
Doxo (hrs)
fac
t(TF
mp2
1)
C 0 1 2 4 6 8 24
1 08 12 19 29 32 3 16
(c)
0 2 4 6 8 2402468
1012
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)
p21
expr
essio
n (a
u)
02468
1012
miR-572Control
02468
1012
miR-93
02468
1012
Model Data
0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24
1 23 46 107124 124 23 28 48 66 53 31 08 16 21 38 50 18 06 06 12 25 27 12
miR-572+ 93Doxo (hrs)
p21
Actin
p21
Actin
p21
Actin
p21
Actin
(d)
Figure 2 Model calibration and validation (a) Model calibration The figures show the relative change of the p21 mRNA and proteinexpression levels after overexpression of the indicated miRNAs (Model model simulation Data experimental data) These data werenormalized to the control group in which the p21 mRNA and protein expression levels were measured when the miRNAs were normallyexpressed (au arbitrary unit) (b) Experimental workflow In the experiments Sk-Mel-147 cells were seeded in six well plates Then maturemiRNA mimics were transfected individually at a concentration of 100 nM (miR-572 and miR-93) or in combination at 50 nM each (miR-572 + miR-93) After 48 hr transfection with miRNA mimics the cells were pulse treated with 250 nM doxorubicin for 1 hour after whichnormal growth medium was replenished The immunblotting were performed to measure p21 expression at 0 2 4 6 8 and 24 hr post-doxorubicin treatment (c) Temporal dynamics of p21 transcriptional function Afterdoxorubicin treatment the expression of p53 a TF ofp21 was measured using immunblotting and these data were used to characterize the transcriptional function of p21 using MATLAB linearinterpolation function (d) Model validation We measured the expression of p21 protein in response to genotoxic stress in the four scenariosas described in the main text The measured data (Data) were compared with the model simulations (Model) The figures (a) (c) and (d) areadapted from our previous publication [4]
119889 [mp21 | miR119894| miR
119895]
119889119905
= 119896co-complex
119894119895
ass sdotmp21 sdotmiR119894sdotmiR119895
minus 119896co-complex
119894119895
deg sdot [mp21 | miR119894| miR
119895]
(10)
To model stronger repression of the target gene bycooperating miRNAs we assumed a stronger association rateconstant for the complex [mp21 | miR
119894| miR
119895] which is
equal to the sum of their individual association rate constants
(119896co-complex119894119895
ass = 119896complex
119894
ass + 119896complex
119895
ass ) Similarly the degradationrates of the complexes [mp21 | miR
119894| miR
119895] were assumed
to be equal to the sum of degradation rate constants ofsingle miRNA binding complexes (119896co-complex
119894119895
deg = 119896complex
119894
deg +
119896complex
119895
deg ) However it has to be noted that these addedequations are an abstract description ofmiRNAcooperativitybecause the details of this mechanism are not yet known
To experimentally validate the capability of our model topredict the relative p21 concentrations regulated by coopera-tivemiRNAs we selectedmiR-572 andmiR-93 as a case study
10 BioMed Research International
Table 3 Initial concentrations of model variables and model parameter values Based on the experimental data the p21-targeting miRNAsverified byWu et al [23] were divided into two groups the translation repression group (marked with asterisk) and the mRNA deadenylationgroup AmiRNAwas classified into themRNA deadenylation group if its overexpression can result in 20 ormore downregulation of the p21mRNA level (ie p21mRNA levelle 08 the basal level is 1) otherwise it was classified into the translation repression group For the translationrepression group only 119896complex119894
ass was estimated and 119896complex119894deg was fixed For the other group both 119896complex119894
deg and 119896complex119894ass were estimatedThe initial
concentrations of p21 and mp21 were set to 1 and this value was used as their basal expression levels During the parameter estimation theinitial concentrations of p21-targeting miRNAs were set to 100 because in the publication [23] the expression levels of p21 and mp21 weremeasured after the individual introduction of the p21-targeting miRNAs with amount of 100 nM Due to the lack of biological information tocharacterize the transcriptional activation function (119891act) of p21 and its targeting miRNAs the corresponding functions were assumed to be1 for simplicity The data is adapted from our previous publication [4]
Initial concentration of variables and TF functionsVariable Description Initial concentration (au)p21 p21 protein 1
mp21 p21 mRNA 1
miR119894=115
p21-targeting miRNAs 100
[mp21 | miR119894=115
] Complexes formed by miR119894and mp21 0
119891act(TFmp21) p21rsquos transcriptional activation function 1119891act(TFmiR119894 (119894=115)) The transcriptional activation function of miR
1198941
Fixed parameter valuesParameter Description Value (hrminus1) Reference119896mp21syn Synthesis rate constant of mp21 01155 fixed119896mp21deg Degradation rate constant of mp21 01155 [24]119896miR119894syn (119894 = 1 15) Synthesis rate constant of miR
11989400289 fixed
119896miR119894deg (119894 = 1 15) Degradation rate constant of miR
11989400289 [25]
119896p21syn Synthesis rate constant of p21 13863 fixed119896p21deg Degradation rate constant of p21 13863 [26]
Estimated parameter values
miRNA (statevariable)
119896complex119894deg
(119894 = 1 15) (hrminus1)
119896complex119894ass
(119894 = 1 15)
(au minus1sdot hrminus1)
119865miR119894cost
(119894 = 1 15)Experimental data of p21 (protein mRNA plusmn SD)
miR-298 (miR1)lowast 01155 00254 34119890 minus 004 (016 1074 plusmn 0025)
miR-208a (miR2)lowast 01155 00041 20119890 minus 003 (051 1192 plusmn 0022)
miR-132 (miR3)lowast 01155 00275 24119890 minus 003 (015 121 plusmn 0147)
miR-28-5p (miR4)lowast 01155 00119 59119890 minus 003 (028 135 plusmn 006)
miR-125-5p (miR5)lowast 01155 00018 18119890 minus 003 (069 085 plusmn 0051)
miR-299-5p (miR6)lowast 01155 00080 18119890 minus 004 (036 095 plusmn 0038)
miR-345 (miR7)lowast 01155 00051 11119890 minus 004 (046 096 plusmn 0039)
miR-93 (miR8) 01564 00235 41119890 minus 014 (017 07776 plusmn 003)
miR-423-3p (miR9) 09118 00055 28119890 minus 009 (044 05102 plusmn 011)
miR-515-3p (miR10) 02098 00253 12119890 minus 013 (016 0616 plusmn 0037)
miR-363 (miR11) 02261 00399 22119890 minus 014 (011 056 plusmn 015)
mR-657 (miR12) 03465 00158 21119890 minus 014 (023 048 plusmn 012)
miR-639 (miR13) 04305 00327 18119890 minus 017 (013 036 plusmn 0084)
miR-572 (miR14) 03039 00360 94119890 minus 023 (012 045 plusmn 0044)
miR-654-3p (miR15) 97485 00024 30119890 minus 014 (064 063 plusmn 0053)
These two miRNAs were chosen because their predictedtarget sites in the p21 31015840 UTR are in close proximity to eachother and thereby they can induce cooperative repression onp21 as suggested in [24] The experiments were performed asfollows
(i) Melanoma cells (Sk-Mel-147) were transfected withthe mature miRNAmimics of the twomiRNAs eitherindividually (100 nM) or in combination (50 nMeach) whereas untreated cells were used as control(Figure 2(b))
BioMed Research International 11
p21
FOXF2
miR-572
Activation
Inhibition120572
Incoherent FFL + cascaded regulation
AF2
miR-93
(a)
miR-93
miR-572
TP53
MYC
p21
FOXF2
Coherent FFL + cascaded regulation
(b)
Saturation
Pulse
p21
(au
)
p21
(au
)Time (hr) Time (hr)
Saturation Pulse
Pulse
Saturation
Saturation
di
(au
)
102
101
102
100
101
10minus1
10010minus2
10minus1
102
101
100
10minus1
10minus2
FOXF2 (au) FOXF2 (au)
TP53
(au
)
1
09
08
07
06
05
04
03
02
01
10minus2 10210110010minus110minus20
AF2120572
(au
)
(c)
Figure 3 Different p21 dynamics in different network motifs We ran simulations to show the different dynamics of p21 for two differentnetwork motifs (a) and (b) Through simulations two dynamical patterns of p21 were identified saturation and pulse ((c) top) For eachnetwork motif the corresponding distributions of the two dynamical patterns were plotted ((c) bottom) For different combinations ofthe transcriptional strengths the normalized distance (119889
119894) between peaks (119901
119894) and steady states (ss
119894) of p21 is determined by the equation
119889119894= (119901119894minus ss119894)119901max 119901max = max(119901
1 119901
119899) If 119889
119894= 0 for the corresponding combination of transcriptional strengths the p21 dynamics is
saturation otherwise it is pulse The regions showing different dynamical patterns of p21 are separated using the white lines
(ii) Next the cells were treated with doxorubicina genotoxic-stress inducing agent The agent canupregulate the expression of p53 which is a knownTFof p21 and therefore it can result in the upregulationof p21 (Figure 2(c))
(iii) After doxorubicin treatment the expression levels ofp21 were measured by immunoblotting at differenttime points (0 2 4 6 8 24 hr) The p21 expressionvalues were normalized based on the p21 expression
level in the control group measured at time point 0 hr(Figure 2(d))
By doing so we obtained the p21 response after genotoxicstress in four scenarios (1) endogenous miRNA expression(Control) (2) overexpression of miR-572 (3) overexpressionof miR-93 and (4) both miRNAs moderately overexpressedThereafter we derived a model of seven ODEs based onthe original equations which was configured according tothe designed experiments making the simulation results
12 BioMed Research International
00001
0001
001
01
1Cell cycle Cell growth DNA damage Senescence Apoptosis
G1G2
MS
p21
expr
essio
n (a
u)
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
log10
wow
Figure 4 p21 expression regulated by cooperative miRNAs for different cellular processesThe associations of themiRNAs with these cellularprocesses were derived from GO terms of their TFs A miRNA was supposed to be expressed (in bold black font) in a cellular process onlyif its TF is related to the corresponding GO term of this process The p21 expression levels are computed for each process with (w)without(wo) considering the cooperative effect among the p21 targeting miRNAs
comparable with the experimental data As shown in theFigure 2(c) the simulations are in good agreement with theexperimental observations The individual overexpression ofmiR-572 or miR-93 led to the reduction of the upregulationof p21 response after genotoxic stress induction The twomiRNAs cause different degrees of repression due to theirdifferent repression efficiencies on p21 Interestingly the com-bined overexpression of both miRNAs induced the strongestdownregulation of p21 and therefore verifying the hypothesisof their cooperative regulation of p21 Above all the resultsnot only validated themodel but also demonstrated the abilityof our method to identify cooperative miRNA pairs for p21
(3) Predictive Simulations As there are abundant of networkmotifs such as FFLs in p21 regulatory network and thesenetwork motifs are important for determining p21 dynamicsit is interesting to investigate the dynamics of p21 in networkmodules where FFLs are involved To do so two networkmodules including both miR-93 and miR-572 and theirTFs were exemplified In Figure 3(a) the network modulecontains an incoherent FFL composed by AF2120572 miR-93 andp21 and a cascaded regulation in which p21 is repressed
by FOXF2 via miR-572 in Figure 3(b) the same cascadedregulation together with a coherent FFL composed of TP53MYC miR-93 and p21 forms another regulatory moduleof p21 By modulating the transcriptional strengths of thetwo miRNAs by their TFs two types of p21 dynamics wereidentified saturation and pulse (Figure 3(c) top) In theformer the p21 expression increases and reaches its steadystate at the highest level in contrast in the latter the p21expression increases to a peak and thereafter drops to asteady state at lower level For the two different networkmodules various combinations of transcriptional strengthsof the two miRNAs lead to different distributions of thetwo p21 dynamical patterns For the network module of anincoherent FFL plus the cascaded regulation (Figure 3(c)bottom left) the saturation pattern appears in two distinctregions one with weakened transcriptional strength of miR-93 and the other with enhanced transcriptional strength ofmiR-93 for the other module (Figure 3(c) bottom right)the saturation pattern only appears in the region in whichthe transcriptional strength of miR-93 is enhanced Takentogether the results showed that for the two different networkmotifs the dynamical pattern of p21 is changing according to
BioMed Research International 13
different combinations of its upstream regulators suggestingthe adaptation of p21 dynamics for different biological con-texts
Furthermore we performed a number of simulations toshow the influence of miRNA regulation on p21 expressionlevels in different cellular processes arranged in a consecutivemanner In this procedure a cell is first in the process ofcell cycle followed by proliferation (cell growth) then thecell responses to DNA damage and enters into the processof senescence and finally apoptosis is initiated As shown inFigure 4 during the cell cycle process the p21 expression islow due to the activation of most of its targeting miRNAswhen the cell starts proliferating the p21 expression declinesto an even lower level because of more activated miRNAsin this process after responding to DNA damage the p21expression soars to a high level which is caused by theactivation of its TFs like p53 and fewer expressed miRNAsalthough the p21 expression keeps at a similar level whilethe cell is undergoing senescence the expressed miRNAs aredifferent from the previous process finally the p21 expressiondecreases again to a low level due to the reemergence ofmost of its targeting miRNAs and the cell enters apoptosisInterestingly the model simulations are also consistent withexperimental observations under non-stressed condition thelow expression level of p21 is needed for cell proliferation theupregulation of p21 happens after response to DNA damagevia p53 and the increased p21 expression further results incell cycle arrest leading to senescence and apoptosis [25]Besides when considering the effect of cooperatingmiRNAsthe p21 expression levels were indistinguishable from theprevious simulation of DNA damage response However forthe other processes p21 expression levels were significantlylower compared to the simulations without considering thecooperative effect of the miRNAs Above all these resultsindicated that selective expression of cooperative miRNAscould be adopted by cells to ensure diverse expression levelsof p21 tomeet the requirements of different cellular processes
3 Conclusions and Discussion
In this paper we presented a systems biology approach com-bining data-drivenmodeling andmodel-driven experimentsto investigate the role of miRNA-mediated repression in generegulatory networksThis approach provides a systematicwayto gain a deeper understanding of the regulation of targetgenes by mutiple and cooperative miRNAs Using the regu-lation of p21 by multiple miRNAs as a case study we showedhow the ODE-basedmodel which is calibrated and validatedby means of experimental data is suitable for predicting thetemporal dynamics of molecular concentrations involved inbiochemical systems
Provided there are sufficiently rich quantitative data setsavaliable to characterize the model the use of the methodol-ogy here shown can be extended to more complex regulatorynetworks involving multiple targets cooperating TFs andmiRNAs and signaling pathways displaying cross-talk viapost-tranlational modifications In this case the criticalelement is the quality and quantitiy of the available data
Insufficiency and low quality of experimental data can causeerrors in the process of model construction and overfitting inparameter estimation can lead to uncertainties in the modelpredictions We believe that the quick development of quan-titative high throughput techniques such as transcriptomicsproteomics and miRNomics will facilitate the constructionand characterization of larger miRNA-mediated regulatorynetworks
Other modeling frameworks than ODE-based modelscan be used to describe biological systems such as prob-abilistic (eg Bayesian) or logical (eg Boolean) modelsImportantly different modeling frameworks have differentproperties and perform well regarding different perspectivesand levels of mechanistic details of biochemical systems [46]For example Bayesian models are helpful in the constructionof connections in signaling networks and can reveal the mostlikely underlying structure of the network in a probabilisticmanner Boolean models use binary values (0 and 1) andlogical gates (AND NOT and OR) to describe activities ofnetwork components and the information flow among themWe believe that in the coming future hybrid models whichconsist of modeling framework and experimental techniquespecific sub-modules will provide the necessary compromisebetween quantitativequalitative accuracy and scalability forthe investigation of large biochemical networks [47]
Disclosures
The authors declare that they have no competing financialinterests
Authorsrsquo Contribution
Julio Vera and Olaf Wolkenhauer are equal contributers
Acknowledgments
The authors would like to acknowledge the funding fromthe German Federal Ministry of Education and Research(BMBF) eBio-miRSys (0316175A to Xin Lai and Julio Vera)eBio-SysMet (0316171 to OlafWolkenhauer) and Gerontosys-ROSAge (0315892A to Olaf Wolkenhauer)
References
[1] A E Pasquinelli ldquoMicroRNAs and their targets recognitionregulation and an emerging reciprocal relationshiprdquo NatureReviews Genetics vol 13 no 4 pp 271ndash282 2012
[2] J Vera X Lai U Schmitz and O Wolkenhauer ldquoMicroRNA-regulated networks the perfect storm for classical molecularbiology the ideal scenario for systems biologyrdquo Advances inExperimental Medicine and Biology vol 774 pp 55ndash76 2013
[3] S Nikolov J Vera U Schmitz and O Wolkenhauer ldquoA model-based strategy to investigate the role of microRNA regulation incancer signalling networksrdquo Theory in Biosciences vol 130 no1 pp 55ndash69 2011
14 BioMed Research International
[4] X Lai U Schmitz S K Gupta et al ldquoComputational analysisof target hub gene repression regulated by multiple and cooper-ative miRNAsrdquoNucleic Acids Research vol 40 no 18 pp 8818ndash8834 2012
[5] X Lai O Wolkenhauer and J Vera ldquoModeling miRNA regu-lation in cancer signaling systems miR-34a regulation of thep53Sirt1 signaling modulerdquoMethods in Molecular Biology vol880 pp 87ndash108 2012
[6] C Jiang Z Xuan F Zhao and M Q Zhang ldquoTRED atranscriptional regulatory element database new entries andother developmentrdquoNucleic Acids Research vol 35 supplement1 pp D137ndashD140 2007
[7] L A Bovolenta M L Acencio and N Lemke ldquoHTRIdban open-access database for experimentally verified humantranscriptional regulation interactionsrdquo BMCGenomics vol 13article 405 2012
[8] VMatys O V Kel-Margoulis E Fricke et al ldquoTRANSFAC andits module TRANSCompel transcriptional gene regulation ineukaryotesrdquo Nucleic Acids Research vol 34 database issue ppD108ndashD110 2006
[9] D Karolchik R Baertsch M Diekhans et al ldquoThe UCSCgenome browser databaserdquo Nucleic Acids Research vol 31 no1 pp 51ndash54 2003
[10] P Alexiou T Vergoulis M Gleditzsch et al ldquomiRGen 20 adatabase of microRNA genomic information and regulationrdquoNucleic Acids Research vol 38 supplement 1 pp D137ndashD1412010
[11] F Xiao Z Zuo G Cai S Kang X Gao and T Li ldquomiRecordsan integrated resource for microRNA-target interactionsrdquoNucleic Acids Research vol 37 supplement 1 pp D105ndashD1102009
[12] S-D Hsu F-M Lin W-Y Wu et al ldquomiRTarBase a databasecurates experimentally validated microRNA-target interac-tionsrdquo Nucleic Acids Research vol 39 supplement 1 pp D163ndashD169 2011
[13] P Sethupathy B Corda and A G Hatzigeorgiou ldquoTarBase acomprehensive database of experimentally supported animalmicroRNA targetsrdquo RNA vol 12 no 2 pp 192ndash197 2006
[14] H Dweep C Sticht P Pandey and N Gretz ldquomiRWalkmdashdatabase prediction of possible miRNA binding sites byldquowalkingrdquo the genes of three genomesrdquo Journal of BiomedicalInformatics vol 44 no 5 pp 839ndash847 2011
[15] T S Keshava Prasad RGoel K Kandasamy et al ldquoHumanpro-tein reference databasemdash2009 updaterdquo Nucleic Acids Researchvol 37 supplement 1 pp D767ndashD772 2009
[16] D Szklarczyk A Franceschini M Kuhn et al ldquoThe STRINGdatabase in 2011 functional interaction networks of proteinsglobally integrated and scoredrdquo Nucleic Acids Research vol 39supplement 1 pp D561ndashD568 2011
[17] D Croft G OrsquoKelly G Wu et al ldquoReactome a database ofreactions pathways and biological processesrdquo Nucleic AcidsResearch vol 39 supplement 1 pp D691ndashD697 2011
[18] N le Novere M Hucka H Mi et al ldquoThe systems biologygraphical notationrdquoNature Biotechnology vol 27 no 8 pp 735ndash741 2009
[19] A Funahashi Y Matsuoka A Jouraku M Morohashi NKikuchi and H Kitano ldquoCellDesigner 35 a versatile modelingtool for biochemical networksrdquo Proceedings of the IEEE vol 96no 8 pp 1254ndash1265 2008
[20] M S Cline M Smoot E Cerami et al ldquoIntegration ofbiological networks and gene expression data using CytoscaperdquoNature Protocols vol 2 no 10 pp 2366ndash2382 2007
[21] D-H Le andY-KKwon ldquoNetDS aCytoscape plugin to analyzethe robustness of dynamics and feedforwardfeedback loopstructures of biological networksrdquo Bioinformatics vol 27 no 19pp 2767ndash2768 2011
[22] E O Voit Computational Analysis of Biochemical SystemsA Practical Guide for Biochemists and Molecular BiologistsCambridge University Press Cambridge UK 2000
[23] S Wu S Huang J Ding et al ldquoMultiple microRNAs modulatep21Cip1Waf1 expression by directly targeting its 31015840 untranslatedregionrdquo Oncogene vol 29 no 15 pp 2302ndash2308 2010
[24] P Saeligtrom B S E Heale O Snoslashve Jr L Aagaard J Alluin andJ J Rossi ldquoDistance constraints between microRNA target sitesdictate efficacy and cooperativityrdquo Nucleic Acids Research vol35 no 7 pp 2333ndash2342 2007
[25] Y-S Jung YQian andXChen ldquoExamination of the expandingpathways for the regulation of p21 expression and activityrdquoCellular Signalling vol 22 no 7 pp 1003ndash1012 2010
[26] W Wang H Furneaux H Cheng et al ldquoHuR regulatesp21 mRNA stabilization by UV lightrdquo Molecular and CellularBiology vol 20 no 3 pp 760ndash769 2000
[27] U Wittig R Kania M Golebiewski et al ldquoSABIO-RKmdashdatabase for biochemical reaction kineticsrdquo Nucleic AcidsResearch vol 40 database issue pp D790ndashD796 2012
[28] R Milo P Jorgensen U Moran G Weber and M SpringerldquoBioNumbersmdashthe database of key numbers in molecular andcell biologyrdquo Nucleic Acids Research vol 38 supplement 1 ppD750ndashD753 2010
[29] I-C Chou and E O Voit ldquoRecent developments in parameterestimation and structure identification of biochemical andgenomic systemsrdquoMathematical Biosciences vol 219 no 2 pp57ndash83 2009
[30] A Saltelli K Chan and E M Scott Sensitivity Analysis JohnWiley amp Sons New York NY USA 1st edition 2000
[31] S Strogatz Nonlinear Dynamics and Chaos Applications toPhysics Biology Chemistry and Engineering With Applicationsto Physics Biology Chemistry and Engineering Westview PressBoulder Colo USA 2000
[32] P Zhou S Cai Z Liu and R Wang ldquoMechanisms generatingbistability and oscillations in microRNA-mediated motifsrdquoPhysical Review E vol 85 no 4 apart 1 Article ID 041916 9pages 2012
[33] S Marino I B Hogue C J Ray and D E Kirschner ldquoAmethodology for performing global uncertainty and sensitivityanalysis in systems biologyrdquo Journal of Theoretical Biology vol254 no 1 pp 178ndash196 2008
[34] L P LimNC Lau PGarrett-Engele et al ldquoMicroarray analysisshows that some microRNAs downregulate large numbers of-target mRNAsrdquo Nature vol 433 no 7027 pp 769ndash773 2005
[35] J Krutzfeldt N Rajewsky R Braich et al ldquoSilencing ofmicroRNAs in vivowith lsquoantagomirsrsquordquoNature vol 438 no 7068pp 685ndash689 2005
[36] MWPfafflGWHorgan andLDempfle ldquoRelative expressionsoftware tool (REST) for group-wise comparison and statisticalanalysis of relative expression results in real-time PCRrdquoNucleicAcids Research vol 30 no 9 p e36 2002
[37] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[38] T Abbas and A Dutta ldquop21 in cancer intricate networks andmultiple activitiesrdquo Nature Reviews Cancer vol 9 no 6 pp400ndash414 2009
[39] J Wang M Lu C Qiu and Q Cui ldquoTransmiR a transcriptionfactor microRNA regulation databaserdquo Nucleic Acids Researchvol 38 supplement 1 pp D119ndashD122 2010
BioMed Research International 15
[40] S Bandyopadhyay andM Bhattacharyya ldquoPuTmiR a databasefor extracting neighboring transcription factors of humanmicroRNAsrdquo BMC Bioinformatics vol 11 article 190 2010
[41] A Le Bechec E Portales-Casamar G Vetter et alldquoMIRNTN a framework integrating transcription factorsmicroRNAs and their targets to identify sub-network motifs ina meta-regulation network modelrdquo BMC Bioinformatics vol12 article 67 2011
[42] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyTheGene Ontology ConsortiumrdquoNature Genetics vol 25 no 1 pp 25ndash29 2000
[43] A I F Vaz and L N Vicente ldquoA particle swarm pattern searchmethod for bound constrained global optimizationrdquo Journal ofGlobal Optimization vol 39 no 2 pp 197ndash219 2007
[44] W H PressNumerical Recipes The Art of Scientific ComputingCambridge University Press Cambridge UK 2007
[45] J G Doench and P A Sharp ldquoSpecificity of microRNA targetselection in translational repressionrdquo Genes and Developmentvol 18 no 5 pp 504ndash511 2004
[46] B Kholodenko M B Yaffe and W Kolch ldquoComputationalapproaches for analyzing information flow in biological net-worksrdquo Science Signaling vol 5 no 220 2012
[47] F M Khan U Schmitz S Nikolov et al ldquoHybrid modeling ofthe crosstalk between signaling and transcriptional networksusing ordinary differential equations and multi-valued logicrdquoBiochimica et Biophysica Acta 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
4 BioMed Research International
provide experimentalists with different levels of confidencethus differing confidences can be reflected using differentvalues for 119878
119898based on the experience of experimentalists
Of note although the confidence scores cannot be directlyconverted into a mathematical model with the help of thesescores we can discard non-reliable putative interactions togenerate the ultimate version of a gene regulatory networkThe final version of the network can be further analyzedto identify regulatory motifs like feedforward loops (FFLs)for example with the help of the Cytoscape plugin NetDS[21] Thereafter the complete network or parts of it canbe converted into a detailed mechanistic model which isdescribed in detail in the following section
(3) Model Construction and Calibration After the construc-tion visualization and refinement of a gene regulatory net-work it is converted into a detailedmechanistic model whichenables the investigation of unanswered biological questionsand validation of hypotheses Ordinary differential equations(ODEs) describe how processes of synthesis biochemicalmodification andor degradation affect the temporal con-centration profile of biochemical species like proteins RNAsand metabolites An ODE model can be constructed usingappropriate kinetic laws such as the law ofmass-action whichstates that the rate of a chemical reaction is proportional tothe probability that the reacting species collideThis collisionprobability is in turn proportional to the concentration ofthe reactants [22] A general representation of ODE-basedmodels using mass-action kinetics is given by the followingequation
119889119909119894
119889119905=
119898
sum
120583=1
119888119894120583sdot 119896120583sdot
119899
prod
119895=1
119909119895
119892120583119895 119894 isin 1 2 119899 (2)
where 119909119894represents state variables which denote the molar
concentration of the 119894th biochemical specie Every biochem-ical reaction 120583 is described as a product of a rate constant(119896120583) and biochemical species (119909
119895 119895 isin 1 2 119899) that are
involved in this reaction 119888119894120583 the so-called stoichiometric
coefficients relate the number of reactant molecules con-sumed to the number of product molecules generated inthe reaction 120583 119892
120583119895denotes kinetic orders which are equal
to the number of species of 119909119894involved in the biochemical
reaction 120583 The rate constants kinetic orders and the initialconditions of state variables are defined as model parametersBesides mass-action kinetics other kinetic rate laws such asMichaelis-Menten kinetics Hill equation and power-laws arealso frequently used in ODE models
After ODEs are formulated the model requires to becalibrated a process by which parameter values are adjustedin order to make model simulations match experimentalobservations as good as possible To do so there are twopossible means characterization of parameter values usingavailable biological information or estimation of parametervalues using optimization methods Some parameter valuescan be directly measured or obtained from literature ordatabases For example the half-life (119905
12) of some molecules
(eg protein) can be measured in vitro via western blot-ting This information can be used to characterize their
degradation rate constants through the equation 119896deg =
(ln 211990512) The database SABIO-RK provides a platform for
modelers of biochemical networks to assemble informationabout reactions and kinetic constants [27] However formost model parameters whose values cannot be measuredin laboratories or be accessed from literature or databasesparameter estimation is a necessary process to characterizetheir values Before running parameter estimation initialparameter values and boundaries should be set within phys-ically plausible ranges To do so the database BioNumbersprovides modelers with key numbers in molecular and cellbiology ranging from cell sizes to metabolite concentrationsfrom reaction rates to generation times from genome sizesto the number of mitochondria in a cell [28] After parameterestimation unknown parameter values are determined usingoptimization methods which can minimize a cost functionthat measures the goodness of fit of the model with respect togiven quantitative experimental data sets Parameter estima-tion using optimization algorithms is an open research fieldin which several methods have been developed accordingto the nature and numerical properties of biological dataanalyzed The discussion for choosing proper optimizationmethods for parameter estimation is beyond the scope ofthis paper but the interested reader is referred to the paperpublished by Chou and Voit [29]
(4) Model Validation and Analysis Usually the model sim-ulations are compared with the experimental data used forthe parameter estimation but a good agreement betweenboth is not enough to guarantee the predictive ability of themodel Therefore it is necessary to validate the model withdata sets that are not used during the parameter estimationThis process is called model validation and can ensure morereliable and accurate model predictions To do so the datagenerated in new experiments or extracted from literature arecompared with model simulations which are obtained afterconfiguring the model according to the new experimentalsettings Once a model is validated it can be used to per-form predictive simulations which are helpful to studythe dynamic properties of biochemical systems guide thedesign of new experiments in the laboratory and formulateadditional hypotheses In addition tools such as sensitivityand bifurcation analysis can be used to study complex proper-ties and behavior of the modeled system Sensitivity analysisis used to evaluate the influence of model parameters (eginitial concentrations of the state variables and rate constants)on model outputs such as the temporal behavior of networkcomponents [30] Bifurcation analysis is employed to detectcontrol parameters (also known as bifurcation parameters)whose variations are able to change drastically the dynamicalproperties of the biochemical system as well as the stabilityof its fixed points [31] The application of these tools tomathematical modeling is beyond the scope of this paper butthe interested reader is referred to the publication of Zhouet al [32] and Marino et al [33]
212 Experiment Methods As mentioned in the previoussections after a model is established it can be calibrated
BioMed Research International 5
and validated using temporal experimental data To do sothe data can either be derived from literature or generatedto calibrate the model by own experiments In case of ODEmodels the most suitable data for model calibration isquantitative time-series data obtained from perturbation orquantitative dose-response experimentsThe experiments inwhich time-series data are measured for different regulators(such as miRNAs and TFs) of a gene regulatory network canbe obtained using the techniques described in the subsectionsbelow
(1) qRT-PCR Quantitative real time polymerase chain reac-tion (qRT-PCR) has been used to identify mRNAs regulatedby overexpression or silencing of a specific miRNA [3435] miRNAs typically exhibit their regulatory effects byassociating with specific 31015840 UTR regions of the mRNAscalled miRNA seed regions [34] This association can leadeither to a temporary inhibition of translation or completedegradation of the mRNA in which case qRT-PCR mediateddetection is beneficial For this the cells are transfectedwith the miRNA and non-targeting control oligonucleotidesat an appropriate concentration using either a lipid basedtransfection reagent or nucleofection The transfected cellsare then incubated for the necessary time periods (eg 24 h48 h 72 h etc) after which lysates are prepared and total RNAis extracted 1000ndash2000 ng of the RNA is then converted intocDNA using a reverse transcription kit Taqman qRT-PCR isperformed using 10ndash20 ng of cDNA and primers labeled withfluorescence probes to detect the transcriptional levels of thetarget mRNA Housekeeping genes like GAPDH and HPRTare used as endogenous controls for data normalizationRelative expression at different time points is determinedby comparing the Ct values of the miRNA transfected cellswith Ct values of non-targeting control transfected cells andexpressed as relative expression [36] However qRT-PCR inspite of being highly sensitive is applicable specifically underconditions of complete or partial degradation of the targetmRNA miRNA-mediated inhibition in translation can bebetter demonstrated by techniques like immunoblotting
(2) Western BlotImmunblotting Western blotting or proteinimmunoblotting is a technique to detect the expression of agene at protein level This technique is particularly useful indetermining the regulatory effects of a miRNA on expressionof a target gene which is temporarily inhibited For thisthe cells are transfected with a miRNA or an antagomiRas mentioned above and cell lysates are prepared at appro-priate time points using protein lysis buffers (eg Radio-Immunoprecipitation Assay buffer or RIPA) containing lysisagents like Dithiothreitol (DTT) and protease inhibitors Theproteins from each sample are quantitated using Bradfordor BCA reagents and compared with bovine serum albumin(BSA) standards for accurate protein estimation 20ndash40 120583g ofprotein is then loaded and resolved on a sodium dodecyl sul-fate polyacrylamide gel (SDS-PAGE) along with a pre-stainedprotein marker The protein bands are then transferred ontoa nitrocellulose membrane followed by incubation with theappropriate primary and secondary antibodies linked tofluorescent dyes or horse radish peroxidase enzyme (HRP)
The protein expression is then analyzed using either fluores-cence or chemiluminiscenceHRP substrates in gel documen-tation systems (eg LI-COR Odyssey) Housekeeping geneslike 120573-actin or 120573-tubulin are used for protein normalizationThe time point for maximum target gene suppression gen-erally varies depending on the number of miRNA bindingsites at the 31015840 UTR of the target gene and the extent ofcomplementarity of the seed region [37] Immunoblotting isa widely used technique to provide confirmatory evidence forthe inhibitory effects ofmiRNA at the protein level but it failsto explain the underlying interaction mechanisms
(3) Reporter Gene Assay As each miRNA can inhibit theexpression of a large number of genes regulation of a par-ticular target gene may either be by direct interaction or bean indirect consequence of it In direct regulation a miRNAbinds to the complementary sequences at the 31015840 UTR of atarget gene and thereby suppresses its expression As a con-sequence of this the expression levels of a number of down-stream genes (indirect targets) are also dysregulated makingit crucial to differentiate between primary and secondarymiRNA targets To determine the interaction specificitya reporter construct (luciferase) with intact or mutated 31015840UTR of the target gene cloned at the 51015840 end is co-transfectedinto the cells along with the miRNA The regulatory effect ofthe miRNA on the target gene expression is then measuredusing the expression of a reporter gene In the absence ofthe appropriate binding sequences (mutated 31015840 UTR) themiRNA cannot suppress the reporter mRNA suggesting thatthe suppressive effect of the miRNA is mediated by a directinteraction The reporter activity can be analyzed at differenttime points such as 24 hr 48 hr and 72 hr to determine thetime dependent suppression of a target gene expression bya miRNA
22 Case Study The Regulation of p21 by Multiple andCooperative miRNAs
221 Construction and Visualization of p21 Regulatory Net-work By using the approach described above we investi-gated the regulation of p21 by its multiple targeting miRNAsp21 also known as cyclin-dependent kinase inhibitor 1(CDKN1A) is a transcriptional target of p53 It is requiredfor proper cell cycle progression and plays a role in cell deathDNA repair senescence and aging (reviewed in [38]) Inter-estingly p21 was the first experimentally validated miRNAtarget hub which is a gene that is simultaneously regulatedby many miRNAs This made it an ideal candidate for a casestudy of our approach [23] To do so we first constructeda p21 regulatory network using the following steps
(a) We extracted miRNA-target interactions from thepublication of Wu et al [23] where a list of predictedp21-regulating miRNAs was subjected to experimen-tal validation
(b) Experimentally verified TFs of p21 were extractedfrom literature and putative TFs having conservedbinding sites in the 5 kb upstream region of the p21open reading frame were extracted from UCSC table
6 BioMed Research International
browser A list of TFs controlling the expression of themiRNAswas constructed using information of exper-imentally proven TF-miRNA interactions extractedfrom TransmiR (release 10) [39] In addition wegenerated a list of putative TFs of miRNAs withbinding sites in the 10 kb upstream region of themiRNA genes using information from the databasesPuTmiR (release 10) [40] and MIRNTN (version121) [41] and from the table of TFs with conservedbinding sites in theUCSC genome browser (hg18) [9]
(c) Information about protein interactions was extractedfrom theHuman Protein ReferenceDatabase (HPRDrelease 90) [15] and the STRING database (release90) [16] Only the experimentally verified p21-proteininteractions were used to construct the network
(d) Additionally we associated the TFs in the network tonine biological processes based on the Gene Ontol-ogy (GO) [42] The corresponding GO terms werecell proliferation cell apoptosis immune responseinflammatory response cell cycle DNA damage cellsenescence DNA repair and cell migration
Next we visualized the network in CellDesigner andcomputed a confidence score for each interaction in thenetwork (Figure 1) The confidence scores provide us withthe reliability of the interactions considered in p21 regulatorynetwork With the help of these scores we discarded non-reliable interactions and constructed the mechanistic modelaccounting for p21 regulation by its targeting miRNAsBesides the interested experimentalists can further use thisinformation to choose reliable interacting candidates of p21for designing relevant experiments The scores for eachinteraction of p21 regulatory network are shown in Table 2
222 Mechanistic Modeling of p21 Regulation byIts Targeting miRNAs
(1) Model Construction After constructing the regulatorynetwork a detailed mechanistic model of ODEs whichdescribes the biochemical reactions underlying the regula-tion of p21 was established We chose the ODE modelingapproach as it is a simple formalism for describing temporaldynamics of biochemical systems and a wide range of toolsare available to explore their properties Precisely the modelconsidered themRNA(mp21 (3)) andprotein (p21 (6)) of themiRNA target hub p21 the p21-targeting miRNAs (miR
119894 119894 isin
(1 15) (5)) and the complexes formed by p21 mRNAand miRNA [mp21 | miR
119894] (4) Altogether the model is
constituted by 32 state variables and 64 parameters
119889mp21119889119905
= 119896mp21syn sdot 119891act (TFmp21)
minusmp21 sdot (119896mp21deg +sum
119894
119896complex
119894
ass sdotmiR119894)
(3)
119889 [mp21 | miR119894]
119889119905
= 119896complex
119894
ass sdotmp21 sdotmiR119894minus 119896
complex119894
deg sdot [mp21 | miR119894]
(4)
119889miR119894
119889119905= 119896
miR119894syn sdot 119891act (TFmiR119894)
