Epigénétique et prédiction phénotypique

Size: px
Start display at page:

Download "Epigénétique et prédiction phénotypique"

Transcription

1 Franz Marc, The Red Bull Epigénétique et prédiction phénotypique H. Kiefer, 14/03/2017 Biologie expérimentale animale et modélisation prédictive A G R I C U L T U R E A L I M E N T A T I O N E N V I R O N N E M E N T

2 Epigenetics and phenotype prediction I. Why can epigenetic analyses help to refine phenotype prediction? II. From twin studies to «Epigenome Wide Association Studies» (EWAS) in Human III. The situation in livestock species with a focus on cattle IV. Conclusion A G R I C U L T U R E A L I M E N T A T I O N E N V I R O N N E M E N T.02

3 Epigenetics and phenotype prediction I. Why can epigenetic analyses help to refine phenotype prediction? II. From twin studies to «Epigenome Wide Association Studies» (EWAS) in Human III. The situation in livestock species with a focus on cattle IV. Conclusion A G R I C U L T U R E A L I M E N T A T I O N E N V I R O N N E M E N T.03

4 Epigenetics : an additional layer of information (1st layer=genetics) Alter gene regulation in a DNA-sequence independent fashion Transmitted to daughter cells through mitosis (meiosis???) 1 DNA methylation (hydroxylation etc) 2 Histone posttranslational modifications >100! 3 Non-coding RNAs A G R I C U L T U R E A L I M E N T A T I O N E N V I R O N N E M E N T.04

5 DNA methylation is a covalent but reversible modification of the genomic DNA De novo methylation (unmethylated CpGs) Maintenance methylation (hemimethylated CpGs).05

6 Factors that control the methylome 2 Status of the cell (tissue, environment, physiology ) 3 Demethylation: active & passive CpG 4 Stochastic factors 1 Sequence Error frequency =

7 Why is DNA methylation convenient for phenotype prediction? Fonctional point-of-view: balance between stability (methylome is less changing than transcriptome) and plasticity Not always correlated to gene activity Long-term memory of past environmental exposures Systemic variations («metastable epiallele») Technical point-of-view: covalent modification information easy to access compared to other epigenetic marks P = G + E + G*E + ε.07

8 The regulatory content of the methylome CpG islands Max variation range: 40% (Schübeler, Eur. J. Clin. Invest. 2015) Promoters of pluripotency and tumor suppressor genes Correlation with transcriptional activity the most interesting for phenotype prediction Tend to disappear??? m C -> T not repaired C -> U repaired Not depicted on this scheme : structural methylation: weak variability because necessary for genome integrity (repression of mobile elements, X inactivation, pericentromeric heterochromatin, usually fully methylated and CpG dense).08

9 DNA methylation and gene regulation (Schübeler, Nature 2015) But the relationships between methylation and expression are not always so simple And the regulatory elements such as promoters and enhancers are not always perfectly delineated.09

10 DNA methylation and phenotype construction Methylation marks supervising cell differentiation and development different between cell types, relatively conserved among individuals One genotype, many cell types.010

11 DNA methylation and phenotype construction Methylation marks supervising cell differentiation and development different between cell types, relatively conserved among individuals One genotype, many cell types Plasticity of some methylation marks (for instance, marks that control metabolism, inflammation ) - Basis of the phenotypic variability that exists between genetically identical animals (twins, clones) - Adaptative response of the genome to environment modifications Laurel et Hardy Bressonvilliers (UCEA-INRA).011

12 Plasticity of DNA methylation is particularly important during reprogramming periods Fertilization and gestation Gametogenesis Wang et al., Cell

13 3 generations potentially targeted by adverse environmental conditions during gestation F1 Maternal environment Nutrition Metabolism Stress Pollution Mastitis. F0 F2 From F3 : transgenerational inheritance.013

14 What about the father? Sperm cells of obese men bear reversible methylation alteration Some controversies about the transmission of these alteration to next generation Donkin et al, Cell metabolism, 2016 For more details about nutrition and the epigenome of next generation Anne Gabory A G R I C U L T U R E A L I M E N T A T I O N E N V I R O N N E M E N T.014

15 Not all the methylation is tissue-specific: metastable epiallele in the mouse 1) Incomplete penetrance of the kinky tail phenotype due to methylation variations at IAP element 2) Methylation variations are systemic (Waterland et al., Genesis 2006).015

