genetics for epidemiologists lecture 7: replication and functional studies

Genetics for Epidemiologists

Lecture 7: Replication and Functional Studies

National Human Genome Research

Institute

National Institutes of

Health

U.S. Department of Health and

Human Services

U.S. Department of Health and Human Services

National Institutes of HealthNational Human Genome Research

InstituteTeri A. Manolio, M.D., Ph.D.

Director, Office of Population Genomics andSenior Advisor to the Director, NHGRI,

for Population Genomics

Topics to be Covered

• Replication – Past challenges– Criteria– Reasons for inability to replicate findings

• Finding the causal variant– Neighboring regions: conservation,

nearby genes– Sequencing – Protein product– Expression studies– Experimental studies: Knockdown,

knockout, knockin

Chanock S, Manolio T, et al, Nature 2007; 447:655-660.

Need for Consensus on What Constitutes Replication: circa

November, 2006• Avalanche of GWA and candidate gene

studies anticipated in near future• Replication held as sine qua non• Likelihood of single study establishing an

association is low until sample sizes increase sufficiently and analytical methods improve substantially

• Common problem of how to interpret confusing and spurious findings

Case in Point: DTNBP1 and Schizophrenia

• First identified as putative schizophrenia-susceptibility gene in Irish pedigrees

• Reported confirmation in several replication studies in independent European samples but reported risk alleles and haplotypes appeared to differ between studies

• Comparison among studies difficult because different marker sets used by each group

• HapMap data and all identified polymorphisms typed in CEPH samples to produce high density reference map

Mutsuddi et al, Am J Hum Genet 2006; 79:903-909.

Phylogenetic Tree of Five Common Haplotypes of DTNBP1


Positively Associated Haplotypes Differ in All Six Studies


Each common DTNBP1 haplotype was tagged by association signal of at least one study, implying there is not one common variant contributing to schizophrenia risk at DTNBP1 locus

How NOT To Do A Replication Study

• Use a different phenotype

• Use different markers

• Mix fine-mapping and replication

• Use different analytic methods (haplotype vs. single marker)

• Use different populations

Case in Point: Odds Ratio for Stroke Associated with PDE4D in

Three Studies

Rosand et al, Nat Genet 2006; 38:1091-1092.

Science, 14 Apr 2006 Science, 12 Jan 2007

• Association between minor allele of rs7566605 near INSIG2 and increased BMI and homozygosity in 923 related Framingham Heart Study (FHS) participants

• Association reproduced in four additional cohorts

• Not seen in fifth cohort

Lyon HN et al, PLoS Genet; 2007 Apr 27;3(4):e61.

• Nine large cohorts from eight populations across multiple ethnicities

• Family-based, population-based, case-control designs

• Association at p < 0.05 in five cohorts but none in three cohorts

• Variability in strength of association over time • Replication both in unrelated (p = 0.046) and

family-based (p = 0.004) samples• Suggests initial finding unlikely to be spurious

but effect likely to be heterogeneousLyon HN et al, PLoS Genet; 2007 Apr 27;3(4):e61.

Lasky-Su J et al, Am J Hum Genet; 2008; 82:849-58.

• Second variant (rs1455832-C) intronic to ROBO1 with age-varying association to BMI over time

• Minor allele homozygote associated with increased BMI diminishing after age 45

• Replicated age-varying association in same direction in five of eight other cohorts totaling 13,584 subjects

• One childhood cohort showed very strong interaction (p < 10-9), four others 0.003 < p < 0.05; overall 10-9

• In all replication cohorts but one, association would not have been detected if testing only for main genetic effect and not for age-by-’5832 interaction

Definition of Robust Initial Finding

• Sufficient statistical power to observe reported effect, which will vary by magnitude of observed effect

• Highly significant analysis using stable method

• Consistent findings using simple, straightforward analytic approach

• Consistent findings in epidemiologically sound study

• Consistent findings overall and within key subgroups of initial study

• Consistent findings in same or highly similar phenotypes

Value of Single/First Study

• Initial study rarely definitive by itself but often represents important discovery tool– If consortium of multiple studies, stronger

• What to do with studies not having option for replication?– Don’t change standards for definitiveness

• Don’t just rely on GWA-- have multiple tools for identifying and understanding associations

• May need different standards for findings of major clinical significance, particularly pharmacogenomic demonstration of adverse effect that would be unethical to try to replicate

Importance of Significance Level

• Should we promulgate a specific number– NO, but in general, smaller is better

• General agreement: range is very broad, higher threshold for difficult to measure phenotype

• Beware of the very smallest• If significance depends on analytic method or

multiple comparison correction, BEWARE• If significance or association depends on

phenotype definition, BEWARE• Randomize the phenotypes and report number

significant at that level• Biologic information may be useful A PRIORI but

a posteriori can come up with almost anything

Importance of Genotyping Quality

• Report results of known study sample duplicates, HapMap or other standard duplicates