minusmiR119894sdot (119896
miR119894deg + 119896
complex119894
ass sdotmp21) (5)
dp21119889119905
= 119896p21syn sdotmp21 + 119896p21deg sdot p21 (6)
mp21Total = mp21 +sum119894
[mp21 | miR119894] (7)
For mp21 processes considered in the model were (i) itssynthesis (119896mp21
syn ) mediated by TFs (119891act(TFmp21)) (ii) itsdegradation (119896mp21
deg ) and (iii) its association with a miRNA(119896complex
119894
ass ) For each miR119894 processes considered were (i)
the synthesis (119896miR119894syn ) mediated by TFs (119891act(TFmiR119894)) (ii) the
degradation (119896miR119894deg ) and (iii) the association with the p21
mRNA target (119896complex119894
ass ) For each [mp21 | miR119894] complex
processes considered were (i) the formation of the complexby a miR
119894and the p21 mRNA (119896complex
119894
ass ) and (ii) the complexdegradation (119896complex
119894
deg ) For p21 processes considered were(i) its synthesis (119896p21syn ) and (ii) its degradation (119896p21deg) Anadditional algebraic equation accounting for the total mea-surable amount of p21 mRNA (mp21Total) was also includedThe SBML file of the model is available for download athttpwwwsbiuni-rostockdeuploadstx templavoilap21TargetHub 03092013xml
(2) Model Calibration and Validation For model calibrationwe fixed some parameter values using published data andestimated the other unknown parameters using the time-series data published by Wu et al [23] in which the p21mRNA (northern plot) and protein levels (western plot)were measured 48 hr after transfection of individual p21-targeting miRNAs into human embryonic kidney 293 cellsThe unknown parameter values were estimated using aniterative method combining global (particle swarm patternsearch) [43] and local (downhill simplex method in multi-dimensions) [44] optimization algorithms For each miR
119894
considered in the model the method minimizes the distancebetween model simulations and experimental data using thefollowing cost function
119865miR119894cost =
[mp21miR119894sim (119905) minusmp21miR119894
exp (119905)]2
(120575mp21119894
)2
+
[p21miR119894sim (119905) minus p21miR119894
exp (119905)]2
(12057511990121
119894)2
119894 isin [1 15]
(8)
BioMed Research International 7
1 2 4 1 2 4 5 6 1 2 3 7
E2F TP53 RELA1 2 4 7 2 3 7 6
STAT5 SRF1 2 3 4 1 2
JUN STAT1
SP1 RUNX1 TFAP2A
Common TFs Modulation
CDKN1A
1 2 1 2 4 1 4
1 4 1 2 3 7
1 2 5 2 4 5 7 1
VDR
CUX1
E2A
RAR BRCA1
TP63 TP73 TBX2
STAT6 SP3 STAT3
STAT2 RUNX2
EHMT2 p21 TFs
1 2 4 5 8 9
CEBP120573NF120581B1
(1) Cell proliferation (6) Cell senescence
(2) Apoptosis (7) Inflammatory response
(3) Immune response (8) DNA repair
(4) Cell cycle control (9) Cell migration
(5) DNA damage
miRNAmiRNA
Translational repressionmiRNA
Translation
Transcription
Association
miR-132 miR-28-5p
miR-299-5p miR-298
miR-345 miR-125a-5p
miR-208a
miRNA
miRNADeadenylation followed by decay
Protein-protein interaction
miR-93 miR-654-3p
miR-657 miR-572
miR-423-3p miR-363
miR-639 miR-515-3p
p21
CCND2 CDK3 GMNN CCNE1 POLD2 CSNK2B C1orf123 CCDC85B
PIM1 CCNB2 KIAA0432 GADD45B HDAC11 ITGB1BP3 TTLL5 SLC25A11
GADD45G CDC7 BCCIP CDK1 GNB2LI RAB1A psmA3 TK1
CDC45 CCNB1 CCNA2 GADD45A MAPK8 TSG101 SET CIZ1
CDK14 CDK5 MDM2 CDC27 RNF144B ESR1 PRKCH RPA1
CDC20 SKP2 CDK2 CDK4 DHX9 PCNA CDC5L CCDC5
CCND3 CDK6 CCNE2 CCNA1 TEX11 MAP3K5 MCM10 Caspase-3
CDC6 CCND1 XRCC6 PARP1 DAPK3 NR4A1 CEBPA
Interacting proteins
AKT1
miRNA genes
5 8 1 2 1 2 4 3 1 2 3 3 2 2 3 3 1 4 1 3 1 4
1 4
NFATC2 YY1 CREB1 CEBPA ARNT
FOXCI1
ELK1 SOX9 FOS PAX5 MAX MEF2A RFX1 JUNB AKT1 GATA3 MYC NRSFEGR1
E47 PPARG POU3F2 FOXF2FOXL1 MZF1 NFIC SRY
FOXC1
HNFA4 NFYA MYF1 BACF1 ATF6 ZID RORA POU2F1
miRNA TFs
11 2 34
12 31 5 11
HNF1A
1 2
RB1
Cellular processes
Figure 1 p21 regulatory network The network contains several layers of regulators of p21 TFs (light blue and red boxes) miRNAs (darkblue and green boxes) and proteins (grey boxes) In each big box there are small boxes which represent individual components of this layerof regulation miRNAs are classified into two groups according to the mechanisms by which the expression of p21 is repressed One groupcauses p21 translation repression (green box) These miRNAs bind to p21 mRNA resulting in the repressed translation of p21 but unchangedmRNA expression level The other group of miRNAs (dark blue box) decreases the stability of p21 mRNA by modifying its structure leadingto mRNA decay and finally the downregulation of p21 TFs are classified into three groups p21 TFs (red) miRNA TFs (yellow) and theircommon TFs (light blue) The p21 interacting-proteins are framed in the grey boxes The purple boxes represent nine processes and the TFsassociated with these processes are indicated in the ellipses above them using corresponding figures This data is adapted from our previouspublication [4]
where mp21miR119894sim (119905) and mp21miR119894
exp (119905) represent the simulatedp21 mRNA and protein expression levels for each miR
119894at
time point 119905 p21miR119894sim (119905) and p21miR119894
exp (119905) represent the mea-sured value for each miR
119894at time point 119905 and their standard
deviations are 120575mp21119894
and 120575p21119894
Here 119905 is the time point (48 hr)after overexpression of the individual miRNAs in embryonickidney 293 cells at which the expression levels of the p21 andits mRNA were measured [23] The model calibration resultsare shown in Figure 2(a) and the obtained parameter valuesare listed in Table 3
Experimental results showed that a stronger repressionof the target gene can occur when two miRNA binding siteson the target mRNA are in close proximity [24 45] To testthe consequences of this hypothesis we predicted cooperativemiRNA pairs for p21 with seed site distances between 13ndash35 nt in the p21 31015840 UTR To substantiate the cooperativeeffect associated with pairs of miRNAs we introduced agroup of new state variables ([mp21 | miR
119894| miR
119894]) into
the original model These state variables account for the
ternary complexes composed of p21 mRNA and two puta-tively cooperating miRNAs (miR
119894and miR
119895) For these new
variables processes considered are (i) the association of p21mRNA with miR
119894and miR
119895into a complex (119896co-complex
119894119895
ass )and (ii) the degradation of the complex (119896co-complex
119894119895
deg ) Afterexpansion the corresponding modified and new ODEs arelisted below
dmp21119889119905
= 119896mp21syn sdot 119891act (TFmp21)
minusmp21 sdot (119896mp21deg +sum
119894
119896complex
119894
ass sdotmiR119894
+sum
119894119895
119896co-complex
119894119895
ass sdotmiR119894sdotmiR119895)
(9)
8 BioMed Research InternationalTa
ble2Th
econ
fidence
scores
forthe
interactions
inp21regulatorynetworkTh
enetworkconsistso
ffou
rtypes
ofinteractionsm
iRNA-
p21TF
-miRNAT
F-p21andp21-p
rotein
(a)miRNA-
p21interactio
nscores
miRNA
miR-125a-5p
miR-132
miR-208a
miR-28-5p
miR-298
miR-299-5p
miR-345
miR-363
Score
073
073
073
080
073
080
073
073
miRNA
miR-423-3p
miR-515-3p
miR-572
miR-639
miR-654-3p
miR-657
miR-93
Score
085
073
073
073
073
080
095
(b)TF
-miRNAinteractionscores
Score
Score
Score
Score
Score
Score
miR-208
miR-132
miR-657
miR-125a-5p
miR-28-5p
miR-93
MAX
027
EGR1
093
TFAP2
A034
EGR1
086
MZF
1028
TFAP2
A034
EGR1
029
CREB
1086
SP1
034
PAX5
024
E47
027
SOX9
024
ARN
T027
ARE
B6027
MZF
1034
NRS
F024
FOXC
1031
GAT
A3
027
RFX1
024
E47
024
MAX
029
ZID
024
FOXI1
024
MZF
1031
ATF6
024
ELK1
027
USF1
027
PPARG
027
SRY
027
SP1
034
YY1
027
EGR2
029
EGR1
034
GAT
A3
034
JUN
029
POU2F1
024
JUNB
024
FOS
024
RELA
029
MZF
1034
POU2F1
024
EGR1
029
POU2F1
024
NFIC
024
miR-654-3p
TP53
034
SRF
027
CEBP
A024
NFIC
027
RELA
027
YY1
027
NFIC
034
CEBP
A029
NFY
A024
miR-345
miR-423-3p
BACH
1024
miR-299-5p
RORA
024
RUNX1
024
SP1
034
BACH
1027
FOXL
1029
CREB
1024
miR-363
STAT
1024
RELA
027
STAT
1027
SRY
024
SRY
027
E2F1
062
E2F1
085
NFA
TC2
024
POU3F2
024
NFA
TC2
024
SRF
029
FOXC
1033
MYC
084
HNF4
A031
STAT
5A027
FOS
024
RUNX1
027
MZF
1034
miR-298
NFY
A024
SRF
027
POU2F1
027
miR-572
miR-639
JUNB
031
POU2F1
024
MEF
2A027
NF120581
B1024
FOXF
2024
NFY
A024
NFIC
031
POU2F1
024
HNF1A
024
(c)TF
-p21
interactionscores
TF(verified)
SP1
SP3
E2F1
Runx
1Ru
nx2
STAT
1E2
AST
AT5
TP53
TP63
TP73
Score
100
100
077
077
077
072
072
067
067
067
067
TF(verified)
STAT
3CU
X1TF
AP2
ABR
CA1
VDR
RARA
CEB
P120572CEB
P120573RB
1Tb
x2EH
MT2
Score
066
066
060
060
060
060
060
060
060
060
060
TF(putative)
NF120581
B1RE
LAST
AT2
STAT
6SR
FScore
044
044
044
044
043
(d)p21-p
rotein
interactionscores
Protein
TP53
PCNA
CASP
3CC
NA1
CCND1
SKP2
BCCI
PCC
NA2
CCND2
CCNE2
AKT
1C1
orf123
CCDC8
5BCC
NB1
CCNB2
CCND3
Score
100
092
087
087
087
087
081
081
081
081
072
072
072
072
072
072
ProteinCC
NE1
CDC4
5CD
C5L
CDC6
CDC7
CDK1
CDK14
CDK2
CDK3
CDK4
CDK6
CEBP
ACI
Z1CS
NK2
A1
CSNK2
BDAPK
3Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
ESR1
GADD45A
GADD45B
GADD45G
GMNN
GNB2
L1HDAC
11ITGB1BP
3MAP3
K5MAPK
8MCM
10PA
RP1
PIM1
POLD
2PS
MA3
RAB1A
Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
SET
SLC2
5A11
STAT
3TE
X11
TK1
TSG101
TTLL
5XR
CC6
CDC2
0CD
C27
CDK5
DHX9
MDM2
NR4
A1
PRKC
HRP
A1
Score
072
072
072
072
072
072
072
072
056
056
056
056
056
056
056
056
Verifi
edexp
erim
entally
verifi
edinteraction
putativ
epredictedinteraction
BioMed Research International 9
Con
trol
miR
-298
miR
-208
am
iR-1
32m
iR-2
8-5p
miR
-125
-5p
miR
-299
-5p
miR
-345
miR
-93
miR
-423
-3p
miR
-515
-3p
miR
-363
miR
-657
miR
-639
miR
-572
miR
-654
-3p
0
05
1
15m
RNA
expr
essio
n (a
u)
Data Model Data ModelC
ontro
lm
iR-2
98m
iR-2
08a
miR
-132
miR
-28-
5pm
iR-1
25-5
pm
iR-2
99-5
pm
iR-3
45m
iR-9
3m
iR-4
23-3
pm
iR-5
15-3
pm
iR-3
63m
iR-6
57m
iR-6
39m
iR-5
72m
iR-6
54-3
p
0
02
04
06
08
1
Prot
ein
expr
essio
n (a
u)
(a)
DoxorubicinCellWell
Control miR-572 miR-93
Measure p21 expression at the time points of
miR-572+ miR-93
0 2 4 6 8 24hr
miR-572miR-93
Immunoblotting
(b)
0 4 8 12 16 20 240
4
8
12
16
20
Time (hr)
p53
Actin
Doxo (hrs)
fac
t(TF
mp2
1)
C 0 1 2 4 6 8 24
1 08 12 19 29 32 3 16
(c)
0 2 4 6 8 2402468
1012
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)
p21
expr
essio
n (a
u)
02468
1012
miR-572Control
02468
1012
miR-93
02468
1012
Model Data
0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24
1 23 46 107124 124 23 28 48 66 53 31 08 16 21 38 50 18 06 06 12 25 27 12
miR-572+ 93Doxo (hrs)
p21
Actin
p21
Actin
p21
Actin
p21
Actin
(d)
Figure 2 Model calibration and validation (a) Model calibration The figures show the relative change of the p21 mRNA and proteinexpression levels after overexpression of the indicated miRNAs (Model model simulation Data experimental data) These data werenormalized to the control group in which the p21 mRNA and protein expression levels were measured when the miRNAs were normallyexpressed (au arbitrary unit) (b) Experimental workflow In the experiments Sk-Mel-147 cells were seeded in six well plates Then maturemiRNA mimics were transfected individually at a concentration of 100 nM (miR-572 and miR-93) or in combination at 50 nM each (miR-572 + miR-93) After 48 hr transfection with miRNA mimics the cells were pulse treated with 250 nM doxorubicin for 1 hour after whichnormal growth medium was replenished The immunblotting were performed to measure p21 expression at 0 2 4 6 8 and 24 hr post-doxorubicin treatment (c) Temporal dynamics of p21 transcriptional function Afterdoxorubicin treatment the expression of p53 a TF ofp21 was measured using immunblotting and these data were used to characterize the transcriptional function of p21 using MATLAB linearinterpolation function (d) Model validation We measured the expression of p21 protein in response to genotoxic stress in the four scenariosas described in the main text The measured data (Data) were compared with the model simulations (Model) The figures (a) (c) and (d) areadapted from our previous publication [4]
119889 [mp21 | miR119894| miR
119895]
119889119905
= 119896co-complex
119894119895
ass sdotmp21 sdotmiR119894sdotmiR119895
minus 119896co-complex
119894119895
deg sdot [mp21 | miR119894| miR
119895]
(10)
To model stronger repression of the target gene bycooperating miRNAs we assumed a stronger association rateconstant for the complex [mp21 | miR
119894| miR
119895] which is
equal to the sum of their individual association rate constants
(119896co-complex119894119895
ass = 119896complex
119894
ass + 119896complex
119895
ass ) Similarly the degradationrates of the complexes [mp21 | miR
119894| miR
119895] were assumed
to be equal to the sum of degradation rate constants ofsingle miRNA binding complexes (119896co-complex
119894119895
deg = 119896complex
119894
deg +
119896complex
119895
deg ) However it has to be noted that these addedequations are an abstract description ofmiRNAcooperativitybecause the details of this mechanism are not yet known
To experimentally validate the capability of our model topredict the relative p21 concentrations regulated by coopera-tivemiRNAs we selectedmiR-572 andmiR-93 as a case study
10 BioMed Research International
Table 3 Initial concentrations of model variables and model parameter values Based on the experimental data the p21-targeting miRNAsverified byWu et al [23] were divided into two groups the translation repression group (marked with asterisk) and the mRNA deadenylationgroup AmiRNAwas classified into themRNA deadenylation group if its overexpression can result in 20 ormore downregulation of the p21mRNA level (ie p21mRNA levelle 08 the basal level is 1) otherwise it was classified into the translation repression group For the translationrepression group only 119896complex119894
ass was estimated and 119896complex119894deg was fixed For the other group both 119896complex119894
deg and 119896complex119894ass were estimatedThe initial
concentrations of p21 and mp21 were set to 1 and this value was used as their basal expression levels During the parameter estimation theinitial concentrations of p21-targeting miRNAs were set to 100 because in the publication [23] the expression levels of p21 and mp21 weremeasured after the individual introduction of the p21-targeting miRNAs with amount of 100 nM Due to the lack of biological information tocharacterize the transcriptional activation function (119891act) of p21 and its targeting miRNAs the corresponding functions were assumed to be1 for simplicity The data is adapted from our previous publication [4]
Initial concentration of variables and TF functionsVariable Description Initial concentration (au)p21 p21 protein 1
mp21 p21 mRNA 1
miR119894=115
p21-targeting miRNAs 100
[mp21 | miR119894=115
] Complexes formed by miR119894and mp21 0
119891act(TFmp21) p21rsquos transcriptional activation function 1119891act(TFmiR119894 (119894=115)) The transcriptional activation function of miR
1198941
Fixed parameter valuesParameter Description Value (hrminus1) Reference119896mp21syn Synthesis rate constant of mp21 01155 fixed119896mp21deg Degradation rate constant of mp21 01155 [24]119896miR119894syn (119894 = 1 15) Synthesis rate constant of miR
11989400289 fixed
119896miR119894deg (119894 = 1 15) Degradation rate constant of miR
11989400289 [25]
119896p21syn Synthesis rate constant of p21 13863 fixed119896p21deg Degradation rate constant of p21 13863 [26]
Estimated parameter values
miRNA (statevariable)
119896complex119894deg
(119894 = 1 15) (hrminus1)
119896complex119894ass
(119894 = 1 15)
(au minus1sdot hrminus1)
119865miR119894cost
(119894 = 1 15)Experimental data of p21 (protein mRNA plusmn SD)
miR-298 (miR1)lowast 01155 00254 34119890 minus 004 (016 1074 plusmn 0025)
miR-208a (miR2)lowast 01155 00041 20119890 minus 003 (051 1192 plusmn 0022)
miR-132 (miR3)lowast 01155 00275 24119890 minus 003 (015 121 plusmn 0147)
miR-28-5p (miR4)lowast 01155 00119 59119890 minus 003 (028 135 plusmn 006)
miR-125-5p (miR5)lowast 01155 00018 18119890 minus 003 (069 085 plusmn 0051)
miR-299-5p (miR6)lowast 01155 00080 18119890 minus 004 (036 095 plusmn 0038)
miR-345 (miR7)lowast 01155 00051 11119890 minus 004 (046 096 plusmn 0039)
miR-93 (miR8) 01564 00235 41119890 minus 014 (017 07776 plusmn 003)
miR-423-3p (miR9) 09118 00055 28119890 minus 009 (044 05102 plusmn 011)
miR-515-3p (miR10) 02098 00253 12119890 minus 013 (016 0616 plusmn 0037)
miR-363 (miR11) 02261 00399 22119890 minus 014 (011 056 plusmn 015)
mR-657 (miR12) 03465 00158 21119890 minus 014 (023 048 plusmn 012)
miR-639 (miR13) 04305 00327 18119890 minus 017 (013 036 plusmn 0084)
miR-572 (miR14) 03039 00360 94119890 minus 023 (012 045 plusmn 0044)
miR-654-3p (miR15) 97485 00024 30119890 minus 014 (064 063 plusmn 0053)
These two miRNAs were chosen because their predictedtarget sites in the p21 31015840 UTR are in close proximity to eachother and thereby they can induce cooperative repression onp21 as suggested in [24] The experiments were performed asfollows
(i) Melanoma cells (Sk-Mel-147) were transfected withthe mature miRNAmimics of the twomiRNAs eitherindividually (100 nM) or in combination (50 nMeach) whereas untreated cells were used as control(Figure 2(b))
BioMed Research International 11
p21
FOXF2
miR-572
Activation
Inhibition120572
Incoherent FFL + cascaded regulation
AF2
miR-93
(a)
miR-93
miR-572
TP53
MYC
p21
FOXF2
Coherent FFL + cascaded regulation
(b)
Saturation
Pulse
p21
(au
)
p21
(au
)Time (hr) Time (hr)
Saturation Pulse
Pulse
Saturation
Saturation
di
(au
)
102
101
102
100
101
10minus1
10010minus2
10minus1
102
101
100
10minus1
10minus2
FOXF2 (au) FOXF2 (au)
TP53
(au
)
1
09
08
07
06
05
04
03
02
01
10minus2 10210110010minus110minus20
AF2120572
(au
)
(c)
Figure 3 Different p21 dynamics in different network motifs We ran simulations to show the different dynamics of p21 for two differentnetwork motifs (a) and (b) Through simulations two dynamical patterns of p21 were identified saturation and pulse ((c) top) For eachnetwork motif the corresponding distributions of the two dynamical patterns were plotted ((c) bottom) For different combinations ofthe transcriptional strengths the normalized distance (119889
119894) between peaks (119901
119894) and steady states (ss
119894) of p21 is determined by the equation
119889119894= (119901119894minus ss119894)119901max 119901max = max(119901
1 119901
119899) If 119889
119894= 0 for the corresponding combination of transcriptional strengths the p21 dynamics is
saturation otherwise it is pulse The regions showing different dynamical patterns of p21 are separated using the white lines
(ii) Next the cells were treated with doxorubicina genotoxic-stress inducing agent The agent canupregulate the expression of p53 which is a knownTFof p21 and therefore it can result in the upregulationof p21 (Figure 2(c))
(iii) After doxorubicin treatment the expression levels ofp21 were measured by immunoblotting at differenttime points (0 2 4 6 8 24 hr) The p21 expressionvalues were normalized based on the p21 expression
level in the control group measured at time point 0 hr(Figure 2(d))
By doing so we obtained the p21 response after genotoxicstress in four scenarios (1) endogenous miRNA expression(Control) (2) overexpression of miR-572 (3) overexpressionof miR-93 and (4) both miRNAs moderately overexpressedThereafter we derived a model of seven ODEs based onthe original equations which was configured according tothe designed experiments making the simulation results
12 BioMed Research International
00001
0001
001
01
1Cell cycle Cell growth DNA damage Senescence Apoptosis
G1G2
MS
p21
expr
essio
n (a
u)
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
log10
wow
Figure 4 p21 expression regulated by cooperative miRNAs for different cellular processesThe associations of themiRNAs with these cellularprocesses were derived from GO terms of their TFs A miRNA was supposed to be expressed (in bold black font) in a cellular process onlyif its TF is related to the corresponding GO term of this process The p21 expression levels are computed for each process with (w)without(wo) considering the cooperative effect among the p21 targeting miRNAs
comparable with the experimental data As shown in theFigure 2(c) the simulations are in good agreement with theexperimental observations The individual overexpression ofmiR-572 or miR-93 led to the reduction of the upregulationof p21 response after genotoxic stress induction The twomiRNAs cause different degrees of repression due to theirdifferent repression efficiencies on p21 Interestingly the com-bined overexpression of both miRNAs induced the strongestdownregulation of p21 and therefore verifying the hypothesisof their cooperative regulation of p21 Above all the resultsnot only validated themodel but also demonstrated the abilityof our method to identify cooperative miRNA pairs for p21
(3) Predictive Simulations As there are abundant of networkmotifs such as FFLs in p21 regulatory network and thesenetwork motifs are important for determining p21 dynamicsit is interesting to investigate the dynamics of p21 in networkmodules where FFLs are involved To do so two networkmodules including both miR-93 and miR-572 and theirTFs were exemplified In Figure 3(a) the network modulecontains an incoherent FFL composed by AF2120572 miR-93 andp21 and a cascaded regulation in which p21 is repressed
by FOXF2 via miR-572 in Figure 3(b) the same cascadedregulation together with a coherent FFL composed of TP53MYC miR-93 and p21 forms another regulatory moduleof p21 By modulating the transcriptional strengths of thetwo miRNAs by their TFs two types of p21 dynamics wereidentified saturation and pulse (Figure 3(c) top) In theformer the p21 expression increases and reaches its steadystate at the highest level in contrast in the latter the p21expression increases to a peak and thereafter drops to asteady state at lower level For the two different networkmodules various combinations of transcriptional strengthsof the two miRNAs lead to different distributions of thetwo p21 dynamical patterns For the network module of anincoherent FFL plus the cascaded regulation (Figure 3(c)bottom left) the saturation pattern appears in two distinctregions one with weakened transcriptional strength of miR-93 and the other with enhanced transcriptional strength ofmiR-93 for the other module (Figure 3(c) bottom right)the saturation pattern only appears in the region in whichthe transcriptional strength of miR-93 is enhanced Takentogether the results showed that for the two different networkmotifs the dynamical pattern of p21 is changing according to
BioMed Research International 13
different combinations of its upstream regulators suggestingthe adaptation of p21 dynamics for different biological con-texts
Furthermore we performed a number of simulations toshow the influence of miRNA regulation on p21 expressionlevels in different cellular processes arranged in a consecutivemanner In this procedure a cell is first in the process ofcell cycle followed by proliferation (cell growth) then thecell responses to DNA damage and enters into the processof senescence and finally apoptosis is initiated As shown inFigure 4 during the cell cycle process the p21 expression islow due to the activation of most of its targeting miRNAswhen the cell starts proliferating the p21 expression declinesto an even lower level because of more activated miRNAsin this process after responding to DNA damage the p21expression soars to a high level which is caused by theactivation of its TFs like p53 and fewer expressed miRNAsalthough the p21 expression keeps at a similar level whilethe cell is undergoing senescence the expressed miRNAs aredifferent from the previous process finally the p21 expressiondecreases again to a low level due to the reemergence ofmost of its targeting miRNAs and the cell enters apoptosisInterestingly the model simulations are also consistent withexperimental observations under non-stressed condition thelow expression level of p21 is needed for cell proliferation theupregulation of p21 happens after response to DNA damagevia p53 and the increased p21 expression further results incell cycle arrest leading to senescence and apoptosis [25]Besides when considering the effect of cooperatingmiRNAsthe p21 expression levels were indistinguishable from theprevious simulation of DNA damage response However forthe other processes p21 expression levels were significantlylower compared to the simulations without considering thecooperative effect of the miRNAs Above all these resultsindicated that selective expression of cooperative miRNAscould be adopted by cells to ensure diverse expression levelsof p21 tomeet the requirements of different cellular processes
3 Conclusions and Discussion
In this paper we presented a systems biology approach com-bining data-drivenmodeling andmodel-driven experimentsto investigate the role of miRNA-mediated repression in generegulatory networksThis approach provides a systematicwayto gain a deeper understanding of the regulation of targetgenes by mutiple and cooperative miRNAs Using the regu-lation of p21 by multiple miRNAs as a case study we showedhow the ODE-basedmodel which is calibrated and validatedby means of experimental data is suitable for predicting thetemporal dynamics of molecular concentrations involved inbiochemical systems
Provided there are sufficiently rich quantitative data setsavaliable to characterize the model the use of the methodol-ogy here shown can be extended to more complex regulatorynetworks involving multiple targets cooperating TFs andmiRNAs and signaling pathways displaying cross-talk viapost-tranlational modifications In this case the criticalelement is the quality and quantitiy of the available data
Insufficiency and low quality of experimental data can causeerrors in the process of model construction and overfitting inparameter estimation can lead to uncertainties in the modelpredictions We believe that the quick development of quan-titative high throughput techniques such as transcriptomicsproteomics and miRNomics will facilitate the constructionand characterization of larger miRNA-mediated regulatorynetworks
Other modeling frameworks than ODE-based modelscan be used to describe biological systems such as prob-abilistic (eg Bayesian) or logical (eg Boolean) modelsImportantly different modeling frameworks have differentproperties and perform well regarding different perspectivesand levels of mechanistic details of biochemical systems [46]For example Bayesian models are helpful in the constructionof connections in signaling networks and can reveal the mostlikely underlying structure of the network in a probabilisticmanner Boolean models use binary values (0 and 1) andlogical gates (AND NOT and OR) to describe activities ofnetwork components and the information flow among themWe believe that in the coming future hybrid models whichconsist of modeling framework and experimental techniquespecific sub-modules will provide the necessary compromisebetween quantitativequalitative accuracy and scalability forthe investigation of large biochemical networks [47]
Disclosures
The authors declare that they have no competing financialinterests
Authorsrsquo Contribution
Julio Vera and Olaf Wolkenhauer are equal contributers
Acknowledgments
The authors would like to acknowledge the funding fromthe German Federal Ministry of Education and Research(BMBF) eBio-miRSys (0316175A to Xin Lai and Julio Vera)eBio-SysMet (0316171 to OlafWolkenhauer) and Gerontosys-ROSAge (0315892A to Olaf Wolkenhauer)
References
[1] A E Pasquinelli ldquoMicroRNAs and their targets recognitionregulation and an emerging reciprocal relationshiprdquo NatureReviews Genetics vol 13 no 4 pp 271ndash282 2012
[2] J Vera X Lai U Schmitz and O Wolkenhauer ldquoMicroRNA-regulated networks the perfect storm for classical molecularbiology the ideal scenario for systems biologyrdquo Advances inExperimental Medicine and Biology vol 774 pp 55ndash76 2013
[3] S Nikolov J Vera U Schmitz and O Wolkenhauer ldquoA model-based strategy to investigate the role of microRNA regulation incancer signalling networksrdquo Theory in Biosciences vol 130 no1 pp 55ndash69 2011
14 BioMed Research International
[4] X Lai U Schmitz S K Gupta et al ldquoComputational analysisof target hub gene repression regulated by multiple and cooper-ative miRNAsrdquoNucleic Acids Research vol 40 no 18 pp 8818ndash8834 2012
[5] X Lai O Wolkenhauer and J Vera ldquoModeling miRNA regu-lation in cancer signaling systems miR-34a regulation of thep53Sirt1 signaling modulerdquoMethods in Molecular Biology vol880 pp 87ndash108 2012
[6] C Jiang Z Xuan F Zhao and M Q Zhang ldquoTRED atranscriptional regulatory element database new entries andother developmentrdquoNucleic Acids Research vol 35 supplement1 pp D137ndashD140 2007
[7] L A Bovolenta M L Acencio and N Lemke ldquoHTRIdban open-access database for experimentally verified humantranscriptional regulation interactionsrdquo BMCGenomics vol 13article 405 2012
[8] VMatys O V Kel-Margoulis E Fricke et al ldquoTRANSFAC andits module TRANSCompel transcriptional gene regulation ineukaryotesrdquo Nucleic Acids Research vol 34 database issue ppD108ndashD110 2006
[9] D Karolchik R Baertsch M Diekhans et al ldquoThe UCSCgenome browser databaserdquo Nucleic Acids Research vol 31 no1 pp 51ndash54 2003
[10] P Alexiou T Vergoulis M Gleditzsch et al ldquomiRGen 20 adatabase of microRNA genomic information and regulationrdquoNucleic Acids Research vol 38 supplement 1 pp D137ndashD1412010
[11] F Xiao Z Zuo G Cai S Kang X Gao and T Li ldquomiRecordsan integrated resource for microRNA-target interactionsrdquoNucleic Acids Research vol 37 supplement 1 pp D105ndashD1102009
[12] S-D Hsu F-M Lin W-Y Wu et al ldquomiRTarBase a databasecurates experimentally validated microRNA-target interac-tionsrdquo Nucleic Acids Research vol 39 supplement 1 pp D163ndashD169 2011
[13] P Sethupathy B Corda and A G Hatzigeorgiou ldquoTarBase acomprehensive database of experimentally supported animalmicroRNA targetsrdquo RNA vol 12 no 2 pp 192ndash197 2006
[14] H Dweep C Sticht P Pandey and N Gretz ldquomiRWalkmdashdatabase prediction of possible miRNA binding sites byldquowalkingrdquo the genes of three genomesrdquo Journal of BiomedicalInformatics vol 44 no 5 pp 839ndash847 2011
[15] T S Keshava Prasad RGoel K Kandasamy et al ldquoHumanpro-tein reference databasemdash2009 updaterdquo Nucleic Acids Researchvol 37 supplement 1 pp D767ndashD772 2009
[16] D Szklarczyk A Franceschini M Kuhn et al ldquoThe STRINGdatabase in 2011 functional interaction networks of proteinsglobally integrated and scoredrdquo Nucleic Acids Research vol 39supplement 1 pp D561ndashD568 2011
[17] D Croft G OrsquoKelly G Wu et al ldquoReactome a database ofreactions pathways and biological processesrdquo Nucleic AcidsResearch vol 39 supplement 1 pp D691ndashD697 2011
[18] N le Novere M Hucka H Mi et al ldquoThe systems biologygraphical notationrdquoNature Biotechnology vol 27 no 8 pp 735ndash741 2009
[19] A Funahashi Y Matsuoka A Jouraku M Morohashi NKikuchi and H Kitano ldquoCellDesigner 35 a versatile modelingtool for biochemical networksrdquo Proceedings of the IEEE vol 96no 8 pp 1254ndash1265 2008
[20] M S Cline M Smoot E Cerami et al ldquoIntegration ofbiological networks and gene expression data using CytoscaperdquoNature Protocols vol 2 no 10 pp 2366ndash2382 2007
[21] D-H Le andY-KKwon ldquoNetDS aCytoscape plugin to analyzethe robustness of dynamics and feedforwardfeedback loopstructures of biological networksrdquo Bioinformatics vol 27 no 19pp 2767ndash2768 2011
[22] E O Voit Computational Analysis of Biochemical SystemsA Practical Guide for Biochemists and Molecular BiologistsCambridge University Press Cambridge UK 2000
[23] S Wu S Huang J Ding et al ldquoMultiple microRNAs modulatep21Cip1Waf1 expression by directly targeting its 31015840 untranslatedregionrdquo Oncogene vol 29 no 15 pp 2302ndash2308 2010
[24] P Saeligtrom B S E Heale O Snoslashve Jr L Aagaard J Alluin andJ J Rossi ldquoDistance constraints between microRNA target sitesdictate efficacy and cooperativityrdquo Nucleic Acids Research vol35 no 7 pp 2333ndash2342 2007
[25] Y-S Jung YQian andXChen ldquoExamination of the expandingpathways for the regulation of p21 expression and activityrdquoCellular Signalling vol 22 no 7 pp 1003ndash1012 2010
[26] W Wang H Furneaux H Cheng et al ldquoHuR regulatesp21 mRNA stabilization by UV lightrdquo Molecular and CellularBiology vol 20 no 3 pp 760ndash769 2000
[27] U Wittig R Kania M Golebiewski et al ldquoSABIO-RKmdashdatabase for biochemical reaction kineticsrdquo Nucleic AcidsResearch vol 40 database issue pp D790ndashD796 2012
[28] R Milo P Jorgensen U Moran G Weber and M SpringerldquoBioNumbersmdashthe database of key numbers in molecular andcell biologyrdquo Nucleic Acids Research vol 38 supplement 1 ppD750ndashD753 2010
[29] I-C Chou and E O Voit ldquoRecent developments in parameterestimation and structure identification of biochemical andgenomic systemsrdquoMathematical Biosciences vol 219 no 2 pp57ndash83 2009
[30] A Saltelli K Chan and E M Scott Sensitivity Analysis JohnWiley amp Sons New York NY USA 1st edition 2000
[31] S Strogatz Nonlinear Dynamics and Chaos Applications toPhysics Biology Chemistry and Engineering With Applicationsto Physics Biology Chemistry and Engineering Westview PressBoulder Colo USA 2000
[32] P Zhou S Cai Z Liu and R Wang ldquoMechanisms generatingbistability and oscillations in microRNA-mediated motifsrdquoPhysical Review E vol 85 no 4 apart 1 Article ID 041916 9pages 2012
[33] S Marino I B Hogue C J Ray and D E Kirschner ldquoAmethodology for performing global uncertainty and sensitivityanalysis in systems biologyrdquo Journal of Theoretical Biology vol254 no 1 pp 178ndash196 2008
[34] L P LimNC Lau PGarrett-Engele et al ldquoMicroarray analysisshows that some microRNAs downregulate large numbers of-target mRNAsrdquo Nature vol 433 no 7027 pp 769ndash773 2005
[35] J Krutzfeldt N Rajewsky R Braich et al ldquoSilencing ofmicroRNAs in vivowith lsquoantagomirsrsquordquoNature vol 438 no 7068pp 685ndash689 2005
[36] MWPfafflGWHorgan andLDempfle ldquoRelative expressionsoftware tool (REST) for group-wise comparison and statisticalanalysis of relative expression results in real-time PCRrdquoNucleicAcids Research vol 30 no 9 p e36 2002
[37] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[38] T Abbas and A Dutta ldquop21 in cancer intricate networks andmultiple activitiesrdquo Nature Reviews Cancer vol 9 no 6 pp400ndash414 2009
[39] J Wang M Lu C Qiu and Q Cui ldquoTransmiR a transcriptionfactor microRNA regulation databaserdquo Nucleic Acids Researchvol 38 supplement 1 pp D119ndashD122 2010
BioMed Research International 15
[40] S Bandyopadhyay andM Bhattacharyya ldquoPuTmiR a databasefor extracting neighboring transcription factors of humanmicroRNAsrdquo BMC Bioinformatics vol 11 article 190 2010
[41] A Le Bechec E Portales-Casamar G Vetter et alldquoMIRNTN a framework integrating transcription factorsmicroRNAs and their targets to identify sub-network motifs ina meta-regulation network modelrdquo BMC Bioinformatics vol12 article 67 2011
[42] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyTheGene Ontology ConsortiumrdquoNature Genetics vol 25 no 1 pp 25ndash29 2000
[43] A I F Vaz and L N Vicente ldquoA particle swarm pattern searchmethod for bound constrained global optimizationrdquo Journal ofGlobal Optimization vol 39 no 2 pp 197ndash219 2007
[44] W H PressNumerical Recipes The Art of Scientific ComputingCambridge University Press Cambridge UK 2007
[45] J G Doench and P A Sharp ldquoSpecificity of microRNA targetselection in translational repressionrdquo Genes and Developmentvol 18 no 5 pp 504ndash511 2004
[46] B Kholodenko M B Yaffe and W Kolch ldquoComputationalapproaches for analyzing information flow in biological net-worksrdquo Science Signaling vol 5 no 220 2012
[47] F M Khan U Schmitz S Nikolov et al ldquoHybrid modeling ofthe crosstalk between signaling and transcriptional networksusing ordinary differential equations and multi-valued logicrdquoBiochimica et Biophysica Acta 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
BioMed Research International 5
and validated using temporal experimental data To do sothe data can either be derived from literature or generatedto calibrate the model by own experiments In case of ODEmodels the most suitable data for model calibration isquantitative time-series data obtained from perturbation orquantitative dose-response experimentsThe experiments inwhich time-series data are measured for different regulators(such as miRNAs and TFs) of a gene regulatory network canbe obtained using the techniques described in the subsectionsbelow
(1) qRT-PCR Quantitative real time polymerase chain reac-tion (qRT-PCR) has been used to identify mRNAs regulatedby overexpression or silencing of a specific miRNA [3435] miRNAs typically exhibit their regulatory effects byassociating with specific 31015840 UTR regions of the mRNAscalled miRNA seed regions [34] This association can leadeither to a temporary inhibition of translation or completedegradation of the mRNA in which case qRT-PCR mediateddetection is beneficial For this the cells are transfectedwith the miRNA and non-targeting control oligonucleotidesat an appropriate concentration using either a lipid basedtransfection reagent or nucleofection The transfected cellsare then incubated for the necessary time periods (eg 24 h48 h 72 h etc) after which lysates are prepared and total RNAis extracted 1000ndash2000 ng of the RNA is then converted intocDNA using a reverse transcription kit Taqman qRT-PCR isperformed using 10ndash20 ng of cDNA and primers labeled withfluorescence probes to detect the transcriptional levels of thetarget mRNA Housekeeping genes like GAPDH and HPRTare used as endogenous controls for data normalizationRelative expression at different time points is determinedby comparing the Ct values of the miRNA transfected cellswith Ct values of non-targeting control transfected cells andexpressed as relative expression [36] However qRT-PCR inspite of being highly sensitive is applicable specifically underconditions of complete or partial degradation of the targetmRNA miRNA-mediated inhibition in translation can bebetter demonstrated by techniques like immunoblotting
(2) Western BlotImmunblotting Western blotting or proteinimmunoblotting is a technique to detect the expression of agene at protein level This technique is particularly useful indetermining the regulatory effects of a miRNA on expressionof a target gene which is temporarily inhibited For thisthe cells are transfected with a miRNA or an antagomiRas mentioned above and cell lysates are prepared at appro-priate time points using protein lysis buffers (eg Radio-Immunoprecipitation Assay buffer or RIPA) containing lysisagents like Dithiothreitol (DTT) and protease inhibitors Theproteins from each sample are quantitated using Bradfordor BCA reagents and compared with bovine serum albumin(BSA) standards for accurate protein estimation 20ndash40 120583g ofprotein is then loaded and resolved on a sodium dodecyl sul-fate polyacrylamide gel (SDS-PAGE) along with a pre-stainedprotein marker The protein bands are then transferred ontoa nitrocellulose membrane followed by incubation with theappropriate primary and secondary antibodies linked tofluorescent dyes or horse radish peroxidase enzyme (HRP)
The protein expression is then analyzed using either fluores-cence or chemiluminiscenceHRP substrates in gel documen-tation systems (eg LI-COR Odyssey) Housekeeping geneslike 120573-actin or 120573-tubulin are used for protein normalizationThe time point for maximum target gene suppression gen-erally varies depending on the number of miRNA bindingsites at the 31015840 UTR of the target gene and the extent ofcomplementarity of the seed region [37] Immunoblotting isa widely used technique to provide confirmatory evidence forthe inhibitory effects ofmiRNA at the protein level but it failsto explain the underlying interaction mechanisms
(3) Reporter Gene Assay As each miRNA can inhibit theexpression of a large number of genes regulation of a par-ticular target gene may either be by direct interaction or bean indirect consequence of it In direct regulation a miRNAbinds to the complementary sequences at the 31015840 UTR of atarget gene and thereby suppresses its expression As a con-sequence of this the expression levels of a number of down-stream genes (indirect targets) are also dysregulated makingit crucial to differentiate between primary and secondarymiRNA targets To determine the interaction specificitya reporter construct (luciferase) with intact or mutated 31015840UTR of the target gene cloned at the 51015840 end is co-transfectedinto the cells along with the miRNA The regulatory effect ofthe miRNA on the target gene expression is then measuredusing the expression of a reporter gene In the absence ofthe appropriate binding sequences (mutated 31015840 UTR) themiRNA cannot suppress the reporter mRNA suggesting thatthe suppressive effect of the miRNA is mediated by a directinteraction The reporter activity can be analyzed at differenttime points such as 24 hr 48 hr and 72 hr to determine thetime dependent suppression of a target gene expression bya miRNA
22 Case Study The Regulation of p21 by Multiple andCooperative miRNAs
221 Construction and Visualization of p21 Regulatory Net-work By using the approach described above we investi-gated the regulation of p21 by its multiple targeting miRNAsp21 also known as cyclin-dependent kinase inhibitor 1(CDKN1A) is a transcriptional target of p53 It is requiredfor proper cell cycle progression and plays a role in cell deathDNA repair senescence and aging (reviewed in [38]) Inter-estingly p21 was the first experimentally validated miRNAtarget hub which is a gene that is simultaneously regulatedby many miRNAs This made it an ideal candidate for a casestudy of our approach [23] To do so we first constructeda p21 regulatory network using the following steps
