The impact of whole genome duplications: insights from Paramecium tetraurelia Genome Annotation • Ab initio gene predictions • Comparative approach • 90,000 ESTs A compact Mac genome • Protein-coding regions: 78% of the genome • Short intergenic regions Average = 352 bp • Introns: Short (average = 25 bp) … … but numerous : 80% of genes contain introns (average = 2.9 introns / gene) Gene content Number of genes Not due to annotation artefacts (control with cDNA data, distribution of protein length, manual 39642 annotated genes curation on chrom. 1, …) 45000 40000 35000 30000 25000 20000 15000 10000 5000 0 40000 39642 37500 27900 24000 20600 14000 12500 10000 6000 2000 11200 11000 9000 5200 26900 24000 28000 Many genes belong to multigenic families Computing Best Reciprocal Hits (BRH) within Paramecium proteins 39 642 proteins SW comparisons + filtering 13 085 pairs of proteins in BRH BRH are found in large duplicated blocs (paralogons). Example: scaffold 1 & 8 Building paralogons • Using a sliding window of size w genes • For each window : – Select a paralogous region if at least p % of w genes are BRH with the sequence • Merging overlapping windows • Add syntenic genes which do not have BRH Whole genome duplication (WGD) Settings : W = 10 p = 61% Coverage : 61.3 Mb (85%) 35 503 genes (90%) Résults : 24 052 genes in 2 copies (68%) 11 451 genes in 1 copie (32%) 51% of ancestral genes are still in 2 copies Progressive loss of gene duplicates Frequency (%) • ~1500 recent pseudogenes (recognizable) • Length distribution of genic and intergenic sequences : relics of more ancient pseudogenes in intergenic regions Single-copy gene Intergenic region encompassing a gene loss Other intergenic regions Sequence length (bp) BRH from supercontig 8 Number of BRH (>3000) remains outside of paralogons Inferring ancestral blocs Paralogous genes Arbitrary order Ancestral blocs Building paralogons with 131 ancestral blocs Intermediary WGD Settings : W = 10 p = 40% Coverage : 31,129 genes (79%) Content before WGD : 20,578 genes 7 996 genes in 2 copies (39%) 12 582 genes in 1 copy (61%) Old WGD Settings : W = 20 p = 30% Coverage : 18,792 genes (47%) Content before WGD : 9,999 genes 1 530 genes in 2 copies (15%) 8 469 genes in 1 copy (85%) Gene content at each WGD 19 552 genes Old WGD x 1.1 21 172 Intermediary WGD Recent WGD x2 (not x 8) x 1.2 26 214 x 1.5 39 642 Protein sequence similarity between duplicates (ohnologs) Recent WGD Intermediary WGD Old WGD Distribution of the rate of synonymous substitution (dS) between ohnologs saturation Old WGD Intermediary WGD Recent WGD dS computed with PAML Recent gene conversion Frequency (%) Distribution of dN/dS Recent WGD dN/dS • => both ohnologs are under strong negative selective pressure • Yet … the fate of most ohnologs is to be pseudogenized ! • => gene-silencing mutations can be tolerated … • … but deleterious mutations affecting the coding sequence of one copy are counterselected (i.e. dominant effect of mutations, despite the presence of a duplicate) • Once a gene has been silenced (e.g. by mutation of regulatory elements), mutations can accumulate in coding regions Gene duplicates are evolutionarily unstable Gene duplication ... Time Pseudogene Ancient paralogs Selective pressure to maintain 2 copies Retention of gene duplicates • Different (non-exclusive) models have been proposed for the retention of gene duplicates: – Robustness against mutations – Functional changes: neo- or sub-functionalization – Dosage constraints • Which are the genes that are preferentially retained after a WGD ? • How does the pattern of gene retention vary with time ? – Compare the pattern of retention after a recent WGD and a more ancient WGD – Paramecium: 3 successive WGDs ! Mutational robustness • Under certain conditions (high mutation rate and very large population size) redundant genes may be maintained by selection acting against double null alleles (Force et al. 1999) • Essential genes (e.g. ribosomal proteins) are more retained than the average • … but most of them are present in more than 2 copies ! • … their high rate of retention may be due to other factors (see later) Functional changes Function: F1F2 Function: F Time ... ... Function: F Function: F’ Neofunctionalization (adaptation) Function: F1 Function: F2 Subfunctionalization (neutral evolution) Functional changes: - changes in gene expression pattern - changes in the encoded protein Force et al. (1999) Prediction of the subfunctionalization model • A gene that has been preserved by subfunctionalization at a given WGD, is less likely to be retained in two copies at a subsequent WGD (Force et al. 1999) F1F2 WGD1 F2 WGD2 F1 F1 F2 F1F2 WGD1 F1F2 F1 WGD2 F2 Test of the subfunctionalization model (1) N=12,582 Retained: 47% Intermediate WGD Retention at the recent WGD ? N=7,996 Retained: 57% • Apparent contradiction with the subfunctionalization model • Due to variations in retention rate between different functional classes ? Test of the subfunctionalization model (2) Old WGD N = 343 gene families Intermediate WGD Retained: 67% Retention at the recent WGD ? Retained: 60% • A gene that has been preserved at a given WGD, is less likely to be retained in two copies at a