16 Metastable epiallele in human Leukocytes (mesoderm) 1) Some loci display a stable methylation in two (>2?) different tissues 2) In Gambia, children conceived during the rainy season (higher content in methyl donor molecules in maternal food) display a higher methylation of metastable epialleles A part of the methylome of blood leukocytes can be used for associations with traits that involve other tissues (Dominguez-Salas et al., Nature Communication 2014).016

17 Why is DNA methylation convenient for phenotype prediction? Fonctional point-of-view: balance between stability (methylome is less changing than transcriptome) and plasticity Not always correlated to gene activity Long-term memory of past environmental exposures Systemic variations («metastable epiallele») Technical point-of-view: covalent modification information easy to access compared to other epigenetic marks P = G + E + G*E + ε.017

18 Conversion of non-methylated cytosines by sodium bisulfite treatment Methylated C are resistant Cheap (chemicals and kits) Quick (1 day) Easy to perform without specialized equipment Small amounts OK Conversion of methylation information into sequence polymorphism Clark et al, Nucleic Acid Research

19 Methylome can be explored (but not analysed) using the same technologies as the genome DNA extraction Bisulfite conversion Amplification PCR/WGA Sequencing or epigenotyping Tissue Native DNA Treated DNA PCR product Frequency C C U T 39% m C m C C C 61% A base-level information as for the genotype Polymorphisme C T For each CpG, methylation % is a quantitative variable: 0 [C/(C+T)]* Cell C C C m C m C m C Tissue

20 Illumina Infinium beadchips Bisulfite-treated genomic DNA Genotyping platform Most of them are curently using Illumina technology Only for Human 27,000 CpGs, 450,000 CpGs Recently: Infinium MethylationEPIC: 850,000 CpGs Infinium DNA methylation assay, Illumina ~30,000,000 CpGs in human genome.020

21 HumanMethylation450 BeadChip: one design to study all human traits It covers 99% of RefSeq genes, with an average of 17 CpG sites per gene region distributed across the promoter, 5'UTR, first exon, gene body, and 3'UTR. It covers 96% of CpG islands, with additional coverage in island shores and the regions flanking them. Further content categories requested by the Consortium include: CpG sites outside of CpG islands Non-CpG methylated sites identified in human stem cells Differentially methylated sites identified in tumor versus normal (multiple forms of cancer) and across several tissue types CpG islands outside of coding regions mirna promoter regions Most recent version: 850K (Infinium MethylationEPIC beadchip) Earlier version: 27K, canceroriented (

22 Enzymatic digestion RRBS method (Gu et al., 2011) End-repair / A tailing / Adapters ligation Size selection MspI cutting frequency Repeated sequences CpG-rich regions ~ 10% of total CpGs CpG desert Fragments size Bisulfite conversion Amplification NGS.0*.022

23 Epigenetics and phenotype prediction I. Why can epigenetic analyses help to refine phenotype prediction? II. From twin studies to «Epigenome Wide Association Studies» (EWAS) in Human III. The situation in livestock species with a focus on cattle IV. Conclusion.023

24 Twins are not epigenetically identical 1) Epigenetic differences increase with age 2) Gene expression differences, which is a proxy for phenotype differences, also increase with age Effect of environment? Stochasticity in the maintenance of methylation? (Fraga et al, PNAS 2005).024

25 EWAS example 1: type I diabetes on twins T1D : autoimmune disease leading to the destruction of insulin secreting cells Monozygotic twins: if one twin is T1D, the other has 50% chance to be T1D too genetic and non genetic factors may be involved Target cells: purified monocytes, 27K %meth affected twin - %meth non affected twins 1) Identification of differentially methylated CpGs in discordant twins 2) Validation in an independent cohort (singletons) Affected before diagnosis Vs non affected Affected after diagnosis Vs non affected After diagnosis Vs before diagnosis Some methylation marks have a pronostic value (Rakyan et al, Plos Genetics 2011).025

26 Example 2: combining GWAS and EWAS to predict BMI and height (I) Two complex traits with different relative contributions of genetics and environment: Height: highly heritable trait BMI: variable heritability (high in childhood, weak in adulthood) EWAS: discovery cohort and validation cohort (n=400~750), blood white cells corrected for cell counts, 450K Discovery cohort P=b 1 CpG 1 + b 2 CpG 2 + b 3 CpG 3 P= sex and age-adjusted BMI and height Validation cohort Estimated P Real P Keep CpGs significantly associated with P (up to 9 using stringent p-value threshold) Estimate coefficients b i Correlation (Shah et al, 2015).026