(a) We extracted miRNA-target interactions from thepublication of Wu et al [23] where a list of predictedp21-regulating miRNAs was subjected to experimen-tal validation
(b) Experimentally verified TFs of p21 were extractedfrom literature and putative TFs having conservedbinding sites in the 5 kb upstream region of the p21open reading frame were extracted from UCSC table
6 BioMed Research International
browser A list of TFs controlling the expression of themiRNAswas constructed using information of exper-imentally proven TF-miRNA interactions extractedfrom TransmiR (release 10) [39] In addition wegenerated a list of putative TFs of miRNAs withbinding sites in the 10 kb upstream region of themiRNA genes using information from the databasesPuTmiR (release 10) [40] and MIRNTN (version121) [41] and from the table of TFs with conservedbinding sites in theUCSC genome browser (hg18) [9]
(c) Information about protein interactions was extractedfrom theHuman Protein ReferenceDatabase (HPRDrelease 90) [15] and the STRING database (release90) [16] Only the experimentally verified p21-proteininteractions were used to construct the network
(d) Additionally we associated the TFs in the network tonine biological processes based on the Gene Ontol-ogy (GO) [42] The corresponding GO terms werecell proliferation cell apoptosis immune responseinflammatory response cell cycle DNA damage cellsenescence DNA repair and cell migration
Next we visualized the network in CellDesigner andcomputed a confidence score for each interaction in thenetwork (Figure 1) The confidence scores provide us withthe reliability of the interactions considered in p21 regulatorynetwork With the help of these scores we discarded non-reliable interactions and constructed the mechanistic modelaccounting for p21 regulation by its targeting miRNAsBesides the interested experimentalists can further use thisinformation to choose reliable interacting candidates of p21for designing relevant experiments The scores for eachinteraction of p21 regulatory network are shown in Table 2
222 Mechanistic Modeling of p21 Regulation byIts Targeting miRNAs
(1) Model Construction After constructing the regulatorynetwork a detailed mechanistic model of ODEs whichdescribes the biochemical reactions underlying the regula-tion of p21 was established We chose the ODE modelingapproach as it is a simple formalism for describing temporaldynamics of biochemical systems and a wide range of toolsare available to explore their properties Precisely the modelconsidered themRNA(mp21 (3)) andprotein (p21 (6)) of themiRNA target hub p21 the p21-targeting miRNAs (miR
119894 119894 isin
(1 15) (5)) and the complexes formed by p21 mRNAand miRNA [mp21 | miR
119894] (4) Altogether the model is
constituted by 32 state variables and 64 parameters
119889mp21119889119905
= 119896mp21syn sdot 119891act (TFmp21)
minusmp21 sdot (119896mp21deg +sum
119894
119896complex
119894
ass sdotmiR119894)
(3)
119889 [mp21 | miR119894]
119889119905
= 119896complex
119894
ass sdotmp21 sdotmiR119894minus 119896
complex119894
deg sdot [mp21 | miR119894]
(4)
119889miR119894
119889119905= 119896
miR119894syn sdot 119891act (TFmiR119894)
minusmiR119894sdot (119896
miR119894deg + 119896
complex119894
ass sdotmp21) (5)
dp21119889119905
= 119896p21syn sdotmp21 + 119896p21deg sdot p21 (6)
mp21Total = mp21 +sum119894
[mp21 | miR119894] (7)
For mp21 processes considered in the model were (i) itssynthesis (119896mp21
syn ) mediated by TFs (119891act(TFmp21)) (ii) itsdegradation (119896mp21
deg ) and (iii) its association with a miRNA(119896complex
119894
ass ) For each miR119894 processes considered were (i)
the synthesis (119896miR119894syn ) mediated by TFs (119891act(TFmiR119894)) (ii) the
degradation (119896miR119894deg ) and (iii) the association with the p21
mRNA target (119896complex119894
ass ) For each [mp21 | miR119894] complex
processes considered were (i) the formation of the complexby a miR
119894and the p21 mRNA (119896complex
119894
ass ) and (ii) the complexdegradation (119896complex
119894
deg ) For p21 processes considered were(i) its synthesis (119896p21syn ) and (ii) its degradation (119896p21deg) Anadditional algebraic equation accounting for the total mea-surable amount of p21 mRNA (mp21Total) was also includedThe SBML file of the model is available for download athttpwwwsbiuni-rostockdeuploadstx templavoilap21TargetHub 03092013xml
(2) Model Calibration and Validation For model calibrationwe fixed some parameter values using published data andestimated the other unknown parameters using the time-series data published by Wu et al [23] in which the p21mRNA (northern plot) and protein levels (western plot)were measured 48 hr after transfection of individual p21-targeting miRNAs into human embryonic kidney 293 cellsThe unknown parameter values were estimated using aniterative method combining global (particle swarm patternsearch) [43] and local (downhill simplex method in multi-dimensions) [44] optimization algorithms For each miR
119894
considered in the model the method minimizes the distancebetween model simulations and experimental data using thefollowing cost function
119865miR119894cost =
[mp21miR119894sim (119905) minusmp21miR119894
exp (119905)]2
(120575mp21119894
)2
+
[p21miR119894sim (119905) minus p21miR119894
exp (119905)]2
(12057511990121
119894)2
119894 isin [1 15]
(8)
BioMed Research International 7
1 2 4 1 2 4 5 6 1 2 3 7
E2F TP53 RELA1 2 4 7 2 3 7 6
STAT5 SRF1 2 3 4 1 2
JUN STAT1
SP1 RUNX1 TFAP2A
Common TFs Modulation
CDKN1A
1 2 1 2 4 1 4
1 4 1 2 3 7
1 2 5 2 4 5 7 1
VDR
CUX1
E2A
RAR BRCA1
TP63 TP73 TBX2
STAT6 SP3 STAT3
STAT2 RUNX2
EHMT2 p21 TFs
1 2 4 5 8 9
CEBP120573NF120581B1
(1) Cell proliferation (6) Cell senescence
(2) Apoptosis (7) Inflammatory response
(3) Immune response (8) DNA repair
(4) Cell cycle control (9) Cell migration
(5) DNA damage
miRNAmiRNA
Translational repressionmiRNA
Translation
Transcription
Association
miR-132 miR-28-5p
miR-299-5p miR-298
miR-345 miR-125a-5p
miR-208a
miRNA
miRNADeadenylation followed by decay
Protein-protein interaction
miR-93 miR-654-3p
miR-657 miR-572
miR-423-3p miR-363
miR-639 miR-515-3p
p21
CCND2 CDK3 GMNN CCNE1 POLD2 CSNK2B C1orf123 CCDC85B
PIM1 CCNB2 KIAA0432 GADD45B HDAC11 ITGB1BP3 TTLL5 SLC25A11
GADD45G CDC7 BCCIP CDK1 GNB2LI RAB1A psmA3 TK1
CDC45 CCNB1 CCNA2 GADD45A MAPK8 TSG101 SET CIZ1
CDK14 CDK5 MDM2 CDC27 RNF144B ESR1 PRKCH RPA1
CDC20 SKP2 CDK2 CDK4 DHX9 PCNA CDC5L CCDC5
CCND3 CDK6 CCNE2 CCNA1 TEX11 MAP3K5 MCM10 Caspase-3
CDC6 CCND1 XRCC6 PARP1 DAPK3 NR4A1 CEBPA
Interacting proteins
AKT1
miRNA genes
5 8 1 2 1 2 4 3 1 2 3 3 2 2 3 3 1 4 1 3 1 4
1 4
NFATC2 YY1 CREB1 CEBPA ARNT
FOXCI1
ELK1 SOX9 FOS PAX5 MAX MEF2A RFX1 JUNB AKT1 GATA3 MYC NRSFEGR1
E47 PPARG POU3F2 FOXF2FOXL1 MZF1 NFIC SRY
FOXC1
HNFA4 NFYA MYF1 BACF1 ATF6 ZID RORA POU2F1
miRNA TFs
11 2 34
12 31 5 11
HNF1A
1 2
RB1
Cellular processes
Figure 1 p21 regulatory network The network contains several layers of regulators of p21 TFs (light blue and red boxes) miRNAs (darkblue and green boxes) and proteins (grey boxes) In each big box there are small boxes which represent individual components of this layerof regulation miRNAs are classified into two groups according to the mechanisms by which the expression of p21 is repressed One groupcauses p21 translation repression (green box) These miRNAs bind to p21 mRNA resulting in the repressed translation of p21 but unchangedmRNA expression level The other group of miRNAs (dark blue box) decreases the stability of p21 mRNA by modifying its structure leadingto mRNA decay and finally the downregulation of p21 TFs are classified into three groups p21 TFs (red) miRNA TFs (yellow) and theircommon TFs (light blue) The p21 interacting-proteins are framed in the grey boxes The purple boxes represent nine processes and the TFsassociated with these processes are indicated in the ellipses above them using corresponding figures This data is adapted from our previouspublication [4]
where mp21miR119894sim (119905) and mp21miR119894
exp (119905) represent the simulatedp21 mRNA and protein expression levels for each miR
119894at
time point 119905 p21miR119894sim (119905) and p21miR119894
exp (119905) represent the mea-sured value for each miR
119894at time point 119905 and their standard
deviations are 120575mp21119894
and 120575p21119894
Here 119905 is the time point (48 hr)after overexpression of the individual miRNAs in embryonickidney 293 cells at which the expression levels of the p21 andits mRNA were measured [23] The model calibration resultsare shown in Figure 2(a) and the obtained parameter valuesare listed in Table 3
Experimental results showed that a stronger repressionof the target gene can occur when two miRNA binding siteson the target mRNA are in close proximity [24 45] To testthe consequences of this hypothesis we predicted cooperativemiRNA pairs for p21 with seed site distances between 13ndash35 nt in the p21 31015840 UTR To substantiate the cooperativeeffect associated with pairs of miRNAs we introduced agroup of new state variables ([mp21 | miR
119894| miR
119894]) into
the original model These state variables account for the
ternary complexes composed of p21 mRNA and two puta-tively cooperating miRNAs (miR
119894and miR
119895) For these new
variables processes considered are (i) the association of p21mRNA with miR
119894and miR
119895into a complex (119896co-complex
119894119895
ass )and (ii) the degradation of the complex (119896co-complex
119894119895
deg ) Afterexpansion the corresponding modified and new ODEs arelisted below
dmp21119889119905
= 119896mp21syn sdot 119891act (TFmp21)
minusmp21 sdot (119896mp21deg +sum
119894
119896complex
119894
ass sdotmiR119894
+sum
119894119895
119896co-complex
119894119895
ass sdotmiR119894sdotmiR119895)
(9)
8 BioMed Research InternationalTa
ble2Th
econ
fidence
scores
forthe
interactions
inp21regulatorynetworkTh
enetworkconsistso
ffou
rtypes
ofinteractionsm
iRNA-
p21TF
-miRNAT
F-p21andp21-p
rotein
(a)miRNA-
p21interactio
nscores
miRNA
miR-125a-5p
miR-132
miR-208a
miR-28-5p
miR-298
miR-299-5p
miR-345
miR-363
Score
073
073
073
080
073
080
073
073
miRNA
miR-423-3p
miR-515-3p
miR-572
miR-639
miR-654-3p
miR-657
miR-93
Score
085
073
073
073
073
080
095
(b)TF
-miRNAinteractionscores
Score
Score
Score
Score
Score
Score
miR-208
miR-132
miR-657
miR-125a-5p
miR-28-5p
miR-93
MAX
027
EGR1
093
TFAP2
A034
EGR1
086
MZF
1028
TFAP2
A034
EGR1
029
CREB
1086
SP1
034
PAX5
024
E47
027
SOX9
024
ARN
T027
ARE
B6027
MZF
1034
NRS
F024
FOXC
1031
GAT
A3
027
RFX1
024
E47
024
MAX
029
ZID
024
FOXI1
024
MZF
1031
ATF6
024
ELK1
027
USF1
027
PPARG
027
SRY
027
SP1
034
YY1
027
EGR2
029
EGR1
034
GAT
A3
034
JUN
029
POU2F1
024
JUNB
024
FOS
024
RELA
029
MZF
1034
POU2F1
024
EGR1
029
POU2F1
024
NFIC
024
miR-654-3p
TP53
034
SRF
027
CEBP
A024
NFIC
027
RELA
027
YY1
027
NFIC
034
CEBP
A029
NFY
A024
miR-345
miR-423-3p
BACH
1024
miR-299-5p
RORA
024
RUNX1
024
SP1
034
BACH
1027
FOXL
1029
CREB
1024
miR-363
STAT
1024
RELA
027
STAT
1027
SRY
024
SRY
027
E2F1
062
E2F1
085
NFA
TC2
024
POU3F2
024
NFA
TC2
024
SRF
029
FOXC
1033
MYC
084
HNF4
A031
STAT
5A027
FOS
024
RUNX1
027
MZF
1034
miR-298
NFY
A024
SRF
027
POU2F1
027
miR-572
miR-639
JUNB
031
POU2F1
024
MEF
2A027
NF120581
B1024
FOXF
2024
NFY
A024
NFIC
031
POU2F1
024
HNF1A
024
(c)TF
-p21
interactionscores
TF(verified)
SP1
SP3
E2F1
Runx
1Ru
nx2
STAT
1E2
AST
AT5
TP53
TP63
TP73
Score
100
100
077
077
077
072
072
067
067
067
067
TF(verified)
STAT
3CU
X1TF
AP2
ABR
CA1
VDR
RARA
CEB
P120572CEB
P120573RB
1Tb
x2EH
MT2
Score
066
066
060
060
060
060
060
060
060
060
060
TF(putative)
NF120581
B1RE
LAST
AT2
STAT
6SR
FScore
044
044
044
044
043
(d)p21-p
rotein
interactionscores
Protein
TP53
PCNA
CASP
3CC
NA1
CCND1
SKP2
BCCI
PCC
NA2
CCND2
CCNE2
AKT
1C1
orf123
CCDC8
5BCC
NB1
CCNB2
CCND3
Score
100
092
087
087
087
087
081
081
081
081
072
072
072
072
072
072
ProteinCC
NE1
CDC4
5CD
C5L
CDC6
CDC7
CDK1
CDK14
CDK2
CDK3
CDK4
CDK6
CEBP
ACI
Z1CS
NK2
A1
CSNK2
BDAPK
3Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
ESR1
GADD45A
GADD45B
GADD45G
GMNN
GNB2
L1HDAC
11ITGB1BP
3MAP3
K5MAPK
8MCM
10PA
RP1
PIM1
POLD
2PS
MA3
RAB1A
Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
SET
SLC2
5A11
STAT
3TE
X11
TK1
TSG101
TTLL
5XR
CC6
CDC2
0CD
C27
CDK5
DHX9
MDM2
NR4
A1
PRKC
HRP
A1
Score
072
072
072
072
072
072
072
072
056
056
056
056
056
056
056
056
Verifi
edexp
erim
entally
verifi
edinteraction
putativ
epredictedinteraction
BioMed Research International 9
Con
trol
miR
-298
miR
-208
am
iR-1
32m
iR-2
8-5p
miR
-125
-5p
miR
-299
-5p
miR
-345
miR
-93
miR
-423
-3p
miR
-515
-3p
miR
-363
miR
-657
miR
-639
miR
-572
miR
-654
-3p
0
05
1
15m
RNA
expr
essio
n (a
u)
Data Model Data ModelC
ontro
lm
iR-2
98m
iR-2
08a
miR
-132
miR
-28-
5pm
iR-1
25-5
pm
iR-2
99-5
pm
iR-3
45m
iR-9
3m
iR-4
23-3
pm
iR-5
15-3
pm
iR-3
63m
iR-6
57m
iR-6
39m
iR-5
72m
iR-6
54-3
p
0
02
04
06
08
1
Prot
ein
expr
essio
n (a
u)
(a)
DoxorubicinCellWell
Control miR-572 miR-93
Measure p21 expression at the time points of
miR-572+ miR-93
0 2 4 6 8 24hr
miR-572miR-93
Immunoblotting
(b)
0 4 8 12 16 20 240
4
8
12
16
20
Time (hr)
p53
Actin
Doxo (hrs)
fac
t(TF
mp2
1)
C 0 1 2 4 6 8 24
1 08 12 19 29 32 3 16
(c)
0 2 4 6 8 2402468
1012
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)
p21
expr
essio
n (a
u)
02468
1012
miR-572Control
02468
1012
miR-93
02468
1012
Model Data
0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24
1 23 46 107124 124 23 28 48 66 53 31 08 16 21 38 50 18 06 06 12 25 27 12
miR-572+ 93Doxo (hrs)
p21
Actin
p21
Actin
p21
Actin
p21
Actin
(d)
Figure 2 Model calibration and validation (a) Model calibration The figures show the relative change of the p21 mRNA and proteinexpression levels after overexpression of the indicated miRNAs (Model model simulation Data experimental data) These data werenormalized to the control group in which the p21 mRNA and protein expression levels were measured when the miRNAs were normallyexpressed (au arbitrary unit) (b) Experimental workflow In the experiments Sk-Mel-147 cells were seeded in six well plates Then maturemiRNA mimics were transfected individually at a concentration of 100 nM (miR-572 and miR-93) or in combination at 50 nM each (miR-572 + miR-93) After 48 hr transfection with miRNA mimics the cells were pulse treated with 250 nM doxorubicin for 1 hour after whichnormal growth medium was replenished The immunblotting were performed to measure p21 expression at 0 2 4 6 8 and 24 hr post-doxorubicin treatment (c) Temporal dynamics of p21 transcriptional function Afterdoxorubicin treatment the expression of p53 a TF ofp21 was measured using immunblotting and these data were used to characterize the transcriptional function of p21 using MATLAB linearinterpolation function (d) Model validation We measured the expression of p21 protein in response to genotoxic stress in the four scenariosas described in the main text The measured data (Data) were compared with the model simulations (Model) The figures (a) (c) and (d) areadapted from our previous publication [4]
119889 [mp21 | miR119894| miR
119895]
119889119905
= 119896co-complex
119894119895
ass sdotmp21 sdotmiR119894sdotmiR119895
minus 119896co-complex
119894119895
deg sdot [mp21 | miR119894| miR
119895]
(10)
To model stronger repression of the target gene bycooperating miRNAs we assumed a stronger association rateconstant for the complex [mp21 | miR
119894| miR
119895] which is
equal to the sum of their individual association rate constants
(119896co-complex119894119895
ass = 119896complex
119894
ass + 119896complex
119895
ass ) Similarly the degradationrates of the complexes [mp21 | miR
119894| miR
119895] were assumed
to be equal to the sum of degradation rate constants ofsingle miRNA binding complexes (119896co-complex
119894119895
deg = 119896complex
119894
deg +
119896complex
119895
deg ) However it has to be noted that these addedequations are an abstract description ofmiRNAcooperativitybecause the details of this mechanism are not yet known
To experimentally validate the capability of our model topredict the relative p21 concentrations regulated by coopera-tivemiRNAs we selectedmiR-572 andmiR-93 as a case study
10 BioMed Research International
Table 3 Initial concentrations of model variables and model parameter values Based on the experimental data the p21-targeting miRNAsverified byWu et al [23] were divided into two groups the translation repression group (marked with asterisk) and the mRNA deadenylationgroup AmiRNAwas classified into themRNA deadenylation group if its overexpression can result in 20 ormore downregulation of the p21mRNA level (ie p21mRNA levelle 08 the basal level is 1) otherwise it was classified into the translation repression group For the translationrepression group only 119896complex119894
ass was estimated and 119896complex119894deg was fixed For the other group both 119896complex119894
deg and 119896complex119894ass were estimatedThe initial
concentrations of p21 and mp21 were set to 1 and this value was used as their basal expression levels During the parameter estimation theinitial concentrations of p21-targeting miRNAs were set to 100 because in the publication [23] the expression levels of p21 and mp21 weremeasured after the individual introduction of the p21-targeting miRNAs with amount of 100 nM Due to the lack of biological information tocharacterize the transcriptional activation function (119891act) of p21 and its targeting miRNAs the corresponding functions were assumed to be1 for simplicity The data is adapted from our previous publication [4]
Initial concentration of variables and TF functionsVariable Description Initial concentration (au)p21 p21 protein 1
mp21 p21 mRNA 1
miR119894=115
p21-targeting miRNAs 100
[mp21 | miR119894=115
] Complexes formed by miR119894and mp21 0
119891act(TFmp21) p21rsquos transcriptional activation function 1119891act(TFmiR119894 (119894=115)) The transcriptional activation function of miR
1198941
Fixed parameter valuesParameter Description Value (hrminus1) Reference119896mp21syn Synthesis rate constant of mp21 01155 fixed119896mp21deg Degradation rate constant of mp21 01155 [24]119896miR119894syn (119894 = 1 15) Synthesis rate constant of miR
11989400289 fixed
119896miR119894deg (119894 = 1 15) Degradation rate constant of miR
11989400289 [25]
119896p21syn Synthesis rate constant of p21 13863 fixed119896p21deg Degradation rate constant of p21 13863 [26]
Estimated parameter values
miRNA (statevariable)
119896complex119894deg
(119894 = 1 15) (hrminus1)
119896complex119894ass
(119894 = 1 15)
(au minus1sdot hrminus1)
119865miR119894cost
(119894 = 1 15)Experimental data of p21 (protein mRNA plusmn SD)
miR-298 (miR1)lowast 01155 00254 34119890 minus 004 (016 1074 plusmn 0025)
miR-208a (miR2)lowast 01155 00041 20119890 minus 003 (051 1192 plusmn 0022)
miR-132 (miR3)lowast 01155 00275 24119890 minus 003 (015 121 plusmn 0147)
miR-28-5p (miR4)lowast 01155 00119 59119890 minus 003 (028 135 plusmn 006)
miR-125-5p (miR5)lowast 01155 00018 18119890 minus 003 (069 085 plusmn 0051)
miR-299-5p (miR6)lowast 01155 00080 18119890 minus 004 (036 095 plusmn 0038)
miR-345 (miR7)lowast 01155 00051 11119890 minus 004 (046 096 plusmn 0039)
miR-93 (miR8) 01564 00235 41119890 minus 014 (017 07776 plusmn 003)
miR-423-3p (miR9) 09118 00055 28119890 minus 009 (044 05102 plusmn 011)
miR-515-3p (miR10) 02098 00253 12119890 minus 013 (016 0616 plusmn 0037)
miR-363 (miR11) 02261 00399 22119890 minus 014 (011 056 plusmn 015)
mR-657 (miR12) 03465 00158 21119890 minus 014 (023 048 plusmn 012)
miR-639 (miR13) 04305 00327 18119890 minus 017 (013 036 plusmn 0084)
miR-572 (miR14) 03039 00360 94119890 minus 023 (012 045 plusmn 0044)
miR-654-3p (miR15) 97485 00024 30119890 minus 014 (064 063 plusmn 0053)
These two miRNAs were chosen because their predictedtarget sites in the p21 31015840 UTR are in close proximity to eachother and thereby they can induce cooperative repression onp21 as suggested in [24] The experiments were performed asfollows
(i) Melanoma cells (Sk-Mel-147) were transfected withthe mature miRNAmimics of the twomiRNAs eitherindividually (100 nM) or in combination (50 nMeach) whereas untreated cells were used as control(Figure 2(b))
BioMed Research International 11
p21
FOXF2
miR-572
Activation
Inhibition120572
Incoherent FFL + cascaded regulation
AF2
miR-93
(a)
miR-93
miR-572
TP53
MYC
p21
FOXF2
Coherent FFL + cascaded regulation
(b)
Saturation
Pulse
p21
(au
)
p21
(au
)Time (hr) Time (hr)
Saturation Pulse
Pulse
Saturation
Saturation
di
(au
)
102
101
102
100
101
10minus1
10010minus2
10minus1
102
101
100
10minus1
10minus2
FOXF2 (au) FOXF2 (au)
TP53
(au
)
1
09
08
07
06
05
04
03
02
01
10minus2 10210110010minus110minus20
AF2120572
(au
)
(c)
Figure 3 Different p21 dynamics in different network motifs We ran simulations to show the different dynamics of p21 for two differentnetwork motifs (a) and (b) Through simulations two dynamical patterns of p21 were identified saturation and pulse ((c) top) For eachnetwork motif the corresponding distributions of the two dynamical patterns were plotted ((c) bottom) For different combinations ofthe transcriptional strengths the normalized distance (119889
119894) between peaks (119901
119894) and steady states (ss
119894) of p21 is determined by the equation
119889119894= (119901119894minus ss119894)119901max 119901max = max(119901
1 119901
119899) If 119889
119894= 0 for the corresponding combination of transcriptional strengths the p21 dynamics is
saturation otherwise it is pulse The regions showing different dynamical patterns of p21 are separated using the white lines
(ii) Next the cells were treated with doxorubicina genotoxic-stress inducing agent The agent canupregulate the expression of p53 which is a knownTFof p21 and therefore it can result in the upregulationof p21 (Figure 2(c))
(iii) After doxorubicin treatment the expression levels ofp21 were measured by immunoblotting at differenttime points (0 2 4 6 8 24 hr) The p21 expressionvalues were normalized based on the p21 expression
level in the control group measured at time point 0 hr(Figure 2(d))
By doing so we obtained the p21 response after genotoxicstress in four scenarios (1) endogenous miRNA expression(Control) (2) overexpression of miR-572 (3) overexpressionof miR-93 and (4) both miRNAs moderately overexpressedThereafter we derived a model of seven ODEs based onthe original equations which was configured according tothe designed experiments making the simulation results
12 BioMed Research International
00001
0001
001
01
1Cell cycle Cell growth DNA damage Senescence Apoptosis
G1G2
MS
p21
expr
essio
n (a
u)
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
log10
wow
Figure 4 p21 expression regulated by cooperative miRNAs for different cellular processesThe associations of themiRNAs with these cellularprocesses were derived from GO terms of their TFs A miRNA was supposed to be expressed (in bold black font) in a cellular process onlyif its TF is related to the corresponding GO term of this process The p21 expression levels are computed for each process with (w)without(wo) considering the cooperative effect among the p21 targeting miRNAs
comparable with the experimental data As shown in theFigure 2(c) the simulations are in good agreement with theexperimental observations The individual overexpression ofmiR-572 or miR-93 led to the reduction of the upregulationof p21 response after genotoxic stress induction The twomiRNAs cause different degrees of repression due to theirdifferent repression efficiencies on p21 Interestingly the com-bined overexpression of both miRNAs induced the strongestdownregulation of p21 and therefore verifying the hypothesisof their cooperative regulation of p21 Above all the resultsnot only validated themodel but also demonstrated the abilityof our method to identify cooperative miRNA pairs for p21
(3) Predictive Simulations As there are abundant of networkmotifs such as FFLs in p21 regulatory network and thesenetwork motifs are important for determining p21 dynamicsit is interesting to investigate the dynamics of p21 in networkmodules where FFLs are involved To do so two networkmodules including both miR-93 and miR-572 and theirTFs were exemplified In Figure 3(a) the network modulecontains an incoherent FFL composed by AF2120572 miR-93 andp21 and a cascaded regulation in which p21 is repressed
by FOXF2 via miR-572 in Figure 3(b) the same cascadedregulation together with a coherent FFL composed of TP53MYC miR-93 and p21 forms another regulatory moduleof p21 By modulating the transcriptional strengths of thetwo miRNAs by their TFs two types of p21 dynamics wereidentified saturation and pulse (Figure 3(c) top) In theformer the p21 expression increases and reaches its steadystate at the highest level in contrast in the latter the p21expression increases to a peak and thereafter drops to asteady state at lower level For the two different networkmodules various combinations of transcriptional strengthsof the two miRNAs lead to different distributions of thetwo p21 dynamical patterns For the network module of anincoherent FFL plus the cascaded regulation (Figure 3(c)bottom left) the saturation pattern appears in two distinctregions one with weakened transcriptional strength of miR-93 and the other with enhanced transcriptional strength ofmiR-93 for the other module (Figure 3(c) bottom right)the saturation pattern only appears in the region in whichthe transcriptional strength of miR-93 is enhanced Takentogether the results showed that for the two different networkmotifs the dynamical pattern of p21 is changing according to
BioMed Research International 13
different combinations of its upstream regulators suggestingthe adaptation of p21 dynamics for different biological con-texts
Furthermore we performed a number of simulations toshow the influence of miRNA regulation on p21 expressionlevels in different cellular processes arranged in a consecutivemanner In this procedure a cell is first in the process ofcell cycle followed by proliferation (cell growth) then thecell responses to DNA damage and enters into the processof senescence and finally apoptosis is initiated As shown inFigure 4 during the cell cycle process the p21 expression islow due to the activation of most of its targeting miRNAswhen the cell starts proliferating the p21 expression declinesto an even lower level because of more activated miRNAsin this process after responding to DNA damage the p21expression soars to a high level which is caused by theactivation of its TFs like p53 and fewer expressed miRNAsalthough the p21 expression keeps at a similar level whilethe cell is undergoing senescence the expressed miRNAs aredifferent from the previous process finally the p21 expressiondecreases again to a low level due to the reemergence ofmost of its targeting miRNAs and the cell enters apoptosisInterestingly the model simulations are also consistent withexperimental observations under non-stressed condition thelow expression level of p21 is needed for cell proliferation theupregulation of p21 happens after response to DNA damagevia p53 and the increased p21 expression further results incell cycle arrest leading to senescence and apoptosis [25]Besides when considering the effect of cooperatingmiRNAsthe p21 expression levels were indistinguishable from theprevious simulation of DNA damage response However forthe other processes p21 expression levels were significantlylower compared to the simulations without considering thecooperative effect of the miRNAs Above all these resultsindicated that selective expression of cooperative miRNAscould be adopted by cells to ensure diverse expression levelsof p21 tomeet the requirements of different cellular processes
3 Conclusions and Discussion
In this paper we presented a systems biology approach com-bining data-drivenmodeling andmodel-driven experimentsto investigate the role of miRNA-mediated repression in generegulatory networksThis approach provides a systematicwayto gain a deeper understanding of the regulation of targetgenes by mutiple and cooperative miRNAs Using the regu-lation of p21 by multiple miRNAs as a case study we showedhow the ODE-basedmodel which is calibrated and validatedby means of experimental data is suitable for predicting thetemporal dynamics of molecular concentrations involved inbiochemical systems
Provided there are sufficiently rich quantitative data setsavaliable to characterize the model the use of the methodol-ogy here shown can be extended to more complex regulatorynetworks involving multiple targets cooperating TFs andmiRNAs and signaling pathways displaying cross-talk viapost-tranlational modifications In this case the criticalelement is the quality and quantitiy of the available data
Insufficiency and low quality of experimental data can causeerrors in the process of model construction and overfitting inparameter estimation can lead to uncertainties in the modelpredictions We believe that the quick development of quan-titative high throughput techniques such as transcriptomicsproteomics and miRNomics will facilitate the constructionand characterization of larger miRNA-mediated regulatorynetworks
Other modeling frameworks than ODE-based modelscan be used to describe biological systems such as prob-abilistic (eg Bayesian) or logical (eg Boolean) modelsImportantly different modeling frameworks have differentproperties and perform well regarding different perspectivesand levels of mechanistic details of biochemical systems [46]For example Bayesian models are helpful in the constructionof connections in signaling networks and can reveal the mostlikely underlying structure of the network in a probabilisticmanner Boolean models use binary values (0 and 1) andlogical gates (AND NOT and OR) to describe activities ofnetwork components and the information flow among themWe believe that in the coming future hybrid models whichconsist of modeling framework and experimental techniquespecific sub-modules will provide the necessary compromisebetween quantitativequalitative accuracy and scalability forthe investigation of large biochemical networks [47]
Disclosures
The authors declare that they have no competing financialinterests
Authorsrsquo Contribution
Julio Vera and Olaf Wolkenhauer are equal contributers
Acknowledgments
The authors would like to acknowledge the funding fromthe German Federal Ministry of Education and Research(BMBF) eBio-miRSys (0316175A to Xin Lai and Julio Vera)eBio-SysMet (0316171 to OlafWolkenhauer) and Gerontosys-ROSAge (0315892A to Olaf Wolkenhauer)
References
[1] A E Pasquinelli ldquoMicroRNAs and their targets recognitionregulation and an emerging reciprocal relationshiprdquo NatureReviews Genetics vol 13 no 4 pp 271ndash282 2012
[2] J Vera X Lai U Schmitz and O Wolkenhauer ldquoMicroRNA-regulated networks the perfect storm for classical molecularbiology the ideal scenario for systems biologyrdquo Advances inExperimental Medicine and Biology vol 774 pp 55ndash76 2013
[3] S Nikolov J Vera U Schmitz and O Wolkenhauer ldquoA model-based strategy to investigate the role of microRNA regulation incancer signalling networksrdquo Theory in Biosciences vol 130 no1 pp 55ndash69 2011
14 BioMed Research International
[4] X Lai U Schmitz S K Gupta et al ldquoComputational analysisof target hub gene repression regulated by multiple and cooper-ative miRNAsrdquoNucleic Acids Research vol 40 no 18 pp 8818ndash8834 2012
[5] X Lai O Wolkenhauer and J Vera ldquoModeling miRNA regu-lation in cancer signaling systems miR-34a regulation of thep53Sirt1 signaling modulerdquoMethods in Molecular Biology vol880 pp 87ndash108 2012
[6] C Jiang Z Xuan F Zhao and M Q Zhang ldquoTRED atranscriptional regulatory element database new entries andother developmentrdquoNucleic Acids Research vol 35 supplement1 pp D137ndashD140 2007
[7] L A Bovolenta M L Acencio and N Lemke ldquoHTRIdban open-access database for experimentally verified humantranscriptional regulation interactionsrdquo BMCGenomics vol 13article 405 2012
[8] VMatys O V Kel-Margoulis E Fricke et al ldquoTRANSFAC andits module TRANSCompel transcriptional gene regulation ineukaryotesrdquo Nucleic Acids Research vol 34 database issue ppD108ndashD110 2006
[9] D Karolchik R Baertsch M Diekhans et al ldquoThe UCSCgenome browser databaserdquo Nucleic Acids Research vol 31 no1 pp 51ndash54 2003
[10] P Alexiou T Vergoulis M Gleditzsch et al ldquomiRGen 20 adatabase of microRNA genomic information and regulationrdquoNucleic Acids Research vol 38 supplement 1 pp D137ndashD1412010
[11] F Xiao Z Zuo G Cai S Kang X Gao and T Li ldquomiRecordsan integrated resource for microRNA-target interactionsrdquoNucleic Acids Research vol 37 supplement 1 pp D105ndashD1102009
[12] S-D Hsu F-M Lin W-Y Wu et al ldquomiRTarBase a databasecurates experimentally validated microRNA-target interac-tionsrdquo Nucleic Acids Research vol 39 supplement 1 pp D163ndashD169 2011
[13] P Sethupathy B Corda and A G Hatzigeorgiou ldquoTarBase acomprehensive database of experimentally supported animalmicroRNA targetsrdquo RNA vol 12 no 2 pp 192ndash197 2006
[14] H Dweep C Sticht P Pandey and N Gretz ldquomiRWalkmdashdatabase prediction of possible miRNA binding sites byldquowalkingrdquo the genes of three genomesrdquo Journal of BiomedicalInformatics vol 44 no 5 pp 839ndash847 2011
[15] T S Keshava Prasad RGoel K Kandasamy et al ldquoHumanpro-tein reference databasemdash2009 updaterdquo Nucleic Acids Researchvol 37 supplement 1 pp D767ndashD772 2009
[16] D Szklarczyk A Franceschini M Kuhn et al ldquoThe STRINGdatabase in 2011 functional interaction networks of proteinsglobally integrated and scoredrdquo Nucleic Acids Research vol 39supplement 1 pp D561ndashD568 2011
[17] D Croft G OrsquoKelly G Wu et al ldquoReactome a database ofreactions pathways and biological processesrdquo Nucleic AcidsResearch vol 39 supplement 1 pp D691ndashD697 2011
[18] N le Novere M Hucka H Mi et al ldquoThe systems biologygraphical notationrdquoNature Biotechnology vol 27 no 8 pp 735ndash741 2009
[19] A Funahashi Y Matsuoka A Jouraku M Morohashi NKikuchi and H Kitano ldquoCellDesigner 35 a versatile modelingtool for biochemical networksrdquo Proceedings of the IEEE vol 96no 8 pp 1254ndash1265 2008
[20] M S Cline M Smoot E Cerami et al ldquoIntegration ofbiological networks and gene expression data using CytoscaperdquoNature Protocols vol 2 no 10 pp 2366ndash2382 2007
[21] D-H Le andY-KKwon ldquoNetDS aCytoscape plugin to analyzethe robustness of dynamics and feedforwardfeedback loopstructures of biological networksrdquo Bioinformatics vol 27 no 19pp 2767ndash2768 2011
[22] E O Voit Computational Analysis of Biochemical SystemsA Practical Guide for Biochemists and Molecular BiologistsCambridge University Press Cambridge UK 2000
[23] S Wu S Huang J Ding et al ldquoMultiple microRNAs modulatep21Cip1Waf1 expression by directly targeting its 31015840 untranslatedregionrdquo Oncogene vol 29 no 15 pp 2302ndash2308 2010
[24] P Saeligtrom B S E Heale O Snoslashve Jr L Aagaard J Alluin andJ J Rossi ldquoDistance constraints between microRNA target sitesdictate efficacy and cooperativityrdquo Nucleic Acids Research vol35 no 7 pp 2333ndash2342 2007
[25] Y-S Jung YQian andXChen ldquoExamination of the expandingpathways for the regulation of p21 expression and activityrdquoCellular Signalling vol 22 no 7 pp 1003ndash1012 2010
[26] W Wang H Furneaux H Cheng et al ldquoHuR regulatesp21 mRNA stabilization by UV lightrdquo Molecular and CellularBiology vol 20 no 3 pp 760ndash769 2000
[27] U Wittig R Kania M Golebiewski et al ldquoSABIO-RKmdashdatabase for biochemical reaction kineticsrdquo Nucleic AcidsResearch vol 40 database issue pp D790ndashD796 2012
[28] R Milo P Jorgensen U Moran G Weber and M SpringerldquoBioNumbersmdashthe database of key numbers in molecular andcell biologyrdquo Nucleic Acids Research vol 38 supplement 1 ppD750ndashD753 2010
[29] I-C Chou and E O Voit ldquoRecent developments in parameterestimation and structure identification of biochemical andgenomic systemsrdquoMathematical Biosciences vol 219 no 2 pp57ndash83 2009
[30] A Saltelli K Chan and E M Scott Sensitivity Analysis JohnWiley amp Sons New York NY USA 1st edition 2000
[31] S Strogatz Nonlinear Dynamics and Chaos Applications toPhysics Biology Chemistry and Engineering With Applicationsto Physics Biology Chemistry and Engineering Westview PressBoulder Colo USA 2000
[32] P Zhou S Cai Z Liu and R Wang ldquoMechanisms generatingbistability and oscillations in microRNA-mediated motifsrdquoPhysical Review E vol 85 no 4 apart 1 Article ID 041916 9pages 2012
[33] S Marino I B Hogue C J Ray and D E Kirschner ldquoAmethodology for performing global uncertainty and sensitivityanalysis in systems biologyrdquo Journal of Theoretical Biology vol254 no 1 pp 178ndash196 2008
[34] L P LimNC Lau PGarrett-Engele et al ldquoMicroarray analysisshows that some microRNAs downregulate large numbers of-target mRNAsrdquo Nature vol 433 no 7027 pp 769ndash773 2005
[35] J Krutzfeldt N Rajewsky R Braich et al ldquoSilencing ofmicroRNAs in vivowith lsquoantagomirsrsquordquoNature vol 438 no 7068pp 685ndash689 2005
[36] MWPfafflGWHorgan andLDempfle ldquoRelative expressionsoftware tool (REST) for group-wise comparison and statisticalanalysis of relative expression results in real-time PCRrdquoNucleicAcids Research vol 30 no 9 p e36 2002
[37] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[38] T Abbas and A Dutta ldquop21 in cancer intricate networks andmultiple activitiesrdquo Nature Reviews Cancer vol 9 no 6 pp400ndash414 2009
[39] J Wang M Lu C Qiu and Q Cui ldquoTransmiR a transcriptionfactor microRNA regulation databaserdquo Nucleic Acids Researchvol 38 supplement 1 pp D119ndashD122 2010
BioMed Research International 15
[40] S Bandyopadhyay andM Bhattacharyya ldquoPuTmiR a databasefor extracting neighboring transcription factors of humanmicroRNAsrdquo BMC Bioinformatics vol 11 article 190 2010
[41] A Le Bechec E Portales-Casamar G Vetter et alldquoMIRNTN a framework integrating transcription factorsmicroRNAs and their targets to identify sub-network motifs ina meta-regulation network modelrdquo BMC Bioinformatics vol12 article 67 2011
[42] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyTheGene Ontology ConsortiumrdquoNature Genetics vol 25 no 1 pp 25ndash29 2000
[43] A I F Vaz and L N Vicente ldquoA particle swarm pattern searchmethod for bound constrained global optimizationrdquo Journal ofGlobal Optimization vol 39 no 2 pp 197ndash219 2007
[44] W H PressNumerical Recipes The Art of Scientific ComputingCambridge University Press Cambridge UK 2007
[45] J G Doench and P A Sharp ldquoSpecificity of microRNA targetselection in translational repressionrdquo Genes and Developmentvol 18 no 5 pp 504ndash511 2004
[46] B Kholodenko M B Yaffe and W Kolch ldquoComputationalapproaches for analyzing information flow in biological net-worksrdquo Science Signaling vol 5 no 220 2012
[47] F M Khan U Schmitz S Nikolov et al ldquoHybrid modeling ofthe crosstalk between signaling and transcriptional networksusing ordinary differential equations and multi-valued logicrdquoBiochimica et Biophysica Acta 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
6 BioMed Research International
browser A list of TFs controlling the expression of themiRNAswas constructed using information of exper-imentally proven TF-miRNA interactions extractedfrom TransmiR (release 10) [39] In addition wegenerated a list of putative TFs of miRNAs withbinding sites in the 10 kb upstream region of themiRNA genes using information from the databasesPuTmiR (release 10) [40] and MIRNTN (version121) [41] and from the table of TFs with conservedbinding sites in theUCSC genome browser (hg18) [9]
(c) Information about protein interactions was extractedfrom theHuman Protein ReferenceDatabase (HPRDrelease 90) [15] and the STRING database (release90) [16] Only the experimentally verified p21-proteininteractions were used to construct the network
(d) Additionally we associated the TFs in the network tonine biological processes based on the Gene Ontol-ogy (GO) [42] The corresponding GO terms werecell proliferation cell apoptosis immune responseinflammatory response cell cycle DNA damage cellsenescence DNA repair and cell migration
Next we visualized the network in CellDesigner andcomputed a confidence score for each interaction in thenetwork (Figure 1) The confidence scores provide us withthe reliability of the interactions considered in p21 regulatorynetwork With the help of these scores we discarded non-reliable interactions and constructed the mechanistic modelaccounting for p21 regulation by its targeting miRNAsBesides the interested experimentalists can further use thisinformation to choose reliable interacting candidates of p21for designing relevant experiments The scores for eachinteraction of p21 regulatory network are shown in Table 2
222 Mechanistic Modeling of p21 Regulation byIts Targeting miRNAs
(1) Model Construction After constructing the regulatorynetwork a detailed mechanistic model of ODEs whichdescribes the biochemical reactions underlying the regula-tion of p21 was established We chose the ODE modelingapproach as it is a simple formalism for describing temporaldynamics of biochemical systems and a wide range of toolsare available to explore their properties Precisely the modelconsidered themRNA(mp21 (3)) andprotein (p21 (6)) of themiRNA target hub p21 the p21-targeting miRNAs (miR
119894 119894 isin
(1 15) (5)) and the complexes formed by p21 mRNAand miRNA [mp21 | miR
