M.P. Cosma 16/09/2010 Circuiti genetici e programmabilità della cellula Pia Cosma CRG, Barcelona Regolazione genica Bressanone GNB 2010 1 M.P. Cosma 16/09/2010 I geni GAL in lievito Saccharomyces cerevisiae G Gal3 Gal80 Tup1 L’attivatore Gal4 Bressanone GNB 2010 2 M.P. Cosma 16/09/2010 Il dominio di legame e quello di attivazione sono separabili Genoma del fago λ Ciclo litico Ciclo litico o lisogenico Ciclo lisogenico Bressanone GNB 2010 3 M.P. Cosma 16/09/2010 Il ciclo litico e il ciclo lisogenico Lo switch da lisogenia a lisi Bressanone GNB 2010 4 M.P. Cosma 16/09/2010 Legame cooperativo del repressore λ al DNA C N Or3 Or2 Or1 Interazione di λ ai siti OR e OL stabilizza il binding e aumenta la repressione Questo porta come conseguenza che la curva che descrive il legame del repressore ai due siti O in funzione della concentrazione del repressore stesso è sigmoide. E questo fa si che piccole variazioni della concentrazione del repressore hanno un effetto importante sulla efficienza di legame al sito. Bressanone GNB 2010 5 M.P. Cosma 16/09/2010 Domini di λ e loro impatto sulla cooperativita’ di legame Eliminare la cooperativita’ ha lo stesso effetto di diminuire la concentrazione dei monomeri di 100 volte Ordered recruitment mechanism TATA TBP or TFIID + TFIIA TFIIB TFIIF-PolII TFIIE TFIIH D A F E H TBP B PolII CTD Bressanone GNB 2010 6 M.P. Cosma 16/09/2010 One step recruitment TATA TBP D E H B F A PolII CTD D A F E H TBP B PolII CTD NO GENERAL MECHANISM Bressanone GNB 2010 7 M.P. Cosma 16/09/2010 Time Factor ordered recruitment Recruitment timing Histone code Specific cofactors Time Factor ordered recruitment Recruitment timing Histone code Specific cofactors Bressanone GNB 2010 8 M.P. Cosma 16/09/2010 HO URS1 URS2 TATA Cdk1 SWI5 URS1 HO URS2 Late Anaphase TATA SWI/SNF URS1 URS2 TATA Telophase SAGA URS1 URS2 TATA SAGA SBF URS1 TATA URS2 SWI/SNF SAGA SBF URS1 SrbMed TATA URS2 Cdk1 G1 SWI/SNF SrbMed SAGA URS1 Pol II SBF TFIIB URS2 TFIIH TATA Synthetic Biology Biologia come tecnologia … il punto di vista dell’Ingegneria Bressanone GNB 2010 9 M.P. Cosma 16/09/2010 Dal lievito ad una cellula del pancreas Telethon Institute of Genetic and Medicine Insulina Ingegnerizzare le nostre cellule per Oscillatore in cellule pancreatiche produrre insulina col ciclo giorno/notte ore Bressanone GNB 2010 10 M.P. Cosma 16/09/2010 Creare un modello in lievito per capire i meccanismi che governano le nostre cellule Modello Biologico (Dispositivi, Parti, DNA) Analisi e Ricostruzione del Sistema Modello del Sistema Modello Matematico IRMA A yeast Synthetic Network for In vivo Reverse-engineering and Modelling Assessment Bressanone GNB 2010 11 M.P. Cosma 16/09/2010 Unravelling Gene Networks ? Different approaches for Reconstructing Gene Networks Reverse Engineering Data Network Structure f(x1) = a1x1 +… anxn ………………….. f(xn) = a1x1 +… anxn Forward Engineering (Modelling) Predictions in untested conditions Network Structure f(x1) = a1x1 +… anxn ………………….. f(xn) = a1x1 +… anxn Targeted Experiments Confirm Hypothesis Validate Network Structure/Behaviour Refine model Bressanone GNB 2010 12 M.P. Cosma 16/09/2010 Aim of the Project Creating an in vivo Benchmark for Testing and Comparing different computational strategies IRMA (In vivo Reverse engineering and Modeling Assessment) 1) Synthetic Network Construction 2) Modelling 3) Reverse Engineering Construction of a Gene Synthetic Network in Saccharomyces Cerevisiae Aim: Perfectly Known Real Network Essential Features • Negligible Influence from Cellular Genes (Isolation) • Modulation of the behaviour of specific components • Choice of Transcriptional Motifs which are peculiar of Natural Networks Lee, TI et al. Science 298, 799-804 (2002) Bressanone GNB 2010 13 M.P. Cosma 16/09/2010 Creating an Isolated System How selecting genes • Non-essential and non redundant genes • Genes involved in different cellular pathways • TF genes sufficient and