Imaging genetics: Adventures in the dopaminergic system Christian Büchel

buechel@uke.uni-hamburg.de

NeuroImage NordHamburg University Medical School Eppendorf

HBM Barcelona2010

Introductory remarks Hypothesis driven association studies

Reward processing– Predictions?

– Genetic influence on predictions

Novelty and memory– The role of DRD4

General remarks

Outline

Imaging genetics - Imaging neuroscience meets genetics

Commonalities Are interested in interindividual differences Battle the multiple comparisons problem in statistical analysis

of their data

What a Geneticist might think about Neuroscientists… They have no clue about methodology in genetics

(eg never heard of Plink) They don’t care about the heritability of their

traits They use ridiciously small sample sizes They stick to boring candidate gene approaches

and will never find out anything exciting

What a Neuroscientist might think about Geneticists … They have no clue about methodology in

neuroimaging They don’t know anything about the brain (i.e. my

ground breaking hypotheses) They advocate whole genome approaches that

nobody is able to interpret They have no clue about the costs of an MR scan Their gold standard is an uncorrected p-value of

~10^-? and think that solves the multiple comparisons problem (havn’t they used FDR before we did?)

Explaining interindividual variance

Volunteer

Simple model : 1-sample t-test Significant deactivation for the whole group in PFC A lot of unexplained interindividual variance Age effects? Gender effects? Personality effects? Genetics

effect?

Activ

atio

n in

PFC

Innate values – sucrose vs quinine

Adapted from K. Berridge

Conditioned reward

3 7

choice anticipation outcome

0time (s)

2 x 2 x 2 factorial design: PROBABILITY (12.5 [26%] – 50% [66%]) MAGNITUDE (1 – 5€) OUTCOME (win – lose)

20€ 15€

20€ 21€

Anticipation phase: Expected reward magnitude & probability

probability high > low

y = 15 mm R

y=3mmmagnitude 5€ > 1€

Rz = 0 mmy = 3 mm R

Yacubian et al., J Neuroscience 2006

Which one would you chose ?

10€ / 70% or 100€ / 50% EV 7 EV 50

Val/Val

Met/Met

Val/Met

Schott et al., 2006 Bertolino et al., 2006

DAT- COMT interactions

DAT reuptake of dopamine Variable number of tandem

repeats (VNTR) polymorphism (40bp) mainly 9R and 10R

10R Probably higher activity

DAT - COMT interactions

COMT degrades dopamine SN polymorphism

(val158met) met158

Low enzyme activity

Ventral striatumBilder et al., 2004

from PFC

Effect of COMT and DAT on predictions

Genetic influence on expected value coding during anticipation

1€/p-lo 1€/p-hi 5€/p-lo

BO

LD s

igna

l (a.

u.)

5€/p-hi

COMTMet/Met

Val/Met

Val/Val

DAT 9R 10R

Yacubian et al., PNAS 2007

Inverted u-shape response

“Phasic DA“

COM

T M

et/M

etDA

T 10

RCO

MT

Met

/Met

DAT

9R

COM

T Va

l/Met

DAT

10R

COM

T Va

l/Met

DAT

9RCO

MT

Val/V

alDA

T 10

RCO

MT

Val/V

alDA

T 9R

Slop

e of

fMR

I res

pons

eSe

nsat

ion

seek

ing

≈

r=-0.77, p<0.05

Reuter et al., Nature Neuroscience 2005

robustness Encourage publication of null results of

imaging genetics data (given adequate methodology e.g. sample size etc.)

As usual, large n is helpful Consider split half testing (e.g. odd-even

samples)

Some thoughts on …

Split half testing

Yacubian et al., PNAS 2007

Odd samples

Even samples

Whole group

“While the sample size in this study was fairly substantial for an imaging study, it is rather small for a genetics study. The reviewer appreciates the logistical problems and cost of a very large scale imaging x genetics study, and their sample size certainly falls within the scope of others of this type. However, the authors should at least acknowledge the possibility that such studies fall into the complex trait category (looking for an effect of allelic variants in the brain induced by a behavioral paradigm is, by definition, complex) and are therefore subject to the type I error problem that has plagued behavioral genetics research.” (the unknown reviewer)

N = 105 Consider stratified sample

Opinions – Sample size

Dopamine D4 receptor polymorphisms and novelty

Novelty and Dopamine Dopamine activity signals unexpected, salient, motivationally-

relevant information mediated via reciprocal dopaminergic projections between

hippocampus, ventral striatum and dopaminergic midbrain The role of the Dopamine D4 receptor

D4 receptor is preferentially expressed in limbic regions, cortex, basal ganglia and midbrain (SN/VTA)

association between novelty seeking and a C to T polymorphism in the DRD4 promoter region (-521C>T; rs1800955) in LD with the exon III VNTR

T allele associated with reduced transcription levels of 40% Study:

N=46, stratified for rs1800955 (DRD4 -521C>T)

Strange et al., in preparation

Experimental paradigm and behavioural data

Behavioural effects Effect only for perceptually salient stimulus (-521C>T)

Strange et al., in preparation

Neuroimaging results

Strange et al., in preparation

Candidate gene vs. whole genome approach Interpretability of the results (cf.

neuroimaging as a mapping technique vs neuroimaging as a neurophysiology tool)

Very strong hypotheses: You can only find what you already know

In between approaches (i.e. reducing genetic dimensionality to signal cascades that might be involved in the process (cf. small volume correction in neuroimaging)

Both can be interesting

Some thoughts on …

Integrated Project FP 6:Reinforcement-related behaviour in normal brain function and psychopathology

Study design Investigate 2000+ 14 years old adolecents across Europe since Dec 2007 Predictive Markers for drug abuse

Neuropsychology, Behavioural testing, personality assessment, environment assessment

Brain function (Reward: MID, Impulsivity: SSRT), Brain structure: T1, DTI Whole genome approach

Berlin, Dresden, Dublin, Hamburg, Mannheim, Nottingham, London, Paris Current status: ~1200 volunteers included

gain-related effects: conjunction Met/Met & Val/Valp<0.001, FWE corrected

Prelim. neuroimaging results: MID task

Sample Val158met (rs4680) Focus on homozygotes

(Met/Met, Val/Val) n=110 (Met/Met) vs. n=115

(Val/Val)

y=10Ventral striatum (peak t=4.86)

Outcome–related activation

Val/Val > Met/Metp<0.001, uncorrected

Peters et al., in preparation

Ventral striatumBilder et al., 2004

from PFC

substructures Imaging genetics: explaining interindividual

variance in activation patterns of a certain brain region by a certain marker / genotype

Make sure that the marker of interest is uncorrelated to – Other markers (e.g. check indicator SNPs on

other chromosomes)“Only five genes were analyzed. In order to identify substructures in a study population to rule out type I error from stratification, a more intensive genomic control analysis is necessary (approximately 50-100 genes)” (from the unknown reviewer)

– But also to other variables (e.g. age, personality) Again, large n is helpful

Some thoughts on …

Combining Imaging and Genetics A very promising approach ( endophenoytpe) As usual there are many pitfalls Field is in a stage of maturation

Interpretability Control for substructure

Candidate vs whole genome approach Both have their merits (data vs hypothesis driven) Ideally have a large sample to do both Entertain immediate approaches: e.g. signalling cascades GWAS: Cooperation with an advanced functional genetics unit is

helpful Sample size

Candidate genes: Stratification from a large pool of genotyped volunteers

Multi-site data acquisition: Feasible for fMRI and sMRI

Summary