Discordance for exposures and disease phenotypes

Transcription

1 Discordance for exposures and disease phenotypes

2 To what extent can easilyaccessible peripheral tissues/cells (e.g. whole blood, saliva, buccal) be used as a proxy for inaccessible tissues/cells? UK Brain bank network Biomarkers Peripheral mechanisms [immune?] But careful interpretation needed.

3 Cellular heterogeneity within a tissue Infection Disease Age Stress Diet Exercise Circadian rhythm Lifestyle exposures many more

4 NeuN (neurons) Neurons Oligo Bulk Astrocytes? Sox10, Olig2 (oligodendrocytes) Pax6, Aldolase C (astrocytes) NeuN+ Sox10+ Bulk NeuN-/Sox10-

5 Controlling for neural cell differences Neuronal cell proportions derived from 5mC data Gene expression data ENO2 neurons CD68 microglia OLIG2 oligodendrocytes GFAP astrocytes CD34 - endothelial Covariate in analysis models

6 Individual cells can exhibit substantial differences even when derived from an apparently homogenous population Stochastic variation Cell-specific pathogenic changes Single-cell transcriptomic and epigenomic profiling

7 Applications of epigenetics to personalised medicine??? Epigenetic biomarkers of disease Epigenetic markers of exposure Mechanisms of disease Refining traditional risk models Epigenetic biomarkers of drug response Modulation of epigenome via environmental, lifestyle, pharmacological interventions

8 Targeting the epigenome for drug discovery Target identification Connectivity mapping of test compounds

9

10 Data resources Fetal brain DNA modifications (total, 5mC, 5hmC): Fetal brain RNA-seq (Nick Bray, Cardiff): Fetal brain mqtl: GoDMC: (available soon) Heritability estimates for DNAm sites: mqtl / SMR (for fine-mapping GWAS regions):

11 Schizophrenia-associated hypermethylation at HDAC4 Dempster et al (in prep)

12 Structural genomic variation Dup15q copy number variation in autism cis-effects: Hyper- and hypo-methylation trans-effects Overlap with idiopathic autism Wong et al (in review)

13 Regulatory variation associated with elevated polygenic risk burden Striatum Cross-brain region model Viana et al (2017) Hannon et al (2018)

14 Study designs that have been used in genetic epidemiology not optimal for epigenomics Mill & Heijmans, Nature Reviews Genetics, 2013

15 Regulatory genomic signatures of clozapine exposure in schizophrenia BUB1 TMCO6 DIS3L2 TMEM194A TMUB2 S100A4 LCK MHC 65

16

17 Clozapine-induced transcriptional networks in the zebrafish brain module 1 (455 genes) Connectivity for each gene defined as the sum of connection strengths with other network genes CONTROL HIGH LOW P = 4.94E-244

18

19 Critical windows of environmental sensitivity in schizophrenia across the life-course Embryo/fetus Rutten B P F, Mill J Schizophr Bull 2009;35:

20

21 Kundacovik et al, Biol Psychiatry Jan 15; 81(2):

22 Gene regulation: relevance to disease Scenario A Scenario B significant expression The DNA sequence determines what specific mrna molecules are synthesized large amounts of protein no expression Epigenetic regulation determines how much of the mrna is made, and where and when it is synthesized Even genes without any disease-causing mutation/polymorphism may be pathogenic no protein

23 Parallel trajectories of 5mC and 5hmC 5hmC male 5hmC female 5mC male 5mC female Interaction between trajectories of 5mC and 5hmC Spiers, Hannon et al (2017)

24 5hmC particularly enriched in brain 5hmC generally depleted in promoter CGIs

25 Not differentiating between different DNA modifications in the brain can majorly confound associations with disease. Standard sodium bisulfite approaches cannot distinguish 5mC from 5hmC may be critical in analysis of CNS phenotypes. 5hmC: corr = 0.45 DNAmod: corr = mC: corr = Smith et al (in review)

26

27

28 Fetal mqtls in schizophrenia-associated regions have larger effects on DNA methylation during neurodevelopment than the in adult brain PFC P = , STR P = , CER P =

29 Clozapine-induced behavioural changes Clozapine exposure had a striking dose-dependent effect on spatial position in the tank General activity Feeding Spawning No increase in stress No hypoxia Social hierarchy maintained No overt physiological effects

30 Clozapine-induced transcriptional changes DMSO Clozapine No gene expression differences were identified between the water control and DMSO vehicle groups. Gene ID Widespread gene expression changes associated with clozapine exposure Gene symbol log2 fold change (low) log2 fold change (high) P-value Human orthologue gene symbol ENSDARG atp2a1l E-06 - ENSDARG odc E-06 ODC1 ENSDARG arrdc3b E-05 - ENSDARG angptl E-05 ANGPTL4 ENSDARG slc16a9b E-05 SLC16A9 / MCT9

31 Clozapine-induced transcriptional changes Dose-dependent changes in gene expression Ornithine decarboxylase is an enzyme which catalyses ornithine to putrescine. Higher blood concentration of ornithine has been reported in the plasma of schizophrenia patients (He et al., 2012). Increased expression of ornithine decarboxylase in the hippocampus of a rat model of schizophrenia (Bernstein et al., 1999).

