Candidate Gene Predicting Health Vulnerabilities In Families

Download Report

Transcript Candidate Gene Predicting Health Vulnerabilities In Families

Strategies for candidate gene x environment research
and opportunities provided by intervention designs
H. Harrington Cleveland
Human Development Family Studies
The Pennsylvania State University
Thanks to NIDA:
PROSPER Pis: Mark Greenberg, Mark Feinberg, Richard Spoth,
GPROSPER collaborators: David Vandenbergh (M-PI), and
Gabriel Scholmer (Post Doctoral Scholar)
Overview
•
•
•
•
Critiques of candidate gene research
Common difficulties in candidate gene research
General guidelines for strengthening cGxE
How gPROSPER leverages the PROSPER
intervention study to investigate cG x E
• Results from three or more analyses
Candidate G X E Research Controversy
• Candidate Gene Research – especially G x E research
(e.g. Caspi et. al., 2003) – has been aggressively
criticized.
• Some (e.g. meta-analysis by Risch et.al., 2009) claim
that findings are consistent with chance.
• Others say that Risch et. al. (2009) is incorrect due to
inclusion of weak studies (see Uher & McGuffin,
2010, and Rutter et al., 2010).
• Common difficulties in candidate gene research
– Poor measurement: few items, single reporters
– Retrospective data: recall bias
– Compound phenotyping
– Passive large scale data sets are vulnerable to rGE:
• reducing exposure to full range of environments across
genotypes (reduces power)
• undercutting interpretation of findings
– Population structure confounds
Compound Phenotyping
• What it is: Combining subgroups of behaviors into one behavioral
measure, either explicitly or because assessments are non-specific
– e.g., ”Amount of monthly drinking” captures social drinking and
coping drinking
• If predictors are etiologically non-specific, such as parental drinking
levels, associations could be due to several causal links (i.e.,
etiological paths).
• However, if predictors are etiologically-specific, such as genes(?),
correlations will be substantially attenuated.
rGE confounds
The problem: Families, peers, experiences are all genetically influenced
The result: G x E “effects” might be due to genetic selection of
environments/experiences
Minimum controls: investigate associations between genes and presumptive
“environmental moderator”
Better solutions:
Random assignment
Quasi-experimental solutions
- school level interventions (e.g., PROSPER)
- regression discontinuity designs
- state level differences in welfare policies
Population Stratification
A 3rd variable confound:
1. Samples may be made up of distinct genetic groups or they have been
genetic mixing of groups in recent past (i.e., racial admixture).
2. Groups may differ in allelic distributions and outcomes
3. Creating spurious associations between alleles and outcomes
4. Classic Study: Knowler et al.
General Finding – Genetic variant negatively associated with Type 2 diabetes in Native
American /Euro sample.
But: Genetic variant more common among European descent cases
Analyses within Native American and Euro groups found no association
Citations: Kwowler, et al. (1988).GM and Type 2 Diabetis Mellitus: An Association in American Indians with Racial Admixture. American Journal of Human Genetics. 43. 520526.
Cardon & Palmer (2003). Population stratification and spurious allelic association. Lancet, 361, 598-604
Moving cGxE Research Forward
We suggest five domains in which candidate
gene research should be evaluated:
• Design
• Measurement
• Theory
• Biological Role
• Population Structure
Design.1
• In epidemiological studies, causality is difficult to determine since
experience-outcome associations may equally reflect causal
environmental influences or self-selection into those environmental
experiences.
• Prevention/intervention trials, via randomization, eliminate nonrandom selection to the environment (i.e. intervention vs. control) and
create a unique opportunity to examine cGxE interactions without rGE
confounds.
• Randomized designs offer substantially more power to detect
interactions in cGxE than other designs (see Bakermans-Kranenberg &
van IJzendoorn, 2015; McClelland & Judd, 1993).
Design.2
• Other quasi-experimental designs should also be leveraged
– Do Regional or cohort differences modify the effect of genes?
– Within-family sibling comparisons
• Do discordant genes predict discordant outcomes?
• Do these patterns vary across families with different
environments?
Theory
• Theory can direct researchers toward constructs to
consider
• It can also help researchers avoid searching for
findings and leveraging chance results
Measurement
The strongest critiques of cGxE research have emerged from
research reviews (e.g. Duncan & Keller, 2011; Risch et al., 2009;
Munafo et al., 2009) that include studies that vary substantially
in measurement quality.
