1. Dissecting the dynamics of transcriptional regulatory networks MRC Laboratory of Molecular Biology Cambridge, UK MRC Laboratory of Molecular Biology Cambridge, UK M. Madan Babu Group leader M. Madan Babu Group leader

2. Overview of research Discovery of novel DNA binding proteins Data integration and function prediction Nuc. Acids. Res (2005)J Bacteriol (2008) C C H H Discovery of transcription factors in pathogens Evolution of a global regulatory hubs Nuc. Acids. Res (2006) Evolution of biological systems Evolution of transcriptional networks Evolution of networks within and across genomes Nature Genetics (2004) J Mol Biol (2006a) Uncovering a distributed architecture in networks PNAS (2008) Structure and dynamics of biological networks Structure, function and regulation of biological systems Principles of network dynamics Science (in press)Nature (2004) Methods to investigate network dynamics Submitted

3. Global dynamics (network dynamics) of regulatory networks Local dynamics (node dynamics) of regulatory networks Structure of the transcriptional regulatory network Outline

4. Networks in Biology Nodes Links Interaction A B Network Proteins Physical Interaction Protein-Protein A B Protein Interaction Metabolites Enzymatic conversion Protein-Metabolite A B Metabolic Transcription factor Target genes Transcriptional Interaction Protein-DNA A B Transcriptional

5. Organization of the transcriptional regulatory network analogy to the organization of the protein structures Basic unit Transcriptional interaction Transcription factor Target gene Basic unit Peptide link between amino acids Local structure - Motifs Patterns of interconnections Local structure Secondary structure Global structure - Scale free network All interactions in a cell Global structure Class/Fold

6. Transcriptional networks are made up of motifs Single input Motif - Co-ordinates expression - Enforces order in expression - Quicker response ArgR ArgD ArgEArgF Multiple input Motif - Integrates different signals - Quicker response TrpRTyrR AroM AroL Network Motif “Patterns of interconnections that recur at different parts and with specific information processing task” Feed Forward Motif - Responds to persistent signal - Filters noise Crp AraC AraBAD Function Shen-Orr et. al. Nature Genetics (2002) & Lee et. al. Science (2002)

7. N (k)  k  1 Scale-free structure Presence of few nodes with many links and many nodes with few links Transcriptional networks are “scale-free” “Scale-free” structure provides robustness to the system Albert & Barabasi, Rev Mod Phys (2002)

8. Summary I – Structure of transcriptional networks Transcriptional networks are made up of motifs that have specific information processing task Transcriptional networks have a scale-free structure which confers robustnessto such systems, with hubs assuming importance Madan Babu M, Luscombe N et. al Current Opinion in Structural Biology (2004)

9. Global dynamics (network dynamics) of regulatory networks Local dynamics (node dynamics) of regulatory networks Structure of the transcriptional regulatory network Outline

10. How does the global structure change in different conditions? Cell cycle Sporulation Stress Static regulatory network in yeast ~ 7000 interactions involving 3000 genes and 142 TFs Across all cellular conditions Global dynamics of the regulatory networks How does the local structure change in different cellular conditions?

11. DNA damage Cell cycle Sporulation Diauxic shift Stress Regulatory program specific transcriptional networks Multi-step Processes Regulatory programs involved in development Binary Processes Regulatory programs involved in survival Pre-sporulation Sporulation Germination E L M G0 G1 S G2 M

12. Temporal dynamics of local structure Network motifs that allow for efficient execution of regulatory steps are preferentially used in different regulatory programs fast acting & direct slow acting & indirect fast acting & direct Multi-step regulatory programs (development) Binary regulatory programs (survival)

13. Temporal dynamics of global structure (hubs) Condition specific hubs Each regulatory program is triggered by specific hubs Permanent hubs Active across all regulatory programs Hubs regulate other hubs to trigger cellular events This suggests a dynamic structure which transfers ‘power’ between hubs to trigger distinct regulatory programs (developmental & survival)

14. Sub-networks re-wire both their local and global structure to respond to cellular conditions efficiently Luscombe N, Madan Babu M et. al Nature (2004) binary conditions more target genes per TF shorter path lengths less inter-regulation between TFs Diauxic shiftDNA damageStress Quick response multi-stage conditions fewer target genes per TF longer path lengths more inter-regulation between TFs Cell cycleSporulation Fidelity in response

15. Summary II – Temporal dynamics of network structure Different regulatory proteins become global regulatory hubs under various cellular conditions. This allows for transition between regulatory programs by switching on condition specific hubs Luscombe N, Madan Babu M et. al Nature (2004) Network motifs are preferentially used by different regulatory programs. This allows for efficient response to different cellular conditions