27 Example 2: combining GWAS and EWAS to predict BMI and height (II) meta-gwas: same approach except that only one huge cohort is used as discovery cohort (n=250,000) 1) Significant contribution of DNA methylation to BMI variance, that is probably underestimated due to the small size of the discovery cohort 2) Epigenetic and genetic effects mostly acted in an additive manner 3) No contribution of DNA methylation to height variance (Shah et al, 2015).027

28 Epigenetics and phenotype prediction I. Why can epigenetic analyses help to refine phenotype prediction? II. From twin studies to «Epigenome Wide Association Studies» (EWAS) in Human III. The situation in livestock species with a focus on cattle IV. Conclusion.028

29 Why is it interesting to considere DNA methylation in livestock species? Because current technologies offer a quantitative and accurate measurement of DNA methylation supplementary layer of information to refine phenotype prediction models based on the genotype by taking account of the environment (especially for traits with low heritability) Some domestic animals (cattle, horses) have a long production career, a long gestation, a long intergeneration interval cumulative effects of environment exposure Several generation are potentially targeted To determine which methylation marks escape epigenetic reprogramming and could be transgenerationnally transmitted To improve genome annotation.029

30 Present situation: mostly speculations Doherty et al, Animal Genetics

31 Why is it difficult to estimate the effects of epigenetic factors in economically important traits? Target tissues difficult to access for a routine analysis (mammary gland, muscle) better to study blood or hair follicle but metastable epiallele not known in livestock species Intensive genetic selection and breed specialization necessity to integrate the genotype and the epigenotype The technology used to obtain methylome must be as cost-efficient as genotyping, for a routine analysis Relationships between epigenetics and phenotype not completely understood difficult to integrate epigenetic information to phenotype prediction models Part of epigenetic marks transmitted to next generation not known difficult to integrate epigenetic information to genetic models Two examples: phenotypic variability in bovine clones and fertility of AI bulls.031

32 Why cattle clones? Farm animals are not inbred populations Intense animal selection based on the relationships between phenotype and genotype Difficulty to estimate the contribution of epigenetic factors to the elaboration of complex traits Cattle clones: a model of phenotypic variations in a context of genetic identity Goddard and Whitelaw, 2014 González-Recio et al., 2015 Keefer,

33 Why cattle clones? Farm animals are not inbred populations Intense animal selection based on the relationships between phenotype and genotype Difficulty to estimate the contribution of epigenetic factors to the elaboration of complex traits Cattle clones: a model of phenotypic variations in a context of genetic identity (Modified from Suvà et al., Science, 2013).033

34 Pathological and normal clones are obtained following nuclear reprogramming of the same donor cells Pathological placentome Normal placentome Hydrallantois, placentation defects, spontaneous abortions Delivery by c-section Large offspring syndrome, respiratory distress, limb deformities, abnormal glycaemia At necropsy, excess size and abnormal appearance of internal organs (here: liver) Clones that survive the perinatal period are mostly clinically normal Chavatte-Palmer et al., 2012 Watanabe, 2013 Hill, 2014 Maiorka et al., 2015 Keefer,

35 964, 972 (GD 266) 955, 979 (GD 267) ell donor 5538 (15 y) 578 (5 y) 428, 449, 460 (6 y) 2353 (7 y) 229 (8 y) 2 (10 y) 5538 (15 y) 3302 (GD 257) 828 (GD 263) 411 (GD 268) 3303, 3304 (GD 273) 406 (Term) 2263 (PD 4) 512 (3 y) 477 (3.5 y) 439 (4 y) 447, 468 (5 y) 248, 437 (6 y) Animals CLONES Fertilization/ Nuc. Transfert Birth (c-section) Adulthood ARTIFICIAL INSEMINATION (AI) CONTROLS Genotype 5538 Genotype 2251 Genotype 029 Perinatal clones Adult clones Perinatal AI controls Adult AI controls.035

36 Identification of 246 age-related differentially methylated regions (age-dmrs) in the liver 3302 (GD 257) 828 (GD 263) 964, 972 (GD 266) 411 (GD 268) 955, 979 (GD 267) 3303, 3304 (GD 273) ell donor 5538 (15 y) 406 (Term) 2263 (PD 4) 512 (3 y) 578 (5 y) 477 (3.5 y) 428, 449, 460 (6 y) 439 (4 y) 2353 (7 y) 447, 468 (5 y) 229 (8 y) 248, 437 (6 y) 2 (10 y) 5538 (15 y) Age Pathology VS Fertilization/ Nuc. Transfert Birth (c-section) Adulthood Age VS Genotype 5538 Genotype 2251 Genotype 029 Perinatal clones Adult clones Perinatal AI controls Adult AI controls.036