• Replicate small number of “significant” SNPs with second technology at some late stage

• May not be needed if nearby SNPs in strong LD show same results

• Strong caveats are needed regarding fallibility of genotyping- Results can change based on genotype calling algorithm- QC filters and consistency of results after

applying them must be described

NCI-NHGRI Working Group Criteria for Positive Replication

• Sufficient sample size to distinguish proposed effect from no effect convincingly

• Same or very similar trait (extension to related trait may increase confidence in finding, such as consistent finding for both dichotomized obesity and continuous BMI)

• Same or very similar population (extension to other populations may also increase confidence in finding, such as consistent association in populations of European, Asian, or even recent African ancestry)


NCI-NHGRI Working Group Criteria for Positive Replication

(continued)• Same inheritance model (dominant, co-

dominant, recessive), though not necessarily same analytic method

• Same gene, same SNP (or SNP in complete LD with prior SNP, r2 = 1), same direction as original finding

• Highly significant association• N.B.: Initial study must adequately

describe these parameters


Criteria for True Non-Replication or “Meaningful Negativity”

• Same as for positive replication (same trait, same gene, same SNP, same direction, same genetic model)

• Must be identical trait and population to claim non-replication

• Powered to appropriate effect size (account for “winner’s curse”)


Replication in Samples of Different Ancestral Origin

• Shorter LD blocks may explain failure of associations identified in European ancestry populations to replicate in recent African ancestry samples

• Shorter LD may permit better localization of risk variants

• Allele frequency differences may also explain lack of replication

Skol et al, Nat Genet 2006; 38:209-13.

Larson, G. The Complete Far Side. 2003.

Narrowing the Association Region…

Flow of Investigation: From Genome-Wide Association to Clinical Translation

Sequencing/Genotyping

Functional Studies

Translational Studies

Initial Genome-Wide

Association (GWA) Studies

Replication/Fine Mapping

Linkage of Chromosome 13q12-13 and MI in 296 Icelandic Families

Helgadottir et al, Nat Genet 2004; 36:233-239.

Fine Mapping of 1-LOD Drop Region Containing ALOX5AP


Most significant in males

Most significant in females

• Single marker-- Two-marker haplotype-- Three-marker haplotype

-- Four-marker haplotype-- Five-marker haplotype

Sequencing of ALOX5AP Gene


Sequencing for GWA: 1,000 Genomes Project

http://www.1000genomes.org/

LD (r2) among 8 TCF7L2 SNPs in Icelandic and West African

Population Samples2906

9699

6992

0271

6744

8222

0833

1514

rs7752906

--0.55

0.66

0.56

0.56

0.67

0.66

0.65

rs1569699

--0.87

0.99

0.98

0.85

0.83

0.83

rs7756992

--0.86

0.86

0.99

0.97

0.96

rs9350271

--1.00

0.86

0.85

0.84

rs9356744

--0.87

0.86

0.85

rs9368222

--0.98

0.97

rs10440833

--0.99

rs6931514

--

Steinthorsdottir et al, Nat Genet 2007; 39:770-75.

LD (r2) among 8 TCF7L2 SNPs in Icelandic and West African

Population Samples2906

9699

6992

0271

6744

8222

0833

1514

rs7752906

--0.55

0.66

0.56

0.56

0.67

0.66

0.65

rs1569699

0.16

--0.87

0.99

0.98

0.85

0.83

0.83

rs7756992

0.32

0.61

--0.86

0.86

0.99

0.97

0.96

rs9350271

0.13

0.72

0.67

--1.00

0.86

0.85

0.84

rs9356744

0.14

0.72

0.67

0.99

--0.87

0.86

0.85

rs9368222

0.12

0.12

0.14

0.10

0.10

--0.98

0.97

rs10440833

0.07

0.07

0.08

0.04

0.04

0.86

--0.99

rs6931514

0.08

0.09

0.10

0.05

0.06

0.76

0.87

--

> 0.9 0.80 – 0.89 0.60-0.79 0.30-0.59 < 0.30


Haiman et al, Nat Genet 2007; 39:638-44.

P-values for 8q24 SNPs Most Strongly Associated with

Prostate Cancer

McPherson et al, Science 2007; 316:1488-91.

CDKN2A/B and Coronary Disease

Zeggini E et al, Science 2007; 316:1336-41.

CDKN2A/B and Type 2 Diabetes


CDKN2A/B and Aortic and Intracranial Aneurysm

WTCCC, Nature 2007; 447:661-78.