Family and developmental researchers are in a unique position
to capitalize on high quality measurement, which can increase
statistical power to a greater extent than increasing sample size
(Manchia, et al., 2013).
Biological Role
The “which gene” decision should be made on a range of levels
that demonstrate a cogent role for the marker vis-à-vis both the
phenotype and environment.
From broad to narrow:
• Association with behavioral outcomes – eg. , DRD4 linked to attention
deficit disorder and novelty-seeking.
• Associations with perceptual processes – eg., DRD4 encodes receptors
in frontal cortex associated with recognizing and paying attention to salient
information in the environment (reviewed by Bromberg-Martin et al., 2010).
•
Associations with neurocognitive functioning – eg., elevated levels of
activity in the striatum (caudate nucleus).
• Molecular level – eg., DRD4 7+ variant has been linked to less effective receptor
signaling) and lower gene expression.
Population Structure
Because allele frequencies can vary across populations,
inattention to population structure (e.g. genetic ancestry) can
lead to spurious results (Knowler, et al., 1988).
The best strategies to address population structure confounds
will vary by sample size, sample diversity, genes examined,
outcomes considered, and the combination of these that pose
threats to internal validity.
We use principle coordinate methods and subsample
confirmation.
Readings: Ziv et al. (2003) and Keller (2014).
gPROSPER builds on the PROSPER intervention study
o Randomized prevention trial design:
*Community-level randomized of interventions removes person-level selection –both general and
rGE– of intervention-related experiences
*14 intervention communities and 14 control communities in Pennsylvania and Iowa
*Random assignment increases G x E statistical power 5 – 10 fold
o
Detailed measurement of environment and outcomes:
Developed to assess early adolescent interventions designed to operate through family and peer
contexts, a primary focus of PROSPER is measuring processes within environmental domains
o
Prospective cohort longitudinal design:
Reduces perception-related rGE by not using long-term retrospective report of environmental
factors (Jaffe & Price, 2007).
Allows for construction of complex longitudinal phenotypes
o
Adequate sample size for Gene x Intervention Research (PROSPER = 8,000)
1,000 genotyped for adolescent and young adult analyses
2,000 genotyped for adolescent-only analyses
550 of hyper-measured family process data genotyped with Affymetrix 300K SNP Exon array
Conceptual model: How genes may influence substance use and
transact with intervention, family, and peer factors
Intervention
Family
context
Peer
context
rGE
GxE
Genes
Substance use
Population Stratification
in gProsper
PC2
PC1
Population Stratification
in gProsper
1 sd from mean of
pc 1 among selfidentified noneuros
PC2
PC1
Common gPROSPER Analytic Framework
Analytic steps:
1. Run models on all available cases
2. Repeat controlling for pop stratification
3. Repeat dropping all non-euros based on PC classification
Results from gPROSPER analyses
Study 1: Genetic moderation of adolescent alcohol use
trajectories and intervention effects on genetic moderation
Study 2: The nicotinic receptor subunit α5 gene, smoking, and
intervention status
Study 3: Interactions between DRD4, intervention, and maternal
involvement in the prediction of alcohol use
Study 1: Genetic moderation of adolescent alcohol use
trajectories and intervention effects on genetic moderation
Adolescent Drinking is Linked to Array of Risks, from School Failure to Sexual
Risks
Early Adolescent Drinking is Linked to Problem Behaviors, Aggression, Family
Dysfunction, and School Failure
Later Adolescence: Common, Less Risky. . . Part of Autonomy Seeking
Measures
Alcohol Use: Assessed across 8 waves: 0 = No, 1 = Yes
- Ever had a drink of alcohol?
- Ever had more than a few sips of alcohol?
- Ever been drunk from drinking alcohol?
Intervention Status: Analytic Sample = 1,932
46.5% of adolescents in the control condition (coded = 0; n = 899)
53.5% in the intervention (coded = 1; n = 1,033)
Alcohol Dehydrogenase (ADH) Genes:. Six SNPs in three ADH Genes – ADH1C, ADH1B, ADH4
Each marker was coded 0 = homozygous non-risk allele, 1 = heterozygous, and 2 = homozygous
risk allele.