37 Identification of 83 cloning-related differentially methylated regions (cloning-dmrs) in the liver 3302 (GD 257) 828 (GD 263) 964, 972 (GD 266) 411 (GD 268) 955, 979 (GD 267) 3303, 3304 (GD 273) ell donor 5538 (15 y) VS 406 (Term) 2263 (PD 4) 512 (3 y) 578 (5 y) 477 (3.5 y) 428, 449, 460 (6 y) 439 (4 y) 2353 (7 y) 447, 468 (5 y) VS 229 (8 y) 248, 437 (6 y) 2 (10 y) 5538 (15 y) Nuclear reprogramming In vitro culture Genotype 5538 Genotype 2251 Genotype 029 Perinatal clones Adult clones Perinatal AI controls Adult AI controls.037

38 Phenotypic alterations in the livers of pathological perinatal clones Perinatal AI Fibrosis (orange staining) Alterations to the rows of hepatocytes ( ) Altered histomorphometrical parameters Perinatal clone Steatosis ( * ) Altered fatty acid composition * * * * Absence of glycogen storage Adult clones are normal.038

39 Multiple Factor Analysis (MFA) Datasets Variables Individuals From: factominer.free.fr/docs/afm.pdf Escofier and Pagès, 1994 Correlations between Esti variables belonging to different datasets.039

40 Integration of variables from three datasets using MFA DMR Epigenetic dataset - age and cloning-related DMRs variables Datasets Variables Individuals Morpho Histomorphometrical parameters variables FA Fatty acid composition variables.040

41 Dim 2 (17.16%) Dim 2 (17.16%) Age has a major effect on phenotype and DNA methylation Perinatal AI Adult clone Adult AI Perinatal clone DMR Group Morpho FA Dim 1 (42.65%) Dim 1 (42.65%) Age Dim1: DMR < Morpho < FA MFA run on all individuals with no missing data for the datasets considered.041

42 Dim 2 (17.16%) Cloning??? Cloning has no significant effect on phenotype and DNA methylation of adults Perinatal AI Adult clone Adult AI Perinatal clone Dim 1 (42.65%) MFA on perinatal animals only MFA run on all individuals with no missing data for the datasets considered.042

43 Dim 2 (20.85%) Dim 2 (20.85%) Cloning has a major effect on both DNA methylation and the phenotype of perinatal animals 3302 FA Perinatal clone Perinatal AI Morpho DMR Group Dim 1 (39.54%) Dim 1 (39.54%) Cloning Dim1: FA < DMR < Morpho MFA run on all individuals with no missing data for the datasets considered.043

44 Dim 2 (17.16%) Dim 2 (20.85%) Two distinct sets of DMRs underlie normal transition to adult life and abnormal liver phenotype Adult clone Adult AI Perinatal AI Perinatal clone 3302 Perinatal clone Perinatal AI Age-related DMRs (246) Dim 1 (42.65%) Cloning-related DMRs (83) Age-related DMRs (246) Dim 1 (39.54%) Cloning-related DMRs (83) DMRs correlated to dim1 (182) DMRs correlated to dim1 (148) Glycogen, lipid & cholesterol metabolism Stress & stimulus responses glycogen & steroid metabolism.044

45 Dim 2 (17.16%) Dim 2 (20.85%) Example : TCF7L2, Wnt pathway effector involved in gluconeogenesis and type 2 diabetes Adult clone Adult AI Perinatal AI Perinatal clone 3302 Perinatal clone Perinatal AI Age-related DMRs (246) Dim 1 (42.65%) Cloning-related DMRs (83) Age-related DMRs (246) Dim 1 (39.54%) Cloning-related DMRs (83) DMRs correlated to dim1 (182) DMRs correlated to dim1 (148) Glycogen, lipid & cholesterol metabolism Stress & stimulus responses glycogen & steroid metabolism Grant et al., 2006 Boj et al., 2012 Ip et al.,

46 NL DHA/AA PL C14:0 (%) Cell area (µm 2 ) Nu CV Shape Factor (%) Nu CV area (%) Methylation at TCF7L2 DMR is correlated to several phenotypic markers of pathology r=-0.89 r=0.61 r= Methylation (%) Decreased glycogen storage in cloned fetuses with low methylation Spearman s rank correlation test r= Methylation (%) r= Methylation (%) Fetus AI Calf AI Fetus clone Calf clone Kiefer et al.,