Genes (+) strandGenes (–) strandConserved regions

9p21 Region Associated with CAD

Functional Studies: Correlation of SNPs with Logical Intermediate

Phenotypes• rs7756992 on 6p22.3 strongly associated

with type 2 diabetes (OR 1.20, p < 8 x 10-

8), resides in intron 5 of CDK5 regulatory subunit associated protein 1-like1 (CDKAL1)

• rs13244434 on 8q24 also associated with T2DM: OR 1.15, p < 4 x 10-6

• Nonsynonymous arginine to tryptophan change in last exon of solute carrier family 30 (zinc transporter), member 8 (SLC30A8)

• Specific to pancreas and expressed in beta cells


Relationship of Diabetes-Associated SNPs with Insulin

Secretion


(CDKAL1)

(SLC30A8)

LTB4 Production of Ionomycin-Stimulated Neutrophils in MI Cases

and Controls


Co-Localization of Gene Product with Histopathologic Changes

• CFH in retina and drusen (macular degeneration)

• GAB2 in dystrophic neurons (Alzheimers disease)

Complement Deposition in Affected Retina

Klein et al, Science 2005; 308:385-89.

Complement deposition in Bruch’s membrane (thin black arrows)Deposition also in choroidal artery (double headed arrow, pt C) and choroidal vein (white arrow, both)Deposition in drusen (*) as well as Bruch’s membrane and choroidal vein

Gab2 Colocalizes with Dystrophic Neurons in LOAD Brain

Reiman et al, Neuron 2007; 54:713-20.

Dystrophic neuron (arrow) and neurites (arrowheads)Tangle-containing neuron (arrow), dystrophic neurites (arrowheads)Tangle-bearing neuron (open arrow), immuno-reactive structures resembling dendrites (arrowheads)Gab2 immunoreactive cell with flame-shaped tangle-like inclusion

Conservation and Expression Studies: Asthma and ORMDL3

Moffatt et al, Nature 2007; 448:470-73.



PSMD3GSDM1ZPBP2IKZF3GSDMLORMDL3

Knockdown and Knockout Studies

• Knockdown of ATG16L1– Associated with Crohn’s disease– Reduces phagocytosis of S. typhimurium

in HeLa cells• Knockdown of GAB2

– Associated with Azheimer’s disease– Increases tau phosphorylation

• Knockout of MLXIPL– Associated with lower triglyceride levels– Knockout shows lower triglyceride levels– Transgenic (knockin) shows higher

levels

Genome-Wide Associations in Crohn’s Disease

Rioux et al, Nat Genet 2007; 39:596-604.

CARD15

IL23R

2q37.1; rs2241880ATG16L1 exon 8 A197T

Identification of IBD1 Locus by Family-Based Linkage and Fine

Mapping

Hugot et al, Nature 2001; 411:599-603.

Sequencing of IBD1 Region for Identification of Potentially Causative

SNPs

Hugot et al, Nature 2001; 411:599-603.

*

1%

10%

100%

1%

10%

100%Basal NF-B

Activation(vs WT)

PGNinducedNF-B

Activation(vs WT)

SNP8 SNP12 SNP13

W15

7R

R13

8Q

V79

3M

E77

8K

A72

5G

R71

3C

R70

3C

R68

4W

E44

1KA

432V

S43

1L

N41

4S

H35

2R

R31

1W

T29

4S

R23

5C

E84

3K

M86

3V

NACHTCARD2 LRRs

127 220 273 617 733 1037

1040

26 124

CARD1

V97

2I

G97

8E

R70

2W

P26

8S

G90

8R

V95

5I

1007

fs

1

L34

8V

N28

9S

D29

1N

A60

2T

A60

2V

558D

LG

R79

0Q

CARD15 Sequence Variants and NF-κB Activation

Chamaillard et al, PNAS 2003; 100:3455-3460.

Gene Expression in Crohn’s Disease

• rs2241880 associated at p < 10-8

• Nonsynonymous amino acid change in exon 8 of autophagy-related 16-like 1 (ATG16L1)

• Autophagy is biologic process involved in protein degradation, antigen processing, absorption of cellular organelles, initiation and regulation of inflammatory response


Expression of ATG16L1 in Human Primary Immune Cells


Knockdown of Endogenous ATG16L1 by siRNA 2 in HeLa

Cells


89%↓

Decreases transcripts

Prevents encapsulation of S.

typhimurium into

autophagosomes

89%↓

siRNA Knockdown of GAB2 Increases Tau Phosphylation without Increasing Total Tau

Reiman et al, Neuron 2007; 54:713-20.

Increased Triglyceride Levels in Mice Expressing Transgenes of

SREBP (Knockin)

Horton et al, J Clin Invest 1998; 101:2331-9.

Post GWA: Finding (Putative) Causal Variants

• Narrowing region with fine mapping, sequencing

• Structure of association region: nearby genes, conservation

• Association with levels of protein product

• Co-localization with histopathologic changes

• Association with expression levels

• Knockdown, knockout

genetics for epidemiologists lecture 7: replication and functional studies

Documents

j hum genet

replication studies

studies rosand et

studiesmutsuddi et

plos genet

nat genet

laskysu j et

heterogeneouslyon hn