32 Clozapine-induced transcriptional networks in the brain module 1 (455 genes) Connectivity for each gene defined as the sum of connection strengths with other network genes Control: , Low exposure: , High exposure:

33 Data resources Fetal brain DNA modifications (total, 5mC, 5hmC): Fetal brain mqtl: GoDMC: Other databases:

34

35 The epigenetic clock : DNA methylation age DNAm age: calculated using the weighted average of DNA methylation levels at 353 CpG sites epigenetic age acceleration: calculated by contrasting DNAm age with chronological age Is accellerated epigenetic aging associated with diseases of aging?

36 but evidence for an accelerated epigenetic clock in neurodegenration. r = 0.89, P < 1.0E-16 N = 1,366 P = CONTROL ALS P =

37 Genomic profiling of autism brain (BA9, BA41/42, cerebellum) consistent DNA methylation differences across cases Differentially regulated autism genes strongly enriched for neurodevelopmental pathways Wong et al (in prep)

38 Methylomic variation associated with dup15q autism (BA9, BA41/42, cerebellum) trans-effects Wong et al (unpublished)

39 Despite striking dup15q ciseffects, highly-correlated DNA methylation alterations in iasd and dup15q patients. Count Color Key and Histogram Value Blue iasd, Green dup15q, Black Control Wong et al (unpublished)

40 SMR results across a GWAS region on chromosome 12 where differential DNAm is associated with with schizophrenia. Hannon et al, in prep

41 Overlap of mqtl in GWAS regions Not sufficient to show causal or even pleiotropic effects

42 Stage 1 b GWAS (Estimated in GWAS) Trait SNP b EWAS = b GWAS /b mqtl Stage 2 b mqtl (Estimated in mqtl study) Tests for association between DNAm and trait Causal SNP b GWAS Trait Causal SNP α DNA methylation b GWAS mqtl datasets blood, human fetal brain, adult brain regions Trait b EWAS = αb GWAS /βb mqtl α tag SNP α b mqtl b EWAS = αb GWAS /αb mqtl DNA methylation tag SNP In development: matched eqtl and H3K27ac/etc QTL datasets (brain) β Causal SNP DNA methylation

43 A suite of epigenetic modifications act to fine-tune genomic function Zhou et al (2011)

44 Stage 1 b GWAS (Estimated in GWAS) Trait SNP b EWAS = b GWAS /b mqtl Stage 2 Pleiotropy no heterogeneity Causal SNP b GWAS b mqtl (Estimated in mqtl study) Trait Linkage - heterogeneity Causal SNP α DNA methylation b GWAS Trait b EWAS = αb GWAS /βb mqtl α tag SNP α b mqtl b EWAS = αb GWAS /αb mqtl DNA methylation tag SNP β Causal SNP DNA methylation

45 Bayesian co-localisation H 0 : there exist no causal variants for either trait; H 1 : there exists a causal variant for one trait only, schizophrenia; H 2 : there exists a causal variant for one trait only, DNA methylation; Bayesian Co-localisation Do the same genetic variants influence risk of schizophrenia and DNA methylation? i.e. do the GWAS signals overlap? H 3 : there exist two distinct causal variants, one for each trait; H 4 : there exists a single causal variant common to both traits.

46 Differential aging effects across different regions of the human brain cortex cerebellum Hannon et al (in preparation)

47 Highly convergent H3K27ac profiles in Alzheimer s Disease cortex 4,162 of 182,065 peaks with FDR < ,687 AD-hypoacteylated peaks [1.5% of total] 1,475 AD-hyperacetylated peaks [0.8% of total] MAPT Bonferroni FDR Hyperphosphorylation of the tau protein precipitates the neurofibrillary tangles associated with the pathogenesis of AD MAPT, APP, PSEN1, PSEN2 GWAS Cortex-specific regions [CR1, TOMM40, enhancer APOE ] domains Marzi et al (Nature Neuroscience, in press)

48 Stochastic autosomal monoallelic expression is common in neural cell lineages 3.4% of all expressed autosomal genes showed random monoallelic expression Location: Extracellular region Plasma membrane Functions: Neurodevelopment al genes Axon Guidance Cell adhesion Jeffries et al (2012) Jeffries et al (2016) Fold enrichment Autism Biallelic Schizophrenia Biallelic Candidate gene lists from evidence based weighted matrices p-values (Fisher s Exact test) = and respectively

49 Erasure and re-establishment of stochastic MA NSC = Neural Stem Cell ipsc = Induced Pluripotent Stem Cell Jeffries et al (in prep)

50 Stochastic monoallelically-expressed loci are associated with promoter hypermethylation

51 The brain is a mosaic of clonal cell-lineages characterized by stochastic patterns of monoallelic gene expression Cellular level AB A B Random inactivation of alleles in different cells A A B AB Different A combinations of expressed alleles in a cell population AB B A B ABDiversity A of possible outcomes (cell fate, disease) AB A B A necessity for correct development? Factor in phenotypic discordance between monozygotic twins?

52 Novel isoform / splicing diversity in the developing human brain Evidence of alternative splicing at 95% of human genes Particularly prevalent in the CNS Developmentally dynamic Functional diversity, potentially antagonistic functional effects SATB2 Jeffries et al (in prep)