119894] (4) Altogether the model is
constituted by 32 state variables and 64 parameters
119889mp21119889119905
= 119896mp21syn sdot 119891act (TFmp21)
minusmp21 sdot (119896mp21deg +sum
119894
119896complex
119894
ass sdotmiR119894)
(3)
119889 [mp21 | miR119894]
119889119905
= 119896complex
119894
ass sdotmp21 sdotmiR119894minus 119896
complex119894
deg sdot [mp21 | miR119894]
(4)
119889miR119894
119889119905= 119896
miR119894syn sdot 119891act (TFmiR119894)
minusmiR119894sdot (119896
miR119894deg + 119896
complex119894
ass sdotmp21) (5)
dp21119889119905
= 119896p21syn sdotmp21 + 119896p21deg sdot p21 (6)
mp21Total = mp21 +sum119894
[mp21 | miR119894] (7)
For mp21 processes considered in the model were (i) itssynthesis (119896mp21
syn ) mediated by TFs (119891act(TFmp21)) (ii) itsdegradation (119896mp21
deg ) and (iii) its association with a miRNA(119896complex
119894
ass ) For each miR119894 processes considered were (i)
the synthesis (119896miR119894syn ) mediated by TFs (119891act(TFmiR119894)) (ii) the
degradation (119896miR119894deg ) and (iii) the association with the p21
mRNA target (119896complex119894
ass ) For each [mp21 | miR119894] complex
processes considered were (i) the formation of the complexby a miR
119894and the p21 mRNA (119896complex
119894
ass ) and (ii) the complexdegradation (119896complex
119894
deg ) For p21 processes considered were(i) its synthesis (119896p21syn ) and (ii) its degradation (119896p21deg) Anadditional algebraic equation accounting for the total mea-surable amount of p21 mRNA (mp21Total) was also includedThe SBML file of the model is available for download athttpwwwsbiuni-rostockdeuploadstx templavoilap21TargetHub 03092013xml
(2) Model Calibration and Validation For model calibrationwe fixed some parameter values using published data andestimated the other unknown parameters using the time-series data published by Wu et al [23] in which the p21mRNA (northern plot) and protein levels (western plot)were measured 48 hr after transfection of individual p21-targeting miRNAs into human embryonic kidney 293 cellsThe unknown parameter values were estimated using aniterative method combining global (particle swarm patternsearch) [43] and local (downhill simplex method in multi-dimensions) [44] optimization algorithms For each miR
119894
considered in the model the method minimizes the distancebetween model simulations and experimental data using thefollowing cost function
119865miR119894cost =
[mp21miR119894sim (119905) minusmp21miR119894
exp (119905)]2
(120575mp21119894
)2
+
[p21miR119894sim (119905) minus p21miR119894
exp (119905)]2
(12057511990121
119894)2
119894 isin [1 15]
(8)
BioMed Research International 7
1 2 4 1 2 4 5 6 1 2 3 7
E2F TP53 RELA1 2 4 7 2 3 7 6
STAT5 SRF1 2 3 4 1 2
JUN STAT1
SP1 RUNX1 TFAP2A
Common TFs Modulation
CDKN1A
1 2 1 2 4 1 4
1 4 1 2 3 7
1 2 5 2 4 5 7 1
VDR
CUX1
E2A
RAR BRCA1
TP63 TP73 TBX2
STAT6 SP3 STAT3
STAT2 RUNX2
EHMT2 p21 TFs
1 2 4 5 8 9
CEBP120573NF120581B1
(1) Cell proliferation (6) Cell senescence
(2) Apoptosis (7) Inflammatory response
(3) Immune response (8) DNA repair
(4) Cell cycle control (9) Cell migration
(5) DNA damage
miRNAmiRNA
Translational repressionmiRNA
Translation
Transcription
Association
miR-132 miR-28-5p
miR-299-5p miR-298
miR-345 miR-125a-5p
miR-208a
miRNA
miRNADeadenylation followed by decay
Protein-protein interaction
miR-93 miR-654-3p
miR-657 miR-572
miR-423-3p miR-363
miR-639 miR-515-3p
p21
CCND2 CDK3 GMNN CCNE1 POLD2 CSNK2B C1orf123 CCDC85B
PIM1 CCNB2 KIAA0432 GADD45B HDAC11 ITGB1BP3 TTLL5 SLC25A11
GADD45G CDC7 BCCIP CDK1 GNB2LI RAB1A psmA3 TK1
CDC45 CCNB1 CCNA2 GADD45A MAPK8 TSG101 SET CIZ1
CDK14 CDK5 MDM2 CDC27 RNF144B ESR1 PRKCH RPA1
CDC20 SKP2 CDK2 CDK4 DHX9 PCNA CDC5L CCDC5
CCND3 CDK6 CCNE2 CCNA1 TEX11 MAP3K5 MCM10 Caspase-3
CDC6 CCND1 XRCC6 PARP1 DAPK3 NR4A1 CEBPA
Interacting proteins
AKT1
miRNA genes
5 8 1 2 1 2 4 3 1 2 3 3 2 2 3 3 1 4 1 3 1 4
1 4
NFATC2 YY1 CREB1 CEBPA ARNT
FOXCI1
ELK1 SOX9 FOS PAX5 MAX MEF2A RFX1 JUNB AKT1 GATA3 MYC NRSFEGR1
E47 PPARG POU3F2 FOXF2FOXL1 MZF1 NFIC SRY
FOXC1
HNFA4 NFYA MYF1 BACF1 ATF6 ZID RORA POU2F1
miRNA TFs
11 2 34
12 31 5 11
HNF1A
1 2
RB1
Cellular processes
Figure 1 p21 regulatory network The network contains several layers of regulators of p21 TFs (light blue and red boxes) miRNAs (darkblue and green boxes) and proteins (grey boxes) In each big box there are small boxes which represent individual components of this layerof regulation miRNAs are classified into two groups according to the mechanisms by which the expression of p21 is repressed One groupcauses p21 translation repression (green box) These miRNAs bind to p21 mRNA resulting in the repressed translation of p21 but unchangedmRNA expression level The other group of miRNAs (dark blue box) decreases the stability of p21 mRNA by modifying its structure leadingto mRNA decay and finally the downregulation of p21 TFs are classified into three groups p21 TFs (red) miRNA TFs (yellow) and theircommon TFs (light blue) The p21 interacting-proteins are framed in the grey boxes The purple boxes represent nine processes and the TFsassociated with these processes are indicated in the ellipses above them using corresponding figures This data is adapted from our previouspublication [4]
where mp21miR119894sim (119905) and mp21miR119894
exp (119905) represent the simulatedp21 mRNA and protein expression levels for each miR
119894at
time point 119905 p21miR119894sim (119905) and p21miR119894
exp (119905) represent the mea-sured value for each miR
119894at time point 119905 and their standard
deviations are 120575mp21119894
and 120575p21119894
Here 119905 is the time point (48 hr)after overexpression of the individual miRNAs in embryonickidney 293 cells at which the expression levels of the p21 andits mRNA were measured [23] The model calibration resultsare shown in Figure 2(a) and the obtained parameter valuesare listed in Table 3
Experimental results showed that a stronger repressionof the target gene can occur when two miRNA binding siteson the target mRNA are in close proximity [24 45] To testthe consequences of this hypothesis we predicted cooperativemiRNA pairs for p21 with seed site distances between 13ndash35 nt in the p21 31015840 UTR To substantiate the cooperativeeffect associated with pairs of miRNAs we introduced agroup of new state variables ([mp21 | miR
119894| miR
119894]) into
the original model These state variables account for the
ternary complexes composed of p21 mRNA and two puta-tively cooperating miRNAs (miR
119894and miR
119895) For these new
variables processes considered are (i) the association of p21mRNA with miR
119894and miR
119895into a complex (119896co-complex
119894119895
ass )and (ii) the degradation of the complex (119896co-complex
119894119895
deg ) Afterexpansion the corresponding modified and new ODEs arelisted below
dmp21119889119905
= 119896mp21syn sdot 119891act (TFmp21)
minusmp21 sdot (119896mp21deg +sum
119894
119896complex
119894
ass sdotmiR119894
+sum
119894119895
119896co-complex
119894119895
ass sdotmiR119894sdotmiR119895)
(9)
8 BioMed Research InternationalTa
ble2Th
econ
fidence
scores
forthe
interactions
inp21regulatorynetworkTh
enetworkconsistso
ffou
rtypes
ofinteractionsm
iRNA-
p21TF
-miRNAT
F-p21andp21-p
rotein
(a)miRNA-
p21interactio
nscores
miRNA
miR-125a-5p
miR-132
miR-208a
miR-28-5p
miR-298
miR-299-5p
miR-345
miR-363
Score
073
073
073
080
073
080
073
073
miRNA
miR-423-3p
miR-515-3p
miR-572
miR-639
miR-654-3p
miR-657
miR-93
Score
085
073
073
073
073
080
095
(b)TF
-miRNAinteractionscores
Score
Score
Score
Score
Score
Score
miR-208
miR-132
miR-657
miR-125a-5p
miR-28-5p
miR-93
MAX
027
EGR1
093
TFAP2
A034
EGR1
086
MZF
1028
TFAP2
A034
EGR1
029
CREB
1086
SP1
034
PAX5
024
E47
027
SOX9
024
ARN
T027
ARE
B6027
MZF
1034
NRS
F024
FOXC
1031
GAT
A3
027
RFX1
024
E47
024
MAX
029
ZID
024
FOXI1
024
MZF
1031
ATF6
024
ELK1
027
USF1
027
PPARG
027
SRY
027
SP1
034
YY1
027
EGR2
029
EGR1
034
GAT
A3
034
JUN
029
POU2F1
024
JUNB
024
FOS
024
RELA
029
MZF
1034
POU2F1
024
EGR1
029
POU2F1
024
NFIC
024
miR-654-3p
TP53
034
SRF
027
CEBP
A024
NFIC
027
RELA
027
YY1
027
NFIC
034
CEBP
A029
NFY
A024
miR-345
miR-423-3p
BACH
1024
miR-299-5p
RORA
024
RUNX1
024
SP1
034
BACH
1027
FOXL
1029
CREB
1024
miR-363
STAT
1024
RELA
027
STAT
1027
SRY
024
SRY
027
E2F1
062
E2F1
085
NFA
TC2
024
POU3F2
024
NFA
TC2
024
SRF
029
FOXC
1033
MYC
084
HNF4
A031
STAT
5A027
FOS
024
RUNX1
027
MZF
1034
miR-298
NFY
A024
SRF
027
POU2F1
027
miR-572
miR-639
JUNB
031
POU2F1
024
MEF
2A027
NF120581
B1024
FOXF
2024
NFY
A024
NFIC
031
POU2F1
024
HNF1A
024
(c)TF
-p21
interactionscores
TF(verified)
SP1
SP3
E2F1
Runx
1Ru
nx2
STAT
1E2
AST
AT5
TP53
TP63
TP73
Score
100
100
077
077
077
072
072
067
067
067
067
TF(verified)
STAT
3CU
X1TF
AP2
ABR
CA1
VDR
RARA
CEB
P120572CEB
P120573RB
1Tb
x2EH
MT2
Score
066
066
060
060
060
060
060
060
060
060
060
TF(putative)
NF120581
B1RE
LAST
AT2
STAT
6SR
FScore
044
044
044
044
043
(d)p21-p
rotein
interactionscores
Protein
TP53
PCNA
CASP
3CC
NA1
CCND1
SKP2
BCCI
PCC
NA2
CCND2
CCNE2
AKT
1C1
orf123
CCDC8
5BCC
NB1
CCNB2
CCND3
Score
100
092
087
087
087
087
081
081
081
081
072
072
072
072
072
072
ProteinCC
NE1
CDC4
5CD
C5L
CDC6
CDC7
CDK1
CDK14
CDK2
CDK3
CDK4
CDK6
CEBP
ACI
Z1CS
NK2
A1
CSNK2
BDAPK
3Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
ESR1
GADD45A
GADD45B
GADD45G
GMNN
GNB2
L1HDAC
11ITGB1BP
3MAP3
K5MAPK
8MCM
10PA
RP1
PIM1
POLD
2PS
MA3
RAB1A
Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
SET
SLC2
5A11
STAT
3TE
X11
TK1
TSG101
TTLL
5XR
CC6
CDC2
0CD
C27
CDK5
DHX9
MDM2
NR4
A1
PRKC
HRP
A1
Score
072
072
072
072
072
072
072
072
056
056
056
056
056
056
056
056
Verifi
edexp
erim
entally
verifi
edinteraction
putativ
epredictedinteraction
BioMed Research International 9
Con
trol
miR
-298
miR
-208
am
iR-1
32m
iR-2
8-5p
miR
-125
-5p
miR
-299
-5p
miR
-345
miR
-93
miR
-423
-3p
miR
-515
-3p
miR
-363
miR
-657
miR
-639
miR
-572
miR
-654
-3p
0
05
1
15m
RNA
expr
essio
n (a
u)
Data Model Data ModelC
ontro
lm
iR-2
98m
iR-2
08a
miR
-132
miR
-28-
5pm
iR-1
25-5
pm
iR-2
99-5
pm
iR-3
45m
iR-9
3m
iR-4
23-3
pm
iR-5
15-3
pm
iR-3
63m
iR-6
57m
iR-6
39m
iR-5
72m
iR-6
54-3
p
0
02
04
06
08
1
Prot
ein
expr
essio
n (a
u)
(a)
DoxorubicinCellWell
Control miR-572 miR-93
Measure p21 expression at the time points of
miR-572+ miR-93
0 2 4 6 8 24hr
miR-572miR-93
Immunoblotting
(b)
0 4 8 12 16 20 240
4
8
12
16
20
Time (hr)
p53
Actin
Doxo (hrs)
fac
t(TF
mp2
1)
C 0 1 2 4 6 8 24
1 08 12 19 29 32 3 16
(c)
0 2 4 6 8 2402468
1012
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)
p21
expr
essio
n (a
u)
02468
1012
miR-572Control
02468
1012
miR-93
02468
1012
Model Data
0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24
1 23 46 107124 124 23 28 48 66 53 31 08 16 21 38 50 18 06 06 12 25 27 12
miR-572+ 93Doxo (hrs)
p21
Actin
p21
Actin
p21
Actin
p21
Actin
(d)
Figure 2 Model calibration and validation (a) Model calibration The figures show the relative change of the p21 mRNA and proteinexpression levels after overexpression of the indicated miRNAs (Model model simulation Data experimental data) These data werenormalized to the control group in which the p21 mRNA and protein expression levels were measured when the miRNAs were normallyexpressed (au arbitrary unit) (b) Experimental workflow In the experiments Sk-Mel-147 cells were seeded in six well plates Then maturemiRNA mimics were transfected individually at a concentration of 100 nM (miR-572 and miR-93) or in combination at 50 nM each (miR-572 + miR-93) After 48 hr transfection with miRNA mimics the cells were pulse treated with 250 nM doxorubicin for 1 hour after whichnormal growth medium was replenished The immunblotting were performed to measure p21 expression at 0 2 4 6 8 and 24 hr post-doxorubicin treatment (c) Temporal dynamics of p21 transcriptional function Afterdoxorubicin treatment the expression of p53 a TF ofp21 was measured using immunblotting and these data were used to characterize the transcriptional function of p21 using MATLAB linearinterpolation function (d) Model validation We measured the expression of p21 protein in response to genotoxic stress in the four scenariosas described in the main text The measured data (Data) were compared with the model simulations (Model) The figures (a) (c) and (d) areadapted from our previous publication [4]
119889 [mp21 | miR119894| miR
119895]
119889119905
= 119896co-complex
119894119895
ass sdotmp21 sdotmiR119894sdotmiR119895
minus 119896co-complex
119894119895
deg sdot [mp21 | miR119894| miR
119895]
(10)
To model stronger repression of the target gene bycooperating miRNAs we assumed a stronger association rateconstant for the complex [mp21 | miR
119894| miR
119895] which is
equal to the sum of their individual association rate constants
(119896co-complex119894119895
ass = 119896complex
119894
ass + 119896complex
119895
ass ) Similarly the degradationrates of the complexes [mp21 | miR
119894| miR
119895] were assumed
to be equal to the sum of degradation rate constants ofsingle miRNA binding complexes (119896co-complex
119894119895
deg = 119896complex
119894
deg +
119896complex
119895
deg ) However it has to be noted that these addedequations are an abstract description ofmiRNAcooperativitybecause the details of this mechanism are not yet known
To experimentally validate the capability of our model topredict the relative p21 concentrations regulated by coopera-tivemiRNAs we selectedmiR-572 andmiR-93 as a case study
10 BioMed Research International
Table 3 Initial concentrations of model variables and model parameter values Based on the experimental data the p21-targeting miRNAsverified byWu et al [23] were divided into two groups the translation repression group (marked with asterisk) and the mRNA deadenylationgroup AmiRNAwas classified into themRNA deadenylation group if its overexpression can result in 20 ormore downregulation of the p21mRNA level (ie p21mRNA levelle 08 the basal level is 1) otherwise it was classified into the translation repression group For the translationrepression group only 119896complex119894
ass was estimated and 119896complex119894deg was fixed For the other group both 119896complex119894
deg and 119896complex119894ass were estimatedThe initial
concentrations of p21 and mp21 were set to 1 and this value was used as their basal expression levels During the parameter estimation theinitial concentrations of p21-targeting miRNAs were set to 100 because in the publication [23] the expression levels of p21 and mp21 weremeasured after the individual introduction of the p21-targeting miRNAs with amount of 100 nM Due to the lack of biological information tocharacterize the transcriptional activation function (119891act) of p21 and its targeting miRNAs the corresponding functions were assumed to be1 for simplicity The data is adapted from our previous publication [4]
Initial concentration of variables and TF functionsVariable Description Initial concentration (au)p21 p21 protein 1
mp21 p21 mRNA 1
miR119894=115
p21-targeting miRNAs 100
[mp21 | miR119894=115
] Complexes formed by miR119894and mp21 0
119891act(TFmp21) p21rsquos transcriptional activation function 1119891act(TFmiR119894 (119894=115)) The transcriptional activation function of miR
1198941
Fixed parameter valuesParameter Description Value (hrminus1) Reference119896mp21syn Synthesis rate constant of mp21 01155 fixed119896mp21deg Degradation rate constant of mp21 01155 [24]119896miR119894syn (119894 = 1 15) Synthesis rate constant of miR
11989400289 fixed
119896miR119894deg (119894 = 1 15) Degradation rate constant of miR
11989400289 [25]
119896p21syn Synthesis rate constant of p21 13863 fixed119896p21deg Degradation rate constant of p21 13863 [26]
Estimated parameter values
miRNA (statevariable)
119896complex119894deg
(119894 = 1 15) (hrminus1)
119896complex119894ass
(119894 = 1 15)
(au minus1sdot hrminus1)
119865miR119894cost
(119894 = 1 15)Experimental data of p21 (protein mRNA plusmn SD)
miR-298 (miR1)lowast 01155 00254 34119890 minus 004 (016 1074 plusmn 0025)
miR-208a (miR2)lowast 01155 00041 20119890 minus 003 (051 1192 plusmn 0022)
miR-132 (miR3)lowast 01155 00275 24119890 minus 003 (015 121 plusmn 0147)
miR-28-5p (miR4)lowast 01155 00119 59119890 minus 003 (028 135 plusmn 006)
miR-125-5p (miR5)lowast 01155 00018 18119890 minus 003 (069 085 plusmn 0051)
miR-299-5p (miR6)lowast 01155 00080 18119890 minus 004 (036 095 plusmn 0038)
miR-345 (miR7)lowast 01155 00051 11119890 minus 004 (046 096 plusmn 0039)
miR-93 (miR8) 01564 00235 41119890 minus 014 (017 07776 plusmn 003)
miR-423-3p (miR9) 09118 00055 28119890 minus 009 (044 05102 plusmn 011)
miR-515-3p (miR10) 02098 00253 12119890 minus 013 (016 0616 plusmn 0037)
miR-363 (miR11) 02261 00399 22119890 minus 014 (011 056 plusmn 015)
mR-657 (miR12) 03465 00158 21119890 minus 014 (023 048 plusmn 012)
miR-639 (miR13) 04305 00327 18119890 minus 017 (013 036 plusmn 0084)
miR-572 (miR14) 03039 00360 94119890 minus 023 (012 045 plusmn 0044)
miR-654-3p (miR15) 97485 00024 30119890 minus 014 (064 063 plusmn 0053)
These two miRNAs were chosen because their predictedtarget sites in the p21 31015840 UTR are in close proximity to eachother and thereby they can induce cooperative repression onp21 as suggested in [24] The experiments were performed asfollows
(i) Melanoma cells (Sk-Mel-147) were transfected withthe mature miRNAmimics of the twomiRNAs eitherindividually (100 nM) or in combination (50 nMeach) whereas untreated cells were used as control(Figure 2(b))
BioMed Research International 11
p21
FOXF2
miR-572
Activation
Inhibition120572
Incoherent FFL + cascaded regulation
AF2
miR-93
(a)
miR-93
miR-572
TP53
MYC
p21
FOXF2
Coherent FFL + cascaded regulation
(b)
Saturation
Pulse
p21
(au
)
p21
(au
)Time (hr) Time (hr)
Saturation Pulse
Pulse
Saturation
Saturation
di
(au
)
102
101
102
100
101
10minus1
10010minus2
10minus1
102
101
100
10minus1
10minus2
FOXF2 (au) FOXF2 (au)
TP53
(au
)
1
09
08
07
06
05
04
03
02
01
10minus2 10210110010minus110minus20
AF2120572
(au
)
(c)
Figure 3 Different p21 dynamics in different network motifs We ran simulations to show the different dynamics of p21 for two differentnetwork motifs (a) and (b) Through simulations two dynamical patterns of p21 were identified saturation and pulse ((c) top) For eachnetwork motif the corresponding distributions of the two dynamical patterns were plotted ((c) bottom) For different combinations ofthe transcriptional strengths the normalized distance (119889
119894) between peaks (119901
119894) and steady states (ss
119894) of p21 is determined by the equation
119889119894= (119901119894minus ss119894)119901max 119901max = max(119901
1 119901
119899) If 119889
119894= 0 for the corresponding combination of transcriptional strengths the p21 dynamics is
saturation otherwise it is pulse The regions showing different dynamical patterns of p21 are separated using the white lines
(ii) Next the cells were treated with doxorubicina genotoxic-stress inducing agent The agent canupregulate the expression of p53 which is a knownTFof p21 and therefore it can result in the upregulationof p21 (Figure 2(c))
(iii) After doxorubicin treatment the expression levels ofp21 were measured by immunoblotting at differenttime points (0 2 4 6 8 24 hr) The p21 expressionvalues were normalized based on the p21 expression
level in the control group measured at time point 0 hr(Figure 2(d))
By doing so we obtained the p21 response after genotoxicstress in four scenarios (1) endogenous miRNA expression(Control) (2) overexpression of miR-572 (3) overexpressionof miR-93 and (4) both miRNAs moderately overexpressedThereafter we derived a model of seven ODEs based onthe original equations which was configured according tothe designed experiments making the simulation results
12 BioMed Research International
00001
0001
001
01
1Cell cycle Cell growth DNA damage Senescence Apoptosis
G1G2
MS
p21
expr
essio
n (a
u)
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
log10
wow
Figure 4 p21 expression regulated by cooperative miRNAs for different cellular processesThe associations of themiRNAs with these cellularprocesses were derived from GO terms of their TFs A miRNA was supposed to be expressed (in bold black font) in a cellular process onlyif its TF is related to the corresponding GO term of this process The p21 expression levels are computed for each process with (w)without(wo) considering the cooperative effect among the p21 targeting miRNAs
comparable with the experimental data As shown in theFigure 2(c) the simulations are in good agreement with theexperimental observations The individual overexpression ofmiR-572 or miR-93 led to the reduction of the upregulationof p21 response after genotoxic stress induction The twomiRNAs cause different degrees of repression due to theirdifferent repression efficiencies on p21 Interestingly the com-bined overexpression of both miRNAs induced the strongestdownregulation of p21 and therefore verifying the hypothesisof their cooperative regulation of p21 Above all the resultsnot only validated themodel but also demonstrated the abilityof our method to identify cooperative miRNA pairs for p21
(3) Predictive Simulations As there are abundant of networkmotifs such as FFLs in p21 regulatory network and thesenetwork motifs are important for determining p21 dynamicsit is interesting to investigate the dynamics of p21 in networkmodules where FFLs are involved To do so two networkmodules including both miR-93 and miR-572 and theirTFs were exemplified In Figure 3(a) the network modulecontains an incoherent FFL composed by AF2120572 miR-93 andp21 and a cascaded regulation in which p21 is repressed
by FOXF2 via miR-572 in Figure 3(b) the same cascadedregulation together with a coherent FFL composed of TP53MYC miR-93 and p21 forms another regulatory moduleof p21 By modulating the transcriptional strengths of thetwo miRNAs by their TFs two types of p21 dynamics wereidentified saturation and pulse (Figure 3(c) top) In theformer the p21 expression increases and reaches its steadystate at the highest level in contrast in the latter the p21expression increases to a peak and thereafter drops to asteady state at lower level For the two different networkmodules various combinations of transcriptional strengthsof the two miRNAs lead to different distributions of thetwo p21 dynamical patterns For the network module of anincoherent FFL plus the cascaded regulation (Figure 3(c)bottom left) the saturation pattern appears in two distinctregions one with weakened transcriptional strength of miR-93 and the other with enhanced transcriptional strength ofmiR-93 for the other module (Figure 3(c) bottom right)the saturation pattern only appears in the region in whichthe transcriptional strength of miR-93 is enhanced Takentogether the results showed that for the two different networkmotifs the dynamical pattern of p21 is changing according to
BioMed Research International 13
different combinations of its upstream regulators suggestingthe adaptation of p21 dynamics for different biological con-texts
Furthermore we performed a number of simulations toshow the influence of miRNA regulation on p21 expressionlevels in different cellular processes arranged in a consecutivemanner In this procedure a cell is first in the process ofcell cycle followed by proliferation (cell growth) then thecell responses to DNA damage and enters into the processof senescence and finally apoptosis is initiated As shown inFigure 4 during the cell cycle process the p21 expression islow due to the activation of most of its targeting miRNAswhen the cell starts proliferating the p21 expression declinesto an even lower level because of more activated miRNAsin this process after responding to DNA damage the p21expression soars to a high level which is caused by theactivation of its TFs like p53 and fewer expressed miRNAsalthough the p21 expression keeps at a similar level whilethe cell is undergoing senescence the expressed miRNAs aredifferent from the previous process finally the p21 expressiondecreases again to a low level due to the reemergence ofmost of its targeting miRNAs and the cell enters apoptosisInterestingly the model simulations are also consistent withexperimental observations under non-stressed condition thelow expression level of p21 is needed for cell proliferation theupregulation of p21 happens after response to DNA damagevia p53 and the increased p21 expression further results incell cycle arrest leading to senescence and apoptosis [25]Besides when considering the effect of cooperatingmiRNAsthe p21 expression levels were indistinguishable from theprevious simulation of DNA damage response However forthe other processes p21 expression levels were significantlylower compared to the simulations without considering thecooperative effect of the miRNAs Above all these resultsindicated that selective expression of cooperative miRNAscould be adopted by cells to ensure diverse expression levelsof p21 tomeet the requirements of different cellular processes
3 Conclusions and Discussion
In this paper we presented a systems biology approach com-bining data-drivenmodeling andmodel-driven experimentsto investigate the role of miRNA-mediated repression in generegulatory networksThis approach provides a systematicwayto gain a deeper understanding of the regulation of targetgenes by mutiple and cooperative miRNAs Using the regu-lation of p21 by multiple miRNAs as a case study we showedhow the ODE-basedmodel which is calibrated and validatedby means of experimental data is suitable for predicting thetemporal dynamics of molecular concentrations involved inbiochemical systems
Provided there are sufficiently rich quantitative data setsavaliable to characterize the model the use of the methodol-ogy here shown can be extended to more complex regulatorynetworks involving multiple targets cooperating TFs andmiRNAs and signaling pathways displaying cross-talk viapost-tranlational modifications In this case the criticalelement is the quality and quantitiy of the available data
Insufficiency and low quality of experimental data can causeerrors in the process of model construction and overfitting inparameter estimation can lead to uncertainties in the modelpredictions We believe that the quick development of quan-titative high throughput techniques such as transcriptomicsproteomics and miRNomics will facilitate the constructionand characterization of larger miRNA-mediated regulatorynetworks
Other modeling frameworks than ODE-based modelscan be used to describe biological systems such as prob-abilistic (eg Bayesian) or logical (eg Boolean) modelsImportantly different modeling frameworks have differentproperties and perform well regarding different perspectivesand levels of mechanistic details of biochemical systems [46]For example Bayesian models are helpful in the constructionof connections in signaling networks and can reveal the mostlikely underlying structure of the network in a probabilisticmanner Boolean models use binary values (0 and 1) andlogical gates (AND NOT and OR) to describe activities ofnetwork components and the information flow among themWe believe that in the coming future hybrid models whichconsist of modeling framework and experimental techniquespecific sub-modules will provide the necessary compromisebetween quantitativequalitative accuracy and scalability forthe investigation of large biochemical networks [47]
Disclosures
The authors declare that they have no competing financialinterests
Authorsrsquo Contribution
Julio Vera and Olaf Wolkenhauer are equal contributers
Acknowledgments
The authors would like to acknowledge the funding fromthe German Federal Ministry of Education and Research(BMBF) eBio-miRSys (0316175A to Xin Lai and Julio Vera)eBio-SysMet (0316171 to OlafWolkenhauer) and Gerontosys-ROSAge (0315892A to Olaf Wolkenhauer)
References
[1] A E Pasquinelli ldquoMicroRNAs and their targets recognitionregulation and an emerging reciprocal relationshiprdquo NatureReviews Genetics vol 13 no 4 pp 271ndash282 2012
[2] J Vera X Lai U Schmitz and O Wolkenhauer ldquoMicroRNA-regulated networks the perfect storm for classical molecularbiology the ideal scenario for systems biologyrdquo Advances inExperimental Medicine and Biology vol 774 pp 55ndash76 2013
[3] S Nikolov J Vera U Schmitz and O Wolkenhauer ldquoA model-based strategy to investigate the role of microRNA regulation incancer signalling networksrdquo Theory in Biosciences vol 130 no1 pp 55ndash69 2011
14 BioMed Research International
[4] X Lai U Schmitz S K Gupta et al ldquoComputational analysisof target hub gene repression regulated by multiple and cooper-ative miRNAsrdquoNucleic Acids Research vol 40 no 18 pp 8818ndash8834 2012
[5] X Lai O Wolkenhauer and J Vera ldquoModeling miRNA regu-lation in cancer signaling systems miR-34a regulation of thep53Sirt1 signaling modulerdquoMethods in Molecular Biology vol880 pp 87ndash108 2012
[6] C Jiang Z Xuan F Zhao and M Q Zhang ldquoTRED atranscriptional regulatory element database new entries andother developmentrdquoNucleic Acids Research vol 35 supplement1 pp D137ndashD140 2007
[7] L A Bovolenta M L Acencio and N Lemke ldquoHTRIdban open-access database for experimentally verified humantranscriptional regulation interactionsrdquo BMCGenomics vol 13article 405 2012
[8] VMatys O V Kel-Margoulis E Fricke et al ldquoTRANSFAC andits module TRANSCompel transcriptional gene regulation ineukaryotesrdquo Nucleic Acids Research vol 34 database issue ppD108ndashD110 2006
[9] D Karolchik R Baertsch M Diekhans et al ldquoThe UCSCgenome browser databaserdquo Nucleic Acids Research vol 31 no1 pp 51ndash54 2003
[10] P Alexiou T Vergoulis M Gleditzsch et al ldquomiRGen 20 adatabase of microRNA genomic information and regulationrdquoNucleic Acids Research vol 38 supplement 1 pp D137ndashD1412010
[11] F Xiao Z Zuo G Cai S Kang X Gao and T Li ldquomiRecordsan integrated resource for microRNA-target interactionsrdquoNucleic Acids Research vol 37 supplement 1 pp D105ndashD1102009
[12] S-D Hsu F-M Lin W-Y Wu et al ldquomiRTarBase a databasecurates experimentally validated microRNA-target interac-tionsrdquo Nucleic Acids Research vol 39 supplement 1 pp D163ndashD169 2011
[13] P Sethupathy B Corda and A G Hatzigeorgiou ldquoTarBase acomprehensive database of experimentally supported animalmicroRNA targetsrdquo RNA vol 12 no 2 pp 192ndash197 2006
[14] H Dweep C Sticht P Pandey and N Gretz ldquomiRWalkmdashdatabase prediction of possible miRNA binding sites byldquowalkingrdquo the genes of three genomesrdquo Journal of BiomedicalInformatics vol 44 no 5 pp 839ndash847 2011
[15] T S Keshava Prasad RGoel K Kandasamy et al ldquoHumanpro-tein reference databasemdash2009 updaterdquo Nucleic Acids Researchvol 37 supplement 1 pp D767ndashD772 2009
[16] D Szklarczyk A Franceschini M Kuhn et al ldquoThe STRINGdatabase in 2011 functional interaction networks of proteinsglobally integrated and scoredrdquo Nucleic Acids Research vol 39supplement 1 pp D561ndashD568 2011
[17] D Croft G OrsquoKelly G Wu et al ldquoReactome a database ofreactions pathways and biological processesrdquo Nucleic AcidsResearch vol 39 supplement 1 pp D691ndashD697 2011
[18] N le Novere M Hucka H Mi et al ldquoThe systems biologygraphical notationrdquoNature Biotechnology vol 27 no 8 pp 735ndash741 2009
[19] A Funahashi Y Matsuoka A Jouraku M Morohashi NKikuchi and H Kitano ldquoCellDesigner 35 a versatile modelingtool for biochemical networksrdquo Proceedings of the IEEE vol 96no 8 pp 1254ndash1265 2008
[20] M S Cline M Smoot E Cerami et al ldquoIntegration ofbiological networks and gene expression data using CytoscaperdquoNature Protocols vol 2 no 10 pp 2366ndash2382 2007
[21] D-H Le andY-KKwon ldquoNetDS aCytoscape plugin to analyzethe robustness of dynamics and feedforwardfeedback loopstructures of biological networksrdquo Bioinformatics vol 27 no 19pp 2767ndash2768 2011
[22] E O Voit Computational Analysis of Biochemical SystemsA Practical Guide for Biochemists and Molecular BiologistsCambridge University Press Cambridge UK 2000
[23] S Wu S Huang J Ding et al ldquoMultiple microRNAs modulatep21Cip1Waf1 expression by directly targeting its 31015840 untranslatedregionrdquo Oncogene vol 29 no 15 pp 2302ndash2308 2010
[24] P Saeligtrom B S E Heale O Snoslashve Jr L Aagaard J Alluin andJ J Rossi ldquoDistance constraints between microRNA target sitesdictate efficacy and cooperativityrdquo Nucleic Acids Research vol35 no 7 pp 2333ndash2342 2007
[25] Y-S Jung YQian andXChen ldquoExamination of the expandingpathways for the regulation of p21 expression and activityrdquoCellular Signalling vol 22 no 7 pp 1003ndash1012 2010
[26] W Wang H Furneaux H Cheng et al ldquoHuR regulatesp21 mRNA stabilization by UV lightrdquo Molecular and CellularBiology vol 20 no 3 pp 760ndash769 2000
[27] U Wittig R Kania M Golebiewski et al ldquoSABIO-RKmdashdatabase for biochemical reaction kineticsrdquo Nucleic AcidsResearch vol 40 database issue pp D790ndashD796 2012
[28] R Milo P Jorgensen U Moran G Weber and M SpringerldquoBioNumbersmdashthe database of key numbers in molecular andcell biologyrdquo Nucleic Acids Research vol 38 supplement 1 ppD750ndashD753 2010
[29] I-C Chou and E O Voit ldquoRecent developments in parameterestimation and structure identification of biochemical andgenomic systemsrdquoMathematical Biosciences vol 219 no 2 pp57ndash83 2009
[30] A Saltelli K Chan and E M Scott Sensitivity Analysis JohnWiley amp Sons New York NY USA 1st edition 2000
[31] S Strogatz Nonlinear Dynamics and Chaos Applications toPhysics Biology Chemistry and Engineering With Applicationsto Physics Biology Chemistry and Engineering Westview PressBoulder Colo USA 2000
[32] P Zhou S Cai Z Liu and R Wang ldquoMechanisms generatingbistability and oscillations in microRNA-mediated motifsrdquoPhysical Review E vol 85 no 4 apart 1 Article ID 041916 9pages 2012
[33] S Marino I B Hogue C J Ray and D E Kirschner ldquoAmethodology for performing global uncertainty and sensitivityanalysis in systems biologyrdquo Journal of Theoretical Biology vol254 no 1 pp 178ndash196 2008
[34] L P LimNC Lau PGarrett-Engele et al ldquoMicroarray analysisshows that some microRNAs downregulate large numbers of-target mRNAsrdquo Nature vol 433 no 7027 pp 769ndash773 2005
[35] J Krutzfeldt N Rajewsky R Braich et al ldquoSilencing ofmicroRNAs in vivowith lsquoantagomirsrsquordquoNature vol 438 no 7068pp 685ndash689 2005
[36] MWPfafflGWHorgan andLDempfle ldquoRelative expressionsoftware tool (REST) for group-wise comparison and statisticalanalysis of relative expression results in real-time PCRrdquoNucleicAcids Research vol 30 no 9 p e36 2002
[37] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[38] T Abbas and A Dutta ldquop21 in cancer intricate networks andmultiple activitiesrdquo Nature Reviews Cancer vol 9 no 6 pp400ndash414 2009
[39] J Wang M Lu C Qiu and Q Cui ldquoTransmiR a transcriptionfactor microRNA regulation databaserdquo Nucleic Acids Researchvol 38 supplement 1 pp D119ndashD122 2010
BioMed Research International 15
[40] S Bandyopadhyay andM Bhattacharyya ldquoPuTmiR a databasefor extracting neighboring transcription factors of humanmicroRNAsrdquo BMC Bioinformatics vol 11 article 190 2010
[41] A Le Bechec E Portales-Casamar G Vetter et alldquoMIRNTN a framework integrating transcription factorsmicroRNAs and their targets to identify sub-network motifs ina meta-regulation network modelrdquo BMC Bioinformatics vol12 article 67 2011
[42] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyTheGene Ontology ConsortiumrdquoNature Genetics vol 25 no 1 pp 25ndash29 2000
[43] A I F Vaz and L N Vicente ldquoA particle swarm pattern searchmethod for bound constrained global optimizationrdquo Journal ofGlobal Optimization vol 39 no 2 pp 197ndash219 2007
[44] W H PressNumerical Recipes The Art of Scientific ComputingCambridge University Press Cambridge UK 2007
[45] J G Doench and P A Sharp ldquoSpecificity of microRNA targetselection in translational repressionrdquo Genes and Developmentvol 18 no 5 pp 504ndash511 2004
[46] B Kholodenko M B Yaffe and W Kolch ldquoComputationalapproaches for analyzing information flow in biological net-worksrdquo Science Signaling vol 5 no 220 2012
[47] F M Khan U Schmitz S Nikolov et al ldquoHybrid modeling ofthe crosstalk between signaling and transcriptional networksusing ordinary differential equations and multi-valued logicrdquoBiochimica et Biophysica Acta 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
BioMed Research International 7
1 2 4 1 2 4 5 6 1 2 3 7
E2F TP53 RELA1 2 4 7 2 3 7 6
STAT5 SRF1 2 3 4 1 2
JUN STAT1
SP1 RUNX1 TFAP2A
Common TFs Modulation
CDKN1A
1 2 1 2 4 1 4
1 4 1 2 3 7
1 2 5 2 4 5 7 1
VDR
CUX1
E2A
RAR BRCA1
TP63 TP73 TBX2
STAT6 SP3 STAT3
STAT2 RUNX2
EHMT2 p21 TFs
1 2 4 5 8 9
CEBP120573NF120581B1
(1) Cell proliferation (6) Cell senescence
(2) Apoptosis (7) Inflammatory response
(3) Immune response (8) DNA repair
(4) Cell cycle control (9) Cell migration
(5) DNA damage
miRNAmiRNA
Translational repressionmiRNA
Translation
Transcription
Association
miR-132 miR-28-5p
miR-299-5p miR-298
miR-345 miR-125a-5p
miR-208a
miRNA
miRNADeadenylation followed by decay
Protein-protein interaction
miR-93 miR-654-3p
miR-657 miR-572
miR-423-3p miR-363
miR-639 miR-515-3p
p21
CCND2 CDK3 GMNN CCNE1 POLD2 CSNK2B C1orf123 CCDC85B
PIM1 CCNB2 KIAA0432 GADD45B HDAC11 ITGB1BP3 TTLL5 SLC25A11
GADD45G CDC7 BCCIP CDK1 GNB2LI RAB1A psmA3 TK1
CDC45 CCNB1 CCNA2 GADD45A MAPK8 TSG101 SET CIZ1
CDK14 CDK5 MDM2 CDC27 RNF144B ESR1 PRKCH RPA1
CDC20 SKP2 CDK2 CDK4 DHX9 PCNA CDC5L CCDC5
CCND3 CDK6 CCNE2 CCNA1 TEX11 MAP3K5 MCM10 Caspase-3
CDC6 CCND1 XRCC6 PARP1 DAPK3 NR4A1 CEBPA
Interacting proteins
AKT1
miRNA genes
5 8 1 2 1 2 4 3 1 2 3 3 2 2 3 3 1 4 1 3 1 4
1 4
NFATC2 YY1 CREB1 CEBPA ARNT
FOXCI1
ELK1 SOX9 FOS PAX5 MAX MEF2A RFX1 JUNB AKT1 GATA3 MYC NRSFEGR1
E47 PPARG POU3F2 FOXF2FOXL1 MZF1 NFIC SRY
FOXC1
HNFA4 NFYA MYF1 BACF1 ATF6 ZID RORA POU2F1
miRNA TFs
11 2 34
12 31 5 11
HNF1A
1 2
RB1
Cellular processes
Figure 1 p21 regulatory network The network contains several layers of regulators of p21 TFs (light blue and red boxes) miRNAs (darkblue and green boxes) and proteins (grey boxes) In each big box there are small boxes which represent individual components of this layerof regulation miRNAs are classified into two groups according to the mechanisms by which the expression of p21 is repressed One groupcauses p21 translation repression (green box) These miRNAs bind to p21 mRNA resulting in the repressed translation of p21 but unchangedmRNA expression level The other group of miRNAs (dark blue box) decreases the stability of p21 mRNA by modifying its structure leadingto mRNA decay and finally the downregulation of p21 TFs are classified into three groups p21 TFs (red) miRNA TFs (yellow) and theircommon TFs (light blue) The p21 interacting-proteins are framed in the grey boxes The purple boxes represent nine processes and the TFsassociated with these processes are indicated in the ellipses above them using corresponding figures This data is adapted from our previouspublication [4]
where mp21miR119894sim (119905) and mp21miR119894
exp (119905) represent the simulatedp21 mRNA and protein expression levels for each miR
119894at
time point 119905 p21miR119894sim (119905) and p21miR119894
exp (119905) represent the mea-sured value for each miR
119894at time point 119905 and their standard
deviations are 120575mp21119894
and 120575p21119894
Here 119905 is the time point (48 hr)after overexpression of the individual miRNAs in embryonickidney 293 cells at which the expression levels of the p21 andits mRNA were measured [23] The model calibration resultsare shown in Figure 2(a) and the obtained parameter valuesare listed in Table 3
Experimental results showed that a stronger repressionof the target gene can occur when two miRNA binding siteson the target mRNA are in close proximity [24 45] To testthe consequences of this hypothesis we predicted cooperativemiRNA pairs for p21 with seed site distances between 13ndash35 nt in the p21 31015840 UTR To substantiate the cooperativeeffect associated with pairs of miRNAs we introduced agroup of new state variables ([mp21 | miR
119894| miR
119894]) into
the original model These state variables account for the
ternary complexes composed of p21 mRNA and two puta-tively cooperating miRNAs (miR
119894and miR
119895) For these new
variables processes considered are (i) the association of p21mRNA with miR
119894and miR
119895into a complex (119896co-complex
119894119895
ass )and (ii) the degradation of the complex (119896co-complex
119894119895
deg ) Afterexpansion the corresponding modified and new ODEs arelisted below
dmp21119889119905
= 119896mp21syn sdot 119891act (TFmp21)
minusmp21 sdot (119896mp21deg +sum
119894
119896complex
119894
ass sdotmiR119894
+sum
119894119895
119896co-complex
119894119895
ass sdotmiR119894sdotmiR119895)
(9)
8 BioMed Research InternationalTa
ble2Th
econ
fidence
scores
forthe
interactions
inp21regulatorynetworkTh
enetworkconsistso
ffou
rtypes
ofinteractionsm
iRNA-
p21TF
-miRNAT
F-p21andp21-p
rotein
(a)miRNA-
p21interactio
nscores
miRNA
miR-125a-5p
miR-132
miR-208a
miR-28-5p
miR-298
miR-299-5p
miR-345
miR-363
Score
073
073
073
080
073
080
073
073
miRNA
miR-423-3p
miR-515-3p
miR-572
miR-639
miR-654-3p
miR-657
miR-93
Score
085
073
073
073
073
080
095
(b)TF
-miRNAinteractionscores
Score
Score
Score
Score
Score
Score
miR-208
miR-132
miR-657
miR-125a-5p
miR-28-5p
miR-93
MAX
027
EGR1
093
TFAP2
A034
EGR1
086
MZF
1028
TFAP2
A034
EGR1
029
CREB
1086
SP1
034
PAX5
024
E47
027
SOX9
024
ARN
T027
ARE
B6027
MZF
1034
NRS
F024
FOXC
1031
GAT
A3
027
RFX1
024
E47
024
MAX
029
ZID
024
FOXI1
024
MZF
1031
ATF6
024
ELK1
027
USF1
027
PPARG
027
SRY
027