subsequent WGD • Difference significant (p<5%), but not very strong • Subfunctionalization is an unlikely evolutionary pathway in species with large population sizes (Lynch Test of the neofunctionalization model • Analysis of gene expression (work in progress) • Analysis of the rate of protein evolution: Outgroup (function F) Ohnolog 1 (function F) Ohnolog 2 (function F’) • Relative rate test (PAML); correction for multiple tests • Frequency of ohnologs with asymetric substitution rates: – Recent WGD (N=2297) : 11% – Intermediate WGD (N=293 ) : 16% • More functional redundancy among recent duplicates • Functional changes account for retention on the long term Fate of neofunctionalized genes at subsequent WGD Intermediate WGD N = 62 Retention at the recent WGD ? Slow copy: 66% retained Fast copy: 26% retained Neofunctionalized genes are more prone to pseudogenization at subsequent WGD Retention for dosage constraints (1): high expression level • Genes that have to be expressed at very high level are often present in multiple copies (e.g. histones) • The loss of one copy is counterselected because it cannot be compensated for by the upregulation of other copies • => More retention among highly expressed genes Retention rates For each WGD, the retention rate for a given gene category is : Proportion of genes retained in duplicates in this category Ratio = Proportion of total genes retained in duplicates Ratio = 1 no specific retention above the mean value for all genes Ratio > 1 over-retained category Ratio < 1 under-retained category Expression versus Retention Retention for dosage constraints (2): the balance hypothesis (Papp et al. 2003) • The relative expression levels of proteins involved in a same functional network have to be controled to ensure the proper stoichiometry of the network • Initially, the loss of one copy is counterselected because it creates an imbalance within the network • On the long term, gene losses may occur because they can be compensated for by the upregulation of other copies Testing the balance hypothesis (1): Genes involved in multi-protein complexes • Protein complexes predicted by homology with yeast: – MIPS database (curation from the litterature) – TAP / MS data (Gavin et al. Nature 2006) Multi-protein complexes Genes involved in the coding of protein complexes are initially over-retained Additive effects of Expression and Inclusion in Complex • Proteins involved in complexes are overretained at the recent WGD • Does this mean that complex stoichiometry tends to be conserved ? Constraint of stoichiometry and fate of duplicates A B complex Complexes with conserved stoichiometry p-value Recent WGD 265 (44%) 74 (68%) 2.6x10-2 4.3x10-4 Intermediary WGD 114 (20%) 43 (43%) 1.5x10-3 2.4x10-4 Old WGD 106 (24%) 26 (43%) 1.2x10-5 2.5x10-3 Number of copy of A MIPS complexes Complexes from Gavin et al. Nature 2006 Number of copy of B Testing the balance hypothesis (2): genes involved in central metabolism Retention of central metabolism gene duplicates Genes involved in the central metabolism are initially overretained and then under-retained (less neofunctionalization ?) Dating genome duplications • Phylogenetic analyses of orthologous genes in other ciliate species => date WGDs relative to speciation events Tetrahymena thermophila P. bursaria P. putrinum P. duboscqui P. polycaryum Old WGD P. nephridiatum P. caudatum P. multimicronucleatum P. jenningsi Complex aurelia: 15 sibling species (same kind of habitat, Intermediate initially thought to WGD correspond to a single species) Recent WGD P. sexaurelia P. pentaurelia P. novaurelia P. primaurelia P. octaurelia P. quadecaurelia P. tredecaurelia P. tetraurelia Paramecium aurelia complex How does WGD relate to speciation? Polyploid paramecia Ptetra Pprim With the kind permission of K. Wolfe Polyploid paramecia Ptetra Pprim Mating, meiosis Dobzhansky-Muller incompatibility by reciprocal gene loss For 1 locus, 1/4 of the offspring is inviable. For n loci, offspring viability is (3/4)n Reproductive isolation Conclusions (1) • At least 3 WGDs in paramecium (probably 4) • WGDs are rare events … that occured recurrently in the evolution of eukaryotes (fungi, animals, plants, ciliates …) • Major impact on the evolution of the gene repertoire Conclusions (2) • Dosage constraints appear as an essential force shaping the gene repertoire after WGD • Functional changes contribute to gene retention on the long term … • … but the fate of the vast majority of genes is to get pseudogenized Conclusions (3) • Relationship between the number of genes and organism complexity – The number of genes is driven by selection … – … and contingency (time since the last WGD) • WGDs may be reponsible for (nonadaptative) explosive radiation of species (Dobzhansky-Muller incompatibility by reciprocal gene loss) • CNRS-UPR2167 - CGM - Gif sur Yvette – Jean Cohen – Linda Sperling • CNRS-UMR8541 – ENS - Paris – Eric Meyer – Mireille Bétermier • CNRS-UMR8125 – IGR - Villejuif – Philippe Dessen • CNRS-UMR5558 – PBIL - Lyon – Laurent Duret – Vincent Daubin • Genoscope - CNRS UMR 8030 – Jean-Marc Aury – Olivier Jaillon – Benjamin Noel – Betina Porcel – Vincent Schachter – Patrick Wincker – Jean Weissenbach