16. Global dynamics (network dynamics) of regulatory networks Local dynamics (node dynamics) of regulatory networks Structure of the transcriptional regulatory network Outline

17. Dynamics in transcription and translation Each node in the network represents several entities (gene, mRNA, and protein) and events (transcription, translation, degradation, etc) that are compressed in both space and time

18. 1 3 4 5 6 78 10 2 9 1 3 4 5 6 78 2 9 3 4 5 6 78 2 9 1 Higher-order organization of transcriptional regulatory networks Flat structureInherent higher-order organizationHierarchical structure

19. Transcription Factor Target gene Hierarchical organization of the yeast regulatory network MET32 OAF1 PIP2 MBP1* MAC1 ARG80ARO80 UME6* HSF1*◄INO2 NDT80 ABF1*◄MCM1◄ SKN7* MAL33 NRG1* HAL9 AZF1 MET31 IXR1 CHA4 ZAP1 ACE2 CUP9 GAT3 HMLALPHA2 PLM2* SMP1 TOS8* YOX1* ADR1 DAL80 GCN4* INO4 PUT3 SOK2* TYE7 AFT1* DAL81 GLN3 LEU3 RAP1*◄ STE12* XBP1 AFT2* FHL1*◄ GTS1 MIG1 REB1*◄ SUT1 YAP1 ARG81 FKH1 GZF3 MIG2 RIM101 SWI4* YAP5* ASH1 FKH2* HAP1 MSN2* ROX1 SWI5 YAP6* CBF1* GAL4 HAP4 MSN4* RPN4 TEC1* YAP7* CIN5* GAT1 HCM1* PHD1* SKO1 TOS4* YHP1 RTG3 HMRA1 RME1 SFL1 SRD1 HMRA2 SIP4 STP2 PDR3 CAT8 SPT23 BAS1 HMLALPHA1 MATALPHA1 PDR8 HMS1 STP1 PHO4 UGA3 GCR1◄ STB4 FZF1 MET4◄ RGT1 YAP3 MET28 STB5 HMS2 PDR1 SUM1 YDR026C MGA1* ACA1 CUP2 CRZ1 CST6 MIG3 YRR1 CAD1 ECM22 LYS14 MSN1 RGM1 YER130C HAC1 MAL13 MSS11 SFP1 YER184C PDC2◄ RTG1 MOT3 RDS1 WAR1 YML081W ARR1 DAT1 URC2USV1 OTU1 RPH1 PHO2 EDS1 FLO8* PPR1 RDR1 RLM1 STP4 THI2 UPC2 YBL054W YDR266C YJL206C YKR064W YRM1 Level 7 Level 6 Level 5 Level 4 Level 3 Level 2 Level 1 Top (25) Core (64) Bottom (59) YPR196W ◄ Essential genes* Regulatory hubs DAL82 IME2 YDR049W Topological Sort Top layer Core layer Bottom layer Target Genes

20. Hierarchical organization of regulatory proteins Regulatory proteins are hierarchically organized into three basic layers Do TFs in the different hierarchical levels have distinct dynamic properties? Target Genes Top layer (29) Core layer (57) Bottom layer (58)

21. Datasets characterizing dynamics of transcription and translation Transcript abundances for yeast grown in YPD (S. cerevisiae) and Edinburgh minimal medium (S. pombe) were determined by using an Affymetrix high density oligonucleotide array. Transcript abundance (Holstege et al., 1998) Transcript half-life (Wang et al., 2002) (Yang et al., 2003) Transcript half-lives were determined by obtaining transcript levels over several minutes after inhibiting transcription. This was done using the temperature sensitive RNA polymerase rpb1-1 mutant S. cerevisiae strain. Transcriptional dynamics Protein half-life (Belle et al, 2006 ) Protein half-lives were determined by first inhibiting protein synthesis via the addition of cyclohexamide and by monitoring the abundance of each TAP-tagged protein in the yeast genome as a function of time. Estimates of the endogenous protein expression levels during log-phase were obtained by TAP- tagging every yeast protein for S. cerevisiae. Protein abundance (Ghaemmaghami et al, 2003) Translational dynamics

22. Regulation of regulatory proteins within the hierarchical framework Regulation of protein abundance and degradation appears to be a major control mechanism by which the steady state levels of TFs are controlled post-transcriptional regulation plays an important role in ensuring the availability of right amounts of each TF within the cell Regulation of transcript abundance or degradation does not appear to be a major control mechanism by which the steady state levels of TFs are controlled