essential to activate transcription of a Promoter AAAAAA TF gene TF Pr gene Key proteins that trigger HO transcription: Swi5p SWI genes Swi4p Swi6p (SBF) Swi5 Swi5p is a zinc finger protein implicated in mother/daughter asymmetry. It accumulates in the cytoplasm during G2 and M phases and enters in both the mother and daughter nuclei during late anaphase. Once into the nuclei Swi5p activates transcription of several genes: HO, CDC6, ASH1, EGT2, SIC1, PCL2, PCL9. Swi4 Swi6 Swi4 Swi6 SBF SBF is a tetramer containing two molecules of Swi4p and two molecules of Swi6p. It binds to SCB elements (CACGAAA) trough Swi4p subunit. It is responsible for the expression of HO only during late G1/S phase. In addition to HO, SBF triggers transcription of other G1/S expressed genes like CLN1, CLN2, PCL1, PCL2. Bressanone GNB 2010 14 M.P. Cosma 16/09/2010 Key proteins that trigger HO transcription: Ash1 Ash1 Ash1p is a protein responsible for HO repression in daughter cells. Its transcription is triggered by entry of Swi5 into the nucleus. It accumulates at the end of anaphase in daughter nuclei. The asymmetric accumulation of Ash1p is due to the activity of SHE genes. SHE genes encode proteins needed for HO expression. She1p is involved in the transport of Ash1p from mother cells into their growing buds. Cbf1 MET16 MET16 encodes a phosphoadenosine-phosphosulfate reductase, an enzyme involved in the sulfate assimilation pathway in yeast. MET16 was first identified in a screen for methionine auxotrophs, as sulfate assimilation is required for methionine biosynthesis in yeast. Transcription of MET16 and other genes required for sulfate assimilation is activated in the absence of methionine. A complex comprising Cbf1p, the transcriptional activator Met4p, and Met28p binds to a UAS in the MET16 promoter Bressanone GNB 2010 15 M.P. Cosma 16/09/2010 IRMA Network Transcriptional regulation Cbf1 HO CBF1 GFP Gal4 MET16 GAL4 GAL10 SWI5 MYC9 Swi5A Protein-protein regulation AA Promoter Gene Tag Gal80 3xFLAGGAL80 ASH1 Ash1 2xHA ASH ASH1 1 CH2OH OH O OH OH OH Galactose Contemporarily Knock-in the new Promoter/Gene cassette and Knock-out endogenous Genes 1 Plasmid Promoter a Gene b Tag Selectable marker 2 PCR Promoter a Gene b Tag 1 Selectable marker PCR product Genomic locus Gene c 2 Recombined allele Gene b Bressanone GNB 2010 16 M.P. Cosma 16/09/2010 IRMA: sequential strains construction Background Strain Strain 1 ∆gal4 ∆gal80 ASH1 locus Strain 4 SWI5 locus ASH1 GAL80 3xFLAG ASH1 ASH1 2xHA Strain 5 Strain 2 Strain 3 SHE2 locus ACE2 locus MET16 GAL4 natMX4 H0 locus HO CBF1 GFP Strain 6: IRMA CBF1 locus GAL10 SWI5 MYC9 IRMA is switched on by Galactose -RT Cbf1 Gal4 Swi5 Ash1 Gal80 Act1 Galactose DIC Glucose 10 µm Fluorescence 10 µm Bressanone GNB 2010 17 M.P. Cosma 16/09/2010 Cell-genes controlled by IRMA are switched on by Galactose -RT Pcl9 Rme1 Cdc6 Sic1 SWI5 targets (regulate M-G1 transition) Pcl2 Met16 Gal10 Enzymes Act1 Perturb the system to gain Information Dynamic vs Static Data Bressanone GNB 2010 18 M.P. Cosma 16/09/2010 Time-series measurements after Single Perturbation Switch on or switch off IRMA by changing medium Time course expression profiles ON Measure mRNA levels at different time points 5 switch-on Timeseries (5h every 20’) 4 switch-off Timeseries (4h every 10’) OFF Steady--State Measurements after Multiple Perturbations Steady Apply single gene perturbation Measure mRNA at steady-state Steady-state expression of IRMA genes after overexpression of each of the 5 genes vs unperturbed status Aim of the Project Creating an in vivo Benchmark for testing and comparing different computational strategies IRMA (In vivo Reverse engineering and Modelling Assessment) 1) Synthetic Network Construction 2) Modelling 3) Reverse Engineering Bressanone GNB 2010 19 M.P. Cosma 16/09/2010 Steps in Model Development Modelling approach (qualitative / mechanistic / ...) ? Experimental data for identification & validation ? Logical Continuous Logical (qualitative PLDEs) Continuous non-linear linear (ODEs) Stochastic 39 IRMA as a Benchmark for Modelling k1 Non-linear Delay Differential Equations model k3 CBF1 k4 GAL4 SWI5 k5 GAL80 Galactose ASH1 k6 5 variables; 33 unknown parameters; steady state assumption for protein levels; nonnon-linear Hill functions to describe transcriptional interactions; phenomenological description of the proteinproteinprotein interaction triggered by the input; input; fixed time delay (100 min). k2 Build a Formal Model (non linear DDEs) 33 unknown Parameters Bressanone GNB 2010 20 M.P. Cosma 16/09/2010 IRMA as a Benchmark for Modelling Measuring connection strength: How can we measure the strength of TF on promoters? TF + promoter Transcritpional complex mRNA + promoter If we consider V= d[mRNA]/dt Substrate = TF Vmax/KM can be used to compare promoter strength Experimentally this means: To express each TF at different concentrations and Measure the rate of transcription of the target promoter IRMA as a Benchmark for Modelling Obtaining different levels of a TF by replacing endogenous promoter 1 Plasmid Selectable marker Inducible promoter 2 PCR Inducible promoter 1 PCR product Genomic locus Promoter TF gene (GAL4) 2 Recombined allele Inducible pr Bressanone GNB 2010 21 M.P. Cosma 16/09/2010 IRMA as a Benchmark for Modelling Promoter strength: finding parameters Met16 expression (2−∆Ct) MET16 promoter 13 Parameters (out of 33) were estimated Data Fitting Cbf1 expression (2−∆Ct) GAL10 promoter (in glucose) Fitting Gal80 expression (2−∆Ct) Gal1o expression (2−∆Ct) Gal80 expression (2−∆Ct) Experimental Data Gal4 expression (2−∆Ct) Gal4 expression (2−∆Ct) IRMA as a Benchmark for Modelling Fitting remaining Parameters on Dynamic Data Switch-on IRMA by changing medium ON Measure mRNA levels every 20’ for 5h Glucose Galactose Solve Model equation to find unknown parameters Data Fitting Bressanone GNB 2010 22 M.P. Cosma 16/09/2010 IRMA as a Benchmark for Modelling Test Model Predictive Power on Dynamic Data Switch off IRMA by changing medium OFF Measure mRNA levels every 10’ for 3h Glucose Galactose Data Prediction IRMA as a Benchmark for Modelling Predicting Static Network Behaviour Overexpress each gene and measure expression at steady-state Prediction Data Gal Bressanone GNB 2010 Glc Gal Glc 23 M.P. Cosma 16/09/2010 Aim of the Project Creating an in vivo Benchmark for testing and comparing different computational strategies IRMA (In vivo Reverse engineering and Modeling Assessment) 1) Synthetic Network Construction 2) Modelling 3) Reverse Engineering Inference Strategies to Unravel Networks Apply a perturbation to the system Measure mRNA levels Learn model t (min) Inference algorithm Overexpression/Silencing Refine Network Local Network for the gene of interest Bressanone GNB 2010 24 M.P. Cosma 16/09/2010 IRMA as a Benchmark for Reverse Engineering BANJO ARACNE NIR and TSNI (Gardner, et al, (Hartemink, A. Nature Biotechnology, 2005.) (Basso et al., Nature Genetics, 2006) Science, 2003; Bansal et al, Bioinformatics, 2006; Della Gatta et al, Genome Research, 2008) DYNAMIC AND STEADY-STATE DYNAMIC AND STEADY-STATE (n-way) (2-way) STEADY-STATE (n-way) Test reliability of the most common Reverse Engineering Strategies ODE-based Algorithms: How they work Linear Model (ODE) 2 6 1 9 = a2x2 + a6x6 +a9x9+ a12x12 12 an indicates the strength of interaction between input (gene n) and output (gene 1) an = 0 ⇒ gene n is not a regulator of gene 1 an >0 ⇒ gene n is an activator of gene 1 an <0 ⇒ gene n is a repressor of gene 1 di Bernardo D et al Nature Biotechnology Vol.23(3) 2005 Bressanone GNB 2010 25 M.P. Cosma 16/09/2010 IRMA as a Benchmark for Reverse Engineering Inference Results Random PPV = 0.4 PPV (Positive Predictive Value)= Right Connections/All Inferred Connections Se (Sensitivity)= Right Connection/True Connections IRMA as a Benchmark for Reverse