Mapping the human interactome: a update
the genomic revolution in numbers
from gene sequence to protein function
large-scale protein interaction mapping yeast two-hybrid binary protein interactions transient AP/MS protein complexes stable different network topology different interactome subspace interrogated >> complementary similar high quality from Yu et al. High-quality binary protein interaction map of the yeast interactome network. Science 2008
MAPPIT validation of Y2H protein network maps
yeast two-hybrid
other two-hybrid methods
other two-hybrid methods
MAPPIT operates in mammalian cells ligand-inducible > extra level of control simple readout > automation
MAPPIT validation of Y2H protein network maps >> CCSB-YI1: 1.809 interactions between 1.278 proteins (estimated interactome size 18.000 +/- 4.500)
MAPPIT validation of Y2H protein network maps WI-2007: 1.816 interactions between 1.496 proteins (estimated interactome size 115.600 +- 26.400)
MAPPIT validation of Y2H protein network maps ~700 full length (bait) x ~700 fragments (prey) 40 fragments per ORF >> 755 interactions between 522 proteins (only 92 previously identified by Y2H !) many interactions that you don’t see in Y2H with full lengths we do see in mappit >> complementary
MAPPIT validation of Y2H protein network maps framework for large-scale Y2H human interactome mapping -validation of available HT-YTH interactome maps: (Vidal & Wanker groups) >> high quality (> literature curated) estimation of interactome size: ~130.000 interactions
MAPPIT validation of Y2H protein network maps framework for large-scale Y2H human interactome mapping -validation of available HT-YTH interactome maps: (Vidal & Wanker groups) >> high quality (> literature curated) estimation of interactome size: ~130.000 interactions -standardized confidence scoring method
empirical confidence score from Braun et al. An experimentally derived confidence score for binary protein-protein interactions. Nature Methods 2009
empirical confidence score from Braun et al. An experimentally derived confidence score for binary protein-protein interactions. Nature Methods 2009
mapping the human interactome 3 year NIH grant Y2H: 16.000 x 16.000 full lenght human ORFs (~ 50% of total matrix of 22.000 x 22.000) interaction toolkit re-test: ~25-30.000 interactions (~10.000/year; ~20% of the map)
what did we learn ?
benchmarking binary interaction mapping methods >> MAPPIT performance is similar to that of the other tested methods from Braun et al. An experimentally derived confidence score for binary protein-protein interactions. Nature Methods 2009
benchmarking binary interaction mapping methods -mappit sees two phosphorylation-dependent interactions (as do PCA, the YFP complementation technique, and remarkably also Y2H); of course this is done in untreated cells and probably other phosphorylation dependent interactions could be seen upon activation of the proper signalling cascade during the assay >> the interaction mapping methods are highly complementary from Braun et al. An experimentally derived confidence score for binary protein-protein interactions. Nature Methods 2009
the ORFeome collection 15.483 full length human ORFs derived from Mammalian Gene Collection (MGC) cloned in Gateway vectors from http://horfdb.dfci.harvard.edu/
MAPPIT for large-scale interactome analysis ? high quality assay access to a large collection of easily transferred cDNAs different and complementary network subspace probed > screening for novel interactions
towards an efficient screening format: reverse transfection
ArrayMAPPIT screening prey (+reporter) plasmid human ORFeome collection transfection reagent reverse transfection mix MAPPIT prey collection -/+ ligand MAPPIT bait cell line luciferase read-out MAPPIT prey array (stable for months !) 24
current screening setup prey collection: 2.000 human ORF preys (GO annotation “signal transduction”) assay format: 96well > 384well automation: Tecan EVO150 (DNA preps) Tecan EVO200/Perkin-Elmer Envision (array production array + assay read-out)
screening for interaction partners of E3 ligase complex adaptors “Specificity module”: SCF – Skp1 + F-box protein ECS – ElonginB/C + SOCS-box protein
SKP1 screen 10-fold cut-off >> 5 hits: 3 known (blue), 3 novel (green); all F-box proteins no other known Skp1 interaction partners in the array 27
Elongin C screen 10-fold induction 5-fold induction 3-fold induction 10-fold cut-off >> 5 hits: 4 known and 1 novel (all SOCS-box proteins) 5-fold cut-off >> 8 additional hits: 4 known interactors (all SOCS-box proteins) 3-fold cut-off >> 14 additional hits: 2 known and 1 novel interactor (all SOCS-box proteins) 6 false negatives 28
Co-IP confirmation SKP1 Elongin C WB anti-E WB anti-Flag mock lysate IP anti-Flag WB anti-Elongin C FBXW11 FBXW9 FBXO46 SOCS2 SPSB2 SPSB4 SKP1 Elongin C
MAPPIT cDNA library screening hIL5Rα Y anti-hIL5Rα anti-PE magnetobead bait LR-F3 CMV hIL5RαΔcyt rPAP1 anti-mIgG-PE FACS sort mEcoR MACS enrichment prey gp130 5’LTR CD90 retroviral prey cDNA library
SKP1 screen 6 known SKP1 interacting proteins Symbol Description Number of clones (fusions) FBXL8 F-box and leucine-rich repeat protein 8 9 (5) FBXL15 F-box and leucine-rich repeat protein 15 1 (1) FBXW5 F-box and WD domain protein 5 12 (5) FBX044 F-box protein 44 9 (7) FBXO2 F-box protein 2 CDCA3 cell division cycle associated 3 FBXL6 F-box and leucine-rich repeat protein 6 3 (1) FBXW9 F-box and WD-40 domain protein 9 5 (3) 6 known SKP1 interacting proteins 2 novel interaction partners (both F-box proteins)
Array versus cDNA library screening cDNA library screening array screening ‘open’: large & diverse prey pool ‘closed’: fixed set of preys labour intensive fast prey identification is tedious position in array determines prey identity
MAPPIT for large-scale interactome analysis ? high quality assay access to a large collection of easily transferred cDNAs different and complementary network subspace probed > > screening for novel interactions mammalian background
yeast two-hybrid interaction maps are static the human interactome is not static but dynamic many protein-protein interactions are conditional or context-dependent require post-translational modifications and/or structural alterations require co-factors, adaptors or regulatory proteins yeast cell doesn’t provide the normal cellular environment for human proteins no accessory proteins no modifications no context-dependent interactions example proteins binding to receptor tail > some only upon activation of receptor sh2 domain binding to tyrosine: only after phosphorylation requires kinase activity
MAPPIT for large-scale interactome analysis ? high quality assay access to a large collection of easily transferred cDNAs different and complementary network subspace probed > > screening for novel interactions mammalian background > > mapping protein network dynamics
mapping dynamic aspects of protein networks ? treatment B treatment C -/+ ligand treatment A MAPPIT bait cell line 36
mapping dynamic aspects of protein interactions: GR signalling cytoplasm nucleus NFkB monomer dimer -gr is a nuclear receptor that in its inactive state sits in the cytoplasm, sequestered by chaperones such as hsp90 -Glucocorticoids act through the glucocorticoid receptor which may remain as a monomer and thereby interact with transcription factors to inhibit transcription of cytokine genes (transrepression >> anti-inflammatory) or they can dimerize and thereby interact with glucocorticoid response elements (GRE) to induce transcription of genes (transactivation) for metabolic enzymes. -p53 and gr have been described to interact and mutually antagonize eachothers transactivational activities p53
MAPPIT can detect these changes in protein interactions
screening for DEX-dependent GR interactions -/+ ligand GR-bait expressing cells 39
screening for DEX-dependent GR interactions GR bait NRIP1: known to be a dex-dependent interactor NCOA4: known ligand-dependent interactor of the androgen receptor PPP5C: known interactors Skp1 bait
screening for DEX-dependent GR interactions + STAT3 – STAT5A – HGMB2 6 stably interacting proteins: STAT3, STAT5A, HGMB2 (known) HBP1, STAT4, SOCS3 GR bait NRIP1: known to be a dex-dependent interactor NCOA4: known ligand-dependent interactor of the androgen receptor PPP5C: known interactors Skp1 bait
screening for DEX-dependent GR interactions + STAT3 – STAT5A – HGMB2 6 DEX-inducible interactions: NRIP1 (known interactor) NCOA4 (AR interactor) FASTK, LPXN, SHC4, DOK3 GR bait NRIP1: known to be a dex-dependent interactor NCOA4: known ligand-dependent interactor of the androgen receptor PPP5C: known interactors Skp1 bait
screening for DEX-dependent GR interactions GR bait Skp1 bait + STAT3 – STAT5A – HGMB2 1 DEX-repressible interaction: PPP5C (known interactor) NRIP1: known to be a dex-dependent interactor NCOA4: known ligand-dependent interactor of the androgen receptor PPP5C: known interactors
screening for DEX-dependent GR interactions
ArrayMAPPIT - further development prey collection: 2.000 human ORF preys > 10.000 (end 09) assay format: 384well > glass slides (?) increase assay sensitivity – decrease assay variability data-management, optimized experimental setup, objective scoring and quality control tracking (StatGent)
CRL Jan Tavernier Dominiek Catteeuw Els Pattyn Delphine Lavens Leentje De Ceuninck Isabel Uyttendaele Celia Bovijn Laura Icardi Margarida Maia Sylvie Seeuws Lennart Zabeau Irma Lemmens Anne-Sophie De Smet Elien Ruyssinck Viola Gesellchen Tim Van Acker Frank Peelman Julie Piessevaux Peter Ulrichts Annick Verhee Joris Wauman José Van der Heyden Nele Vanderroost Dieter Defever CCSB Marc Vidal & co
screening on glass slides IL5Ra-staining assay on glass slides cell lines stably expressing endosialin bait reverse transfection of prey (rPAP-IL5RaDcyt reporter endogenous) stimulation with leptin staining with IRDye labelled antibody & scanning with LI-COR Odyssey >> optimization (sensitivity, diffusion) & down-scaling (microarray) IL5RaDcyt SVT prey p85 prey RNF41 prey TRIP13 prey LR
screening for interaction partners of RNF41 GR bait Skp1 bait