23. 5 3 19 6 8 2 12 1 7 1 3 10 4 9 0 3 1 8 Population of cells No. of copies of Protein i Frequency No of copies Almost no noise Frequency No of copies Noisy Frequency No of copies Very Noisy Noise in protein levels in a population of cells Noise in a population of cells can be beneficial where phenotypic diversity could be advantageous but detrimental if homogeneity and fidelity in cellular behaviour is required

24. Noise levels of regulatory proteins Regulatory proteins in the top layer are more noisy than the ones in the core or bottom layer More noisyLess noisy Target Genes Top layer Core layer Bottom layer

25. Implication Underlying network Noise in TF expression may permit differential utilization of the same underlying regulatory network in different individuals of a population Differential utilization of the same underlying network by different individuals in a population of cells Individual 2 Individual 3 Individual 4 Individual 1

26. Noise in expression of a master regulator of sporulation in yeast Gene network controlling sporulation High variability in the expression of top-level TFs in a population of cells may confer a selective advantage as this permits at least some members in a population to respond quickly to changing conditions Nachman et al Cell (2008)

27. Summary III – local dynamics of regulatory networks Our results suggest that the core- and bottom-level TFs are more tightly regulated at the post-transcriptional level rather than at the transcriptional level itself Our findings suggest that the interplay between the inherent hierarchy of the network and the dynamics of the TFs permits differential utilization of the same underlying network in distinct members of a population

28. Conclusions Global dynamics Local dynamics Different regulatory programs preferentially use different network motifs Top, core and bottom level TFs display distinct patterns in protein abundances and half-lives Condition specific hubs allow for transition between regulatory programs Top level TFs are more noisy than the core or bottom level TFs Regulatory networks are dynamic within the timescale of an organism and the interplay between the global and local dynamics makes the network both robust and adaptable to changing conditions May permit robust response to changing conditions by efficiently re-wiring the active regulatory network May provide flexibility in response in a population of cells, making the organism more adaptable to changing conditions

29. Sarah Teichmann MRC-LMB, UK MRC - Laboratory of Molecular Biology, UK National Institutes of Health, USA Schlumberger and Darwin College, Cambridge, UK L Aravind Teresa Przytycka NCBI, NIH, USA Nick Luscombe EBI, Hinxton, UK Mark Gerstein Haiyuan Yu Mike Snyder Yale University, USA Acknowledgements Raja Jothi S Balaji NCBI, NIH, USA Arthur Wuster Joerg Gsponer MRC-LMB, UK Joshua Grochow Chicago University, USA

30. 4a. Run iterative leaf-removal algorithm Layer 3 Layer 4 1 Layer 1 9 Layer 2 2 7,8 3,4, 5,6 11 10 5. Invert the levels Layer 4 1 Layer 2 10 3,4, 5,6 Layer 1 9 Layer 3 7,8 2 11 1. Identify strongly connected components (SCCs) 9 3 5 4 6 1 2 78 11 10 2. Obtain Directed Acyclic Graph (DAG) by collapsing the SCCs 9 3,4, 5,6 1 7,8 2 11 10 3. Construct the transpose of DAG (reverse the direction of edges) 10 3,4, 5,6 1 7,8 2 11 9 hierarchy 6. Combine to d de ce u Layer 3 Layer 4 1 Layer 1 9 Layer 2 2 3,4, 5,6 10 7,8 11 Layer 3 Layer 4 1 Layer 1 7 Layer 2 2 10 11 7. Construct the final hierarchy 345 89 6 4b. Run iterative leaf-removal algorithm Layer 4 9 Layer 2 10 3,4, 5,6 Layer 1 1 Layer 3 7,8 2 11 4b. Run iterative leaf-removal algorithm Topological Sort: An approach to infer hierarchical organization in networks Identify strongly connected components (SCCs) Obtain Directed Acyclic Graph (DAG) by collapsing the SCCs Run iterative leaf-removal algorithm Construct the transpose of DAG (reverse the direction of edges) Run iterative leaf-removal algorithm Invert the layers Combine to deduce the hierarchy Construct the final hierarchy

31. Applicable on cyclic networks Consistent ordering Scalable Allows for ambiguity (e.g. Mouse) Food web network 1 2 3 4 5 6 2-4 Leaf-removal algorithmBFS-level algorithm Topological sort predator prey 1 2 3 4 1 2 3 4 ? Methods to infer hierarchical organization in networks