The two markers for each gene were averaged to create three gene scores, range 0 to 2; where a
score of 2 indicates an individual carried 4 risk alleles across the two SNPs in a given gene.
Analyses: Piecewise Latent Growth Models: Evaluate alcohol use changes
and levels across early and late adolescence
Step 1: Models estimated:
Growth in alcohol use during early adolescence (S1; waves 1 to 5)
Growth in alcohol use during late adolescence (S2; waves 5 to 8)
Intercept at 9th grade (wave 5; I)
Step 2: Adding ADH Genes to determine their associations with Slope1,
Slope 2, and Intercept
Step 3: Adding Intervention by Genes Interactions to Investigate Whether
Genes’ Associations with Slope1, Slope 2, and Intercept are Modified by
Intervention
Table 1. Piecewise Growth Alcohol Model Results for Early vs. Later Adolescent Drinking
Model
Early Adolescence Slope
1) Unconditional
2) ADH Main Effects
ADH1C
ADH1B
ADH4
Later Adolescence Slope
9th Grade Intercept
.308(.007)*
.236(.009)*
1.405(.027)*
-.036(.013)*
.046(.022)*
.007(.012)
.046(.022)*
.007(.012)
-.008(.029)
-.007(.015)
.170(.082)*
.004(.042)
3) Multiple Group
Control / Intervention
ADH1C
-.066(.020)*/
.009(.018)
-.015(.026)/
-.015(.024)
-.218(.074)*/
.028(.065)
ADH1B
.050(.033) /
.046(.030)
.005(.043) /
-.023(.039)
.200(.125)/
.158(.108)
ADH4
.020(.017)/
-.006 (.016)
-.012(.022) /
.001(.021)
.037(.062)/
- .030(.058)
Table 3. Piecewise Growth Alcohol Model Results for Early vs. Later Adolescent Drinking
Model
Early Adolescence Slope
Later Adolescence Slope
9th Grade Intercept
1) Unconditional
.308(.007)*
.236(.009)*
1.405(.027)*
2) ADH Main Effects
ADH1C
ADH1B
ADH4
-.036(.013)*
.046(.022)*
.007(.012)
.046(.022)*
-.008(.029)
-.007(.015)
.007(.012)
.170(.082)*
.004(.042)
3) Multiple Group
Control / Intervention
ADH1C
-.066(.020)*/
.009(.018)
-.015(.026)/
-.015(.024)
-.218(.074)*/
.028(.065)
ADH1B
.050(.033) /
.046(.030)
.005(.043) /
-.023(.039)
.200(.125)/
.158(.108)
ADH4
.020(.017)/
-.006 (.016)
-.012(.022) /
.001(.021)
.037(.062)/
-.030(.058)
ADH1C
Control Group
-.015ns
Alcohol Initiation
-.218*
-.066*
W1
W2
W3
W4
W5
Time
W5
W6
W7
W8
Intervention Group
ADH1C
.028ns
Alcohol Initiation
.015ns
.009ns
W1
W2
W3
W4
W5
Time
W5
W6
W7
W8
Study 2: The Nicotinic Receptor Subunit α5 Gene,
Smoking, and Intervention
Measures
CHRNA5 Gene
One of several nicotinic genes related to smoking
rs16969968 is known as “Mr. Big”
Linked to Cigarettes-per-day, Nicotine Dependence (DSM4), and Craving
High School Smoking
Average “Past Month Smoking” over 9th-12th grades of past month score
where 0=never, 1=ever but not past month, 2=one to a few times, 3=once a
week or more
Controls show an expected increase in smoking with greater # of the risk allele (A)
Control
1.6
1.4
1.2
Past Month
Smoking in
High School
Control
1
0.8
0.6
0.4
0.2
0
G/G
G/A
Genotype at rs16969968 [increasing # of risk alleles (A) to the right]
Vandenbergh et al., N&TR (In press) doi: 10.1093/ntr/ntv095, PMID: 25941207
A/A
Controls show an expected increase in smoking with greater # of the risk alleles (A)
But there is no increase in smoking in the intervention group (p < 0.05)
1.6
1.4
Control
1.2
Past Month
Smoking in
High School
1
Intervention
0.8
0.6
0.4
0.2
0
G/G
G/A
Genotype at rs16969968 [increasing # of risk alleles (A) to the right]
Vandenbergh et al., N&TR (In press) doi: 10.1093/ntr/ntv095, PMID: 25941207
A/A
Study 3: Interactions between DRD4, intervention,
and maternal involvement in the prediction of
alcohol use
Intervention Experience:
1. Decrease contextual pressures and opportunities to use
2. Increase peer resistance skills
Maternal Involvement:
Index of parental investment
DRD4:
Provides index of genetic variability in susceptabilty to experience
DRD4 findings: Interacting with both family factors and interventions
1. Maternal insensitivity at 10 months predicts externalizing at 39 months (r = .61)
among DRD4 7+ children; but not among7- youth (Bakermans-Kranenberg & van
IJzendoorn, 2006).