47 Conclusion about clones Equivalent to twin studies in human First demonstration of an association between DNA methylation and phenotypic variations in cattle at a genome-wide scale The effect of genotype on the pathology can be excluded, since adults produced from the same somatic cells have a normal phenotype But very small cohorts, because it is difficult and expensive to obtain clones!!! No economical traits!!! A proof of concept that DNA methylation is associated to phenotypic markers, and could therefore be used for phenotype production Kiefer et al.,

48 Bull semen is an important product for artificial insemination (AI) industry and for breeders 6.5 millions AI performed in France in 2013 (ruminants), 110 million AI performed every year worldwide AI allows the diffusion of valuable genotypes Important national and international competition between AI companies Success of AI = important issue for breeders Our partner ALLICE is an association of french AI companies. >100,000 breeders are using ALLICE s services.048

49 Subfertile bulls are an important issue for the AI industry Early genomic selection of candidates no more «testing», no estimation of the fertility before commercialization After puberty, only bulls with abnormal sperm parameters are excluded In case sperm parameters are ± normal, commercialization and field fertility estimation once enough females have been inseminated (nonreturn rates, NRR) New predictive markers of fertility must be identified A G R I C U L T U R E A L I M E N T A T I O N E N V I R O N N E M E N T.049

50 Field fertility = NRR 56 days post AI Prediction of bull fertility using sperm functional markers Most significant parameters : Intracellular ROS Acrosome integrity DNA compaction Mitochondrial activity Viability Velocity Morphological anomlies Fertility predicted from functional analyses on frozen semen A G R I C U L T U R E A L I M E N T A T I O N E N V I R O N N E M E N T Sellem et al., Theriogenology

51 The differentiation of sperm cells requires an unique epigenetic reprogramming Compaction of paternal genome Morphologic features mobility Protection against oxydation (epididymis, female genital tract) Carry an epigenetic information important for fertilization, embryo development and offspring phenotype Use epigenetic marks to predict bull fertility A G R I C U L T U R E A L I M E N T A T I O N E N V I R O N N E M E N T.051

52 Catalogue of the molecular markers related to semen quality DNA methylation Semen of 58 AI bulls characterized on a functional point-of-view Modifications Histones/protamines Small ncrnas Proteomic/ Lipidomic Sébastien Fritz Didier Boichard 2 breeds (Holstein & Montbéliard) 3 fertility groups (n=8-10/group) Fpos: assessed as fertile according to their genotype; field fertility in agreement Fneg: assessed as subfertile according to their genotype; field fertility in agreement Deviant: assessed as fertile according to their genotype BUT subfertile Embryo development Semen of 8 bulls with «extreme» epigenetic marks.052

53 Identification of CpGs whose methylation status varies according to fertility and breed.053

54 This list of DMCs allows an accurate prediction of the fertility group in each breed Ex: Holstein Design of a custom Illumina Infinium beadchip and EWAS on 500 bulls RO-PLS R package.054

55 Why this project will path the way to phenotype prediction using DNA methylation data Aim=new tool and new service for a better characterization of AI bulls Support from the private sector important cohorts can be provided (~500 for validation, different from experimental cohorts), to answer to problems encountered on field by practitioners Targeted animals = AI bulls, well characterized from a genetic point of view large scale integration between genotype and epigenotype Different levels of phenotypes available: field (NRR), cell level (characterization of the ejaculate), molecular level (proteomics, lipidomics ) High value animals systematic epigenotyping should be considered Sperm cells is both the target cell type for male fertility and the material for analyses The data content will also allow to answer to basic scientific questions, such as intergenerational transmission of epigenetic marks and phenotype.055

56 Epigenetics and phenotype prediction I. Why can epigenetic analyses help to refine phenotype prediction? II. From twin studies to «Epigenome Wide Association Studies» (EWAS) in Human III. The situation in livestock species with a focus on cattle IV. Conclusion.056

57 At which scale can epigenetics improve phenotype prediction? P = G + E + G*E + ε Individual Population Evolution Genetic information Genetic information??? Epigenetic information Epigenetic information Use epigenetic information to improve economically important traits on high value animals with a long period of environment exposure.057

58 Another point of view Ibeagha-Awemu & Zhao,

59 Epigenetics and Evolution: controversies.059