SP1
034
YY1
027
EGR2
029
EGR1
034
GAT
A3
034
JUN
029
POU2F1
024
JUNB
024
FOS
024
RELA
029
MZF
1034
POU2F1
024
EGR1
029
POU2F1
024
NFIC
024
miR-654-3p
TP53
034
SRF
027
CEBP
A024
NFIC
027
RELA
027
YY1
027
NFIC
034
CEBP
A029
NFY
A024
miR-345
miR-423-3p
BACH
1024
miR-299-5p
RORA
024
RUNX1
024
SP1
034
BACH
1027
FOXL
1029
CREB
1024
miR-363
STAT
1024
RELA
027
STAT
1027
SRY
024
SRY
027
E2F1
062
E2F1
085
NFA
TC2
024
POU3F2
024
NFA
TC2
024
SRF
029
FOXC
1033
MYC
084
HNF4
A031
STAT
5A027
FOS
024
RUNX1
027
MZF
1034
miR-298
NFY
A024
SRF
027
POU2F1
027
miR-572
miR-639
JUNB
031
POU2F1
024
MEF
2A027
NF120581
B1024
FOXF
2024
NFY
A024
NFIC
031
POU2F1
024
HNF1A
024
(c)TF
-p21
interactionscores
TF(verified)
SP1
SP3
E2F1
Runx
1Ru
nx2
STAT
1E2
AST
AT5
TP53
TP63
TP73
Score
100
100
077
077
077
072
072
067
067
067
067
TF(verified)
STAT
3CU
X1TF
AP2
ABR
CA1
VDR
RARA
CEB
P120572CEB
P120573RB
1Tb
x2EH
MT2
Score
066
066
060
060
060
060
060
060
060
060
060
TF(putative)
NF120581
B1RE
LAST
AT2
STAT
6SR
FScore
044
044
044
044
043
(d)p21-p
rotein
interactionscores
Protein
TP53
PCNA
CASP
3CC
NA1
CCND1
SKP2
BCCI
PCC
NA2
CCND2
CCNE2
AKT
1C1
orf123
CCDC8
5BCC
NB1
CCNB2
CCND3
Score
100
092
087
087
087
087
081
081
081
081
072
072
072
072
072
072
ProteinCC
NE1
CDC4
5CD
C5L
CDC6
CDC7
CDK1
CDK14
CDK2
CDK3
CDK4
CDK6
CEBP
ACI
Z1CS
NK2
A1
CSNK2
BDAPK
3Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
ESR1
GADD45A
GADD45B
GADD45G
GMNN
GNB2
L1HDAC
11ITGB1BP
3MAP3
K5MAPK
8MCM
10PA
RP1
PIM1
POLD
2PS
MA3
RAB1A
Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
SET
SLC2
5A11
STAT
3TE
X11
TK1
TSG101
TTLL
5XR
CC6
CDC2
0CD
C27
CDK5
DHX9
MDM2
NR4
A1
PRKC
HRP
A1
Score
072
072
072
072
072
072
072
072
056
056
056
056
056
056
056
056
Verifi
edexp
erim
entally
verifi
edinteraction
putativ
epredictedinteraction
BioMed Research International 9
Con
trol
miR
-298
miR
-208
am
iR-1
32m
iR-2
8-5p
miR
-125
-5p
miR
-299
-5p
miR
-345
miR
-93
miR
-423
-3p
miR
-515
-3p
miR
-363
miR
-657
miR
-639
miR
-572
miR
-654
-3p
0
05
1
15m
RNA
expr
essio
n (a
u)
Data Model Data ModelC
ontro
lm
iR-2
98m
iR-2
08a
miR
-132
miR
-28-
5pm
iR-1
25-5
pm
iR-2
99-5
pm
iR-3
45m
iR-9
3m
iR-4
23-3
pm
iR-5
15-3
pm
iR-3
63m
iR-6
57m
iR-6
39m
iR-5
72m
iR-6
54-3
p
0
02
04
06
08
1
Prot
ein
expr
essio
n (a
u)
(a)
DoxorubicinCellWell
Control miR-572 miR-93
Measure p21 expression at the time points of
miR-572+ miR-93
0 2 4 6 8 24hr
miR-572miR-93
Immunoblotting
(b)
0 4 8 12 16 20 240
4
8
12
16
20
Time (hr)
p53
Actin
Doxo (hrs)
fac
t(TF
mp2
1)
C 0 1 2 4 6 8 24
1 08 12 19 29 32 3 16
(c)
0 2 4 6 8 2402468
1012
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)
p21
expr
essio
n (a
u)
02468
1012
miR-572Control
02468
1012
miR-93
02468
1012
Model Data
0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24
1 23 46 107124 124 23 28 48 66 53 31 08 16 21 38 50 18 06 06 12 25 27 12
miR-572+ 93Doxo (hrs)
p21
Actin
p21
Actin
p21
Actin
p21
Actin
(d)
Figure 2 Model calibration and validation (a) Model calibration The figures show the relative change of the p21 mRNA and proteinexpression levels after overexpression of the indicated miRNAs (Model model simulation Data experimental data) These data werenormalized to the control group in which the p21 mRNA and protein expression levels were measured when the miRNAs were normallyexpressed (au arbitrary unit) (b) Experimental workflow In the experiments Sk-Mel-147 cells were seeded in six well plates Then maturemiRNA mimics were transfected individually at a concentration of 100 nM (miR-572 and miR-93) or in combination at 50 nM each (miR-572 + miR-93) After 48 hr transfection with miRNA mimics the cells were pulse treated with 250 nM doxorubicin for 1 hour after whichnormal growth medium was replenished The immunblotting were performed to measure p21 expression at 0 2 4 6 8 and 24 hr post-doxorubicin treatment (c) Temporal dynamics of p21 transcriptional function Afterdoxorubicin treatment the expression of p53 a TF ofp21 was measured using immunblotting and these data were used to characterize the transcriptional function of p21 using MATLAB linearinterpolation function (d) Model validation We measured the expression of p21 protein in response to genotoxic stress in the four scenariosas described in the main text The measured data (Data) were compared with the model simulations (Model) The figures (a) (c) and (d) areadapted from our previous publication [4]
119889 [mp21 | miR119894| miR
119895]
119889119905
= 119896co-complex
119894119895
ass sdotmp21 sdotmiR119894sdotmiR119895
minus 119896co-complex
119894119895
deg sdot [mp21 | miR119894| miR
119895]
(10)
To model stronger repression of the target gene bycooperating miRNAs we assumed a stronger association rateconstant for the complex [mp21 | miR
119894| miR
119895] which is
equal to the sum of their individual association rate constants
(119896co-complex119894119895
ass = 119896complex
119894
ass + 119896complex
119895
ass ) Similarly the degradationrates of the complexes [mp21 | miR
119894| miR
119895] were assumed
to be equal to the sum of degradation rate constants ofsingle miRNA binding complexes (119896co-complex
119894119895
deg = 119896complex
119894
deg +
119896complex
119895
deg ) However it has to be noted that these addedequations are an abstract description ofmiRNAcooperativitybecause the details of this mechanism are not yet known
To experimentally validate the capability of our model topredict the relative p21 concentrations regulated by coopera-tivemiRNAs we selectedmiR-572 andmiR-93 as a case study
10 BioMed Research International
Table 3 Initial concentrations of model variables and model parameter values Based on the experimental data the p21-targeting miRNAsverified byWu et al [23] were divided into two groups the translation repression group (marked with asterisk) and the mRNA deadenylationgroup AmiRNAwas classified into themRNA deadenylation group if its overexpression can result in 20 ormore downregulation of the p21mRNA level (ie p21mRNA levelle 08 the basal level is 1) otherwise it was classified into the translation repression group For the translationrepression group only 119896complex119894
ass was estimated and 119896complex119894deg was fixed For the other group both 119896complex119894
deg and 119896complex119894ass were estimatedThe initial
concentrations of p21 and mp21 were set to 1 and this value was used as their basal expression levels During the parameter estimation theinitial concentrations of p21-targeting miRNAs were set to 100 because in the publication [23] the expression levels of p21 and mp21 weremeasured after the individual introduction of the p21-targeting miRNAs with amount of 100 nM Due to the lack of biological information tocharacterize the transcriptional activation function (119891act) of p21 and its targeting miRNAs the corresponding functions were assumed to be1 for simplicity The data is adapted from our previous publication [4]
Initial concentration of variables and TF functionsVariable Description Initial concentration (au)p21 p21 protein 1
mp21 p21 mRNA 1
miR119894=115
p21-targeting miRNAs 100
[mp21 | miR119894=115
] Complexes formed by miR119894and mp21 0
119891act(TFmp21) p21rsquos transcriptional activation function 1119891act(TFmiR119894 (119894=115)) The transcriptional activation function of miR
1198941
Fixed parameter valuesParameter Description Value (hrminus1) Reference119896mp21syn Synthesis rate constant of mp21 01155 fixed119896mp21deg Degradation rate constant of mp21 01155 [24]119896miR119894syn (119894 = 1 15) Synthesis rate constant of miR
11989400289 fixed
119896miR119894deg (119894 = 1 15) Degradation rate constant of miR
11989400289 [25]
119896p21syn Synthesis rate constant of p21 13863 fixed119896p21deg Degradation rate constant of p21 13863 [26]
Estimated parameter values
miRNA (statevariable)
119896complex119894deg
(119894 = 1 15) (hrminus1)
119896complex119894ass
(119894 = 1 15)
(au minus1sdot hrminus1)
119865miR119894cost
(119894 = 1 15)Experimental data of p21 (protein mRNA plusmn SD)
miR-298 (miR1)lowast 01155 00254 34119890 minus 004 (016 1074 plusmn 0025)
miR-208a (miR2)lowast 01155 00041 20119890 minus 003 (051 1192 plusmn 0022)
miR-132 (miR3)lowast 01155 00275 24119890 minus 003 (015 121 plusmn 0147)
miR-28-5p (miR4)lowast 01155 00119 59119890 minus 003 (028 135 plusmn 006)
miR-125-5p (miR5)lowast 01155 00018 18119890 minus 003 (069 085 plusmn 0051)
miR-299-5p (miR6)lowast 01155 00080 18119890 minus 004 (036 095 plusmn 0038)
miR-345 (miR7)lowast 01155 00051 11119890 minus 004 (046 096 plusmn 0039)
miR-93 (miR8) 01564 00235 41119890 minus 014 (017 07776 plusmn 003)
miR-423-3p (miR9) 09118 00055 28119890 minus 009 (044 05102 plusmn 011)
miR-515-3p (miR10) 02098 00253 12119890 minus 013 (016 0616 plusmn 0037)
miR-363 (miR11) 02261 00399 22119890 minus 014 (011 056 plusmn 015)
mR-657 (miR12) 03465 00158 21119890 minus 014 (023 048 plusmn 012)
miR-639 (miR13) 04305 00327 18119890 minus 017 (013 036 plusmn 0084)
miR-572 (miR14) 03039 00360 94119890 minus 023 (012 045 plusmn 0044)
miR-654-3p (miR15) 97485 00024 30119890 minus 014 (064 063 plusmn 0053)
These two miRNAs were chosen because their predictedtarget sites in the p21 31015840 UTR are in close proximity to eachother and thereby they can induce cooperative repression onp21 as suggested in [24] The experiments were performed asfollows
(i) Melanoma cells (Sk-Mel-147) were transfected withthe mature miRNAmimics of the twomiRNAs eitherindividually (100 nM) or in combination (50 nMeach) whereas untreated cells were used as control(Figure 2(b))
BioMed Research International 11
p21
FOXF2
miR-572
Activation
Inhibition120572
Incoherent FFL + cascaded regulation
AF2
miR-93
(a)
miR-93
miR-572
TP53
MYC
p21
FOXF2
Coherent FFL + cascaded regulation
(b)
Saturation
Pulse
p21
(au
)
p21
(au
)Time (hr) Time (hr)
Saturation Pulse
Pulse
Saturation
Saturation
di
(au
)
102
101
102
100
101
10minus1
10010minus2
10minus1
102
101
100
10minus1
10minus2
FOXF2 (au) FOXF2 (au)
TP53
(au
)
1
09
08
07
06
05
04
03
02
01
10minus2 10210110010minus110minus20
AF2120572
(au
)
(c)
Figure 3 Different p21 dynamics in different network motifs We ran simulations to show the different dynamics of p21 for two differentnetwork motifs (a) and (b) Through simulations two dynamical patterns of p21 were identified saturation and pulse ((c) top) For eachnetwork motif the corresponding distributions of the two dynamical patterns were plotted ((c) bottom) For different combinations ofthe transcriptional strengths the normalized distance (119889
119894) between peaks (119901
119894) and steady states (ss
119894) of p21 is determined by the equation
119889119894= (119901119894minus ss119894)119901max 119901max = max(119901
1 119901
119899) If 119889
119894= 0 for the corresponding combination of transcriptional strengths the p21 dynamics is
saturation otherwise it is pulse The regions showing different dynamical patterns of p21 are separated using the white lines
(ii) Next the cells were treated with doxorubicina genotoxic-stress inducing agent The agent canupregulate the expression of p53 which is a knownTFof p21 and therefore it can result in the upregulationof p21 (Figure 2(c))
(iii) After doxorubicin treatment the expression levels ofp21 were measured by immunoblotting at differenttime points (0 2 4 6 8 24 hr) The p21 expressionvalues were normalized based on the p21 expression
level in the control group measured at time point 0 hr(Figure 2(d))
By doing so we obtained the p21 response after genotoxicstress in four scenarios (1) endogenous miRNA expression(Control) (2) overexpression of miR-572 (3) overexpressionof miR-93 and (4) both miRNAs moderately overexpressedThereafter we derived a model of seven ODEs based onthe original equations which was configured according tothe designed experiments making the simulation results
12 BioMed Research International
00001
0001
001
01
1Cell cycle Cell growth DNA damage Senescence Apoptosis
G1G2
MS
p21
expr
essio
n (a
u)
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
log10
wow
Figure 4 p21 expression regulated by cooperative miRNAs for different cellular processesThe associations of themiRNAs with these cellularprocesses were derived from GO terms of their TFs A miRNA was supposed to be expressed (in bold black font) in a cellular process onlyif its TF is related to the corresponding GO term of this process The p21 expression levels are computed for each process with (w)without(wo) considering the cooperative effect among the p21 targeting miRNAs
comparable with the experimental data As shown in theFigure 2(c) the simulations are in good agreement with theexperimental observations The individual overexpression ofmiR-572 or miR-93 led to the reduction of the upregulationof p21 response after genotoxic stress induction The twomiRNAs cause different degrees of repression due to theirdifferent repression efficiencies on p21 Interestingly the com-bined overexpression of both miRNAs induced the strongestdownregulation of p21 and therefore verifying the hypothesisof their cooperative regulation of p21 Above all the resultsnot only validated themodel but also demonstrated the abilityof our method to identify cooperative miRNA pairs for p21
(3) Predictive Simulations As there are abundant of networkmotifs such as FFLs in p21 regulatory network and thesenetwork motifs are important for determining p21 dynamicsit is interesting to investigate the dynamics of p21 in networkmodules where FFLs are involved To do so two networkmodules including both miR-93 and miR-572 and theirTFs were exemplified In Figure 3(a) the network modulecontains an incoherent FFL composed by AF2120572 miR-93 andp21 and a cascaded regulation in which p21 is repressed
by FOXF2 via miR-572 in Figure 3(b) the same cascadedregulation together with a coherent FFL composed of TP53MYC miR-93 and p21 forms another regulatory moduleof p21 By modulating the transcriptional strengths of thetwo miRNAs by their TFs two types of p21 dynamics wereidentified saturation and pulse (Figure 3(c) top) In theformer the p21 expression increases and reaches its steadystate at the highest level in contrast in the latter the p21expression increases to a peak and thereafter drops to asteady state at lower level For the two different networkmodules various combinations of transcriptional strengthsof the two miRNAs lead to different distributions of thetwo p21 dynamical patterns For the network module of anincoherent FFL plus the cascaded regulation (Figure 3(c)bottom left) the saturation pattern appears in two distinctregions one with weakened transcriptional strength of miR-93 and the other with enhanced transcriptional strength ofmiR-93 for the other module (Figure 3(c) bottom right)the saturation pattern only appears in the region in whichthe transcriptional strength of miR-93 is enhanced Takentogether the results showed that for the two different networkmotifs the dynamical pattern of p21 is changing according to
BioMed Research International 13
different combinations of its upstream regulators suggestingthe adaptation of p21 dynamics for different biological con-texts
Furthermore we performed a number of simulations toshow the influence of miRNA regulation on p21 expressionlevels in different cellular processes arranged in a consecutivemanner In this procedure a cell is first in the process ofcell cycle followed by proliferation (cell growth) then thecell responses to DNA damage and enters into the processof senescence and finally apoptosis is initiated As shown inFigure 4 during the cell cycle process the p21 expression islow due to the activation of most of its targeting miRNAswhen the cell starts proliferating the p21 expression declinesto an even lower level because of more activated miRNAsin this process after responding to DNA damage the p21expression soars to a high level which is caused by theactivation of its TFs like p53 and fewer expressed miRNAsalthough the p21 expression keeps at a similar level whilethe cell is undergoing senescence the expressed miRNAs aredifferent from the previous process finally the p21 expressiondecreases again to a low level due to the reemergence ofmost of its targeting miRNAs and the cell enters apoptosisInterestingly the model simulations are also consistent withexperimental observations under non-stressed condition thelow expression level of p21 is needed for cell proliferation theupregulation of p21 happens after response to DNA damagevia p53 and the increased p21 expression further results incell cycle arrest leading to senescence and apoptosis [25]Besides when considering the effect of cooperatingmiRNAsthe p21 expression levels were indistinguishable from theprevious simulation of DNA damage response However forthe other processes p21 expression levels were significantlylower compared to the simulations without considering thecooperative effect of the miRNAs Above all these resultsindicated that selective expression of cooperative miRNAscould be adopted by cells to ensure diverse expression levelsof p21 tomeet the requirements of different cellular processes
3 Conclusions and Discussion
In this paper we presented a systems biology approach com-bining data-drivenmodeling andmodel-driven experimentsto investigate the role of miRNA-mediated repression in generegulatory networksThis approach provides a systematicwayto gain a deeper understanding of the regulation of targetgenes by mutiple and cooperative miRNAs Using the regu-lation of p21 by multiple miRNAs as a case study we showedhow the ODE-basedmodel which is calibrated and validatedby means of experimental data is suitable for predicting thetemporal dynamics of molecular concentrations involved inbiochemical systems
Provided there are sufficiently rich quantitative data setsavaliable to characterize the model the use of the methodol-ogy here shown can be extended to more complex regulatorynetworks involving multiple targets cooperating TFs andmiRNAs and signaling pathways displaying cross-talk viapost-tranlational modifications In this case the criticalelement is the quality and quantitiy of the available data
Insufficiency and low quality of experimental data can causeerrors in the process of model construction and overfitting inparameter estimation can lead to uncertainties in the modelpredictions We believe that the quick development of quan-titative high throughput techniques such as transcriptomicsproteomics and miRNomics will facilitate the constructionand characterization of larger miRNA-mediated regulatorynetworks
Other modeling frameworks than ODE-based modelscan be used to describe biological systems such as prob-abilistic (eg Bayesian) or logical (eg Boolean) modelsImportantly different modeling frameworks have differentproperties and perform well regarding different perspectivesand levels of mechanistic details of biochemical systems [46]For example Bayesian models are helpful in the constructionof connections in signaling networks and can reveal the mostlikely underlying structure of the network in a probabilisticmanner Boolean models use binary values (0 and 1) andlogical gates (AND NOT and OR) to describe activities ofnetwork components and the information flow among themWe believe that in the coming future hybrid models whichconsist of modeling framework and experimental techniquespecific sub-modules will provide the necessary compromisebetween quantitativequalitative accuracy and scalability forthe investigation of large biochemical networks [47]
Disclosures
The authors declare that they have no competing financialinterests
Authorsrsquo Contribution
Julio Vera and Olaf Wolkenhauer are equal contributers
Acknowledgments
The authors would like to acknowledge the funding fromthe German Federal Ministry of Education and Research(BMBF) eBio-miRSys (0316175A to Xin Lai and Julio Vera)eBio-SysMet (0316171 to OlafWolkenhauer) and Gerontosys-ROSAge (0315892A to Olaf Wolkenhauer)
References
[1] A E Pasquinelli ldquoMicroRNAs and their targets recognitionregulation and an emerging reciprocal relationshiprdquo NatureReviews Genetics vol 13 no 4 pp 271ndash282 2012
[2] J Vera X Lai U Schmitz and O Wolkenhauer ldquoMicroRNA-regulated networks the perfect storm for classical molecularbiology the ideal scenario for systems biologyrdquo Advances inExperimental Medicine and Biology vol 774 pp 55ndash76 2013
[3] S Nikolov J Vera U Schmitz and O Wolkenhauer ldquoA model-based strategy to investigate the role of microRNA regulation incancer signalling networksrdquo Theory in Biosciences vol 130 no1 pp 55ndash69 2011
14 BioMed Research International
[4] X Lai U Schmitz S K Gupta et al ldquoComputational analysisof target hub gene repression regulated by multiple and cooper-ative miRNAsrdquoNucleic Acids Research vol 40 no 18 pp 8818ndash8834 2012
[5] X Lai O Wolkenhauer and J Vera ldquoModeling miRNA regu-lation in cancer signaling systems miR-34a regulation of thep53Sirt1 signaling modulerdquoMethods in Molecular Biology vol880 pp 87ndash108 2012
[6] C Jiang Z Xuan F Zhao and M Q Zhang ldquoTRED atranscriptional regulatory element database new entries andother developmentrdquoNucleic Acids Research vol 35 supplement1 pp D137ndashD140 2007
[7] L A Bovolenta M L Acencio and N Lemke ldquoHTRIdban open-access database for experimentally verified humantranscriptional regulation interactionsrdquo BMCGenomics vol 13article 405 2012
[8] VMatys O V Kel-Margoulis E Fricke et al ldquoTRANSFAC andits module TRANSCompel transcriptional gene regulation ineukaryotesrdquo Nucleic Acids Research vol 34 database issue ppD108ndashD110 2006
[9] D Karolchik R Baertsch M Diekhans et al ldquoThe UCSCgenome browser databaserdquo Nucleic Acids Research vol 31 no1 pp 51ndash54 2003
[10] P Alexiou T Vergoulis M Gleditzsch et al ldquomiRGen 20 adatabase of microRNA genomic information and regulationrdquoNucleic Acids Research vol 38 supplement 1 pp D137ndashD1412010
[11] F Xiao Z Zuo G Cai S Kang X Gao and T Li ldquomiRecordsan integrated resource for microRNA-target interactionsrdquoNucleic Acids Research vol 37 supplement 1 pp D105ndashD1102009
[12] S-D Hsu F-M Lin W-Y Wu et al ldquomiRTarBase a databasecurates experimentally validated microRNA-target interac-tionsrdquo Nucleic Acids Research vol 39 supplement 1 pp D163ndashD169 2011
[13] P Sethupathy B Corda and A G Hatzigeorgiou ldquoTarBase acomprehensive database of experimentally supported animalmicroRNA targetsrdquo RNA vol 12 no 2 pp 192ndash197 2006
[14] H Dweep C Sticht P Pandey and N Gretz ldquomiRWalkmdashdatabase prediction of possible miRNA binding sites byldquowalkingrdquo the genes of three genomesrdquo Journal of BiomedicalInformatics vol 44 no 5 pp 839ndash847 2011
[15] T S Keshava Prasad RGoel K Kandasamy et al ldquoHumanpro-tein reference databasemdash2009 updaterdquo Nucleic Acids Researchvol 37 supplement 1 pp D767ndashD772 2009
[16] D Szklarczyk A Franceschini M Kuhn et al ldquoThe STRINGdatabase in 2011 functional interaction networks of proteinsglobally integrated and scoredrdquo Nucleic Acids Research vol 39supplement 1 pp D561ndashD568 2011
[17] D Croft G OrsquoKelly G Wu et al ldquoReactome a database ofreactions pathways and biological processesrdquo Nucleic AcidsResearch vol 39 supplement 1 pp D691ndashD697 2011
[18] N le Novere M Hucka H Mi et al ldquoThe systems biologygraphical notationrdquoNature Biotechnology vol 27 no 8 pp 735ndash741 2009
[19] A Funahashi Y Matsuoka A Jouraku M Morohashi NKikuchi and H Kitano ldquoCellDesigner 35 a versatile modelingtool for biochemical networksrdquo Proceedings of the IEEE vol 96no 8 pp 1254ndash1265 2008
[20] M S Cline M Smoot E Cerami et al ldquoIntegration ofbiological networks and gene expression data using CytoscaperdquoNature Protocols vol 2 no 10 pp 2366ndash2382 2007
[21] D-H Le andY-KKwon ldquoNetDS aCytoscape plugin to analyzethe robustness of dynamics and feedforwardfeedback loopstructures of biological networksrdquo Bioinformatics vol 27 no 19pp 2767ndash2768 2011
[22] E O Voit Computational Analysis of Biochemical SystemsA Practical Guide for Biochemists and Molecular BiologistsCambridge University Press Cambridge UK 2000
[23] S Wu S Huang J Ding et al ldquoMultiple microRNAs modulatep21Cip1Waf1 expression by directly targeting its 31015840 untranslatedregionrdquo Oncogene vol 29 no 15 pp 2302ndash2308 2010
[24] P Saeligtrom B S E Heale O Snoslashve Jr L Aagaard J Alluin andJ J Rossi ldquoDistance constraints between microRNA target sitesdictate efficacy and cooperativityrdquo Nucleic Acids Research vol35 no 7 pp 2333ndash2342 2007
[25] Y-S Jung YQian andXChen ldquoExamination of the expandingpathways for the regulation of p21 expression and activityrdquoCellular Signalling vol 22 no 7 pp 1003ndash1012 2010
[26] W Wang H Furneaux H Cheng et al ldquoHuR regulatesp21 mRNA stabilization by UV lightrdquo Molecular and CellularBiology vol 20 no 3 pp 760ndash769 2000
[27] U Wittig R Kania M Golebiewski et al ldquoSABIO-RKmdashdatabase for biochemical reaction kineticsrdquo Nucleic AcidsResearch vol 40 database issue pp D790ndashD796 2012
[28] R Milo P Jorgensen U Moran G Weber and M SpringerldquoBioNumbersmdashthe database of key numbers in molecular andcell biologyrdquo Nucleic Acids Research vol 38 supplement 1 ppD750ndashD753 2010
[29] I-C Chou and E O Voit ldquoRecent developments in parameterestimation and structure identification of biochemical andgenomic systemsrdquoMathematical Biosciences vol 219 no 2 pp57ndash83 2009
[30] A Saltelli K Chan and E M Scott Sensitivity Analysis JohnWiley amp Sons New York NY USA 1st edition 2000
[31] S Strogatz Nonlinear Dynamics and Chaos Applications toPhysics Biology Chemistry and Engineering With Applicationsto Physics Biology Chemistry and Engineering Westview PressBoulder Colo USA 2000
[32] P Zhou S Cai Z Liu and R Wang ldquoMechanisms generatingbistability and oscillations in microRNA-mediated motifsrdquoPhysical Review E vol 85 no 4 apart 1 Article ID 041916 9pages 2012
[33] S Marino I B Hogue C J Ray and D E Kirschner ldquoAmethodology for performing global uncertainty and sensitivityanalysis in systems biologyrdquo Journal of Theoretical Biology vol254 no 1 pp 178ndash196 2008
[34] L P LimNC Lau PGarrett-Engele et al ldquoMicroarray analysisshows that some microRNAs downregulate large numbers of-target mRNAsrdquo Nature vol 433 no 7027 pp 769ndash773 2005
[35] J Krutzfeldt N Rajewsky R Braich et al ldquoSilencing ofmicroRNAs in vivowith lsquoantagomirsrsquordquoNature vol 438 no 7068pp 685ndash689 2005
[36] MWPfafflGWHorgan andLDempfle ldquoRelative expressionsoftware tool (REST) for group-wise comparison and statisticalanalysis of relative expression results in real-time PCRrdquoNucleicAcids Research vol 30 no 9 p e36 2002
[37] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[38] T Abbas and A Dutta ldquop21 in cancer intricate networks andmultiple activitiesrdquo Nature Reviews Cancer vol 9 no 6 pp400ndash414 2009
[39] J Wang M Lu C Qiu and Q Cui ldquoTransmiR a transcriptionfactor microRNA regulation databaserdquo Nucleic Acids Researchvol 38 supplement 1 pp D119ndashD122 2010
BioMed Research International 15
[40] S Bandyopadhyay andM Bhattacharyya ldquoPuTmiR a databasefor extracting neighboring transcription factors of humanmicroRNAsrdquo BMC Bioinformatics vol 11 article 190 2010
[41] A Le Bechec E Portales-Casamar G Vetter et alldquoMIRNTN a framework integrating transcription factorsmicroRNAs and their targets to identify sub-network motifs ina meta-regulation network modelrdquo BMC Bioinformatics vol12 article 67 2011
[42] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyTheGene Ontology ConsortiumrdquoNature Genetics vol 25 no 1 pp 25ndash29 2000
[43] A I F Vaz and L N Vicente ldquoA particle swarm pattern searchmethod for bound constrained global optimizationrdquo Journal ofGlobal Optimization vol 39 no 2 pp 197ndash219 2007
[44] W H PressNumerical Recipes The Art of Scientific ComputingCambridge University Press Cambridge UK 2007
[45] J G Doench and P A Sharp ldquoSpecificity of microRNA targetselection in translational repressionrdquo Genes and Developmentvol 18 no 5 pp 504ndash511 2004
[46] B Kholodenko M B Yaffe and W Kolch ldquoComputationalapproaches for analyzing information flow in biological net-worksrdquo Science Signaling vol 5 no 220 2012
[47] F M Khan U Schmitz S Nikolov et al ldquoHybrid modeling ofthe crosstalk between signaling and transcriptional networksusing ordinary differential equations and multi-valued logicrdquoBiochimica et Biophysica Acta 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
8 BioMed Research InternationalTa
ble2Th
econ
fidence
scores
forthe
interactions
inp21regulatorynetworkTh
enetworkconsistso
ffou
rtypes
ofinteractionsm
iRNA-
p21TF
-miRNAT
F-p21andp21-p
rotein
(a)miRNA-
p21interactio
nscores
miRNA
miR-125a-5p
miR-132
miR-208a
miR-28-5p
miR-298
miR-299-5p
miR-345
miR-363
Score
073
073
073
080
073
080
073
073
miRNA
miR-423-3p
miR-515-3p
miR-572
miR-639
miR-654-3p
miR-657
miR-93
Score
085
073
073
073
073
080
095
(b)TF
-miRNAinteractionscores
Score
Score
Score
Score
Score
Score
miR-208
miR-132
miR-657
miR-125a-5p
miR-28-5p
miR-93
MAX
027
EGR1
093
TFAP2
A034
EGR1
086
MZF
1028
TFAP2
A034
EGR1
029
CREB
1086
SP1
034
PAX5
024
E47
027
SOX9
024
ARN
T027
ARE
B6027
MZF
1034
NRS
F024
FOXC
1031
GAT
A3
027
RFX1
024
E47
024
MAX
029
ZID
024
FOXI1
024
MZF
1031
ATF6
024
ELK1
027
USF1
027
PPARG
027
SRY
027
SP1
034
YY1
027
EGR2
029
EGR1
034
GAT
A3
034
JUN
029
POU2F1
024
JUNB
024
FOS
024
RELA
029
MZF
1034
POU2F1
024
EGR1
029
POU2F1
024
NFIC
024
miR-654-3p
TP53
034
SRF
027
CEBP
A024
NFIC
027
RELA
027
YY1
027
NFIC
034
CEBP
A029
NFY
A024
miR-345
miR-423-3p
BACH
1024
miR-299-5p
RORA
024
RUNX1
024
SP1
034
BACH
1027
FOXL
1029
CREB
1024
miR-363
STAT
1024
RELA
027
STAT
1027
SRY
024
SRY
027
E2F1
062
E2F1
085
NFA
TC2
024
POU3F2
024
NFA
TC2
024
SRF
029
FOXC
1033
MYC
084
HNF4
A031
STAT
5A027
FOS
024
RUNX1
027
MZF
1034
miR-298
NFY
A024
SRF
027
POU2F1
027
miR-572
miR-639
JUNB
031
POU2F1
024
MEF
2A027
NF120581
B1024
FOXF
2024
NFY
A024
NFIC
031
POU2F1
024
HNF1A
024
(c)TF
-p21
interactionscores
TF(verified)
SP1
SP3
E2F1
Runx
1Ru
nx2
STAT
1E2
AST
AT5
TP53
TP63
TP73
Score
100
100
077
077
077
072
072
067
067
067
067
TF(verified)
STAT
3CU
X1TF
AP2
ABR
CA1
VDR
RARA
CEB
P120572CEB
P120573RB
1Tb
x2EH
MT2
Score
066
066
060
060
060
060
060
060
060
060
060
TF(putative)
NF120581
B1RE
LAST
AT2
STAT
6SR
FScore
044
044
044
044
043
(d)p21-p
rotein
interactionscores
Protein
TP53
PCNA
CASP
3CC
NA1
CCND1
SKP2
BCCI
PCC
NA2
CCND2
CCNE2
AKT
1C1
orf123
CCDC8
5BCC
NB1
CCNB2
CCND3
Score
100
092
087
087
087
087
081
081
081
081
072
072
072
072
072
072
ProteinCC
NE1
CDC4
5CD
C5L
CDC6
CDC7
CDK1
CDK14
CDK2
CDK3
CDK4
CDK6
CEBP
ACI
Z1CS
NK2
A1
CSNK2
BDAPK
3Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
ESR1
GADD45A
GADD45B
GADD45G
GMNN
GNB2
L1HDAC
11ITGB1BP
3MAP3
K5MAPK
8MCM
10PA
RP1
PIM1
POLD
2PS
MA3
RAB1A
Score
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
072
Protein
SET
SLC2
5A11
STAT
3TE
X11
TK1
TSG101
TTLL
5XR
CC6
CDC2
0CD
C27
CDK5
DHX9
MDM2
NR4
A1
PRKC
HRP
A1
Score
072
072
072
072
072
072
072
072
056
056
056
056
056
056
056
056
Verifi
edexp
erim
entally
verifi
edinteraction
putativ
epredictedinteraction
BioMed Research International 9
Con
trol
miR
-298
miR
-208
am
iR-1
32m
iR-2
8-5p
miR
-125
-5p
miR
-299
-5p
miR
-345
miR
-93
miR
-423
-3p
miR
-515
-3p
miR
-363
miR
-657
miR
-639
miR
-572
miR
-654
-3p
0
05
1
15m
RNA
expr
essio
n (a
u)
Data Model Data ModelC
ontro
lm
iR-2
98m
iR-2
08a
miR
-132
miR
-28-
5pm
iR-1
25-5
pm
iR-2
99-5
pm
iR-3
45m
iR-9
3m
iR-4
23-3
pm
iR-5
15-3
pm
iR-3
63m
iR-6
57m
iR-6
39m
iR-5
72m
iR-6
54-3
p
0
02
04
06
08
1
Prot
ein
expr
essio
n (a
u)
(a)
DoxorubicinCellWell
Control miR-572 miR-93
Measure p21 expression at the time points of
miR-572+ miR-93
0 2 4 6 8 24hr
miR-572miR-93
Immunoblotting
(b)
0 4 8 12 16 20 240
4
8
12
16
20
Time (hr)
p53
Actin
Doxo (hrs)
fac
t(TF
mp2
1)
C 0 1 2 4 6 8 24
1 08 12 19 29 32 3 16
(c)
0 2 4 6 8 2402468
1012
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)
p21
expr
essio
n (a
u)
02468
1012
miR-572Control
02468
1012
miR-93
02468
1012
Model Data
0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24
1 23 46 107124 124 23 28 48 66 53 31 08 16 21 38 50 18 06 06 12 25 27 12
miR-572+ 93Doxo (hrs)
p21
Actin
p21
Actin
p21
Actin
p21
Actin
(d)
Figure 2 Model calibration and validation (a) Model calibration The figures show the relative change of the p21 mRNA and proteinexpression levels after overexpression of the indicated miRNAs (Model model simulation Data experimental data) These data werenormalized to the control group in which the p21 mRNA and protein expression levels were measured when the miRNAs were normallyexpressed (au arbitrary unit) (b) Experimental workflow In the experiments Sk-Mel-147 cells were seeded in six well plates Then maturemiRNA mimics were transfected individually at a concentration of 100 nM (miR-572 and miR-93) or in combination at 50 nM each (miR-572 + miR-93) After 48 hr transfection with miRNA mimics the cells were pulse treated with 250 nM doxorubicin for 1 hour after whichnormal growth medium was replenished The immunblotting were performed to measure p21 expression at 0 2 4 6 8 and 24 hr post-doxorubicin treatment (c) Temporal dynamics of p21 transcriptional function Afterdoxorubicin treatment the expression of p53 a TF ofp21 was measured using immunblotting and these data were used to characterize the transcriptional function of p21 using MATLAB linearinterpolation function (d) Model validation We measured the expression of p21 protein in response to genotoxic stress in the four scenariosas described in the main text The measured data (Data) were compared with the model simulations (Model) The figures (a) (c) and (d) areadapted from our previous publication [4]
119889 [mp21 | miR119894| miR
119895]
119889119905
= 119896co-complex
119894119895
ass sdotmp21 sdotmiR119894sdotmiR119895
minus 119896co-complex
119894119895
deg sdot [mp21 | miR119894| miR
119895]
(10)
To model stronger repression of the target gene bycooperating miRNAs we assumed a stronger association rateconstant for the complex [mp21 | miR
119894| miR
119895] which is
equal to the sum of their individual association rate constants
(119896co-complex119894119895
ass = 119896complex
119894
ass + 119896complex
119895
ass ) Similarly the degradationrates of the complexes [mp21 | miR
119894| miR
119895] were assumed
to be equal to the sum of degradation rate constants ofsingle miRNA binding complexes (119896co-complex
119894119895
deg = 119896complex
119894
deg +
119896complex
119895
deg ) However it has to be noted that these addedequations are an abstract description ofmiRNAcooperativitybecause the details of this mechanism are not yet known
To experimentally validate the capability of our model topredict the relative p21 concentrations regulated by coopera-tivemiRNAs we selectedmiR-572 andmiR-93 as a case study
10 BioMed Research International
Table 3 Initial concentrations of model variables and model parameter values Based on the experimental data the p21-targeting miRNAsverified byWu et al [23] were divided into two groups the translation repression group (marked with asterisk) and the mRNA deadenylationgroup AmiRNAwas classified into themRNA deadenylation group if its overexpression can result in 20 ormore downregulation of the p21mRNA level (ie p21mRNA levelle 08 the basal level is 1) otherwise it was classified into the translation repression group For the translationrepression group only 119896complex119894
ass was estimated and 119896complex119894deg was fixed For the other group both 119896complex119894
deg and 119896complex119894ass were estimatedThe initial
concentrations of p21 and mp21 were set to 1 and this value was used as their basal expression levels During the parameter estimation theinitial concentrations of p21-targeting miRNAs were set to 100 because in the publication [23] the expression levels of p21 and mp21 weremeasured after the individual introduction of the p21-targeting miRNAs with amount of 100 nM Due to the lack of biological information tocharacterize the transcriptional activation function (119891act) of p21 and its targeting miRNAs the corresponding functions were assumed to be1 for simplicity The data is adapted from our previous publication [4]
Initial concentration of variables and TF functionsVariable Description Initial concentration (au)p21 p21 protein 1
mp21 p21 mRNA 1
miR119894=115
p21-targeting miRNAs 100
[mp21 | miR119894=115
] Complexes formed by miR119894and mp21 0
119891act(TFmp21) p21rsquos transcriptional activation function 1119891act(TFmiR119894 (119894=115)) The transcriptional activation function of miR
1198941
Fixed parameter valuesParameter Description Value (hrminus1) Reference119896mp21syn Synthesis rate constant of mp21 01155 fixed119896mp21deg Degradation rate constant of mp21 01155 [24]119896miR119894syn (119894 = 1 15) Synthesis rate constant of miR
11989400289 fixed
119896miR119894deg (119894 = 1 15) Degradation rate constant of miR
11989400289 [25]
119896p21syn Synthesis rate constant of p21 13863 fixed119896p21deg Degradation rate constant of p21 13863 [26]
Estimated parameter values
miRNA (statevariable)
119896complex119894deg
(119894 = 1 15) (hrminus1)
119896complex119894ass
(119894 = 1 15)
(au minus1sdot hrminus1)
119865miR119894cost
(119894 = 1 15)Experimental data of p21 (protein mRNA plusmn SD)
miR-298 (miR1)lowast 01155 00254 34119890 minus 004 (016 1074 plusmn 0025)
miR-208a (miR2)lowast 01155 00041 20119890 minus 003 (051 1192 plusmn 0022)
miR-132 (miR3)lowast 01155 00275 24119890 minus 003 (015 121 plusmn 0147)
miR-28-5p (miR4)lowast 01155 00119 59119890 minus 003 (028 135 plusmn 006)
miR-125-5p (miR5)lowast 01155 00018 18119890 minus 003 (069 085 plusmn 0051)
miR-299-5p (miR6)lowast 01155 00080 18119890 minus 004 (036 095 plusmn 0038)
miR-345 (miR7)lowast 01155 00051 11119890 minus 004 (046 096 plusmn 0039)
miR-93 (miR8) 01564 00235 41119890 minus 014 (017 07776 plusmn 003)
miR-423-3p (miR9) 09118 00055 28119890 minus 009 (044 05102 plusmn 011)
miR-515-3p (miR10) 02098 00253 12119890 minus 013 (016 0616 plusmn 0037)
miR-363 (miR11) 02261 00399 22119890 minus 014 (011 056 plusmn 015)
mR-657 (miR12) 03465 00158 21119890 minus 014 (023 048 plusmn 012)
miR-639 (miR13) 04305 00327 18119890 minus 017 (013 036 plusmn 0084)
miR-572 (miR14) 03039 00360 94119890 minus 023 (012 045 plusmn 0044)
miR-654-3p (miR15) 97485 00024 30119890 minus 014 (064 063 plusmn 0053)
These two miRNAs were chosen because their predictedtarget sites in the p21 31015840 UTR are in close proximity to eachother and thereby they can induce cooperative repression onp21 as suggested in [24] The experiments were performed asfollows
(i) Melanoma cells (Sk-Mel-147) were transfected withthe mature miRNAmimics of the twomiRNAs eitherindividually (100 nM) or in combination (50 nMeach) whereas untreated cells were used as control(Figure 2(b))
BioMed Research International 11
p21
FOXF2
miR-572
Activation
Inhibition120572
Incoherent FFL + cascaded regulation
AF2
miR-93
(a)
miR-93
miR-572
TP53
MYC
p21
FOXF2
Coherent FFL + cascaded regulation
(b)
Saturation
Pulse
p21
(au
)
p21
(au
)Time (hr) Time (hr)
Saturation Pulse
Pulse
Saturation
Saturation
di
(au
)
102
101
102
100
101
10minus1
10010minus2
10minus1
102
101
100
10minus1
10minus2
FOXF2 (au) FOXF2 (au)
TP53
(au
)
1
09
08
07
06
05
04
03
02
01
10minus2 10210110010minus110minus20
AF2120572
(au
)
(c)
Figure 3 Different p21 dynamics in different network motifs We ran simulations to show the different dynamics of p21 for two differentnetwork motifs (a) and (b) Through simulations two dynamical patterns of p21 were identified saturation and pulse ((c) top) For eachnetwork motif the corresponding distributions of the two dynamical patterns were plotted ((c) bottom) For different combinations ofthe transcriptional strengths the normalized distance (119889
119894) between peaks (119901
119894) and steady states (ss
119894) of p21 is determined by the equation
119889119894= (119901119894minus ss119894)119901max 119901max = max(119901
1 119901
119899) If 119889
119894= 0 for the corresponding combination of transcriptional strengths the p21 dynamics is
saturation otherwise it is pulse The regions showing different dynamical patterns of p21 are separated using the white lines
(ii) Next the cells were treated with doxorubicina genotoxic-stress inducing agent The agent canupregulate the expression of p53 which is a knownTFof p21 and therefore it can result in the upregulationof p21 (Figure 2(c))
(iii) After doxorubicin treatment the expression levels ofp21 were measured by immunoblotting at differenttime points (0 2 4 6 8 24 hr) The p21 expressionvalues were normalized based on the p21 expression
level in the control group measured at time point 0 hr(Figure 2(d))
By doing so we obtained the p21 response after genotoxicstress in four scenarios (1) endogenous miRNA expression(Control) (2) overexpression of miR-572 (3) overexpressionof miR-93 and (4) both miRNAs moderately overexpressedThereafter we derived a model of seven ODEs based onthe original equations which was configured according tothe designed experiments making the simulation results
12 BioMed Research International
00001
0001
001
01
1Cell cycle Cell growth DNA damage Senescence Apoptosis
G1G2
MS
p21
expr
essio
n (a
u)
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
log10
wow
Figure 4 p21 expression regulated by cooperative miRNAs for different cellular processesThe associations of themiRNAs with these cellularprocesses were derived from GO terms of their TFs A miRNA was supposed to be expressed (in bold black font) in a cellular process onlyif its TF is related to the corresponding GO term of this process The p21 expression levels are computed for each process with (w)without(wo) considering the cooperative effect among the p21 targeting miRNAs