2. Greater relationship between overall parental quality and lab-measured sensation
seeking behavior for DRD4 7+ youth (r = .58) than among DRD4 7- youth (r =.22;
Sheese, Voelker, Rothbart, & Posner, 2007).
3. Maternal attachment predicted altruistic behaviors among seven year olds
who were 7+ compared to those who were DRD4 7- (Bakermans-Kranenberg and van
IJzendoorn , 2011)
4. Parenting program aimed at affecting child behavior through enhancing maternal
sensitivity and positive discipline strategies reduced externalizing more
for DRD4 7+ children than those who were 7- (Bakermans-Kranenburg et al., 2008).
Predicting 9th grade alcohol use by grade 6 Mother Activity and Intervention Status
1. Significant
main effect
Three-way
Interaction
Int*MCA*D4
-.47(.21)*
Predicting 9th grade alcohol use by grade 6 Mother Activity and Intervention Status
1. Significant
main effect
Three-way
Interaction
Int*MCA*D4
-.47(.21)*
2. Significant 3way interaction
Intervention Effects of the Association between 6th grade Maternal Activities
and Alcohol Use Initiation by 9th Grade among DRD4 7+ Adolescents
.06
-.06
-.24
-.23
Thanks to Gabriel Schlomer
Intervention Effects of the Association between 6th grade Maternal Activities
and Alcohol Use Initiation by 9th Grade among DRD4 7+ Adolescents
.06
-.06
-.24
-.23
Thanks to Gabriel Schlomer
Intervention
effect on
DRD4 7+
carriers with
high maternal
activity
Intervention Effects of the Association between 6th grade Maternal Activities
and Alcohol Use Initiation by 9th Grade among DRD4 7+ Adolescents
.06
-.06
-.24
-.23
Thanks to Gabriel Schlomer
Intervention
effect on
DRD4 7+
carriers with
high maternal
activity
p < .05
DRD4 Results
1.8
Control
1.6
Intervention
DRD4 7R-
1.4
1.2
1
0.8
0.6
0.4
0.2
0
Low
Mean
High
MomAct
Involvement
1.8
DRD4 7R+
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
Low
Mean
MomAct
Involvement
High
DRD4 Results
1.8
Control
1.6
Intervention
DRD4 7R-
1.4
1.2
1
0.8
0.6
0.4
0.2
0
Low
Mean
High
MomAct
Involvement
1.8
DRD4 7R+
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
Low
Mean
MomAct
Involvement
High
DRD4 Results
1.8
Control
1.6
Intervention
DRD4 7R-
1.4
1.2
1
0.8
0.6
0.4
0.2
0
Low
Mean
High
MomAct
Involvement
1.8
DRD4 7R+
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
Low
Mean
MomAct
Involvement
High
DRD4 Results
1.8
Control
1.6
Intervention
DRD4 7R-
1.4
2-way
interaction
between DRD4
and maternal
involvement in
intervention
group
1.2
1
0.8
0.6
0.4
0.2
0
Low
Mean
High
MomAct
Involvement
1.8
1.6
DRD4 7R+
1.4
1.2
1
0.8
0.6
0.4
0.2
0
Low
Mean
MomAct
Involvement
High
Intervention Effects of the Association between 6th grade Maternal Activities
and Alcohol Use Initiation by 9th Grade among DRD4 7+ Adolescents
.06
-.06
-.24
-.23
Thanks to Gabriel Schlomer
Intervention
effect on
DRD4 7+
carriers with
high maternal
activity
p < .05
Population stratification sensitivity analysis
3-way
2-way Int*MA, 7+
7+ intervention
-.469*
-.289*
-.228*
PC-control (n = 511) -.446*
-.272*
-.221*
-.512*
-.342*
-.220*
Full (n = 545)
PC-drop (n = 470)
• Bonus findings . . .