32. Yeast transcriptional regulatory network Transcription Factor (TF) Target gene (TG) 156 TFs (regulatory proteins) 4,403 TGs (proteins being regulated) 12,702 links (regulatory interactions) In-degree (#incoming connections) = 2.9 On average, each gene is regulated by ~3 TFs Out-degree (#outgoing connections) = 81.4 On average, each TF regulates ~80 genes

33. y = 43.706x -1.4426 R 2 = 0.87 y = 1E-08x 3 - 8E-06x 2 + 0.0015x + 0.0122 R 2 = 0.47 y = 145.13x -1.8078 R 2 = 0.92 y = 1.8904x -0.7092 R 2 = 0.57 Topological properties of regulatory proteins within the hierarchical framework Target Genes Top Core Bottom p < 10 -4

34. Target Genes Top Core Bottom p < 0.0644 p < 10 -4 p < 0.041 p < 3x10 -3 Genetic and Evolutionary aspects of regulatory proteins within the hierarchical framework

35. Do different proteins become hubs under different conditions? Is it the same protein that acts as a regulatory hub across multiple conditions? Cell cycleSporulationDiauxic shiftDNA damageStress Temporal dynamics of global structure Condition specific networks are all scale-free Internal & multi-step regulatory programs (Development) External & binary regulatory programs (Survival)

36. Target Genes Top layer Core layer Bottom layer Transcription Factor (TF) Target gene (TG) Top Layer (29) Core Layer (57) Bottom Layer (58) Unclassified (12) Higher-order organization in yeast regulatory network

39. MAC1NDT80ARG81 MCM1◄MAL33IXR1INO4GLN3LEU3GTS1MIG1MIG2HMS1 HSF1*◄INO2HAL9HAP1HAP4HCM1*HMRA1HMRA2HMLALPHA2HMS2 UME6*SUT1SWI4*SWI5TEC1*TOS4*STP2STP1UGA3SUM1 ZAP1YOX1*XBP1YAP5*YAP6*YAP7*YHP1YDR026CYAP3 OAF1PIP2TOS8*TYE7YAP1RPN4SPT23PDC2◄ PDR3MATALPHA1IME1MET28MIG3LYS14MSN1MAL13MSS11MOT3 ARO80ABF1*◄CHA4ACE2DAL80AFT1*CBF1*CIN5*ARR1DAT1 SKN7*SMP1SOK2*STE12*ROX1SKO1RTG3SIP4RTG1RPH1 PLM2*PUT3RAP1*◄REB1*◄RIM101PHD1*RME1PHO4RGT1PHO2 HMLALPHA1BAS1GCR1◄ACA1CUP2CRZ1CST6CAD1ECM22EDS1 GAT3GCN4*FHL1*◄FKH1GZF3GAL4GAT1FZF1HAC1FLO8* SFL1SRD1PDR8STB4RGM1SFP1RDS1PPR1RDR1RLM1 YRR1YER130CYER184CYML081WYJL206CYKR064WYRM1YPR196W ARG80AZF1CUP9ADR1DAL81AFT2*ASH1FKH2*CAT8DAL82 MET32MBP1*NRG1*MET31MSN2*MSN4*MET4◄PDR1MGA1*OTU1 STB5WAR1URC2USV1STP4THI2UPC2YBL054WYDR266CYDR049W Level 4 Level 3 Level 2 Level 1 ◄ Essential genes* Regulatory hubsTFs with zero in-degree A BFS-level algorithm on Yeast transcription network

40. NDT80 AFT2 UME6 AFT1 OAF1 YML081WSTB4RGM1YER130C Level 1 Level 2 Level 3 Level 4 Level 2 Level 3 GAT1 ARG81 CUP9 YAP6XBP1 IME1 RDS1ACA1 Level 2 Level 3 Level 4 Level 2 Strongly Connected Component B An instance of inaccurate hierarchy as inferred by the BFS-level algorithm

41. “Scale-free” networks exhibit robustness Robustness – The ability of complex systems to maintain their function even when the structure of the system changes significantly Tolerant to random removal of nodes (mutations) Vulnerable to targeted attack of hubs (mutations) – Drug targets? Hubs are crucial components in such networks

42. Temporal dynamics of global structure (hubs) CCSPDSDDSRTF 25045203015 Swi6 Condition specific hubs Permanent hubs Connectivity profile: No of regulated genes Existence of condition specific hubs Each regulatory program is triggered by specific hubs Presence of permanent hubs across all regulatory programs

43. Hubs regulate other hubs to trigger cellular events This suggests a dynamic structure which transfers ‘power’ between hubs to trigger distinct regulatory programs (developmental & survival)