comparable with the experimental data As shown in theFigure 2(c) the simulations are in good agreement with theexperimental observations The individual overexpression ofmiR-572 or miR-93 led to the reduction of the upregulationof p21 response after genotoxic stress induction The twomiRNAs cause different degrees of repression due to theirdifferent repression efficiencies on p21 Interestingly the com-bined overexpression of both miRNAs induced the strongestdownregulation of p21 and therefore verifying the hypothesisof their cooperative regulation of p21 Above all the resultsnot only validated themodel but also demonstrated the abilityof our method to identify cooperative miRNA pairs for p21
(3) Predictive Simulations As there are abundant of networkmotifs such as FFLs in p21 regulatory network and thesenetwork motifs are important for determining p21 dynamicsit is interesting to investigate the dynamics of p21 in networkmodules where FFLs are involved To do so two networkmodules including both miR-93 and miR-572 and theirTFs were exemplified In Figure 3(a) the network modulecontains an incoherent FFL composed by AF2120572 miR-93 andp21 and a cascaded regulation in which p21 is repressed
by FOXF2 via miR-572 in Figure 3(b) the same cascadedregulation together with a coherent FFL composed of TP53MYC miR-93 and p21 forms another regulatory moduleof p21 By modulating the transcriptional strengths of thetwo miRNAs by their TFs two types of p21 dynamics wereidentified saturation and pulse (Figure 3(c) top) In theformer the p21 expression increases and reaches its steadystate at the highest level in contrast in the latter the p21expression increases to a peak and thereafter drops to asteady state at lower level For the two different networkmodules various combinations of transcriptional strengthsof the two miRNAs lead to different distributions of thetwo p21 dynamical patterns For the network module of anincoherent FFL plus the cascaded regulation (Figure 3(c)bottom left) the saturation pattern appears in two distinctregions one with weakened transcriptional strength of miR-93 and the other with enhanced transcriptional strength ofmiR-93 for the other module (Figure 3(c) bottom right)the saturation pattern only appears in the region in whichthe transcriptional strength of miR-93 is enhanced Takentogether the results showed that for the two different networkmotifs the dynamical pattern of p21 is changing according to
BioMed Research International 13
different combinations of its upstream regulators suggestingthe adaptation of p21 dynamics for different biological con-texts
Furthermore we performed a number of simulations toshow the influence of miRNA regulation on p21 expressionlevels in different cellular processes arranged in a consecutivemanner In this procedure a cell is first in the process ofcell cycle followed by proliferation (cell growth) then thecell responses to DNA damage and enters into the processof senescence and finally apoptosis is initiated As shown inFigure 4 during the cell cycle process the p21 expression islow due to the activation of most of its targeting miRNAswhen the cell starts proliferating the p21 expression declinesto an even lower level because of more activated miRNAsin this process after responding to DNA damage the p21expression soars to a high level which is caused by theactivation of its TFs like p53 and fewer expressed miRNAsalthough the p21 expression keeps at a similar level whilethe cell is undergoing senescence the expressed miRNAs aredifferent from the previous process finally the p21 expressiondecreases again to a low level due to the reemergence ofmost of its targeting miRNAs and the cell enters apoptosisInterestingly the model simulations are also consistent withexperimental observations under non-stressed condition thelow expression level of p21 is needed for cell proliferation theupregulation of p21 happens after response to DNA damagevia p53 and the increased p21 expression further results incell cycle arrest leading to senescence and apoptosis [25]Besides when considering the effect of cooperatingmiRNAsthe p21 expression levels were indistinguishable from theprevious simulation of DNA damage response However forthe other processes p21 expression levels were significantlylower compared to the simulations without considering thecooperative effect of the miRNAs Above all these resultsindicated that selective expression of cooperative miRNAscould be adopted by cells to ensure diverse expression levelsof p21 tomeet the requirements of different cellular processes
3 Conclusions and Discussion
In this paper we presented a systems biology approach com-bining data-drivenmodeling andmodel-driven experimentsto investigate the role of miRNA-mediated repression in generegulatory networksThis approach provides a systematicwayto gain a deeper understanding of the regulation of targetgenes by mutiple and cooperative miRNAs Using the regu-lation of p21 by multiple miRNAs as a case study we showedhow the ODE-basedmodel which is calibrated and validatedby means of experimental data is suitable for predicting thetemporal dynamics of molecular concentrations involved inbiochemical systems
Provided there are sufficiently rich quantitative data setsavaliable to characterize the model the use of the methodol-ogy here shown can be extended to more complex regulatorynetworks involving multiple targets cooperating TFs andmiRNAs and signaling pathways displaying cross-talk viapost-tranlational modifications In this case the criticalelement is the quality and quantitiy of the available data
Insufficiency and low quality of experimental data can causeerrors in the process of model construction and overfitting inparameter estimation can lead to uncertainties in the modelpredictions We believe that the quick development of quan-titative high throughput techniques such as transcriptomicsproteomics and miRNomics will facilitate the constructionand characterization of larger miRNA-mediated regulatorynetworks
Other modeling frameworks than ODE-based modelscan be used to describe biological systems such as prob-abilistic (eg Bayesian) or logical (eg Boolean) modelsImportantly different modeling frameworks have differentproperties and perform well regarding different perspectivesand levels of mechanistic details of biochemical systems [46]For example Bayesian models are helpful in the constructionof connections in signaling networks and can reveal the mostlikely underlying structure of the network in a probabilisticmanner Boolean models use binary values (0 and 1) andlogical gates (AND NOT and OR) to describe activities ofnetwork components and the information flow among themWe believe that in the coming future hybrid models whichconsist of modeling framework and experimental techniquespecific sub-modules will provide the necessary compromisebetween quantitativequalitative accuracy and scalability forthe investigation of large biochemical networks [47]
Disclosures
The authors declare that they have no competing financialinterests
Authorsrsquo Contribution
Julio Vera and Olaf Wolkenhauer are equal contributers
Acknowledgments
The authors would like to acknowledge the funding fromthe German Federal Ministry of Education and Research(BMBF) eBio-miRSys (0316175A to Xin Lai and Julio Vera)eBio-SysMet (0316171 to OlafWolkenhauer) and Gerontosys-ROSAge (0315892A to Olaf Wolkenhauer)
References
[1] A E Pasquinelli ldquoMicroRNAs and their targets recognitionregulation and an emerging reciprocal relationshiprdquo NatureReviews Genetics vol 13 no 4 pp 271ndash282 2012
[2] J Vera X Lai U Schmitz and O Wolkenhauer ldquoMicroRNA-regulated networks the perfect storm for classical molecularbiology the ideal scenario for systems biologyrdquo Advances inExperimental Medicine and Biology vol 774 pp 55ndash76 2013
[3] S Nikolov J Vera U Schmitz and O Wolkenhauer ldquoA model-based strategy to investigate the role of microRNA regulation incancer signalling networksrdquo Theory in Biosciences vol 130 no1 pp 55ndash69 2011
14 BioMed Research International
[4] X Lai U Schmitz S K Gupta et al ldquoComputational analysisof target hub gene repression regulated by multiple and cooper-ative miRNAsrdquoNucleic Acids Research vol 40 no 18 pp 8818ndash8834 2012
[5] X Lai O Wolkenhauer and J Vera ldquoModeling miRNA regu-lation in cancer signaling systems miR-34a regulation of thep53Sirt1 signaling modulerdquoMethods in Molecular Biology vol880 pp 87ndash108 2012
[6] C Jiang Z Xuan F Zhao and M Q Zhang ldquoTRED atranscriptional regulatory element database new entries andother developmentrdquoNucleic Acids Research vol 35 supplement1 pp D137ndashD140 2007
[7] L A Bovolenta M L Acencio and N Lemke ldquoHTRIdban open-access database for experimentally verified humantranscriptional regulation interactionsrdquo BMCGenomics vol 13article 405 2012
[8] VMatys O V Kel-Margoulis E Fricke et al ldquoTRANSFAC andits module TRANSCompel transcriptional gene regulation ineukaryotesrdquo Nucleic Acids Research vol 34 database issue ppD108ndashD110 2006
[9] D Karolchik R Baertsch M Diekhans et al ldquoThe UCSCgenome browser databaserdquo Nucleic Acids Research vol 31 no1 pp 51ndash54 2003
[10] P Alexiou T Vergoulis M Gleditzsch et al ldquomiRGen 20 adatabase of microRNA genomic information and regulationrdquoNucleic Acids Research vol 38 supplement 1 pp D137ndashD1412010
[11] F Xiao Z Zuo G Cai S Kang X Gao and T Li ldquomiRecordsan integrated resource for microRNA-target interactionsrdquoNucleic Acids Research vol 37 supplement 1 pp D105ndashD1102009
[12] S-D Hsu F-M Lin W-Y Wu et al ldquomiRTarBase a databasecurates experimentally validated microRNA-target interac-tionsrdquo Nucleic Acids Research vol 39 supplement 1 pp D163ndashD169 2011
[13] P Sethupathy B Corda and A G Hatzigeorgiou ldquoTarBase acomprehensive database of experimentally supported animalmicroRNA targetsrdquo RNA vol 12 no 2 pp 192ndash197 2006
[14] H Dweep C Sticht P Pandey and N Gretz ldquomiRWalkmdashdatabase prediction of possible miRNA binding sites byldquowalkingrdquo the genes of three genomesrdquo Journal of BiomedicalInformatics vol 44 no 5 pp 839ndash847 2011
[15] T S Keshava Prasad RGoel K Kandasamy et al ldquoHumanpro-tein reference databasemdash2009 updaterdquo Nucleic Acids Researchvol 37 supplement 1 pp D767ndashD772 2009
[16] D Szklarczyk A Franceschini M Kuhn et al ldquoThe STRINGdatabase in 2011 functional interaction networks of proteinsglobally integrated and scoredrdquo Nucleic Acids Research vol 39supplement 1 pp D561ndashD568 2011
[17] D Croft G OrsquoKelly G Wu et al ldquoReactome a database ofreactions pathways and biological processesrdquo Nucleic AcidsResearch vol 39 supplement 1 pp D691ndashD697 2011
[18] N le Novere M Hucka H Mi et al ldquoThe systems biologygraphical notationrdquoNature Biotechnology vol 27 no 8 pp 735ndash741 2009
[19] A Funahashi Y Matsuoka A Jouraku M Morohashi NKikuchi and H Kitano ldquoCellDesigner 35 a versatile modelingtool for biochemical networksrdquo Proceedings of the IEEE vol 96no 8 pp 1254ndash1265 2008
[20] M S Cline M Smoot E Cerami et al ldquoIntegration ofbiological networks and gene expression data using CytoscaperdquoNature Protocols vol 2 no 10 pp 2366ndash2382 2007
[21] D-H Le andY-KKwon ldquoNetDS aCytoscape plugin to analyzethe robustness of dynamics and feedforwardfeedback loopstructures of biological networksrdquo Bioinformatics vol 27 no 19pp 2767ndash2768 2011
[22] E O Voit Computational Analysis of Biochemical SystemsA Practical Guide for Biochemists and Molecular BiologistsCambridge University Press Cambridge UK 2000
[23] S Wu S Huang J Ding et al ldquoMultiple microRNAs modulatep21Cip1Waf1 expression by directly targeting its 31015840 untranslatedregionrdquo Oncogene vol 29 no 15 pp 2302ndash2308 2010
[24] P Saeligtrom B S E Heale O Snoslashve Jr L Aagaard J Alluin andJ J Rossi ldquoDistance constraints between microRNA target sitesdictate efficacy and cooperativityrdquo Nucleic Acids Research vol35 no 7 pp 2333ndash2342 2007
[25] Y-S Jung YQian andXChen ldquoExamination of the expandingpathways for the regulation of p21 expression and activityrdquoCellular Signalling vol 22 no 7 pp 1003ndash1012 2010
[26] W Wang H Furneaux H Cheng et al ldquoHuR regulatesp21 mRNA stabilization by UV lightrdquo Molecular and CellularBiology vol 20 no 3 pp 760ndash769 2000
[27] U Wittig R Kania M Golebiewski et al ldquoSABIO-RKmdashdatabase for biochemical reaction kineticsrdquo Nucleic AcidsResearch vol 40 database issue pp D790ndashD796 2012
[28] R Milo P Jorgensen U Moran G Weber and M SpringerldquoBioNumbersmdashthe database of key numbers in molecular andcell biologyrdquo Nucleic Acids Research vol 38 supplement 1 ppD750ndashD753 2010
[29] I-C Chou and E O Voit ldquoRecent developments in parameterestimation and structure identification of biochemical andgenomic systemsrdquoMathematical Biosciences vol 219 no 2 pp57ndash83 2009
[30] A Saltelli K Chan and E M Scott Sensitivity Analysis JohnWiley amp Sons New York NY USA 1st edition 2000
[31] S Strogatz Nonlinear Dynamics and Chaos Applications toPhysics Biology Chemistry and Engineering With Applicationsto Physics Biology Chemistry and Engineering Westview PressBoulder Colo USA 2000
[32] P Zhou S Cai Z Liu and R Wang ldquoMechanisms generatingbistability and oscillations in microRNA-mediated motifsrdquoPhysical Review E vol 85 no 4 apart 1 Article ID 041916 9pages 2012
[33] S Marino I B Hogue C J Ray and D E Kirschner ldquoAmethodology for performing global uncertainty and sensitivityanalysis in systems biologyrdquo Journal of Theoretical Biology vol254 no 1 pp 178ndash196 2008
[34] L P LimNC Lau PGarrett-Engele et al ldquoMicroarray analysisshows that some microRNAs downregulate large numbers of-target mRNAsrdquo Nature vol 433 no 7027 pp 769ndash773 2005
[35] J Krutzfeldt N Rajewsky R Braich et al ldquoSilencing ofmicroRNAs in vivowith lsquoantagomirsrsquordquoNature vol 438 no 7068pp 685ndash689 2005
[36] MWPfafflGWHorgan andLDempfle ldquoRelative expressionsoftware tool (REST) for group-wise comparison and statisticalanalysis of relative expression results in real-time PCRrdquoNucleicAcids Research vol 30 no 9 p e36 2002
[37] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[38] T Abbas and A Dutta ldquop21 in cancer intricate networks andmultiple activitiesrdquo Nature Reviews Cancer vol 9 no 6 pp400ndash414 2009
[39] J Wang M Lu C Qiu and Q Cui ldquoTransmiR a transcriptionfactor microRNA regulation databaserdquo Nucleic Acids Researchvol 38 supplement 1 pp D119ndashD122 2010
BioMed Research International 15
[40] S Bandyopadhyay andM Bhattacharyya ldquoPuTmiR a databasefor extracting neighboring transcription factors of humanmicroRNAsrdquo BMC Bioinformatics vol 11 article 190 2010
[41] A Le Bechec E Portales-Casamar G Vetter et alldquoMIRNTN a framework integrating transcription factorsmicroRNAs and their targets to identify sub-network motifs ina meta-regulation network modelrdquo BMC Bioinformatics vol12 article 67 2011
[42] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyTheGene Ontology ConsortiumrdquoNature Genetics vol 25 no 1 pp 25ndash29 2000
[43] A I F Vaz and L N Vicente ldquoA particle swarm pattern searchmethod for bound constrained global optimizationrdquo Journal ofGlobal Optimization vol 39 no 2 pp 197ndash219 2007
[44] W H PressNumerical Recipes The Art of Scientific ComputingCambridge University Press Cambridge UK 2007
[45] J G Doench and P A Sharp ldquoSpecificity of microRNA targetselection in translational repressionrdquo Genes and Developmentvol 18 no 5 pp 504ndash511 2004
[46] B Kholodenko M B Yaffe and W Kolch ldquoComputationalapproaches for analyzing information flow in biological net-worksrdquo Science Signaling vol 5 no 220 2012
[47] F M Khan U Schmitz S Nikolov et al ldquoHybrid modeling ofthe crosstalk between signaling and transcriptional networksusing ordinary differential equations and multi-valued logicrdquoBiochimica et Biophysica Acta 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
BioMed Research International 9
Con
trol
miR
-298
miR
-208
am
iR-1
32m
iR-2
8-5p
miR
-125
-5p
miR
-299
-5p
miR
-345
miR
-93
miR
-423
-3p
miR
-515
-3p
miR
-363
miR
-657
miR
-639
miR
-572
miR
-654
-3p
0
05
1
15m
RNA
expr
essio
n (a
u)
Data Model Data ModelC
ontro
lm
iR-2
98m
iR-2
08a
miR
-132
miR
-28-
5pm
iR-1
25-5
pm
iR-2
99-5
pm
iR-3
45m
iR-9
3m
iR-4
23-3
pm
iR-5
15-3
pm
iR-3
63m
iR-6
57m
iR-6
39m
iR-5
72m
iR-6
54-3
p
0
02
04
06
08
1
Prot
ein
expr
essio
n (a
u)
(a)
DoxorubicinCellWell
Control miR-572 miR-93
Measure p21 expression at the time points of
miR-572+ miR-93
0 2 4 6 8 24hr
miR-572miR-93
Immunoblotting
(b)
0 4 8 12 16 20 240
4
8
12
16
20
Time (hr)
p53
Actin
Doxo (hrs)
fac
t(TF
mp2
1)
C 0 1 2 4 6 8 24
1 08 12 19 29 32 3 16
(c)
0 2 4 6 8 2402468
1012
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)0 2 4 6 8 24
Time (hr)
p21
expr
essio
n (a
u)
02468
1012
miR-572Control
02468
1012
miR-93
02468
1012
Model Data
0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24 0 2 4 6 8 24
1 23 46 107124 124 23 28 48 66 53 31 08 16 21 38 50 18 06 06 12 25 27 12
miR-572+ 93Doxo (hrs)
p21
Actin
p21
Actin
p21
Actin
p21
Actin
(d)
Figure 2 Model calibration and validation (a) Model calibration The figures show the relative change of the p21 mRNA and proteinexpression levels after overexpression of the indicated miRNAs (Model model simulation Data experimental data) These data werenormalized to the control group in which the p21 mRNA and protein expression levels were measured when the miRNAs were normallyexpressed (au arbitrary unit) (b) Experimental workflow In the experiments Sk-Mel-147 cells were seeded in six well plates Then maturemiRNA mimics were transfected individually at a concentration of 100 nM (miR-572 and miR-93) or in combination at 50 nM each (miR-572 + miR-93) After 48 hr transfection with miRNA mimics the cells were pulse treated with 250 nM doxorubicin for 1 hour after whichnormal growth medium was replenished The immunblotting were performed to measure p21 expression at 0 2 4 6 8 and 24 hr post-doxorubicin treatment (c) Temporal dynamics of p21 transcriptional function Afterdoxorubicin treatment the expression of p53 a TF ofp21 was measured using immunblotting and these data were used to characterize the transcriptional function of p21 using MATLAB linearinterpolation function (d) Model validation We measured the expression of p21 protein in response to genotoxic stress in the four scenariosas described in the main text The measured data (Data) were compared with the model simulations (Model) The figures (a) (c) and (d) areadapted from our previous publication [4]
119889 [mp21 | miR119894| miR
119895]
119889119905
= 119896co-complex
119894119895
ass sdotmp21 sdotmiR119894sdotmiR119895
minus 119896co-complex
119894119895
deg sdot [mp21 | miR119894| miR
119895]
(10)
To model stronger repression of the target gene bycooperating miRNAs we assumed a stronger association rateconstant for the complex [mp21 | miR
119894| miR
119895] which is
equal to the sum of their individual association rate constants
(119896co-complex119894119895
ass = 119896complex
119894
ass + 119896complex
119895
ass ) Similarly the degradationrates of the complexes [mp21 | miR
119894| miR
119895] were assumed
to be equal to the sum of degradation rate constants ofsingle miRNA binding complexes (119896co-complex
119894119895
deg = 119896complex
119894
deg +
119896complex
119895
deg ) However it has to be noted that these addedequations are an abstract description ofmiRNAcooperativitybecause the details of this mechanism are not yet known
To experimentally validate the capability of our model topredict the relative p21 concentrations regulated by coopera-tivemiRNAs we selectedmiR-572 andmiR-93 as a case study
10 BioMed Research International
Table 3 Initial concentrations of model variables and model parameter values Based on the experimental data the p21-targeting miRNAsverified byWu et al [23] were divided into two groups the translation repression group (marked with asterisk) and the mRNA deadenylationgroup AmiRNAwas classified into themRNA deadenylation group if its overexpression can result in 20 ormore downregulation of the p21mRNA level (ie p21mRNA levelle 08 the basal level is 1) otherwise it was classified into the translation repression group For the translationrepression group only 119896complex119894
ass was estimated and 119896complex119894deg was fixed For the other group both 119896complex119894
deg and 119896complex119894ass were estimatedThe initial
concentrations of p21 and mp21 were set to 1 and this value was used as their basal expression levels During the parameter estimation theinitial concentrations of p21-targeting miRNAs were set to 100 because in the publication [23] the expression levels of p21 and mp21 weremeasured after the individual introduction of the p21-targeting miRNAs with amount of 100 nM Due to the lack of biological information tocharacterize the transcriptional activation function (119891act) of p21 and its targeting miRNAs the corresponding functions were assumed to be1 for simplicity The data is adapted from our previous publication [4]
Initial concentration of variables and TF functionsVariable Description Initial concentration (au)p21 p21 protein 1
mp21 p21 mRNA 1
miR119894=115
p21-targeting miRNAs 100
[mp21 | miR119894=115
] Complexes formed by miR119894and mp21 0
119891act(TFmp21) p21rsquos transcriptional activation function 1119891act(TFmiR119894 (119894=115)) The transcriptional activation function of miR
1198941
Fixed parameter valuesParameter Description Value (hrminus1) Reference119896mp21syn Synthesis rate constant of mp21 01155 fixed119896mp21deg Degradation rate constant of mp21 01155 [24]119896miR119894syn (119894 = 1 15) Synthesis rate constant of miR
11989400289 fixed
119896miR119894deg (119894 = 1 15) Degradation rate constant of miR
11989400289 [25]
119896p21syn Synthesis rate constant of p21 13863 fixed119896p21deg Degradation rate constant of p21 13863 [26]
Estimated parameter values
miRNA (statevariable)
119896complex119894deg
(119894 = 1 15) (hrminus1)
119896complex119894ass
(119894 = 1 15)
(au minus1sdot hrminus1)
119865miR119894cost
(119894 = 1 15)Experimental data of p21 (protein mRNA plusmn SD)
miR-298 (miR1)lowast 01155 00254 34119890 minus 004 (016 1074 plusmn 0025)
miR-208a (miR2)lowast 01155 00041 20119890 minus 003 (051 1192 plusmn 0022)
miR-132 (miR3)lowast 01155 00275 24119890 minus 003 (015 121 plusmn 0147)
miR-28-5p (miR4)lowast 01155 00119 59119890 minus 003 (028 135 plusmn 006)
miR-125-5p (miR5)lowast 01155 00018 18119890 minus 003 (069 085 plusmn 0051)
miR-299-5p (miR6)lowast 01155 00080 18119890 minus 004 (036 095 plusmn 0038)
miR-345 (miR7)lowast 01155 00051 11119890 minus 004 (046 096 plusmn 0039)
miR-93 (miR8) 01564 00235 41119890 minus 014 (017 07776 plusmn 003)
miR-423-3p (miR9) 09118 00055 28119890 minus 009 (044 05102 plusmn 011)
miR-515-3p (miR10) 02098 00253 12119890 minus 013 (016 0616 plusmn 0037)
miR-363 (miR11) 02261 00399 22119890 minus 014 (011 056 plusmn 015)
mR-657 (miR12) 03465 00158 21119890 minus 014 (023 048 plusmn 012)
miR-639 (miR13) 04305 00327 18119890 minus 017 (013 036 plusmn 0084)
miR-572 (miR14) 03039 00360 94119890 minus 023 (012 045 plusmn 0044)
miR-654-3p (miR15) 97485 00024 30119890 minus 014 (064 063 plusmn 0053)
These two miRNAs were chosen because their predictedtarget sites in the p21 31015840 UTR are in close proximity to eachother and thereby they can induce cooperative repression onp21 as suggested in [24] The experiments were performed asfollows
(i) Melanoma cells (Sk-Mel-147) were transfected withthe mature miRNAmimics of the twomiRNAs eitherindividually (100 nM) or in combination (50 nMeach) whereas untreated cells were used as control(Figure 2(b))
BioMed Research International 11
p21
FOXF2
miR-572
Activation
Inhibition120572
Incoherent FFL + cascaded regulation
AF2
miR-93
(a)
miR-93
miR-572
TP53
MYC
p21
FOXF2
Coherent FFL + cascaded regulation
(b)
Saturation
Pulse
p21
(au
)
p21
(au
)Time (hr) Time (hr)
Saturation Pulse
Pulse
Saturation
Saturation
di
(au
)
102
101
102
100
101
10minus1
10010minus2
10minus1
102
101
100
10minus1
10minus2
FOXF2 (au) FOXF2 (au)
TP53
(au
)
1
09
08
07
06
05
04
03
02
01
10minus2 10210110010minus110minus20
AF2120572
(au
)
(c)
Figure 3 Different p21 dynamics in different network motifs We ran simulations to show the different dynamics of p21 for two differentnetwork motifs (a) and (b) Through simulations two dynamical patterns of p21 were identified saturation and pulse ((c) top) For eachnetwork motif the corresponding distributions of the two dynamical patterns were plotted ((c) bottom) For different combinations ofthe transcriptional strengths the normalized distance (119889
119894) between peaks (119901
119894) and steady states (ss
119894) of p21 is determined by the equation
119889119894= (119901119894minus ss119894)119901max 119901max = max(119901
1 119901
119899) If 119889
119894= 0 for the corresponding combination of transcriptional strengths the p21 dynamics is
saturation otherwise it is pulse The regions showing different dynamical patterns of p21 are separated using the white lines
(ii) Next the cells were treated with doxorubicina genotoxic-stress inducing agent The agent canupregulate the expression of p53 which is a knownTFof p21 and therefore it can result in the upregulationof p21 (Figure 2(c))
(iii) After doxorubicin treatment the expression levels ofp21 were measured by immunoblotting at differenttime points (0 2 4 6 8 24 hr) The p21 expressionvalues were normalized based on the p21 expression
level in the control group measured at time point 0 hr(Figure 2(d))
By doing so we obtained the p21 response after genotoxicstress in four scenarios (1) endogenous miRNA expression(Control) (2) overexpression of miR-572 (3) overexpressionof miR-93 and (4) both miRNAs moderately overexpressedThereafter we derived a model of seven ODEs based onthe original equations which was configured according tothe designed experiments making the simulation results
12 BioMed Research International
00001
0001
001
01
1Cell cycle Cell growth DNA damage Senescence Apoptosis
G1G2
MS
p21
expr
essio
n (a
u)
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
log10
wow
Figure 4 p21 expression regulated by cooperative miRNAs for different cellular processesThe associations of themiRNAs with these cellularprocesses were derived from GO terms of their TFs A miRNA was supposed to be expressed (in bold black font) in a cellular process onlyif its TF is related to the corresponding GO term of this process The p21 expression levels are computed for each process with (w)without(wo) considering the cooperative effect among the p21 targeting miRNAs
comparable with the experimental data As shown in theFigure 2(c) the simulations are in good agreement with theexperimental observations The individual overexpression ofmiR-572 or miR-93 led to the reduction of the upregulationof p21 response after genotoxic stress induction The twomiRNAs cause different degrees of repression due to theirdifferent repression efficiencies on p21 Interestingly the com-bined overexpression of both miRNAs induced the strongestdownregulation of p21 and therefore verifying the hypothesisof their cooperative regulation of p21 Above all the resultsnot only validated themodel but also demonstrated the abilityof our method to identify cooperative miRNA pairs for p21
(3) Predictive Simulations As there are abundant of networkmotifs such as FFLs in p21 regulatory network and thesenetwork motifs are important for determining p21 dynamicsit is interesting to investigate the dynamics of p21 in networkmodules where FFLs are involved To do so two networkmodules including both miR-93 and miR-572 and theirTFs were exemplified In Figure 3(a) the network modulecontains an incoherent FFL composed by AF2120572 miR-93 andp21 and a cascaded regulation in which p21 is repressed
by FOXF2 via miR-572 in Figure 3(b) the same cascadedregulation together with a coherent FFL composed of TP53MYC miR-93 and p21 forms another regulatory moduleof p21 By modulating the transcriptional strengths of thetwo miRNAs by their TFs two types of p21 dynamics wereidentified saturation and pulse (Figure 3(c) top) In theformer the p21 expression increases and reaches its steadystate at the highest level in contrast in the latter the p21expression increases to a peak and thereafter drops to asteady state at lower level For the two different networkmodules various combinations of transcriptional strengthsof the two miRNAs lead to different distributions of thetwo p21 dynamical patterns For the network module of anincoherent FFL plus the cascaded regulation (Figure 3(c)bottom left) the saturation pattern appears in two distinctregions one with weakened transcriptional strength of miR-93 and the other with enhanced transcriptional strength ofmiR-93 for the other module (Figure 3(c) bottom right)the saturation pattern only appears in the region in whichthe transcriptional strength of miR-93 is enhanced Takentogether the results showed that for the two different networkmotifs the dynamical pattern of p21 is changing according to
BioMed Research International 13
different combinations of its upstream regulators suggestingthe adaptation of p21 dynamics for different biological con-texts
Furthermore we performed a number of simulations toshow the influence of miRNA regulation on p21 expressionlevels in different cellular processes arranged in a consecutivemanner In this procedure a cell is first in the process ofcell cycle followed by proliferation (cell growth) then thecell responses to DNA damage and enters into the processof senescence and finally apoptosis is initiated As shown inFigure 4 during the cell cycle process the p21 expression islow due to the activation of most of its targeting miRNAswhen the cell starts proliferating the p21 expression declinesto an even lower level because of more activated miRNAsin this process after responding to DNA damage the p21expression soars to a high level which is caused by theactivation of its TFs like p53 and fewer expressed miRNAsalthough the p21 expression keeps at a similar level whilethe cell is undergoing senescence the expressed miRNAs aredifferent from the previous process finally the p21 expressiondecreases again to a low level due to the reemergence ofmost of its targeting miRNAs and the cell enters apoptosisInterestingly the model simulations are also consistent withexperimental observations under non-stressed condition thelow expression level of p21 is needed for cell proliferation theupregulation of p21 happens after response to DNA damagevia p53 and the increased p21 expression further results incell cycle arrest leading to senescence and apoptosis [25]Besides when considering the effect of cooperatingmiRNAsthe p21 expression levels were indistinguishable from theprevious simulation of DNA damage response However forthe other processes p21 expression levels were significantlylower compared to the simulations without considering thecooperative effect of the miRNAs Above all these resultsindicated that selective expression of cooperative miRNAscould be adopted by cells to ensure diverse expression levelsof p21 tomeet the requirements of different cellular processes
3 Conclusions and Discussion
In this paper we presented a systems biology approach com-bining data-drivenmodeling andmodel-driven experimentsto investigate the role of miRNA-mediated repression in generegulatory networksThis approach provides a systematicwayto gain a deeper understanding of the regulation of targetgenes by mutiple and cooperative miRNAs Using the regu-lation of p21 by multiple miRNAs as a case study we showedhow the ODE-basedmodel which is calibrated and validatedby means of experimental data is suitable for predicting thetemporal dynamics of molecular concentrations involved inbiochemical systems
Provided there are sufficiently rich quantitative data setsavaliable to characterize the model the use of the methodol-ogy here shown can be extended to more complex regulatorynetworks involving multiple targets cooperating TFs andmiRNAs and signaling pathways displaying cross-talk viapost-tranlational modifications In this case the criticalelement is the quality and quantitiy of the available data
Insufficiency and low quality of experimental data can causeerrors in the process of model construction and overfitting inparameter estimation can lead to uncertainties in the modelpredictions We believe that the quick development of quan-titative high throughput techniques such as transcriptomicsproteomics and miRNomics will facilitate the constructionand characterization of larger miRNA-mediated regulatorynetworks
Other modeling frameworks than ODE-based modelscan be used to describe biological systems such as prob-abilistic (eg Bayesian) or logical (eg Boolean) modelsImportantly different modeling frameworks have differentproperties and perform well regarding different perspectivesand levels of mechanistic details of biochemical systems [46]For example Bayesian models are helpful in the constructionof connections in signaling networks and can reveal the mostlikely underlying structure of the network in a probabilisticmanner Boolean models use binary values (0 and 1) andlogical gates (AND NOT and OR) to describe activities ofnetwork components and the information flow among themWe believe that in the coming future hybrid models whichconsist of modeling framework and experimental techniquespecific sub-modules will provide the necessary compromisebetween quantitativequalitative accuracy and scalability forthe investigation of large biochemical networks [47]
Disclosures
The authors declare that they have no competing financialinterests
Authorsrsquo Contribution
Julio Vera and Olaf Wolkenhauer are equal contributers
Acknowledgments
The authors would like to acknowledge the funding fromthe German Federal Ministry of Education and Research(BMBF) eBio-miRSys (0316175A to Xin Lai and Julio Vera)eBio-SysMet (0316171 to OlafWolkenhauer) and Gerontosys-ROSAge (0315892A to Olaf Wolkenhauer)
References
[1] A E Pasquinelli ldquoMicroRNAs and their targets recognitionregulation and an emerging reciprocal relationshiprdquo NatureReviews Genetics vol 13 no 4 pp 271ndash282 2012
[2] J Vera X Lai U Schmitz and O Wolkenhauer ldquoMicroRNA-regulated networks the perfect storm for classical molecularbiology the ideal scenario for systems biologyrdquo Advances inExperimental Medicine and Biology vol 774 pp 55ndash76 2013
[3] S Nikolov J Vera U Schmitz and O Wolkenhauer ldquoA model-based strategy to investigate the role of microRNA regulation incancer signalling networksrdquo Theory in Biosciences vol 130 no1 pp 55ndash69 2011
14 BioMed Research International
[4] X Lai U Schmitz S K Gupta et al ldquoComputational analysisof target hub gene repression regulated by multiple and cooper-ative miRNAsrdquoNucleic Acids Research vol 40 no 18 pp 8818ndash8834 2012
[5] X Lai O Wolkenhauer and J Vera ldquoModeling miRNA regu-lation in cancer signaling systems miR-34a regulation of thep53Sirt1 signaling modulerdquoMethods in Molecular Biology vol880 pp 87ndash108 2012
[6] C Jiang Z Xuan F Zhao and M Q Zhang ldquoTRED atranscriptional regulatory element database new entries andother developmentrdquoNucleic Acids Research vol 35 supplement1 pp D137ndashD140 2007
[7] L A Bovolenta M L Acencio and N Lemke ldquoHTRIdban open-access database for experimentally verified humantranscriptional regulation interactionsrdquo BMCGenomics vol 13article 405 2012
[8] VMatys O V Kel-Margoulis E Fricke et al ldquoTRANSFAC andits module TRANSCompel transcriptional gene regulation ineukaryotesrdquo Nucleic Acids Research vol 34 database issue ppD108ndashD110 2006
[9] D Karolchik R Baertsch M Diekhans et al ldquoThe UCSCgenome browser databaserdquo Nucleic Acids Research vol 31 no1 pp 51ndash54 2003
[10] P Alexiou T Vergoulis M Gleditzsch et al ldquomiRGen 20 adatabase of microRNA genomic information and regulationrdquoNucleic Acids Research vol 38 supplement 1 pp D137ndashD1412010
[11] F Xiao Z Zuo G Cai S Kang X Gao and T Li ldquomiRecordsan integrated resource for microRNA-target interactionsrdquoNucleic Acids Research vol 37 supplement 1 pp D105ndashD1102009
[12] S-D Hsu F-M Lin W-Y Wu et al ldquomiRTarBase a databasecurates experimentally validated microRNA-target interac-tionsrdquo Nucleic Acids Research vol 39 supplement 1 pp D163ndashD169 2011
[13] P Sethupathy B Corda and A G Hatzigeorgiou ldquoTarBase acomprehensive database of experimentally supported animalmicroRNA targetsrdquo RNA vol 12 no 2 pp 192ndash197 2006
[14] H Dweep C Sticht P Pandey and N Gretz ldquomiRWalkmdashdatabase prediction of possible miRNA binding sites byldquowalkingrdquo the genes of three genomesrdquo Journal of BiomedicalInformatics vol 44 no 5 pp 839ndash847 2011
[15] T S Keshava Prasad RGoel K Kandasamy et al ldquoHumanpro-tein reference databasemdash2009 updaterdquo Nucleic Acids Researchvol 37 supplement 1 pp D767ndashD772 2009
[16] D Szklarczyk A Franceschini M Kuhn et al ldquoThe STRINGdatabase in 2011 functional interaction networks of proteinsglobally integrated and scoredrdquo Nucleic Acids Research vol 39supplement 1 pp D561ndashD568 2011
[17] D Croft G OrsquoKelly G Wu et al ldquoReactome a database ofreactions pathways and biological processesrdquo Nucleic AcidsResearch vol 39 supplement 1 pp D691ndashD697 2011
[18] N le Novere M Hucka H Mi et al ldquoThe systems biologygraphical notationrdquoNature Biotechnology vol 27 no 8 pp 735ndash741 2009
[19] A Funahashi Y Matsuoka A Jouraku M Morohashi NKikuchi and H Kitano ldquoCellDesigner 35 a versatile modelingtool for biochemical networksrdquo Proceedings of the IEEE vol 96no 8 pp 1254ndash1265 2008
[20] M S Cline M Smoot E Cerami et al ldquoIntegration ofbiological networks and gene expression data using CytoscaperdquoNature Protocols vol 2 no 10 pp 2366ndash2382 2007
[21] D-H Le andY-KKwon ldquoNetDS aCytoscape plugin to analyzethe robustness of dynamics and feedforwardfeedback loopstructures of biological networksrdquo Bioinformatics vol 27 no 19pp 2767ndash2768 2011
[22] E O Voit Computational Analysis of Biochemical SystemsA Practical Guide for Biochemists and Molecular BiologistsCambridge University Press Cambridge UK 2000
[23] S Wu S Huang J Ding et al ldquoMultiple microRNAs modulatep21Cip1Waf1 expression by directly targeting its 31015840 untranslatedregionrdquo Oncogene vol 29 no 15 pp 2302ndash2308 2010
[24] P Saeligtrom B S E Heale O Snoslashve Jr L Aagaard J Alluin andJ J Rossi ldquoDistance constraints between microRNA target sitesdictate efficacy and cooperativityrdquo Nucleic Acids Research vol35 no 7 pp 2333ndash2342 2007
[25] Y-S Jung YQian andXChen ldquoExamination of the expandingpathways for the regulation of p21 expression and activityrdquoCellular Signalling vol 22 no 7 pp 1003ndash1012 2010
[26] W Wang H Furneaux H Cheng et al ldquoHuR regulatesp21 mRNA stabilization by UV lightrdquo Molecular and CellularBiology vol 20 no 3 pp 760ndash769 2000
[27] U Wittig R Kania M Golebiewski et al ldquoSABIO-RKmdashdatabase for biochemical reaction kineticsrdquo Nucleic AcidsResearch vol 40 database issue pp D790ndashD796 2012
[28] R Milo P Jorgensen U Moran G Weber and M SpringerldquoBioNumbersmdashthe database of key numbers in molecular andcell biologyrdquo Nucleic Acids Research vol 38 supplement 1 ppD750ndashD753 2010
[29] I-C Chou and E O Voit ldquoRecent developments in parameterestimation and structure identification of biochemical andgenomic systemsrdquoMathematical Biosciences vol 219 no 2 pp57ndash83 2009
[30] A Saltelli K Chan and E M Scott Sensitivity Analysis JohnWiley amp Sons New York NY USA 1st edition 2000
[31] S Strogatz Nonlinear Dynamics and Chaos Applications toPhysics Biology Chemistry and Engineering With Applicationsto Physics Biology Chemistry and Engineering Westview PressBoulder Colo USA 2000
[32] P Zhou S Cai Z Liu and R Wang ldquoMechanisms generatingbistability and oscillations in microRNA-mediated motifsrdquoPhysical Review E vol 85 no 4 apart 1 Article ID 041916 9pages 2012
[33] S Marino I B Hogue C J Ray and D E Kirschner ldquoAmethodology for performing global uncertainty and sensitivityanalysis in systems biologyrdquo Journal of Theoretical Biology vol254 no 1 pp 178ndash196 2008
[34] L P LimNC Lau PGarrett-Engele et al ldquoMicroarray analysisshows that some microRNAs downregulate large numbers of-target mRNAsrdquo Nature vol 433 no 7027 pp 769ndash773 2005
[35] J Krutzfeldt N Rajewsky R Braich et al ldquoSilencing ofmicroRNAs in vivowith lsquoantagomirsrsquordquoNature vol 438 no 7068pp 685ndash689 2005
[36] MWPfafflGWHorgan andLDempfle ldquoRelative expressionsoftware tool (REST) for group-wise comparison and statisticalanalysis of relative expression results in real-time PCRrdquoNucleicAcids Research vol 30 no 9 p e36 2002
[37] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[38] T Abbas and A Dutta ldquop21 in cancer intricate networks andmultiple activitiesrdquo Nature Reviews Cancer vol 9 no 6 pp400ndash414 2009
[39] J Wang M Lu C Qiu and Q Cui ldquoTransmiR a transcriptionfactor microRNA regulation databaserdquo Nucleic Acids Researchvol 38 supplement 1 pp D119ndashD122 2010
BioMed Research International 15