Schlomer, et al. (2015). Developmental differences in early adolescent aggression: A
gene x environment x intervention analysis. Journal of Youth and Adolescence.
Control, 7-
Control, 7+
Intervention, 7-
Intervention, 7+
Investigating Genetic Moderation of Threat Appraisals
Grych & Fincham, 1990
Cummings & Cummings, 1988
Davies, Cummings, & Winter, 2004
Fosco & Grych, 2008
Interparental
Conflict
Externalizing
Threat
Appraisals
Internalizing
Well-being
.21
Interparental
Conflict
.32
-.19
Interparental
Positivity
Adolescent
Perception IPC
.27
Threat
Appraisals
-.19
.33
DRD4 7-
Adolescent
Perception IPP
.28
-.08
.20
Interparental
Conflict
.36
-.24
-.06
Interparental
Positivity
Adolescent
Perception IPC
.33
Adolescent
Perception IPP
Internalizing
Problems
DRD4 7+
.05
Threat
Appraisals
-.22
.30
Internalizing
Problems
Conclusions
1. Intervention designs provide an important
opportunity to examine GxE interactions.
2. Results confirm the co-active nature of both genes
and environments.
3. Using well-characterized candidate genes provides
an opportunity to understand for whom interventions
work and perhaps why and how interventions work.
Next Steps
• Begin to characterize genetic variance in more
biologically thoughtful ways
– Using multi-locus gene scores
– Using combinatorial approaches to examine gene
networks
• Examine how genes interact with social
networks across adolescence to create risk
and promote well-being
Multi-locus Genetic Scores: Neurotransmitter System Approach
Dopaminergic:
Cholinergic:
Catechol-O-Methyl Transferase
Dopamine Transporter
Dopamine Receptor
Dopamine Receptor
Dopamine Receptor
Dopamine Beta-Hydroxylase
Dopamine Receptor
Amine oxidase (MAOB)
Cholinergic receptor, muscarinic 2
Cholinergic receptor, nicotinic, alpha 4 subunit
Cholinergic receptor, nicotinic, alpha 7 subunit
Neuronal nicotinic acetylcholine receptor beta 2
Cholinergic receptor, nicotinic, alpha 5
Cholinergic receptor, nicotinic, alpha 3
Cholinergic receptor, nicotinic, beta 4
Cholinergic receptor, nicotinic, beta 3
Cholinergic receptor, nicotinic, alpha 6
Cytochrome P450 2A6
Serotonergic:
Vasopressin/Oxytocin:
Monoamine Oxidase A
Serotonin Transporter
5-Hydroxytryptamine (serotonin) receptor 1B
5-Hydroxytryptamine (serotonin) receptor 2A
Neuronal tryptophan hydroxylase
Oxytocin receptor
Arginine vasopressin-neurophysin II
Arginine vasopressin receptor 1A
Arginine vasopressin receptor 1B
CD38 (cyclic ADP ribose hydrolase)
GABAergic:
Alcohol Metabolism:
Gamma-aminobutyric acid A receptor, alpha 1
Gamma-aminobutyric acid A receptor, alpha 6
Gamma-aminobutyric acid A receptor, beta
Gamma-aminobutyric acid A receptor, beta 2
Gamma-aminobutyric acid A receptor, alpha 2
Alcohol Dehydrogenase 1B
Alcohol Dehydrogenase 1C
Alcohol Dehydrogenase 2
Alcohol Dehydrogenase 4
Alcohol Dehydrogenase 5
Opioid:
Cannabinoid, other Genes, and AIMs:
Opioid receptor, kappa 1
Opioid receptor mu 1, isoform MOR-1X
Beta-neoendorphin-dynorphin preprotein
Proenkephalin (PENK)
Cannabinoid Receptor 1 (brain)
Fatty Acid amide hydrolase (FAAH)
Brain Derived Neurotrophic Factor
cAMP Resp Elem Binding Prot 1 (CREB)
FBJ murine osteosarcoma viral oncogene homolog B (FOSB)
DNA cytosine methyltransferase 3 alpha
• Thank You