Gene-Environment

Interaction

AGES workshop, Fall 2012

Lindon Eaves – Introduction
Tim York – Lots of slides
Sarah Medland – Making it work
 Genotype x Environment
Interaction

• Genetic control of sensitivity to the

environment
• Environmental control of gene expression
 The Biometrical Genetic
Approach:
GxE =
Genetic Control of “Sensitivity”
to environment
See e.g. Mather and Jinks.
Like any other phenotype –
individual differences in sensitivity
may be genetic
 Alternative Approach
GxE =
“Environmental modulation” of
path from genotype to
phenotype
See e.g. Purcell
“a,c,e modulated by covariate”
 CARE!
Results depend on scale – often
remove GxE (and other
“interactions” – e.g. epistasis) by
change in scale

Often try to choose units so model

is additive/linear and simple.
“Simplicity is truth” (?)
 “Environment”
• Micro-environment (unmeasured, random)
“E” and “C” in twin models
• Macro-environment (measured, ?”fixed”)
SES, life events, exposure, smoking
• “Independent” (of genotype)
• “Correlated with genotype – of individual
(“active/evocative”) or relative (“passive”)
 Animals, Plants and
Microorgansims
• GxE common property of genetic systems
• Ranking of lines/strains changes over environments: ?“genetic
correlation across environments”
• Polygenic: genes affecting response to E widely distributed across
genome (Caligari and Mather)
• GxE usually small relative to main effects of G and E (“power”?)
• Can select separately for means and slopes (“different genes cause
GxE”)
• Some genes affect response to specific environmental features (Na,
K, rain, diet, separation)
• Some genes affect response to overall quality of environment
(“stress”)
• GxE adaptive – own genetic architecture (A,D etc) – genetic
vulnerability, resilience, sensation-seeking, harm avoidance (rGE)
• GxE scale dependent – choose units of measurement
Genotype x Environment
Interaction
 Where does GxE go in “ACE”
model?
• AxE -> “E”
• AxC -> “A”

See Jinks, JL and Fulker, DW (1970) Psych.

Bulletin
Contributions of Genetic, Shared Environment,
Genotype x Shared Environment Interaction
Effects to Twin/Sib Resemblance

Genotype x Shared
Shared Additive Environment
Environment Genetic Effects Interaction

MZ Pairs 1 1 1x1=1

DZ Pairs/Full Sibs 1 ½ 1x½=½

In other words—if gene-(shared) environment interaction is not explicitly

modeled, it will be subsumed into the A term in the classic twin model.
 Contributions of Genetic, Unshared
Environment, Genotype x Unshared
Environment Interaction Effects to Twin/Sib
Resemblance
Genotype x
Unshared Unshared
(Unique) Additive Environment
Environment Genetic Effects Interaction

MZ Pairs 0 1 0x1=0

DZ Pairs/Full Sibs 0 ½ 0x½=0

If gene-(unshared) environment interaction is not explicitly modeled,

it will be subsumed into the E term in the classic twin model.
 Testing for GxE and GxC
• Plot absolute within MZ pair differences
against pair means (CARE!!)
• Plot absolute within DZ pair differences
against shared measured environment
 Basic Model (1960’s)
Yi |E = gi + biE + di

Yi |E = phenotype of ith genotype in

environment E
gi = main effect (“intercept”) of ith genotype
bi = slope (“sensitivity”) of ith genotype in
response to environment
(gi , bi) ~ N[m,G]
 GxE: A biometrical genetic model

A A M
1 2 g2
g1
a b

E P b=m + g1A1 + g2A2

e
ith phenotype in kth level of measured environment

Pi|Mk= aA1i + (m + g1A1i + g2A2i)Mk+ eEik

= aA1i + (m + g1A1i + g2A2i)Mk+ eEik

= ck+ (a + g1Mk) A1i + g2MkA2i+ eEik

 Expected twin covariance
conditional on environments of first and second twins

Cov(Pi1|Mk ,Pi2|Ml ) =

[a2 + ag1(Mk+Ml)+ (g12+g22)MkMl]r

+ Cov(Ei1k,Ei2l)

r = genetic correlation (1 in MZs)

 No GxE: Genetic covariances and correlations
No G x E: Genetic Covariance as Function of Environment No G x E: Genetic Correlation as Function of Environment
Gene

Gene
ti

tic Co
c Co
varia

rrelat
nce

ion
nt

nt
me

me
n

n
viro

viro
En

En
Tw Tw
in 1 in 1
in 2

in 2
En En
viro viro
Tw

Tw
nm nm
e nt e nt

alpha=0.5, gamma1=0.0, gamma2=0.0 alpha=0.5, gamma1=0.0, gamma2=0.0

 Scalar GxE: Genetic covariances and correlations
Scalar G x E: Genetic Covariance as Function of Environment Scalar G x E: Genetic Correlation as Function of Environment
Gene