[40] S Bandyopadhyay andM Bhattacharyya ldquoPuTmiR a databasefor extracting neighboring transcription factors of humanmicroRNAsrdquo BMC Bioinformatics vol 11 article 190 2010
[41] A Le Bechec E Portales-Casamar G Vetter et alldquoMIRNTN a framework integrating transcription factorsmicroRNAs and their targets to identify sub-network motifs ina meta-regulation network modelrdquo BMC Bioinformatics vol12 article 67 2011
[42] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyTheGene Ontology ConsortiumrdquoNature Genetics vol 25 no 1 pp 25ndash29 2000
[43] A I F Vaz and L N Vicente ldquoA particle swarm pattern searchmethod for bound constrained global optimizationrdquo Journal ofGlobal Optimization vol 39 no 2 pp 197ndash219 2007
[44] W H PressNumerical Recipes The Art of Scientific ComputingCambridge University Press Cambridge UK 2007
[45] J G Doench and P A Sharp ldquoSpecificity of microRNA targetselection in translational repressionrdquo Genes and Developmentvol 18 no 5 pp 504ndash511 2004
[46] B Kholodenko M B Yaffe and W Kolch ldquoComputationalapproaches for analyzing information flow in biological net-worksrdquo Science Signaling vol 5 no 220 2012
[47] F M Khan U Schmitz S Nikolov et al ldquoHybrid modeling ofthe crosstalk between signaling and transcriptional networksusing ordinary differential equations and multi-valued logicrdquoBiochimica et Biophysica Acta 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
10 BioMed Research International
Table 3 Initial concentrations of model variables and model parameter values Based on the experimental data the p21-targeting miRNAsverified byWu et al [23] were divided into two groups the translation repression group (marked with asterisk) and the mRNA deadenylationgroup AmiRNAwas classified into themRNA deadenylation group if its overexpression can result in 20 ormore downregulation of the p21mRNA level (ie p21mRNA levelle 08 the basal level is 1) otherwise it was classified into the translation repression group For the translationrepression group only 119896complex119894
ass was estimated and 119896complex119894deg was fixed For the other group both 119896complex119894
deg and 119896complex119894ass were estimatedThe initial
concentrations of p21 and mp21 were set to 1 and this value was used as their basal expression levels During the parameter estimation theinitial concentrations of p21-targeting miRNAs were set to 100 because in the publication [23] the expression levels of p21 and mp21 weremeasured after the individual introduction of the p21-targeting miRNAs with amount of 100 nM Due to the lack of biological information tocharacterize the transcriptional activation function (119891act) of p21 and its targeting miRNAs the corresponding functions were assumed to be1 for simplicity The data is adapted from our previous publication [4]
Initial concentration of variables and TF functionsVariable Description Initial concentration (au)p21 p21 protein 1
mp21 p21 mRNA 1
miR119894=115
p21-targeting miRNAs 100
[mp21 | miR119894=115
] Complexes formed by miR119894and mp21 0
119891act(TFmp21) p21rsquos transcriptional activation function 1119891act(TFmiR119894 (119894=115)) The transcriptional activation function of miR
1198941
Fixed parameter valuesParameter Description Value (hrminus1) Reference119896mp21syn Synthesis rate constant of mp21 01155 fixed119896mp21deg Degradation rate constant of mp21 01155 [24]119896miR119894syn (119894 = 1 15) Synthesis rate constant of miR
11989400289 fixed
119896miR119894deg (119894 = 1 15) Degradation rate constant of miR
11989400289 [25]
119896p21syn Synthesis rate constant of p21 13863 fixed119896p21deg Degradation rate constant of p21 13863 [26]
Estimated parameter values
miRNA (statevariable)
119896complex119894deg
(119894 = 1 15) (hrminus1)
119896complex119894ass
(119894 = 1 15)
(au minus1sdot hrminus1)
119865miR119894cost
(119894 = 1 15)Experimental data of p21 (protein mRNA plusmn SD)
miR-298 (miR1)lowast 01155 00254 34119890 minus 004 (016 1074 plusmn 0025)
miR-208a (miR2)lowast 01155 00041 20119890 minus 003 (051 1192 plusmn 0022)
miR-132 (miR3)lowast 01155 00275 24119890 minus 003 (015 121 plusmn 0147)
miR-28-5p (miR4)lowast 01155 00119 59119890 minus 003 (028 135 plusmn 006)
miR-125-5p (miR5)lowast 01155 00018 18119890 minus 003 (069 085 plusmn 0051)
miR-299-5p (miR6)lowast 01155 00080 18119890 minus 004 (036 095 plusmn 0038)
miR-345 (miR7)lowast 01155 00051 11119890 minus 004 (046 096 plusmn 0039)
miR-93 (miR8) 01564 00235 41119890 minus 014 (017 07776 plusmn 003)
miR-423-3p (miR9) 09118 00055 28119890 minus 009 (044 05102 plusmn 011)
miR-515-3p (miR10) 02098 00253 12119890 minus 013 (016 0616 plusmn 0037)
miR-363 (miR11) 02261 00399 22119890 minus 014 (011 056 plusmn 015)
mR-657 (miR12) 03465 00158 21119890 minus 014 (023 048 plusmn 012)
miR-639 (miR13) 04305 00327 18119890 minus 017 (013 036 plusmn 0084)
miR-572 (miR14) 03039 00360 94119890 minus 023 (012 045 plusmn 0044)
miR-654-3p (miR15) 97485 00024 30119890 minus 014 (064 063 plusmn 0053)
These two miRNAs were chosen because their predictedtarget sites in the p21 31015840 UTR are in close proximity to eachother and thereby they can induce cooperative repression onp21 as suggested in [24] The experiments were performed asfollows
(i) Melanoma cells (Sk-Mel-147) were transfected withthe mature miRNAmimics of the twomiRNAs eitherindividually (100 nM) or in combination (50 nMeach) whereas untreated cells were used as control(Figure 2(b))
BioMed Research International 11
p21
FOXF2
miR-572
Activation
Inhibition120572
Incoherent FFL + cascaded regulation
AF2
miR-93
(a)
miR-93
miR-572
TP53
MYC
p21
FOXF2
Coherent FFL + cascaded regulation
(b)
Saturation
Pulse
p21
(au
)
p21
(au
)Time (hr) Time (hr)
Saturation Pulse
Pulse
Saturation
Saturation
di
(au
)
102
101
102
100
101
10minus1
10010minus2
10minus1
102
101
100
10minus1
10minus2
FOXF2 (au) FOXF2 (au)
TP53
(au
)
1
09
08
07
06
05
04
03
02
01
10minus2 10210110010minus110minus20
AF2120572
(au
)
(c)
Figure 3 Different p21 dynamics in different network motifs We ran simulations to show the different dynamics of p21 for two differentnetwork motifs (a) and (b) Through simulations two dynamical patterns of p21 were identified saturation and pulse ((c) top) For eachnetwork motif the corresponding distributions of the two dynamical patterns were plotted ((c) bottom) For different combinations ofthe transcriptional strengths the normalized distance (119889
119894) between peaks (119901
119894) and steady states (ss
119894) of p21 is determined by the equation
119889119894= (119901119894minus ss119894)119901max 119901max = max(119901
1 119901
119899) If 119889
119894= 0 for the corresponding combination of transcriptional strengths the p21 dynamics is
saturation otherwise it is pulse The regions showing different dynamical patterns of p21 are separated using the white lines
(ii) Next the cells were treated with doxorubicina genotoxic-stress inducing agent The agent canupregulate the expression of p53 which is a knownTFof p21 and therefore it can result in the upregulationof p21 (Figure 2(c))
(iii) After doxorubicin treatment the expression levels ofp21 were measured by immunoblotting at differenttime points (0 2 4 6 8 24 hr) The p21 expressionvalues were normalized based on the p21 expression
level in the control group measured at time point 0 hr(Figure 2(d))
By doing so we obtained the p21 response after genotoxicstress in four scenarios (1) endogenous miRNA expression(Control) (2) overexpression of miR-572 (3) overexpressionof miR-93 and (4) both miRNAs moderately overexpressedThereafter we derived a model of seven ODEs based onthe original equations which was configured according tothe designed experiments making the simulation results
12 BioMed Research International
00001
0001
001
01
1Cell cycle Cell growth DNA damage Senescence Apoptosis
G1G2
MS
p21
expr
essio
n (a
u)
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
log10
wow
Figure 4 p21 expression regulated by cooperative miRNAs for different cellular processesThe associations of themiRNAs with these cellularprocesses were derived from GO terms of their TFs A miRNA was supposed to be expressed (in bold black font) in a cellular process onlyif its TF is related to the corresponding GO term of this process The p21 expression levels are computed for each process with (w)without(wo) considering the cooperative effect among the p21 targeting miRNAs
comparable with the experimental data As shown in theFigure 2(c) the simulations are in good agreement with theexperimental observations The individual overexpression ofmiR-572 or miR-93 led to the reduction of the upregulationof p21 response after genotoxic stress induction The twomiRNAs cause different degrees of repression due to theirdifferent repression efficiencies on p21 Interestingly the com-bined overexpression of both miRNAs induced the strongestdownregulation of p21 and therefore verifying the hypothesisof their cooperative regulation of p21 Above all the resultsnot only validated themodel but also demonstrated the abilityof our method to identify cooperative miRNA pairs for p21
(3) Predictive Simulations As there are abundant of networkmotifs such as FFLs in p21 regulatory network and thesenetwork motifs are important for determining p21 dynamicsit is interesting to investigate the dynamics of p21 in networkmodules where FFLs are involved To do so two networkmodules including both miR-93 and miR-572 and theirTFs were exemplified In Figure 3(a) the network modulecontains an incoherent FFL composed by AF2120572 miR-93 andp21 and a cascaded regulation in which p21 is repressed
by FOXF2 via miR-572 in Figure 3(b) the same cascadedregulation together with a coherent FFL composed of TP53MYC miR-93 and p21 forms another regulatory moduleof p21 By modulating the transcriptional strengths of thetwo miRNAs by their TFs two types of p21 dynamics wereidentified saturation and pulse (Figure 3(c) top) In theformer the p21 expression increases and reaches its steadystate at the highest level in contrast in the latter the p21expression increases to a peak and thereafter drops to asteady state at lower level For the two different networkmodules various combinations of transcriptional strengthsof the two miRNAs lead to different distributions of thetwo p21 dynamical patterns For the network module of anincoherent FFL plus the cascaded regulation (Figure 3(c)bottom left) the saturation pattern appears in two distinctregions one with weakened transcriptional strength of miR-93 and the other with enhanced transcriptional strength ofmiR-93 for the other module (Figure 3(c) bottom right)the saturation pattern only appears in the region in whichthe transcriptional strength of miR-93 is enhanced Takentogether the results showed that for the two different networkmotifs the dynamical pattern of p21 is changing according to
BioMed Research International 13
different combinations of its upstream regulators suggestingthe adaptation of p21 dynamics for different biological con-texts
Furthermore we performed a number of simulations toshow the influence of miRNA regulation on p21 expressionlevels in different cellular processes arranged in a consecutivemanner In this procedure a cell is first in the process ofcell cycle followed by proliferation (cell growth) then thecell responses to DNA damage and enters into the processof senescence and finally apoptosis is initiated As shown inFigure 4 during the cell cycle process the p21 expression islow due to the activation of most of its targeting miRNAswhen the cell starts proliferating the p21 expression declinesto an even lower level because of more activated miRNAsin this process after responding to DNA damage the p21expression soars to a high level which is caused by theactivation of its TFs like p53 and fewer expressed miRNAsalthough the p21 expression keeps at a similar level whilethe cell is undergoing senescence the expressed miRNAs aredifferent from the previous process finally the p21 expressiondecreases again to a low level due to the reemergence ofmost of its targeting miRNAs and the cell enters apoptosisInterestingly the model simulations are also consistent withexperimental observations under non-stressed condition thelow expression level of p21 is needed for cell proliferation theupregulation of p21 happens after response to DNA damagevia p53 and the increased p21 expression further results incell cycle arrest leading to senescence and apoptosis [25]Besides when considering the effect of cooperatingmiRNAsthe p21 expression levels were indistinguishable from theprevious simulation of DNA damage response However forthe other processes p21 expression levels were significantlylower compared to the simulations without considering thecooperative effect of the miRNAs Above all these resultsindicated that selective expression of cooperative miRNAscould be adopted by cells to ensure diverse expression levelsof p21 tomeet the requirements of different cellular processes
3 Conclusions and Discussion
In this paper we presented a systems biology approach com-bining data-drivenmodeling andmodel-driven experimentsto investigate the role of miRNA-mediated repression in generegulatory networksThis approach provides a systematicwayto gain a deeper understanding of the regulation of targetgenes by mutiple and cooperative miRNAs Using the regu-lation of p21 by multiple miRNAs as a case study we showedhow the ODE-basedmodel which is calibrated and validatedby means of experimental data is suitable for predicting thetemporal dynamics of molecular concentrations involved inbiochemical systems
Provided there are sufficiently rich quantitative data setsavaliable to characterize the model the use of the methodol-ogy here shown can be extended to more complex regulatorynetworks involving multiple targets cooperating TFs andmiRNAs and signaling pathways displaying cross-talk viapost-tranlational modifications In this case the criticalelement is the quality and quantitiy of the available data
Insufficiency and low quality of experimental data can causeerrors in the process of model construction and overfitting inparameter estimation can lead to uncertainties in the modelpredictions We believe that the quick development of quan-titative high throughput techniques such as transcriptomicsproteomics and miRNomics will facilitate the constructionand characterization of larger miRNA-mediated regulatorynetworks
Other modeling frameworks than ODE-based modelscan be used to describe biological systems such as prob-abilistic (eg Bayesian) or logical (eg Boolean) modelsImportantly different modeling frameworks have differentproperties and perform well regarding different perspectivesand levels of mechanistic details of biochemical systems [46]For example Bayesian models are helpful in the constructionof connections in signaling networks and can reveal the mostlikely underlying structure of the network in a probabilisticmanner Boolean models use binary values (0 and 1) andlogical gates (AND NOT and OR) to describe activities ofnetwork components and the information flow among themWe believe that in the coming future hybrid models whichconsist of modeling framework and experimental techniquespecific sub-modules will provide the necessary compromisebetween quantitativequalitative accuracy and scalability forthe investigation of large biochemical networks [47]
Disclosures
The authors declare that they have no competing financialinterests
Authorsrsquo Contribution
Julio Vera and Olaf Wolkenhauer are equal contributers
Acknowledgments
The authors would like to acknowledge the funding fromthe German Federal Ministry of Education and Research(BMBF) eBio-miRSys (0316175A to Xin Lai and Julio Vera)eBio-SysMet (0316171 to OlafWolkenhauer) and Gerontosys-ROSAge (0315892A to Olaf Wolkenhauer)
References
[1] A E Pasquinelli ldquoMicroRNAs and their targets recognitionregulation and an emerging reciprocal relationshiprdquo NatureReviews Genetics vol 13 no 4 pp 271ndash282 2012
[2] J Vera X Lai U Schmitz and O Wolkenhauer ldquoMicroRNA-regulated networks the perfect storm for classical molecularbiology the ideal scenario for systems biologyrdquo Advances inExperimental Medicine and Biology vol 774 pp 55ndash76 2013
[3] S Nikolov J Vera U Schmitz and O Wolkenhauer ldquoA model-based strategy to investigate the role of microRNA regulation incancer signalling networksrdquo Theory in Biosciences vol 130 no1 pp 55ndash69 2011
14 BioMed Research International
[4] X Lai U Schmitz S K Gupta et al ldquoComputational analysisof target hub gene repression regulated by multiple and cooper-ative miRNAsrdquoNucleic Acids Research vol 40 no 18 pp 8818ndash8834 2012
[5] X Lai O Wolkenhauer and J Vera ldquoModeling miRNA regu-lation in cancer signaling systems miR-34a regulation of thep53Sirt1 signaling modulerdquoMethods in Molecular Biology vol880 pp 87ndash108 2012
[6] C Jiang Z Xuan F Zhao and M Q Zhang ldquoTRED atranscriptional regulatory element database new entries andother developmentrdquoNucleic Acids Research vol 35 supplement1 pp D137ndashD140 2007
[7] L A Bovolenta M L Acencio and N Lemke ldquoHTRIdban open-access database for experimentally verified humantranscriptional regulation interactionsrdquo BMCGenomics vol 13article 405 2012
[8] VMatys O V Kel-Margoulis E Fricke et al ldquoTRANSFAC andits module TRANSCompel transcriptional gene regulation ineukaryotesrdquo Nucleic Acids Research vol 34 database issue ppD108ndashD110 2006
[9] D Karolchik R Baertsch M Diekhans et al ldquoThe UCSCgenome browser databaserdquo Nucleic Acids Research vol 31 no1 pp 51ndash54 2003
[10] P Alexiou T Vergoulis M Gleditzsch et al ldquomiRGen 20 adatabase of microRNA genomic information and regulationrdquoNucleic Acids Research vol 38 supplement 1 pp D137ndashD1412010
[11] F Xiao Z Zuo G Cai S Kang X Gao and T Li ldquomiRecordsan integrated resource for microRNA-target interactionsrdquoNucleic Acids Research vol 37 supplement 1 pp D105ndashD1102009
[12] S-D Hsu F-M Lin W-Y Wu et al ldquomiRTarBase a databasecurates experimentally validated microRNA-target interac-tionsrdquo Nucleic Acids Research vol 39 supplement 1 pp D163ndashD169 2011
[13] P Sethupathy B Corda and A G Hatzigeorgiou ldquoTarBase acomprehensive database of experimentally supported animalmicroRNA targetsrdquo RNA vol 12 no 2 pp 192ndash197 2006
[14] H Dweep C Sticht P Pandey and N Gretz ldquomiRWalkmdashdatabase prediction of possible miRNA binding sites byldquowalkingrdquo the genes of three genomesrdquo Journal of BiomedicalInformatics vol 44 no 5 pp 839ndash847 2011
[15] T S Keshava Prasad RGoel K Kandasamy et al ldquoHumanpro-tein reference databasemdash2009 updaterdquo Nucleic Acids Researchvol 37 supplement 1 pp D767ndashD772 2009
[16] D Szklarczyk A Franceschini M Kuhn et al ldquoThe STRINGdatabase in 2011 functional interaction networks of proteinsglobally integrated and scoredrdquo Nucleic Acids Research vol 39supplement 1 pp D561ndashD568 2011
[17] D Croft G OrsquoKelly G Wu et al ldquoReactome a database ofreactions pathways and biological processesrdquo Nucleic AcidsResearch vol 39 supplement 1 pp D691ndashD697 2011
[18] N le Novere M Hucka H Mi et al ldquoThe systems biologygraphical notationrdquoNature Biotechnology vol 27 no 8 pp 735ndash741 2009
[19] A Funahashi Y Matsuoka A Jouraku M Morohashi NKikuchi and H Kitano ldquoCellDesigner 35 a versatile modelingtool for biochemical networksrdquo Proceedings of the IEEE vol 96no 8 pp 1254ndash1265 2008
[20] M S Cline M Smoot E Cerami et al ldquoIntegration ofbiological networks and gene expression data using CytoscaperdquoNature Protocols vol 2 no 10 pp 2366ndash2382 2007
[21] D-H Le andY-KKwon ldquoNetDS aCytoscape plugin to analyzethe robustness of dynamics and feedforwardfeedback loopstructures of biological networksrdquo Bioinformatics vol 27 no 19pp 2767ndash2768 2011
[22] E O Voit Computational Analysis of Biochemical SystemsA Practical Guide for Biochemists and Molecular BiologistsCambridge University Press Cambridge UK 2000
[23] S Wu S Huang J Ding et al ldquoMultiple microRNAs modulatep21Cip1Waf1 expression by directly targeting its 31015840 untranslatedregionrdquo Oncogene vol 29 no 15 pp 2302ndash2308 2010
[24] P Saeligtrom B S E Heale O Snoslashve Jr L Aagaard J Alluin andJ J Rossi ldquoDistance constraints between microRNA target sitesdictate efficacy and cooperativityrdquo Nucleic Acids Research vol35 no 7 pp 2333ndash2342 2007
[25] Y-S Jung YQian andXChen ldquoExamination of the expandingpathways for the regulation of p21 expression and activityrdquoCellular Signalling vol 22 no 7 pp 1003ndash1012 2010
[26] W Wang H Furneaux H Cheng et al ldquoHuR regulatesp21 mRNA stabilization by UV lightrdquo Molecular and CellularBiology vol 20 no 3 pp 760ndash769 2000
[27] U Wittig R Kania M Golebiewski et al ldquoSABIO-RKmdashdatabase for biochemical reaction kineticsrdquo Nucleic AcidsResearch vol 40 database issue pp D790ndashD796 2012
[28] R Milo P Jorgensen U Moran G Weber and M SpringerldquoBioNumbersmdashthe database of key numbers in molecular andcell biologyrdquo Nucleic Acids Research vol 38 supplement 1 ppD750ndashD753 2010
[29] I-C Chou and E O Voit ldquoRecent developments in parameterestimation and structure identification of biochemical andgenomic systemsrdquoMathematical Biosciences vol 219 no 2 pp57ndash83 2009
[30] A Saltelli K Chan and E M Scott Sensitivity Analysis JohnWiley amp Sons New York NY USA 1st edition 2000
[31] S Strogatz Nonlinear Dynamics and Chaos Applications toPhysics Biology Chemistry and Engineering With Applicationsto Physics Biology Chemistry and Engineering Westview PressBoulder Colo USA 2000
[32] P Zhou S Cai Z Liu and R Wang ldquoMechanisms generatingbistability and oscillations in microRNA-mediated motifsrdquoPhysical Review E vol 85 no 4 apart 1 Article ID 041916 9pages 2012
[33] S Marino I B Hogue C J Ray and D E Kirschner ldquoAmethodology for performing global uncertainty and sensitivityanalysis in systems biologyrdquo Journal of Theoretical Biology vol254 no 1 pp 178ndash196 2008
[34] L P LimNC Lau PGarrett-Engele et al ldquoMicroarray analysisshows that some microRNAs downregulate large numbers of-target mRNAsrdquo Nature vol 433 no 7027 pp 769ndash773 2005
[35] J Krutzfeldt N Rajewsky R Braich et al ldquoSilencing ofmicroRNAs in vivowith lsquoantagomirsrsquordquoNature vol 438 no 7068pp 685ndash689 2005
[36] MWPfafflGWHorgan andLDempfle ldquoRelative expressionsoftware tool (REST) for group-wise comparison and statisticalanalysis of relative expression results in real-time PCRrdquoNucleicAcids Research vol 30 no 9 p e36 2002
[37] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[38] T Abbas and A Dutta ldquop21 in cancer intricate networks andmultiple activitiesrdquo Nature Reviews Cancer vol 9 no 6 pp400ndash414 2009
[39] J Wang M Lu C Qiu and Q Cui ldquoTransmiR a transcriptionfactor microRNA regulation databaserdquo Nucleic Acids Researchvol 38 supplement 1 pp D119ndashD122 2010
BioMed Research International 15
[40] S Bandyopadhyay andM Bhattacharyya ldquoPuTmiR a databasefor extracting neighboring transcription factors of humanmicroRNAsrdquo BMC Bioinformatics vol 11 article 190 2010
[41] A Le Bechec E Portales-Casamar G Vetter et alldquoMIRNTN a framework integrating transcription factorsmicroRNAs and their targets to identify sub-network motifs ina meta-regulation network modelrdquo BMC Bioinformatics vol12 article 67 2011
[42] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyTheGene Ontology ConsortiumrdquoNature Genetics vol 25 no 1 pp 25ndash29 2000
[43] A I F Vaz and L N Vicente ldquoA particle swarm pattern searchmethod for bound constrained global optimizationrdquo Journal ofGlobal Optimization vol 39 no 2 pp 197ndash219 2007
[44] W H PressNumerical Recipes The Art of Scientific ComputingCambridge University Press Cambridge UK 2007
[45] J G Doench and P A Sharp ldquoSpecificity of microRNA targetselection in translational repressionrdquo Genes and Developmentvol 18 no 5 pp 504ndash511 2004
[46] B Kholodenko M B Yaffe and W Kolch ldquoComputationalapproaches for analyzing information flow in biological net-worksrdquo Science Signaling vol 5 no 220 2012
[47] F M Khan U Schmitz S Nikolov et al ldquoHybrid modeling ofthe crosstalk between signaling and transcriptional networksusing ordinary differential equations and multi-valued logicrdquoBiochimica et Biophysica Acta 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
BioMed Research International 11
p21
FOXF2
miR-572
Activation
Inhibition120572
Incoherent FFL + cascaded regulation
AF2
miR-93
(a)
miR-93
miR-572
TP53
MYC
p21
FOXF2
Coherent FFL + cascaded regulation
(b)
Saturation
Pulse
p21
(au
)
p21
(au
)Time (hr) Time (hr)
Saturation Pulse
Pulse
Saturation
Saturation
di
(au
)
102
101
102
100
101
10minus1
10010minus2
10minus1
102
101
100
10minus1
10minus2
FOXF2 (au) FOXF2 (au)
TP53
(au
)
1
09
08
07
06
05
04
03
02
01
10minus2 10210110010minus110minus20
AF2120572
(au
)
(c)
Figure 3 Different p21 dynamics in different network motifs We ran simulations to show the different dynamics of p21 for two differentnetwork motifs (a) and (b) Through simulations two dynamical patterns of p21 were identified saturation and pulse ((c) top) For eachnetwork motif the corresponding distributions of the two dynamical patterns were plotted ((c) bottom) For different combinations ofthe transcriptional strengths the normalized distance (119889
119894) between peaks (119901
119894) and steady states (ss
119894) of p21 is determined by the equation
119889119894= (119901119894minus ss119894)119901max 119901max = max(119901
1 119901
119899) If 119889
119894= 0 for the corresponding combination of transcriptional strengths the p21 dynamics is
saturation otherwise it is pulse The regions showing different dynamical patterns of p21 are separated using the white lines
(ii) Next the cells were treated with doxorubicina genotoxic-stress inducing agent The agent canupregulate the expression of p53 which is a knownTFof p21 and therefore it can result in the upregulationof p21 (Figure 2(c))
(iii) After doxorubicin treatment the expression levels ofp21 were measured by immunoblotting at differenttime points (0 2 4 6 8 24 hr) The p21 expressionvalues were normalized based on the p21 expression
level in the control group measured at time point 0 hr(Figure 2(d))
By doing so we obtained the p21 response after genotoxicstress in four scenarios (1) endogenous miRNA expression(Control) (2) overexpression of miR-572 (3) overexpressionof miR-93 and (4) both miRNAs moderately overexpressedThereafter we derived a model of seven ODEs based onthe original equations which was configured according tothe designed experiments making the simulation results
12 BioMed Research International
00001
0001
001
01
1Cell cycle Cell growth DNA damage Senescence Apoptosis
G1G2
MS
p21
expr
essio
n (a
u)
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
log10
wow
Figure 4 p21 expression regulated by cooperative miRNAs for different cellular processesThe associations of themiRNAs with these cellularprocesses were derived from GO terms of their TFs A miRNA was supposed to be expressed (in bold black font) in a cellular process onlyif its TF is related to the corresponding GO term of this process The p21 expression levels are computed for each process with (w)without(wo) considering the cooperative effect among the p21 targeting miRNAs
comparable with the experimental data As shown in theFigure 2(c) the simulations are in good agreement with theexperimental observations The individual overexpression ofmiR-572 or miR-93 led to the reduction of the upregulationof p21 response after genotoxic stress induction The twomiRNAs cause different degrees of repression due to theirdifferent repression efficiencies on p21 Interestingly the com-bined overexpression of both miRNAs induced the strongestdownregulation of p21 and therefore verifying the hypothesisof their cooperative regulation of p21 Above all the resultsnot only validated themodel but also demonstrated the abilityof our method to identify cooperative miRNA pairs for p21
(3) Predictive Simulations As there are abundant of networkmotifs such as FFLs in p21 regulatory network and thesenetwork motifs are important for determining p21 dynamicsit is interesting to investigate the dynamics of p21 in networkmodules where FFLs are involved To do so two networkmodules including both miR-93 and miR-572 and theirTFs were exemplified In Figure 3(a) the network modulecontains an incoherent FFL composed by AF2120572 miR-93 andp21 and a cascaded regulation in which p21 is repressed
by FOXF2 via miR-572 in Figure 3(b) the same cascadedregulation together with a coherent FFL composed of TP53MYC miR-93 and p21 forms another regulatory moduleof p21 By modulating the transcriptional strengths of thetwo miRNAs by their TFs two types of p21 dynamics wereidentified saturation and pulse (Figure 3(c) top) In theformer the p21 expression increases and reaches its steadystate at the highest level in contrast in the latter the p21expression increases to a peak and thereafter drops to asteady state at lower level For the two different networkmodules various combinations of transcriptional strengthsof the two miRNAs lead to different distributions of thetwo p21 dynamical patterns For the network module of anincoherent FFL plus the cascaded regulation (Figure 3(c)bottom left) the saturation pattern appears in two distinctregions one with weakened transcriptional strength of miR-93 and the other with enhanced transcriptional strength ofmiR-93 for the other module (Figure 3(c) bottom right)the saturation pattern only appears in the region in whichthe transcriptional strength of miR-93 is enhanced Takentogether the results showed that for the two different networkmotifs the dynamical pattern of p21 is changing according to
BioMed Research International 13
different combinations of its upstream regulators suggestingthe adaptation of p21 dynamics for different biological con-texts
Furthermore we performed a number of simulations toshow the influence of miRNA regulation on p21 expressionlevels in different cellular processes arranged in a consecutivemanner In this procedure a cell is first in the process ofcell cycle followed by proliferation (cell growth) then thecell responses to DNA damage and enters into the processof senescence and finally apoptosis is initiated As shown inFigure 4 during the cell cycle process the p21 expression islow due to the activation of most of its targeting miRNAswhen the cell starts proliferating the p21 expression declinesto an even lower level because of more activated miRNAsin this process after responding to DNA damage the p21expression soars to a high level which is caused by theactivation of its TFs like p53 and fewer expressed miRNAsalthough the p21 expression keeps at a similar level whilethe cell is undergoing senescence the expressed miRNAs aredifferent from the previous process finally the p21 expressiondecreases again to a low level due to the reemergence ofmost of its targeting miRNAs and the cell enters apoptosisInterestingly the model simulations are also consistent withexperimental observations under non-stressed condition thelow expression level of p21 is needed for cell proliferation theupregulation of p21 happens after response to DNA damagevia p53 and the increased p21 expression further results incell cycle arrest leading to senescence and apoptosis [25]Besides when considering the effect of cooperatingmiRNAsthe p21 expression levels were indistinguishable from theprevious simulation of DNA damage response However forthe other processes p21 expression levels were significantlylower compared to the simulations without considering thecooperative effect of the miRNAs Above all these resultsindicated that selective expression of cooperative miRNAscould be adopted by cells to ensure diverse expression levelsof p21 tomeet the requirements of different cellular processes
3 Conclusions and Discussion
In this paper we presented a systems biology approach com-bining data-drivenmodeling andmodel-driven experimentsto investigate the role of miRNA-mediated repression in generegulatory networksThis approach provides a systematicwayto gain a deeper understanding of the regulation of targetgenes by mutiple and cooperative miRNAs Using the regu-lation of p21 by multiple miRNAs as a case study we showedhow the ODE-basedmodel which is calibrated and validatedby means of experimental data is suitable for predicting thetemporal dynamics of molecular concentrations involved inbiochemical systems
Provided there are sufficiently rich quantitative data setsavaliable to characterize the model the use of the methodol-ogy here shown can be extended to more complex regulatorynetworks involving multiple targets cooperating TFs andmiRNAs and signaling pathways displaying cross-talk viapost-tranlational modifications In this case the criticalelement is the quality and quantitiy of the available data
Insufficiency and low quality of experimental data can causeerrors in the process of model construction and overfitting inparameter estimation can lead to uncertainties in the modelpredictions We believe that the quick development of quan-titative high throughput techniques such as transcriptomicsproteomics and miRNomics will facilitate the constructionand characterization of larger miRNA-mediated regulatorynetworks
Other modeling frameworks than ODE-based modelscan be used to describe biological systems such as prob-abilistic (eg Bayesian) or logical (eg Boolean) modelsImportantly different modeling frameworks have differentproperties and perform well regarding different perspectivesand levels of mechanistic details of biochemical systems [46]For example Bayesian models are helpful in the constructionof connections in signaling networks and can reveal the mostlikely underlying structure of the network in a probabilisticmanner Boolean models use binary values (0 and 1) andlogical gates (AND NOT and OR) to describe activities ofnetwork components and the information flow among themWe believe that in the coming future hybrid models whichconsist of modeling framework and experimental techniquespecific sub-modules will provide the necessary compromisebetween quantitativequalitative accuracy and scalability forthe investigation of large biochemical networks [47]
Disclosures
The authors declare that they have no competing financialinterests
Authorsrsquo Contribution
Julio Vera and Olaf Wolkenhauer are equal contributers
Acknowledgments
The authors would like to acknowledge the funding fromthe German Federal Ministry of Education and Research(BMBF) eBio-miRSys (0316175A to Xin Lai and Julio Vera)eBio-SysMet (0316171 to OlafWolkenhauer) and Gerontosys-ROSAge (0315892A to Olaf Wolkenhauer)
References
[1] A E Pasquinelli ldquoMicroRNAs and their targets recognitionregulation and an emerging reciprocal relationshiprdquo NatureReviews Genetics vol 13 no 4 pp 271ndash282 2012
[2] J Vera X Lai U Schmitz and O Wolkenhauer ldquoMicroRNA-regulated networks the perfect storm for classical molecularbiology the ideal scenario for systems biologyrdquo Advances inExperimental Medicine and Biology vol 774 pp 55ndash76 2013
[3] S Nikolov J Vera U Schmitz and O Wolkenhauer ldquoA model-based strategy to investigate the role of microRNA regulation incancer signalling networksrdquo Theory in Biosciences vol 130 no1 pp 55ndash69 2011
14 BioMed Research International
[4] X Lai U Schmitz S K Gupta et al ldquoComputational analysisof target hub gene repression regulated by multiple and cooper-ative miRNAsrdquoNucleic Acids Research vol 40 no 18 pp 8818ndash8834 2012
[5] X Lai O Wolkenhauer and J Vera ldquoModeling miRNA regu-lation in cancer signaling systems miR-34a regulation of thep53Sirt1 signaling modulerdquoMethods in Molecular Biology vol880 pp 87ndash108 2012
[6] C Jiang Z Xuan F Zhao and M Q Zhang ldquoTRED atranscriptional regulatory element database new entries andother developmentrdquoNucleic Acids Research vol 35 supplement1 pp D137ndashD140 2007
[7] L A Bovolenta M L Acencio and N Lemke ldquoHTRIdban open-access database for experimentally verified humantranscriptional regulation interactionsrdquo BMCGenomics vol 13article 405 2012
[8] VMatys O V Kel-Margoulis E Fricke et al ldquoTRANSFAC andits module TRANSCompel transcriptional gene regulation ineukaryotesrdquo Nucleic Acids Research vol 34 database issue ppD108ndashD110 2006
[9] D Karolchik R Baertsch M Diekhans et al ldquoThe UCSCgenome browser databaserdquo Nucleic Acids Research vol 31 no1 pp 51ndash54 2003
[10] P Alexiou T Vergoulis M Gleditzsch et al ldquomiRGen 20 adatabase of microRNA genomic information and regulationrdquoNucleic Acids Research vol 38 supplement 1 pp D137ndashD1412010
[11] F Xiao Z Zuo G Cai S Kang X Gao and T Li ldquomiRecordsan integrated resource for microRNA-target interactionsrdquoNucleic Acids Research vol 37 supplement 1 pp D105ndashD1102009
[12] S-D Hsu F-M Lin W-Y Wu et al ldquomiRTarBase a databasecurates experimentally validated microRNA-target interac-tionsrdquo Nucleic Acids Research vol 39 supplement 1 pp D163ndashD169 2011
[13] P Sethupathy B Corda and A G Hatzigeorgiou ldquoTarBase acomprehensive database of experimentally supported animalmicroRNA targetsrdquo RNA vol 12 no 2 pp 192ndash197 2006
[14] H Dweep C Sticht P Pandey and N Gretz ldquomiRWalkmdashdatabase prediction of possible miRNA binding sites byldquowalkingrdquo the genes of three genomesrdquo Journal of BiomedicalInformatics vol 44 no 5 pp 839ndash847 2011
[15] T S Keshava Prasad RGoel K Kandasamy et al ldquoHumanpro-tein reference databasemdash2009 updaterdquo Nucleic Acids Researchvol 37 supplement 1 pp D767ndashD772 2009
[16] D Szklarczyk A Franceschini M Kuhn et al ldquoThe STRINGdatabase in 2011 functional interaction networks of proteinsglobally integrated and scoredrdquo Nucleic Acids Research vol 39supplement 1 pp D561ndashD568 2011
[17] D Croft G OrsquoKelly G Wu et al ldquoReactome a database ofreactions pathways and biological processesrdquo Nucleic AcidsResearch vol 39 supplement 1 pp D691ndashD697 2011
[18] N le Novere M Hucka H Mi et al ldquoThe systems biologygraphical notationrdquoNature Biotechnology vol 27 no 8 pp 735ndash741 2009
[19] A Funahashi Y Matsuoka A Jouraku M Morohashi NKikuchi and H Kitano ldquoCellDesigner 35 a versatile modelingtool for biochemical networksrdquo Proceedings of the IEEE vol 96no 8 pp 1254ndash1265 2008
[20] M S Cline M Smoot E Cerami et al ldquoIntegration ofbiological networks and gene expression data using CytoscaperdquoNature Protocols vol 2 no 10 pp 2366ndash2382 2007
[21] D-H Le andY-KKwon ldquoNetDS aCytoscape plugin to analyzethe robustness of dynamics and feedforwardfeedback loopstructures of biological networksrdquo Bioinformatics vol 27 no 19pp 2767ndash2768 2011
[22] E O Voit Computational Analysis of Biochemical SystemsA Practical Guide for Biochemists and Molecular BiologistsCambridge University Press Cambridge UK 2000
[23] S Wu S Huang J Ding et al ldquoMultiple microRNAs modulatep21Cip1Waf1 expression by directly targeting its 31015840 untranslatedregionrdquo Oncogene vol 29 no 15 pp 2302ndash2308 2010
[24] P Saeligtrom B S E Heale O Snoslashve Jr L Aagaard J Alluin andJ J Rossi ldquoDistance constraints between microRNA target sitesdictate efficacy and cooperativityrdquo Nucleic Acids Research vol35 no 7 pp 2333ndash2342 2007
[25] Y-S Jung YQian andXChen ldquoExamination of the expandingpathways for the regulation of p21 expression and activityrdquoCellular Signalling vol 22 no 7 pp 1003ndash1012 2010
[26] W Wang H Furneaux H Cheng et al ldquoHuR regulatesp21 mRNA stabilization by UV lightrdquo Molecular and CellularBiology vol 20 no 3 pp 760ndash769 2000
[27] U Wittig R Kania M Golebiewski et al ldquoSABIO-RKmdashdatabase for biochemical reaction kineticsrdquo Nucleic AcidsResearch vol 40 database issue pp D790ndashD796 2012
[28] R Milo P Jorgensen U Moran G Weber and M SpringerldquoBioNumbersmdashthe database of key numbers in molecular andcell biologyrdquo Nucleic Acids Research vol 38 supplement 1 ppD750ndashD753 2010
[29] I-C Chou and E O Voit ldquoRecent developments in parameterestimation and structure identification of biochemical andgenomic systemsrdquoMathematical Biosciences vol 219 no 2 pp57ndash83 2009
[30] A Saltelli K Chan and E M Scott Sensitivity Analysis JohnWiley amp Sons New York NY USA 1st edition 2000
[31] S Strogatz Nonlinear Dynamics and Chaos Applications toPhysics Biology Chemistry and Engineering With Applicationsto Physics Biology Chemistry and Engineering Westview PressBoulder Colo USA 2000
[32] P Zhou S Cai Z Liu and R Wang ldquoMechanisms generatingbistability and oscillations in microRNA-mediated motifsrdquoPhysical Review E vol 85 no 4 apart 1 Article ID 041916 9pages 2012
[33] S Marino I B Hogue C J Ray and D E Kirschner ldquoAmethodology for performing global uncertainty and sensitivityanalysis in systems biologyrdquo Journal of Theoretical Biology vol254 no 1 pp 178ndash196 2008
[34] L P LimNC Lau PGarrett-Engele et al ldquoMicroarray analysisshows that some microRNAs downregulate large numbers of-target mRNAsrdquo Nature vol 433 no 7027 pp 769ndash773 2005
[35] J Krutzfeldt N Rajewsky R Braich et al ldquoSilencing ofmicroRNAs in vivowith lsquoantagomirsrsquordquoNature vol 438 no 7068pp 685ndash689 2005
[36] MWPfafflGWHorgan andLDempfle ldquoRelative expressionsoftware tool (REST) for group-wise comparison and statisticalanalysis of relative expression results in real-time PCRrdquoNucleicAcids Research vol 30 no 9 p e36 2002
[37] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[38] T Abbas and A Dutta ldquop21 in cancer intricate networks andmultiple activitiesrdquo Nature Reviews Cancer vol 9 no 6 pp400ndash414 2009