Engineering Inferring Protein-Protein Interactions Consider genes involved in protein-protein interactions as one Random PPV = 0.5 Bressanone GNB 2010 26 M.P. Cosma 16/09/2010 Conclusions IRMA: a unique tool to Benchmark Reverse Engineering and Modelling Strategies • IRMA: Real system Ground-truth known Gold-standard dataset Plasmid for gene single perturbation available All proteins are tagged In-vivo: new experimental and measurements techniques can be tested Small network- but can be extended Telethon Institute of Genetics and Medicine, Naples Irene Cantone Lucia Marucci Diego di Bernardo CRG, Reprogramming and Regeneration lab, Barcelona Bressanone GNB 2010 27 M.P. Cosma 16/09/2010 Dall’ovocita fecondato all’ individuo Pre-impianto 2-cellule 1-cellula 8-cellule Post-impianto morula blastocisti ectoderma endoderma mesoderma Cosa sono e come si isolano le cellule staminali? zigote ICM Piastra di cellule staminali Dallo zigote alla blastocisti blastocisti Una cellula staminale è: totipotente si propaga indefinitamente diventa ogni cellula del nostro corpo Cellule staminali embrionali Bressanone GNB 2010 28 M.P. Cosma 16/09/2010 Differenziazione si intende gli insiemi di eventi che portano alla specializzazione di una cellula staminale in tutte le cellule dell’organismo. Cellule staminali Cellule pancreatiche Cellule ossee Cellule neuronali Beating cardiomyocytes Bressanone GNB 2010 29 M.P. Cosma 16/09/2010 Rete di controllo dei geni responsabili del mantenimento della pluripotenza nelle cellule ES Tcl1 ectoderma Pou5f1 Lhx5 Zic1 Otx1 OCT4 Hesx1 Nanog Stat3 Nanog SOX2 Hoxb1 mesoderma Rex1 Sall4 Hand1 Eomes Tbx3 Rest trofoectoderma Myf5 Sox2 T Tcf3 endoderma Dax1 Foxa2 pluripotenza Gata6 specializzazione Reprogramming (increase in potency, dedifferentiation) ? Pluripotency Differentiation Stem cell Somatic Cell Silencing of somatic cell markers Demethylation of embryonic stem cell specific promoters Expression of embryonic stem cell markers (nanog, Oct4…). Epigenetic modifications Bressanone GNB 2010 30 M.P. Cosma 16/09/2010 Riprogrammare in laboratorio… Clonazione o transfer nucleare Cellula differenziata clone ovocita Cellule staminali Animali clonati Riprogrammazione attraverso specifici geni Oct4, Sox2, c-Myc and Klf4 Oct4 Sox2 Myc Klf4 iPS Bressanone GNB 2010 31 M.P. Cosma 16/09/2010 Waddington valley Rete semi-sintetica costruita in cellule di mammifero per teorizzare il meccanismo di riprogrammazione cellulare dox dox OCT4 SOX2 KLF4 ON C-MYC ON Pluripotent genes Sox2 OCT4 promoter Oct4 Nanog promoter Nanog Nanog promoter Bressanone GNB 2010 OFF SOX2 promoter Differentiation genes ON GFP 32 M.P. Cosma 16/09/2010 Rete semisintetica Dinamica del processo Il modello deterministico prevede che le cellule somatiche siano ugualmente predisposte alla riprogrammazione verso la pluripotenza con una latenza fissa, definita come il tempo assoluto affinché la cellula somatica origini una cellula figlia riprogrammata positiva al marker fluorescente. Il modello stocastico al contrario prevede che non tutte le cellule somatiche abbiano il potenziale di essere riprogrammate con tempi di latenza diversi dovuti all’esistenza di un rumore di fondo casuale. Modello Stocastico riprogrammazione di una data cellula somatica come un unico processo che avviene in dipendenza del valore di una costante (k) che indica la predisposizione intrinseca della cellula. Le dinamiche di riprogrammazione ottenute analizzando i dati sperimentali sono risultate consistenti con la simulazioni del modello stocastico [Hanna J et al., 2009]. Bressanone GNB 2010 33 M.P. Cosma 16/09/2010 IRMA as a Benchmark for Modelling Predicting Static Network Behaviour Bressanone GNB 2010 34 M.P. Cosma 16/09/2010 IRMA as a Benchmark for Reverse Engineering Bayesian Inference Results Random PPV = 0.4 IRMA as a Benchmark for Reverse Engineering Mutual Information Inference Results Random PPV = 0.7 Bressanone GNB 2010 35