Gene
tic Co

tic
varia

Corre
nce

lation

nt
ent

me
m

n
ron

viro
vi

En
En

Tw Tw
in 1 in

in 2
in 2

En 1E
viro n vi

Tw
Tw

n ron
me me
n t nt

alpha=0.5, gamma1=0.4, gamma2=0.0 alpha=0.5, gamma1=0.4, gamma2=0.0

 Non-Scalar GxE: Genetic covariances and
correlations
Non-scalar G x E: Genetic Covariance as Function of Environment Non-scalar G x E: Genetic Correlation as Function of Environment

Gene
Gene

tic
tic

Corre
Cova
rianc

lation
e

nt
ent

me
m

n
ron

viro
vi

En
En

Tw Tw
in 1 in 1

in 2
in 2

En En
viro viro

Tw
Tw

nm nm
e nt en
t

alpha=0.5, gamma1=0.0, gamma2=0.4 alpha=0.5, gamma1=0.0, gamma2=0.4

 Mixed GxE: Genetic covariances and correlations

Mixed G x E: Genetic Covariance as Function of Environment Mixed G x E: Genetic Correlation as Function of Environment

Gene
Gene

tic
tic Co

Corre
varia

lation
nce

nt

nt
me

me
n

ron
viro

vi
En

En
Tw Tw
in 1 in 1
in 2

in 2
En En
viro viro
Tw

Tw
nm n me
e nt n t

alpha=0.5, gamma1=0.4, gamma2=0.3 alpha=0.5, gamma1=0.4, gamma2=0.3

 Note:

Can’t separate components of

“mixed” GxE unless you have
same or related genotypes (MZ
and/or DZ pairs) in concordant
and discordant environments.

[c.f. Analysis of sex limited gene

effects].
 Modulation of gene
expression by measured
environment
[and environmental modulation of
non-genetic paths]
 Ways to Model Gene-Environment
Interaction in Twin Data
• Multiple Group Models
– (parallel to testing for sex effects using
multiple groups)
 Sex Effects
Females Males
1.0 (MZ) / .5 (DZ) 1.0 1.0 (MZ) / .5 (DZ) 1.0

A1 C1 E1 A2 C2 E2 A1 C1 E1 A2 C2 E2

aF cF aF cF eF cM cM
eF aM eM aM eM

P1 P2 P1 P2
 Sex Effects
Females Males
1.0 (MZ) / .5 (DZ) 1.0 1.0 (MZ) / .5 (DZ) 1.0

A1 C1 E1 A2 C2 E2 A1 C1 E1 A2 C2 E2

aF cF aF cF eF cM cM
eF aM eM aM eM

P1 P2 P1 P2

aF = aM ? cF = cM ? eF = eM ?
 GxE Effects
Urban Rural
1.0 (MZ) / .5 (DZ) 1.0 1.0 (MZ) / .5 (DZ) 1.0

A1 C1 E1 A2 C2 E2 A1 C1 E1 A2 C2 E2

aF cF aF cF eF cM cM
eF aM eM aM eM

P1 P2 P1 P2

aU = aR ? cU = cR ? eU = eR ?
 Problem:
• Many environments of interest do not fall
into groups
– Regional alcohol sales
– Parental warmth
– Parental monitoring
– Socioeconomic status

• Grouping these variables into high/low categories

loses a lot of information
 Standard model
• Means vector
(m m

• Covariance matrix
a + c + e
2 2 2

 
 Za 2 + c 2 2
a +c +e 
2 2

 Model-fitting approach to GxE

A C E A C E

a c e a c e

m m
Twin 1 Twin 2
M M
 Model-fitting approach to GxE

A C E A C E

a+bXM c e a+bXM c e

m m
Twin 1 Twin 2
M M
 Individual specific moderators

A C E A C E

a+bXM1 c e a+bXM2 c e

m m
Twin 1 Twin 2
M M
 E x E interactions

A C E A C E

c+bYM1 c+bYM2
a+bXM1 a+bXM2
e+bZM1 e+bZM2

m m
Twin 1 Twin 2
M M
 ‘Definition variables’ in Mx
• General definition: Definition variables
are variables that may vary per subject
and that are not dependent variables

• In Mx: The specific value of the def var

for a specific individual is read into a
matrix in Mx when analyzing the data of
that particular individual
 ‘Definition variables’ in Mx
create dynamic var/cov structure

• Common uses:
1. To model changes in variance
components as function of some
variable (e.g., age, SES, etc)
2. As covariates/effects on the means (e.g.
age and sex)
 1.0 (MZ) / .5 (DZ) 1.0

• Classic Twin Model:

A1 C1 E1 A2 C2 E2
Var (T) = a2 + c2 + e2 a c e a c e

Twin 1 Twin 2

• Moderation Model:
Var (T) = A C E

(a + βXM)2 + (c + βYM)2 + (e + βZM)2 e + βZM

a + βXM c + βyM

m + βMM
T
Purcell 2002,
Twin Research
 Var (T) = (a + βXM)2 + (c + βYM)2 (e + βZM)2

Where M is the value of the moderator and

Significance of βX indicates genetic moderation
Significance of βY indicates common environmental
moderation
Significance of βZ indicates unique environmental moderation

BM indicates a main effect of the A C E

moderator on the mean e + βZM

a + βXM c + βyM

m + βMM
T
 Additional Things to Consider
• Unstandardized versus standardized
effects
 Additional Things to Consider
Unstandardized versus standardized effects

ENVIRONMENT 1 ENVIRONMENT 2
Unstandardized Standardized Unstandardized Standardized
Variance Variance Variance Variance

Genetic 60 0.60 60 0.30

Common
environmental 35 0.35 70 0.35
Unique
environmental 5 0.05 70 0.35

Total variance 100 200

Environment may not
modulate all the genes
C.f. Biometrical genetic model –
Different genes may control main
effect and sensitivity/slope
 AS AU

aM aS + βXSM aU + βXUM

βXS indicates moderation

of shared genetic effects

BXU indicates moderation

of unique genetic effects
on trait of interest
M T
 Genotype-Environment
Covariance/Correlation (rGE)
see e.g. RB Cattell (1965)
 Gene-environment Interaction
• Genetic control of sensitivity to the environment
• Environmental control of gene expression

Gene-environment Correlation
• Genetic control of exposure to the environment
• Different genotypes select or create different
environments
• Different genotypes are exposed to correlated
environments (e.g. sibling effects, maternal
effects)
• Environments select on basis of genotype
(Stratification, Mate choice)
 This complicates interpretation of
GxE effects
• If there is a correlation between the moderator
(environment) of interest and the outcome,
and you find a GxE effect, it’s not clear if:

– The environment is moderating the effects of genes

or
– Trait-influencing genes are simply more likely to be
present in that environment
 Ways to deal with rGE
• Limit study to moderators that aren’t correlated with outcome
– Pro: easy
– Con: not very satisfying

• Moderator in means model will remove from the covariance genetic

effects shared by trait and moderator
– Pro: Any interaction detected will be moderation of the trait specific
genetic effects
– Con: Will fail to detect GxE interaction if the moderated genetic
component is shared by the outcome and moderator

• Explicitly model rGE using a bivariate framework

– Pro: explicitly models rGE
– Con: Power to detect BXU decreases with increasing rGE; difficulty
converging
Getting it to work
SARAH!!!!
 Practical (1):
Using Multiple Group Models to
test for GxE
Adding Covariates to Means Model

A C E A C E

a c e a c e

m+bMM1 m+bMM2
Twin 1 Twin 2
M M
 Matrix Letters as Specified in Mx
Script

A C E A C E

a+bXM1 c+bYM1 c+bYM2

c+cM*D1 e+b M a+bXM2 c+cM*D2 e+b M
a +aM*D1 Z 1 Z 2
a+aM*D2 e+eM*D2
e+eM*D1

Twin 1 Twin 2
M M
m+bMM1 m+bMM2
mu+b*D1 Main effects and moderating effects mu+b*D2
 Practical (2):
Using Definition Variables
to test for GxE
 Matrix Letters as Specified in Mx
Script

A C E A C E

a+bXM1 c+bYM1 c+bYM2

c+cM*D1 e+b M a+bXM2 c+cM*D2 e+b M
a +aM*D1 Z 1 Z 2
a+aM*D2 e+eM*D2
e+eM*D1

Twin 1 Twin 2
M M
m m
mu mu
 ‘Definition variables’ in Mx
create dynamic var/cov structure

• Common uses:
1. To model changes in variance
components as function of some
variable (e.g., age, SES, etc)
2. As covariates/effects on the means (e.g.
age and sex)
 Definition variables used as
covariates
General model with age and sex as covariates:
yi = a + b1(agei) + b2 (sexi) + e

Where yi is the observed score of individual i, a

is the intercept or grand mean, b1 is the
regression weight of age, agei is the age of
individual i, b2 is the deviation of males (if sex
is coded 0= female; 1=male), sexi is the sex of
individual i, and e is the residual that is not
explained by the covariates (and can be
decomposed further into ACE etc).
 Allowing for a main effect of X
• Means vector
(m + bX1i m + bX 2 i 

• Covariance matrix
a + c + e
2 2 2

 
 Za 2 + c 2 2
a +c +e 
2 2

 Common uses of definition
variables in the means model

• Incorporating covariates (sex, age, etc)

• Testing the effect of SNPs (association)

• In the context of GxE, controlling for rGE