[39] J Wang M Lu C Qiu and Q Cui ldquoTransmiR a transcriptionfactor microRNA regulation databaserdquo Nucleic Acids Researchvol 38 supplement 1 pp D119ndashD122 2010
BioMed Research International 15
[40] S Bandyopadhyay andM Bhattacharyya ldquoPuTmiR a databasefor extracting neighboring transcription factors of humanmicroRNAsrdquo BMC Bioinformatics vol 11 article 190 2010
[41] A Le Bechec E Portales-Casamar G Vetter et alldquoMIRNTN a framework integrating transcription factorsmicroRNAs and their targets to identify sub-network motifs ina meta-regulation network modelrdquo BMC Bioinformatics vol12 article 67 2011
[42] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyTheGene Ontology ConsortiumrdquoNature Genetics vol 25 no 1 pp 25ndash29 2000
[43] A I F Vaz and L N Vicente ldquoA particle swarm pattern searchmethod for bound constrained global optimizationrdquo Journal ofGlobal Optimization vol 39 no 2 pp 197ndash219 2007
[44] W H PressNumerical Recipes The Art of Scientific ComputingCambridge University Press Cambridge UK 2007
[45] J G Doench and P A Sharp ldquoSpecificity of microRNA targetselection in translational repressionrdquo Genes and Developmentvol 18 no 5 pp 504ndash511 2004
[46] B Kholodenko M B Yaffe and W Kolch ldquoComputationalapproaches for analyzing information flow in biological net-worksrdquo Science Signaling vol 5 no 220 2012
[47] F M Khan U Schmitz S Nikolov et al ldquoHybrid modeling ofthe crosstalk between signaling and transcriptional networksusing ordinary differential equations and multi-valued logicrdquoBiochimica et Biophysica Acta 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
12 BioMed Research International
00001
0001
001
01
1Cell cycle Cell growth DNA damage Senescence Apoptosis
G1G2
MS
p21
expr
essio
n (a
u)
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
miR-298miR-208amiR-132miR-28-5pmiR-125a-5pmiR-299-5pmiR-345miR-93miR-423-3pmiR-363miR-657miR-654-3p
log10
wow
Figure 4 p21 expression regulated by cooperative miRNAs for different cellular processesThe associations of themiRNAs with these cellularprocesses were derived from GO terms of their TFs A miRNA was supposed to be expressed (in bold black font) in a cellular process onlyif its TF is related to the corresponding GO term of this process The p21 expression levels are computed for each process with (w)without(wo) considering the cooperative effect among the p21 targeting miRNAs
comparable with the experimental data As shown in theFigure 2(c) the simulations are in good agreement with theexperimental observations The individual overexpression ofmiR-572 or miR-93 led to the reduction of the upregulationof p21 response after genotoxic stress induction The twomiRNAs cause different degrees of repression due to theirdifferent repression efficiencies on p21 Interestingly the com-bined overexpression of both miRNAs induced the strongestdownregulation of p21 and therefore verifying the hypothesisof their cooperative regulation of p21 Above all the resultsnot only validated themodel but also demonstrated the abilityof our method to identify cooperative miRNA pairs for p21
(3) Predictive Simulations As there are abundant of networkmotifs such as FFLs in p21 regulatory network and thesenetwork motifs are important for determining p21 dynamicsit is interesting to investigate the dynamics of p21 in networkmodules where FFLs are involved To do so two networkmodules including both miR-93 and miR-572 and theirTFs were exemplified In Figure 3(a) the network modulecontains an incoherent FFL composed by AF2120572 miR-93 andp21 and a cascaded regulation in which p21 is repressed
by FOXF2 via miR-572 in Figure 3(b) the same cascadedregulation together with a coherent FFL composed of TP53MYC miR-93 and p21 forms another regulatory moduleof p21 By modulating the transcriptional strengths of thetwo miRNAs by their TFs two types of p21 dynamics wereidentified saturation and pulse (Figure 3(c) top) In theformer the p21 expression increases and reaches its steadystate at the highest level in contrast in the latter the p21expression increases to a peak and thereafter drops to asteady state at lower level For the two different networkmodules various combinations of transcriptional strengthsof the two miRNAs lead to different distributions of thetwo p21 dynamical patterns For the network module of anincoherent FFL plus the cascaded regulation (Figure 3(c)bottom left) the saturation pattern appears in two distinctregions one with weakened transcriptional strength of miR-93 and the other with enhanced transcriptional strength ofmiR-93 for the other module (Figure 3(c) bottom right)the saturation pattern only appears in the region in whichthe transcriptional strength of miR-93 is enhanced Takentogether the results showed that for the two different networkmotifs the dynamical pattern of p21 is changing according to
BioMed Research International 13
different combinations of its upstream regulators suggestingthe adaptation of p21 dynamics for different biological con-texts
Furthermore we performed a number of simulations toshow the influence of miRNA regulation on p21 expressionlevels in different cellular processes arranged in a consecutivemanner In this procedure a cell is first in the process ofcell cycle followed by proliferation (cell growth) then thecell responses to DNA damage and enters into the processof senescence and finally apoptosis is initiated As shown inFigure 4 during the cell cycle process the p21 expression islow due to the activation of most of its targeting miRNAswhen the cell starts proliferating the p21 expression declinesto an even lower level because of more activated miRNAsin this process after responding to DNA damage the p21expression soars to a high level which is caused by theactivation of its TFs like p53 and fewer expressed miRNAsalthough the p21 expression keeps at a similar level whilethe cell is undergoing senescence the expressed miRNAs aredifferent from the previous process finally the p21 expressiondecreases again to a low level due to the reemergence ofmost of its targeting miRNAs and the cell enters apoptosisInterestingly the model simulations are also consistent withexperimental observations under non-stressed condition thelow expression level of p21 is needed for cell proliferation theupregulation of p21 happens after response to DNA damagevia p53 and the increased p21 expression further results incell cycle arrest leading to senescence and apoptosis [25]Besides when considering the effect of cooperatingmiRNAsthe p21 expression levels were indistinguishable from theprevious simulation of DNA damage response However forthe other processes p21 expression levels were significantlylower compared to the simulations without considering thecooperative effect of the miRNAs Above all these resultsindicated that selective expression of cooperative miRNAscould be adopted by cells to ensure diverse expression levelsof p21 tomeet the requirements of different cellular processes
3 Conclusions and Discussion
In this paper we presented a systems biology approach com-bining data-drivenmodeling andmodel-driven experimentsto investigate the role of miRNA-mediated repression in generegulatory networksThis approach provides a systematicwayto gain a deeper understanding of the regulation of targetgenes by mutiple and cooperative miRNAs Using the regu-lation of p21 by multiple miRNAs as a case study we showedhow the ODE-basedmodel which is calibrated and validatedby means of experimental data is suitable for predicting thetemporal dynamics of molecular concentrations involved inbiochemical systems
Provided there are sufficiently rich quantitative data setsavaliable to characterize the model the use of the methodol-ogy here shown can be extended to more complex regulatorynetworks involving multiple targets cooperating TFs andmiRNAs and signaling pathways displaying cross-talk viapost-tranlational modifications In this case the criticalelement is the quality and quantitiy of the available data
Insufficiency and low quality of experimental data can causeerrors in the process of model construction and overfitting inparameter estimation can lead to uncertainties in the modelpredictions We believe that the quick development of quan-titative high throughput techniques such as transcriptomicsproteomics and miRNomics will facilitate the constructionand characterization of larger miRNA-mediated regulatorynetworks
Other modeling frameworks than ODE-based modelscan be used to describe biological systems such as prob-abilistic (eg Bayesian) or logical (eg Boolean) modelsImportantly different modeling frameworks have differentproperties and perform well regarding different perspectivesand levels of mechanistic details of biochemical systems [46]For example Bayesian models are helpful in the constructionof connections in signaling networks and can reveal the mostlikely underlying structure of the network in a probabilisticmanner Boolean models use binary values (0 and 1) andlogical gates (AND NOT and OR) to describe activities ofnetwork components and the information flow among themWe believe that in the coming future hybrid models whichconsist of modeling framework and experimental techniquespecific sub-modules will provide the necessary compromisebetween quantitativequalitative accuracy and scalability forthe investigation of large biochemical networks [47]
Disclosures
The authors declare that they have no competing financialinterests
Authorsrsquo Contribution
Julio Vera and Olaf Wolkenhauer are equal contributers
Acknowledgments
The authors would like to acknowledge the funding fromthe German Federal Ministry of Education and Research(BMBF) eBio-miRSys (0316175A to Xin Lai and Julio Vera)eBio-SysMet (0316171 to OlafWolkenhauer) and Gerontosys-ROSAge (0315892A to Olaf Wolkenhauer)
References
[1] A E Pasquinelli ldquoMicroRNAs and their targets recognitionregulation and an emerging reciprocal relationshiprdquo NatureReviews Genetics vol 13 no 4 pp 271ndash282 2012
[2] J Vera X Lai U Schmitz and O Wolkenhauer ldquoMicroRNA-regulated networks the perfect storm for classical molecularbiology the ideal scenario for systems biologyrdquo Advances inExperimental Medicine and Biology vol 774 pp 55ndash76 2013
[3] S Nikolov J Vera U Schmitz and O Wolkenhauer ldquoA model-based strategy to investigate the role of microRNA regulation incancer signalling networksrdquo Theory in Biosciences vol 130 no1 pp 55ndash69 2011
14 BioMed Research International
[4] X Lai U Schmitz S K Gupta et al ldquoComputational analysisof target hub gene repression regulated by multiple and cooper-ative miRNAsrdquoNucleic Acids Research vol 40 no 18 pp 8818ndash8834 2012
[5] X Lai O Wolkenhauer and J Vera ldquoModeling miRNA regu-lation in cancer signaling systems miR-34a regulation of thep53Sirt1 signaling modulerdquoMethods in Molecular Biology vol880 pp 87ndash108 2012
[6] C Jiang Z Xuan F Zhao and M Q Zhang ldquoTRED atranscriptional regulatory element database new entries andother developmentrdquoNucleic Acids Research vol 35 supplement1 pp D137ndashD140 2007
[7] L A Bovolenta M L Acencio and N Lemke ldquoHTRIdban open-access database for experimentally verified humantranscriptional regulation interactionsrdquo BMCGenomics vol 13article 405 2012
[8] VMatys O V Kel-Margoulis E Fricke et al ldquoTRANSFAC andits module TRANSCompel transcriptional gene regulation ineukaryotesrdquo Nucleic Acids Research vol 34 database issue ppD108ndashD110 2006
[9] D Karolchik R Baertsch M Diekhans et al ldquoThe UCSCgenome browser databaserdquo Nucleic Acids Research vol 31 no1 pp 51ndash54 2003
[10] P Alexiou T Vergoulis M Gleditzsch et al ldquomiRGen 20 adatabase of microRNA genomic information and regulationrdquoNucleic Acids Research vol 38 supplement 1 pp D137ndashD1412010
[11] F Xiao Z Zuo G Cai S Kang X Gao and T Li ldquomiRecordsan integrated resource for microRNA-target interactionsrdquoNucleic Acids Research vol 37 supplement 1 pp D105ndashD1102009
[12] S-D Hsu F-M Lin W-Y Wu et al ldquomiRTarBase a databasecurates experimentally validated microRNA-target interac-tionsrdquo Nucleic Acids Research vol 39 supplement 1 pp D163ndashD169 2011
[13] P Sethupathy B Corda and A G Hatzigeorgiou ldquoTarBase acomprehensive database of experimentally supported animalmicroRNA targetsrdquo RNA vol 12 no 2 pp 192ndash197 2006
[14] H Dweep C Sticht P Pandey and N Gretz ldquomiRWalkmdashdatabase prediction of possible miRNA binding sites byldquowalkingrdquo the genes of three genomesrdquo Journal of BiomedicalInformatics vol 44 no 5 pp 839ndash847 2011
[15] T S Keshava Prasad RGoel K Kandasamy et al ldquoHumanpro-tein reference databasemdash2009 updaterdquo Nucleic Acids Researchvol 37 supplement 1 pp D767ndashD772 2009
[16] D Szklarczyk A Franceschini M Kuhn et al ldquoThe STRINGdatabase in 2011 functional interaction networks of proteinsglobally integrated and scoredrdquo Nucleic Acids Research vol 39supplement 1 pp D561ndashD568 2011
[17] D Croft G OrsquoKelly G Wu et al ldquoReactome a database ofreactions pathways and biological processesrdquo Nucleic AcidsResearch vol 39 supplement 1 pp D691ndashD697 2011
[18] N le Novere M Hucka H Mi et al ldquoThe systems biologygraphical notationrdquoNature Biotechnology vol 27 no 8 pp 735ndash741 2009
[19] A Funahashi Y Matsuoka A Jouraku M Morohashi NKikuchi and H Kitano ldquoCellDesigner 35 a versatile modelingtool for biochemical networksrdquo Proceedings of the IEEE vol 96no 8 pp 1254ndash1265 2008
[20] M S Cline M Smoot E Cerami et al ldquoIntegration ofbiological networks and gene expression data using CytoscaperdquoNature Protocols vol 2 no 10 pp 2366ndash2382 2007
[21] D-H Le andY-KKwon ldquoNetDS aCytoscape plugin to analyzethe robustness of dynamics and feedforwardfeedback loopstructures of biological networksrdquo Bioinformatics vol 27 no 19pp 2767ndash2768 2011
[22] E O Voit Computational Analysis of Biochemical SystemsA Practical Guide for Biochemists and Molecular BiologistsCambridge University Press Cambridge UK 2000
[23] S Wu S Huang J Ding et al ldquoMultiple microRNAs modulatep21Cip1Waf1 expression by directly targeting its 31015840 untranslatedregionrdquo Oncogene vol 29 no 15 pp 2302ndash2308 2010
[24] P Saeligtrom B S E Heale O Snoslashve Jr L Aagaard J Alluin andJ J Rossi ldquoDistance constraints between microRNA target sitesdictate efficacy and cooperativityrdquo Nucleic Acids Research vol35 no 7 pp 2333ndash2342 2007
[25] Y-S Jung YQian andXChen ldquoExamination of the expandingpathways for the regulation of p21 expression and activityrdquoCellular Signalling vol 22 no 7 pp 1003ndash1012 2010
[26] W Wang H Furneaux H Cheng et al ldquoHuR regulatesp21 mRNA stabilization by UV lightrdquo Molecular and CellularBiology vol 20 no 3 pp 760ndash769 2000
[27] U Wittig R Kania M Golebiewski et al ldquoSABIO-RKmdashdatabase for biochemical reaction kineticsrdquo Nucleic AcidsResearch vol 40 database issue pp D790ndashD796 2012
[28] R Milo P Jorgensen U Moran G Weber and M SpringerldquoBioNumbersmdashthe database of key numbers in molecular andcell biologyrdquo Nucleic Acids Research vol 38 supplement 1 ppD750ndashD753 2010
[29] I-C Chou and E O Voit ldquoRecent developments in parameterestimation and structure identification of biochemical andgenomic systemsrdquoMathematical Biosciences vol 219 no 2 pp57ndash83 2009
[30] A Saltelli K Chan and E M Scott Sensitivity Analysis JohnWiley amp Sons New York NY USA 1st edition 2000
[31] S Strogatz Nonlinear Dynamics and Chaos Applications toPhysics Biology Chemistry and Engineering With Applicationsto Physics Biology Chemistry and Engineering Westview PressBoulder Colo USA 2000
[32] P Zhou S Cai Z Liu and R Wang ldquoMechanisms generatingbistability and oscillations in microRNA-mediated motifsrdquoPhysical Review E vol 85 no 4 apart 1 Article ID 041916 9pages 2012
[33] S Marino I B Hogue C J Ray and D E Kirschner ldquoAmethodology for performing global uncertainty and sensitivityanalysis in systems biologyrdquo Journal of Theoretical Biology vol254 no 1 pp 178ndash196 2008
[34] L P LimNC Lau PGarrett-Engele et al ldquoMicroarray analysisshows that some microRNAs downregulate large numbers of-target mRNAsrdquo Nature vol 433 no 7027 pp 769ndash773 2005
[35] J Krutzfeldt N Rajewsky R Braich et al ldquoSilencing ofmicroRNAs in vivowith lsquoantagomirsrsquordquoNature vol 438 no 7068pp 685ndash689 2005
[36] MWPfafflGWHorgan andLDempfle ldquoRelative expressionsoftware tool (REST) for group-wise comparison and statisticalanalysis of relative expression results in real-time PCRrdquoNucleicAcids Research vol 30 no 9 p e36 2002
[37] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[38] T Abbas and A Dutta ldquop21 in cancer intricate networks andmultiple activitiesrdquo Nature Reviews Cancer vol 9 no 6 pp400ndash414 2009
[39] J Wang M Lu C Qiu and Q Cui ldquoTransmiR a transcriptionfactor microRNA regulation databaserdquo Nucleic Acids Researchvol 38 supplement 1 pp D119ndashD122 2010
BioMed Research International 15
[40] S Bandyopadhyay andM Bhattacharyya ldquoPuTmiR a databasefor extracting neighboring transcription factors of humanmicroRNAsrdquo BMC Bioinformatics vol 11 article 190 2010
[41] A Le Bechec E Portales-Casamar G Vetter et alldquoMIRNTN a framework integrating transcription factorsmicroRNAs and their targets to identify sub-network motifs ina meta-regulation network modelrdquo BMC Bioinformatics vol12 article 67 2011
[42] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyTheGene Ontology ConsortiumrdquoNature Genetics vol 25 no 1 pp 25ndash29 2000
[43] A I F Vaz and L N Vicente ldquoA particle swarm pattern searchmethod for bound constrained global optimizationrdquo Journal ofGlobal Optimization vol 39 no 2 pp 197ndash219 2007
[44] W H PressNumerical Recipes The Art of Scientific ComputingCambridge University Press Cambridge UK 2007
[45] J G Doench and P A Sharp ldquoSpecificity of microRNA targetselection in translational repressionrdquo Genes and Developmentvol 18 no 5 pp 504ndash511 2004
[46] B Kholodenko M B Yaffe and W Kolch ldquoComputationalapproaches for analyzing information flow in biological net-worksrdquo Science Signaling vol 5 no 220 2012
[47] F M Khan U Schmitz S Nikolov et al ldquoHybrid modeling ofthe crosstalk between signaling and transcriptional networksusing ordinary differential equations and multi-valued logicrdquoBiochimica et Biophysica Acta 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
BioMed Research International 13
different combinations of its upstream regulators suggestingthe adaptation of p21 dynamics for different biological con-texts
Furthermore we performed a number of simulations toshow the influence of miRNA regulation on p21 expressionlevels in different cellular processes arranged in a consecutivemanner In this procedure a cell is first in the process ofcell cycle followed by proliferation (cell growth) then thecell responses to DNA damage and enters into the processof senescence and finally apoptosis is initiated As shown inFigure 4 during the cell cycle process the p21 expression islow due to the activation of most of its targeting miRNAswhen the cell starts proliferating the p21 expression declinesto an even lower level because of more activated miRNAsin this process after responding to DNA damage the p21expression soars to a high level which is caused by theactivation of its TFs like p53 and fewer expressed miRNAsalthough the p21 expression keeps at a similar level whilethe cell is undergoing senescence the expressed miRNAs aredifferent from the previous process finally the p21 expressiondecreases again to a low level due to the reemergence ofmost of its targeting miRNAs and the cell enters apoptosisInterestingly the model simulations are also consistent withexperimental observations under non-stressed condition thelow expression level of p21 is needed for cell proliferation theupregulation of p21 happens after response to DNA damagevia p53 and the increased p21 expression further results incell cycle arrest leading to senescence and apoptosis [25]Besides when considering the effect of cooperatingmiRNAsthe p21 expression levels were indistinguishable from theprevious simulation of DNA damage response However forthe other processes p21 expression levels were significantlylower compared to the simulations without considering thecooperative effect of the miRNAs Above all these resultsindicated that selective expression of cooperative miRNAscould be adopted by cells to ensure diverse expression levelsof p21 tomeet the requirements of different cellular processes
3 Conclusions and Discussion
In this paper we presented a systems biology approach com-bining data-drivenmodeling andmodel-driven experimentsto investigate the role of miRNA-mediated repression in generegulatory networksThis approach provides a systematicwayto gain a deeper understanding of the regulation of targetgenes by mutiple and cooperative miRNAs Using the regu-lation of p21 by multiple miRNAs as a case study we showedhow the ODE-basedmodel which is calibrated and validatedby means of experimental data is suitable for predicting thetemporal dynamics of molecular concentrations involved inbiochemical systems
Provided there are sufficiently rich quantitative data setsavaliable to characterize the model the use of the methodol-ogy here shown can be extended to more complex regulatorynetworks involving multiple targets cooperating TFs andmiRNAs and signaling pathways displaying cross-talk viapost-tranlational modifications In this case the criticalelement is the quality and quantitiy of the available data
Insufficiency and low quality of experimental data can causeerrors in the process of model construction and overfitting inparameter estimation can lead to uncertainties in the modelpredictions We believe that the quick development of quan-titative high throughput techniques such as transcriptomicsproteomics and miRNomics will facilitate the constructionand characterization of larger miRNA-mediated regulatorynetworks
Other modeling frameworks than ODE-based modelscan be used to describe biological systems such as prob-abilistic (eg Bayesian) or logical (eg Boolean) modelsImportantly different modeling frameworks have differentproperties and perform well regarding different perspectivesand levels of mechanistic details of biochemical systems [46]For example Bayesian models are helpful in the constructionof connections in signaling networks and can reveal the mostlikely underlying structure of the network in a probabilisticmanner Boolean models use binary values (0 and 1) andlogical gates (AND NOT and OR) to describe activities ofnetwork components and the information flow among themWe believe that in the coming future hybrid models whichconsist of modeling framework and experimental techniquespecific sub-modules will provide the necessary compromisebetween quantitativequalitative accuracy and scalability forthe investigation of large biochemical networks [47]
Disclosures
The authors declare that they have no competing financialinterests
Authorsrsquo Contribution
Julio Vera and Olaf Wolkenhauer are equal contributers
Acknowledgments
The authors would like to acknowledge the funding fromthe German Federal Ministry of Education and Research(BMBF) eBio-miRSys (0316175A to Xin Lai and Julio Vera)eBio-SysMet (0316171 to OlafWolkenhauer) and Gerontosys-ROSAge (0315892A to Olaf Wolkenhauer)
References
[1] A E Pasquinelli ldquoMicroRNAs and their targets recognitionregulation and an emerging reciprocal relationshiprdquo NatureReviews Genetics vol 13 no 4 pp 271ndash282 2012
[2] J Vera X Lai U Schmitz and O Wolkenhauer ldquoMicroRNA-regulated networks the perfect storm for classical molecularbiology the ideal scenario for systems biologyrdquo Advances inExperimental Medicine and Biology vol 774 pp 55ndash76 2013
[3] S Nikolov J Vera U Schmitz and O Wolkenhauer ldquoA model-based strategy to investigate the role of microRNA regulation incancer signalling networksrdquo Theory in Biosciences vol 130 no1 pp 55ndash69 2011
14 BioMed Research International
[4] X Lai U Schmitz S K Gupta et al ldquoComputational analysisof target hub gene repression regulated by multiple and cooper-ative miRNAsrdquoNucleic Acids Research vol 40 no 18 pp 8818ndash8834 2012
[5] X Lai O Wolkenhauer and J Vera ldquoModeling miRNA regu-lation in cancer signaling systems miR-34a regulation of thep53Sirt1 signaling modulerdquoMethods in Molecular Biology vol880 pp 87ndash108 2012
[6] C Jiang Z Xuan F Zhao and M Q Zhang ldquoTRED atranscriptional regulatory element database new entries andother developmentrdquoNucleic Acids Research vol 35 supplement1 pp D137ndashD140 2007
[7] L A Bovolenta M L Acencio and N Lemke ldquoHTRIdban open-access database for experimentally verified humantranscriptional regulation interactionsrdquo BMCGenomics vol 13article 405 2012
[8] VMatys O V Kel-Margoulis E Fricke et al ldquoTRANSFAC andits module TRANSCompel transcriptional gene regulation ineukaryotesrdquo Nucleic Acids Research vol 34 database issue ppD108ndashD110 2006
[9] D Karolchik R Baertsch M Diekhans et al ldquoThe UCSCgenome browser databaserdquo Nucleic Acids Research vol 31 no1 pp 51ndash54 2003
[10] P Alexiou T Vergoulis M Gleditzsch et al ldquomiRGen 20 adatabase of microRNA genomic information and regulationrdquoNucleic Acids Research vol 38 supplement 1 pp D137ndashD1412010
[11] F Xiao Z Zuo G Cai S Kang X Gao and T Li ldquomiRecordsan integrated resource for microRNA-target interactionsrdquoNucleic Acids Research vol 37 supplement 1 pp D105ndashD1102009
[12] S-D Hsu F-M Lin W-Y Wu et al ldquomiRTarBase a databasecurates experimentally validated microRNA-target interac-tionsrdquo Nucleic Acids Research vol 39 supplement 1 pp D163ndashD169 2011
[13] P Sethupathy B Corda and A G Hatzigeorgiou ldquoTarBase acomprehensive database of experimentally supported animalmicroRNA targetsrdquo RNA vol 12 no 2 pp 192ndash197 2006
[14] H Dweep C Sticht P Pandey and N Gretz ldquomiRWalkmdashdatabase prediction of possible miRNA binding sites byldquowalkingrdquo the genes of three genomesrdquo Journal of BiomedicalInformatics vol 44 no 5 pp 839ndash847 2011
[15] T S Keshava Prasad RGoel K Kandasamy et al ldquoHumanpro-tein reference databasemdash2009 updaterdquo Nucleic Acids Researchvol 37 supplement 1 pp D767ndashD772 2009
[16] D Szklarczyk A Franceschini M Kuhn et al ldquoThe STRINGdatabase in 2011 functional interaction networks of proteinsglobally integrated and scoredrdquo Nucleic Acids Research vol 39supplement 1 pp D561ndashD568 2011
[17] D Croft G OrsquoKelly G Wu et al ldquoReactome a database ofreactions pathways and biological processesrdquo Nucleic AcidsResearch vol 39 supplement 1 pp D691ndashD697 2011
[18] N le Novere M Hucka H Mi et al ldquoThe systems biologygraphical notationrdquoNature Biotechnology vol 27 no 8 pp 735ndash741 2009
[19] A Funahashi Y Matsuoka A Jouraku M Morohashi NKikuchi and H Kitano ldquoCellDesigner 35 a versatile modelingtool for biochemical networksrdquo Proceedings of the IEEE vol 96no 8 pp 1254ndash1265 2008
[20] M S Cline M Smoot E Cerami et al ldquoIntegration ofbiological networks and gene expression data using CytoscaperdquoNature Protocols vol 2 no 10 pp 2366ndash2382 2007
[21] D-H Le andY-KKwon ldquoNetDS aCytoscape plugin to analyzethe robustness of dynamics and feedforwardfeedback loopstructures of biological networksrdquo Bioinformatics vol 27 no 19pp 2767ndash2768 2011
[22] E O Voit Computational Analysis of Biochemical SystemsA Practical Guide for Biochemists and Molecular BiologistsCambridge University Press Cambridge UK 2000
[23] S Wu S Huang J Ding et al ldquoMultiple microRNAs modulatep21Cip1Waf1 expression by directly targeting its 31015840 untranslatedregionrdquo Oncogene vol 29 no 15 pp 2302ndash2308 2010
[24] P Saeligtrom B S E Heale O Snoslashve Jr L Aagaard J Alluin andJ J Rossi ldquoDistance constraints between microRNA target sitesdictate efficacy and cooperativityrdquo Nucleic Acids Research vol35 no 7 pp 2333ndash2342 2007
[25] Y-S Jung YQian andXChen ldquoExamination of the expandingpathways for the regulation of p21 expression and activityrdquoCellular Signalling vol 22 no 7 pp 1003ndash1012 2010
[26] W Wang H Furneaux H Cheng et al ldquoHuR regulatesp21 mRNA stabilization by UV lightrdquo Molecular and CellularBiology vol 20 no 3 pp 760ndash769 2000
[27] U Wittig R Kania M Golebiewski et al ldquoSABIO-RKmdashdatabase for biochemical reaction kineticsrdquo Nucleic AcidsResearch vol 40 database issue pp D790ndashD796 2012
[28] R Milo P Jorgensen U Moran G Weber and M SpringerldquoBioNumbersmdashthe database of key numbers in molecular andcell biologyrdquo Nucleic Acids Research vol 38 supplement 1 ppD750ndashD753 2010
[29] I-C Chou and E O Voit ldquoRecent developments in parameterestimation and structure identification of biochemical andgenomic systemsrdquoMathematical Biosciences vol 219 no 2 pp57ndash83 2009
[30] A Saltelli K Chan and E M Scott Sensitivity Analysis JohnWiley amp Sons New York NY USA 1st edition 2000
[31] S Strogatz Nonlinear Dynamics and Chaos Applications toPhysics Biology Chemistry and Engineering With Applicationsto Physics Biology Chemistry and Engineering Westview PressBoulder Colo USA 2000
[32] P Zhou S Cai Z Liu and R Wang ldquoMechanisms generatingbistability and oscillations in microRNA-mediated motifsrdquoPhysical Review E vol 85 no 4 apart 1 Article ID 041916 9pages 2012
[33] S Marino I B Hogue C J Ray and D E Kirschner ldquoAmethodology for performing global uncertainty and sensitivityanalysis in systems biologyrdquo Journal of Theoretical Biology vol254 no 1 pp 178ndash196 2008
[34] L P LimNC Lau PGarrett-Engele et al ldquoMicroarray analysisshows that some microRNAs downregulate large numbers of-target mRNAsrdquo Nature vol 433 no 7027 pp 769ndash773 2005
[35] J Krutzfeldt N Rajewsky R Braich et al ldquoSilencing ofmicroRNAs in vivowith lsquoantagomirsrsquordquoNature vol 438 no 7068pp 685ndash689 2005
[36] MWPfafflGWHorgan andLDempfle ldquoRelative expressionsoftware tool (REST) for group-wise comparison and statisticalanalysis of relative expression results in real-time PCRrdquoNucleicAcids Research vol 30 no 9 p e36 2002
[37] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[38] T Abbas and A Dutta ldquop21 in cancer intricate networks andmultiple activitiesrdquo Nature Reviews Cancer vol 9 no 6 pp400ndash414 2009
[39] J Wang M Lu C Qiu and Q Cui ldquoTransmiR a transcriptionfactor microRNA regulation databaserdquo Nucleic Acids Researchvol 38 supplement 1 pp D119ndashD122 2010
BioMed Research International 15
[40] S Bandyopadhyay andM Bhattacharyya ldquoPuTmiR a databasefor extracting neighboring transcription factors of humanmicroRNAsrdquo BMC Bioinformatics vol 11 article 190 2010
[41] A Le Bechec E Portales-Casamar G Vetter et alldquoMIRNTN a framework integrating transcription factorsmicroRNAs and their targets to identify sub-network motifs ina meta-regulation network modelrdquo BMC Bioinformatics vol12 article 67 2011
[42] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyTheGene Ontology ConsortiumrdquoNature Genetics vol 25 no 1 pp 25ndash29 2000
[43] A I F Vaz and L N Vicente ldquoA particle swarm pattern searchmethod for bound constrained global optimizationrdquo Journal ofGlobal Optimization vol 39 no 2 pp 197ndash219 2007
[44] W H PressNumerical Recipes The Art of Scientific ComputingCambridge University Press Cambridge UK 2007
[45] J G Doench and P A Sharp ldquoSpecificity of microRNA targetselection in translational repressionrdquo Genes and Developmentvol 18 no 5 pp 504ndash511 2004
[46] B Kholodenko M B Yaffe and W Kolch ldquoComputationalapproaches for analyzing information flow in biological net-worksrdquo Science Signaling vol 5 no 220 2012
[47] F M Khan U Schmitz S Nikolov et al ldquoHybrid modeling ofthe crosstalk between signaling and transcriptional networksusing ordinary differential equations and multi-valued logicrdquoBiochimica et Biophysica Acta 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
14 BioMed Research International
[4] X Lai U Schmitz S K Gupta et al ldquoComputational analysisof target hub gene repression regulated by multiple and cooper-ative miRNAsrdquoNucleic Acids Research vol 40 no 18 pp 8818ndash8834 2012
[5] X Lai O Wolkenhauer and J Vera ldquoModeling miRNA regu-lation in cancer signaling systems miR-34a regulation of thep53Sirt1 signaling modulerdquoMethods in Molecular Biology vol880 pp 87ndash108 2012
[6] C Jiang Z Xuan F Zhao and M Q Zhang ldquoTRED atranscriptional regulatory element database new entries andother developmentrdquoNucleic Acids Research vol 35 supplement1 pp D137ndashD140 2007
[7] L A Bovolenta M L Acencio and N Lemke ldquoHTRIdban open-access database for experimentally verified humantranscriptional regulation interactionsrdquo BMCGenomics vol 13article 405 2012
[8] VMatys O V Kel-Margoulis E Fricke et al ldquoTRANSFAC andits module TRANSCompel transcriptional gene regulation ineukaryotesrdquo Nucleic Acids Research vol 34 database issue ppD108ndashD110 2006
[9] D Karolchik R Baertsch M Diekhans et al ldquoThe UCSCgenome browser databaserdquo Nucleic Acids Research vol 31 no1 pp 51ndash54 2003
[10] P Alexiou T Vergoulis M Gleditzsch et al ldquomiRGen 20 adatabase of microRNA genomic information and regulationrdquoNucleic Acids Research vol 38 supplement 1 pp D137ndashD1412010
[11] F Xiao Z Zuo G Cai S Kang X Gao and T Li ldquomiRecordsan integrated resource for microRNA-target interactionsrdquoNucleic Acids Research vol 37 supplement 1 pp D105ndashD1102009
[12] S-D Hsu F-M Lin W-Y Wu et al ldquomiRTarBase a databasecurates experimentally validated microRNA-target interac-tionsrdquo Nucleic Acids Research vol 39 supplement 1 pp D163ndashD169 2011
[13] P Sethupathy B Corda and A G Hatzigeorgiou ldquoTarBase acomprehensive database of experimentally supported animalmicroRNA targetsrdquo RNA vol 12 no 2 pp 192ndash197 2006
[14] H Dweep C Sticht P Pandey and N Gretz ldquomiRWalkmdashdatabase prediction of possible miRNA binding sites byldquowalkingrdquo the genes of three genomesrdquo Journal of BiomedicalInformatics vol 44 no 5 pp 839ndash847 2011
[15] T S Keshava Prasad RGoel K Kandasamy et al ldquoHumanpro-tein reference databasemdash2009 updaterdquo Nucleic Acids Researchvol 37 supplement 1 pp D767ndashD772 2009
[16] D Szklarczyk A Franceschini M Kuhn et al ldquoThe STRINGdatabase in 2011 functional interaction networks of proteinsglobally integrated and scoredrdquo Nucleic Acids Research vol 39supplement 1 pp D561ndashD568 2011
[17] D Croft G OrsquoKelly G Wu et al ldquoReactome a database ofreactions pathways and biological processesrdquo Nucleic AcidsResearch vol 39 supplement 1 pp D691ndashD697 2011
[18] N le Novere M Hucka H Mi et al ldquoThe systems biologygraphical notationrdquoNature Biotechnology vol 27 no 8 pp 735ndash741 2009
[19] A Funahashi Y Matsuoka A Jouraku M Morohashi NKikuchi and H Kitano ldquoCellDesigner 35 a versatile modelingtool for biochemical networksrdquo Proceedings of the IEEE vol 96no 8 pp 1254ndash1265 2008
[20] M S Cline M Smoot E Cerami et al ldquoIntegration ofbiological networks and gene expression data using CytoscaperdquoNature Protocols vol 2 no 10 pp 2366ndash2382 2007
[21] D-H Le andY-KKwon ldquoNetDS aCytoscape plugin to analyzethe robustness of dynamics and feedforwardfeedback loopstructures of biological networksrdquo Bioinformatics vol 27 no 19pp 2767ndash2768 2011
[22] E O Voit Computational Analysis of Biochemical SystemsA Practical Guide for Biochemists and Molecular BiologistsCambridge University Press Cambridge UK 2000
[23] S Wu S Huang J Ding et al ldquoMultiple microRNAs modulatep21Cip1Waf1 expression by directly targeting its 31015840 untranslatedregionrdquo Oncogene vol 29 no 15 pp 2302ndash2308 2010
[24] P Saeligtrom B S E Heale O Snoslashve Jr L Aagaard J Alluin andJ J Rossi ldquoDistance constraints between microRNA target sitesdictate efficacy and cooperativityrdquo Nucleic Acids Research vol35 no 7 pp 2333ndash2342 2007
[25] Y-S Jung YQian andXChen ldquoExamination of the expandingpathways for the regulation of p21 expression and activityrdquoCellular Signalling vol 22 no 7 pp 1003ndash1012 2010
[26] W Wang H Furneaux H Cheng et al ldquoHuR regulatesp21 mRNA stabilization by UV lightrdquo Molecular and CellularBiology vol 20 no 3 pp 760ndash769 2000
[27] U Wittig R Kania M Golebiewski et al ldquoSABIO-RKmdashdatabase for biochemical reaction kineticsrdquo Nucleic AcidsResearch vol 40 database issue pp D790ndashD796 2012
[28] R Milo P Jorgensen U Moran G Weber and M SpringerldquoBioNumbersmdashthe database of key numbers in molecular andcell biologyrdquo Nucleic Acids Research vol 38 supplement 1 ppD750ndashD753 2010
[29] I-C Chou and E O Voit ldquoRecent developments in parameterestimation and structure identification of biochemical andgenomic systemsrdquoMathematical Biosciences vol 219 no 2 pp57ndash83 2009
[30] A Saltelli K Chan and E M Scott Sensitivity Analysis JohnWiley amp Sons New York NY USA 1st edition 2000
[31] S Strogatz Nonlinear Dynamics and Chaos Applications toPhysics Biology Chemistry and Engineering With Applicationsto Physics Biology Chemistry and Engineering Westview PressBoulder Colo USA 2000
[32] P Zhou S Cai Z Liu and R Wang ldquoMechanisms generatingbistability and oscillations in microRNA-mediated motifsrdquoPhysical Review E vol 85 no 4 apart 1 Article ID 041916 9pages 2012
[33] S Marino I B Hogue C J Ray and D E Kirschner ldquoAmethodology for performing global uncertainty and sensitivityanalysis in systems biologyrdquo Journal of Theoretical Biology vol254 no 1 pp 178ndash196 2008
[34] L P LimNC Lau PGarrett-Engele et al ldquoMicroarray analysisshows that some microRNAs downregulate large numbers of-target mRNAsrdquo Nature vol 433 no 7027 pp 769ndash773 2005
[35] J Krutzfeldt N Rajewsky R Braich et al ldquoSilencing ofmicroRNAs in vivowith lsquoantagomirsrsquordquoNature vol 438 no 7068pp 685ndash689 2005
[36] MWPfafflGWHorgan andLDempfle ldquoRelative expressionsoftware tool (REST) for group-wise comparison and statisticalanalysis of relative expression results in real-time PCRrdquoNucleicAcids Research vol 30 no 9 p e36 2002
[37] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[38] T Abbas and A Dutta ldquop21 in cancer intricate networks andmultiple activitiesrdquo Nature Reviews Cancer vol 9 no 6 pp400ndash414 2009
[39] J Wang M Lu C Qiu and Q Cui ldquoTransmiR a transcriptionfactor microRNA regulation databaserdquo Nucleic Acids Researchvol 38 supplement 1 pp D119ndashD122 2010
BioMed Research International 15
[40] S Bandyopadhyay andM Bhattacharyya ldquoPuTmiR a databasefor extracting neighboring transcription factors of humanmicroRNAsrdquo BMC Bioinformatics vol 11 article 190 2010
[41] A Le Bechec E Portales-Casamar G Vetter et alldquoMIRNTN a framework integrating transcription factorsmicroRNAs and their targets to identify sub-network motifs ina meta-regulation network modelrdquo BMC Bioinformatics vol12 article 67 2011
[42] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyTheGene Ontology ConsortiumrdquoNature Genetics vol 25 no 1 pp 25ndash29 2000
[43] A I F Vaz and L N Vicente ldquoA particle swarm pattern searchmethod for bound constrained global optimizationrdquo Journal ofGlobal Optimization vol 39 no 2 pp 197ndash219 2007
[44] W H PressNumerical Recipes The Art of Scientific ComputingCambridge University Press Cambridge UK 2007
[45] J G Doench and P A Sharp ldquoSpecificity of microRNA targetselection in translational repressionrdquo Genes and Developmentvol 18 no 5 pp 504ndash511 2004
[46] B Kholodenko M B Yaffe and W Kolch ldquoComputationalapproaches for analyzing information flow in biological net-worksrdquo Science Signaling vol 5 no 220 2012
[47] F M Khan U Schmitz S Nikolov et al ldquoHybrid modeling ofthe crosstalk between signaling and transcriptional networksusing ordinary differential equations and multi-valued logicrdquoBiochimica et Biophysica Acta 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
BioMed Research International 15
[40] S Bandyopadhyay andM Bhattacharyya ldquoPuTmiR a databasefor extracting neighboring transcription factors of humanmicroRNAsrdquo BMC Bioinformatics vol 11 article 190 2010
[41] A Le Bechec E Portales-Casamar G Vetter et alldquoMIRNTN a framework integrating transcription factorsmicroRNAs and their targets to identify sub-network motifs ina meta-regulation network modelrdquo BMC Bioinformatics vol12 article 67 2011
[42] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyTheGene Ontology ConsortiumrdquoNature Genetics vol 25 no 1 pp 25ndash29 2000
[43] A I F Vaz and L N Vicente ldquoA particle swarm pattern searchmethod for bound constrained global optimizationrdquo Journal ofGlobal Optimization vol 39 no 2 pp 197ndash219 2007
[44] W H PressNumerical Recipes The Art of Scientific ComputingCambridge University Press Cambridge UK 2007
[45] J G Doench and P A Sharp ldquoSpecificity of microRNA targetselection in translational repressionrdquo Genes and Developmentvol 18 no 5 pp 504ndash511 2004
[46] B Kholodenko M B Yaffe and W Kolch ldquoComputationalapproaches for analyzing information flow in biological net-worksrdquo Science Signaling vol 5 no 220 2012
[47] F M Khan U Schmitz S Nikolov et al ldquoHybrid modeling ofthe crosstalk between signaling and transcriptional networksusing ordinary differential equations and multi-valued logicrdquoBiochimica et Biophysica Acta 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology