Two stories 1) reconstruction the evolution of a complex 2) Adding qualitative labels to predicted interactions Paulien Smits & Thijs Ettema Department of Paediatrics, NCMD

Introduction – MRPs Human mitoribosome –2 rRNAs, encoded by mtDNA –79 MRPs, encoded by nDNA Select candidate MRPs for genetic disease –Conservation –Function –Location 55S 28S 39S 12S 16S Science at a Distance

Objectives Detection of MRPs Orthology relations between MRPs from different species New human MRPs based on comparison with MRPs in other species Specific functions of MRPs based on comparison with MRPs in other species Extra domains in MRPs Find MRP associated proteins

New orthology relations (profile-to-profile) Human MRPYeast MRP MRPS25Mrp49 MRPS33Rsm27 MRPL9Mrpl50 MRPL24Mrpl40 MRPL40Mrpl28 MRPL45Mba1 MRPL53Mrpl44 Human MRPBacterial MRP MRPS24S3 MRPL47L29

New mammalian MRPs: Rsm22 Small subunit protein in yeast mitoribosome Orthologs in eukaryotes and prokaryotes Homologous to rRNA methylase S. pombe: fusion protein Rsm22+Cox11 Yeast: Cox11 attached to mitoribosome  Rsm22 is novel mammal MRP with a rRNA methylase function

New mammalian MRPs: Mrp10 Small subunit protein in yeast mitoribosome Yeast mutant has mitochondrial translation defect Orthologs in eukaryotes Distant homology with Cox19  Mrp10 orthologs in Mammals are novel candidate MRPs

Proteome data available Smits et al, NAR 2007

Origins of supernumerary subunits MRPL43, MRPS25 & complex I subunit

MRPL39 & threonyl-tRNA synthetase Origins of supernumerary subunits

MRPL43, MRPS25 & complex I subunit MRPL39 & threonyl-tRNA synthetase MRPL44, dsRNA-binding proteins Origins of supernumerary subunits

MRPL43, MRPS25 & complex I subunit MRPL39 & threonyl-tRNA synthetase MRPL44, dsRNA-binding proteins Mrp1, Rsm26 & superoxide dismutase

Triplication of the S18 protein in the metazoa Where do the supernumerary subunits come from?

One new, metazoa specific protein of the Large subunit (L48) has been obtained by duplication of a protein from the small subunit (S10) Where do the supernumerary subunits come from?

Addition of « new » paralogous subunits in the large and the small subunit in the metazoa Where do the supernumerary subunits come from?

Addition of a new subunit (L45 / MBA1) that is homologous to TIM44 (protein import) and bacterial proteins of unknown function

Homology between Mba1/MRPL45 and TIM44 Dolezal P, Likic V, Tachezy J, Lithgow T. Evolution of the molecular machines for protein import into mitochondria. Science 2006;313:314-8

MRPL45, Mba1 & Tim44 Mba1 is physically associated with LSU Transcription of Mba1 and MRPs is co-regulated Function of MRPL45 unknown COG4395 (MRPL45&Tim44) has similar phylogenetic distribution as COG3175 (Cox11)  Alpha-proteobacterial Tim44 is ancestor of MRPL45 and yeast ortholog Mba1, losing the N- terminus and acquiring a function in translation and COX assembly as a constituent of the mitoribosome

Extra domains

MRP interactors Translation Protein import Acyl carrier proteins Other “hypothetical gene”, essential in bacteria, Mitochondrial phenotype in yeast

Conclusions Established orthology relations between bacterial, fungal and metazoa specific ribosomal proteins Highly dynamic evolution of a mitochondrial protein complex 2 Potential novel human MRPs Homologies show diverse origins of supernumerary MRPs Some MRPs have extra domains Identification of novel MRP interactors

Acknowledgements Paulien Smits Thijs Ettema Bert van den Heuvel Jan Smeitink

Exploration of the omics evidence landscape to distinguish metabolic from physical interactions Vera van Noort Berend Snel Martijn Huynen Vera van Noort Berend Snel Martijn Huynen

Interactome Networks Important to know not only that two proteins interact but also how “the cell” “the network” the genome Snel Bork Huynen PNAS

Genomic data sets Comprehensive complex purification data (Krogan, Gavin) Shared Synthetic lethality Co-regulation (ChIP-on-chip) Co-expression Conserved co-expression (orthologous, paralogous, four species) Gene Neighborhood conservation (STRING pink) Gene CoOccurrence (STRING pink)

Complex purifications Fuse query protein with a hook Pull down hook from in vivo extracts Identify proteins that co-purify Socio-Affinity score

Synthetic lethality One knock-out not lethal, second knock- out not lethal, knock- out both lethal Points to complementary pathways Shared synthetic lethality points to same pathway

Objective: distinguish physical from metabolic in omics data We integrate omics data sets for the budding yeast S.cerevisiae because of many high quality data sets as well as classical knowledge about protein functions We construct two separate reference sets: one for physical interactions and one for metabolic interactions. Physical interactions (Mips complexes) –Remove cytosolic ribosomes –Remove “possible”, “hypothetical”, “predicted” –Remove “other” Metabolic interactions (KEGG pathways < 2000) –Remove paralogs –Remove interactions between same EC numbers –Remove interactions that are already physical

Metabolic and Physical accuracy Positive metabolicNegative metabolicPositive physicalNegative physical in binTP metaFP metaTP physFP phys A meta =TP meta / (TP meta + FP meta + TP phys + FP phys) A phys=TP phys / (TP meta + FP meta + TP phys + FP phys) A total = A meta + A phys

Physical and metabolic accuracy No single data set

Differential accuracy Good at predicting metabolic + bad at predicting physical interactions Positive metabolicNegative metabolicPositive physicalNegative physical in binTP metaFP metaTP physFP phys A meta =TP meta / (TP meta + FP meta + TP phys + FP phys) A phys=TP phys / (TP meta + FP meta + TP phys + FP phys) A total = A meta + A phys A diff = A meta – A phys

Evidence Landscape 1 Absence of physical interactions Metabolic relations in areas where proteomic approaches report no co- purification while strong indications for co-regulation. Logical in hindsight? We should not only use integrations based on the top scoring proteins but also use non-scoring proteins. Need physical protein interaction data sets where the nulls are really true nulls rather than the absence of results Absence of physical interactions Metabolic relations in areas where proteomic approaches report no co- purification while strong indications for co-regulation. Logical in hindsight? We should not only use integrations based on the top scoring proteins but also use non-scoring proteins. Need physical protein interaction data sets where the nulls are really true nulls rather than the absence of results Krogan Gavin Krogan+Gavin CoExp2Sp

Evidence Landscape 2 Krogan+Gavin CoExp2Sp Krogan+Gavin sTF*CoExp CoOcc GeNe CoExp2Sp

Network PPI C: 0.53, k 4.1 Met C: 0.031, k 2.0 Threonine biosynthesis Some pathway links between complexes

Conclusion & Discussion We can in principle distinguish metabolic and physical interactions, if 2 reference sets, if comprehensive Yet sparse (problem for multi-dimensional) Novel ways of integration and more types of omics data will allow extraction of more qualitative predictions on the nature of protein interactions

Acknowledgements EMBL –Peer Bork –Lars Juhl Jensen –Christian von Mering Department of Biology, Utrecht University –Berend Snel