Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls

The Wellcome Trust Case Control Consortium, Nature, 2007Presented by Group 4: Jessica Larson, Irene Shui, and Lucia Sobrin

Outline

Introduction Methods Case-control structure Population stratification Data analysis

Results Diabetes (Type II) Crohn’s disease Rheumatoid arthritis Coronary artery disease

Discussion and Conclusion

Introduction

Several common (complex) diseases with evidence for heritability but incomplete knowledge of causal genes

Genome-wide association studies (GWAS) would help ‘unlock’ the genetic basis for common diseases Requires large sample sizes (for sufficient power) HapMap resource

This study validates the GWA method

Introduction, continued

WTCCC combines 50 research groups throughout the UK Large selection of cases and controls Seven common diseases:

Type II diabetes (T2D) Crohn’s disease (CD) Coronary artery disease (CAD) Rheumatoid arthritis (RA) Type I diabetes (T1D) Hypertension (HT) Bipolar disorder (BD)

Multiple diseases studied so that WTCCC could look at differences between the diseases themselves (not just between cases and controls for each disease)

Cases and Controls

2,000 cases for each disease 3,000 shared controls

1500 from 1958 British Birth Cohort1500 from UK Blood Services

Two Control Groups

Purpose: To assess possible bias in ascertaining control samples

Shared Controls

Potential Issues Misclassification bias Inflation of type 1 error rate from failure to match

on socio-demographic variables This study provides compelling case for the

suitability and efficiency of this design in Britain.

Population Stratification Only included self-identified white Europeans

Further excluded 153 individuals with evidence of recent non-European ancestry. Still possible heterogeneity; waves of immigration

Analyzed allele frequency differences in 12 geographic regions 13 genomic regions with strong geographic variation (NW/SE

axis; London set apart) Geographic correlation not apparent in 7 diseases studied

Principal components analysis Conclude that population stratification not much of a

problem once individuals with non-European ancestry excluded Adjusting for principal components and stratifying by geographic

region did not make a big difference in overdispersion; p-values with and without structure correction were similar

Figure 2, Wellcome Trust Case Control Consortium, 2007

NADSYN1 (11q13—possible role in prevention of pellagra)

TLR1 (4p14 toll-like receptor 1—possible role in biology of TB and leprosy)

LCT (Iactase digestion)

HLA (Major histocompatibility complex)

LCT

4p14 HLA

Previously implicated in Europeans

11 df test for differences in allele frequency between geographic regions

SNP genotyping and Data Analysis GeneChip 500K Affymetrix arrays Gene-calling algorithm CHIAMO For polymorphic SNPs

Trend tests General genotype tests between cases and controls Sex-differentiation test

Loci affecting more than one disease, combine the cases vs. the controls CAD+HT+T2D (metabolic overlap) RA+T1D (known to share common loci) CD+RA+T1D (autoimmune diseases)

Data Analysis

Significance levels were chosen not to directly correct for multiple tests (to obtain a ‘genome-wide significance level’), but to still have a low FDR Strong: regions with at least one SNP’s P-val<5x10^-7

(Table 3) Single disease: 21 signals Sex diff: RA Combined cases: RA+T1D 25 total

12 of which previously described Rest have been confirmed, except one

Moderate: 5x10^-7< P-val< 1 x10^-5 (Table 4) Nominal: 1x10^-5< P-val< 1 x10^-4 (Supplementary Table

7)

Notes on interpretation of this data

Replication needed Failure to detect an association does not

mean that a given gene is unassociated with disease

Help define regions of interest, cannot clearly identify causal genes

Overall Results

Figure 4, Wellcome Trust Case Control Consortium, 2007

Type 2 Diabetes

Detected all three previously widely replicated associations TCF7L2

SNP with strongest etiological claims not on Affy chip, but imputation analysis confirms it is the SNP with strongest association effect

PPARG and KCNJ11 (p~0.001 for both) Genuine disease susceptibility genes can generate

signals in GWS that would not attract immediate attention

Type 2 Diabetes

Compared to French GWAS Findings Confirms finding on Chromosome 10 Three other findings cannot be replicated

One SNP is poorly covered by Affy chip and extensive recombination in region limits data imputation

Two other SNPs cannot be confirmed by either genotyped or imputed SNPs from the WTCCC

Crohn’s Disease

Common form of chronic inflammatory bowel disease

Pathogenesis poorly understood Dysregulated immune response to intestinal

bacterial and possibly defects in mucosal barrier function or bacterial clearance

Genetic predisposition is strong (lambda-s 17-35; twin studies: 50% concordance in monozygotic vs 10% in dizygotic twins)

Crohn’s Disease GWAS Results

Previously defined susceptibility loci (6) all replicated Four new strong association signals (p-value <5X10-7)

Successfully replicated in other studies Eight less strong evidence for association markers (p-value

>5X10-7 and <1X10-5) Several with biological candidacy

Majority of associations modest RR<2 Functional mechanism: autophagy

Newly identified susceptibility gene (IRGM) proposed to control the spread of intracellular pathogens by autophagy (ATG16L1 also involved in autophagy)

Possible functional mechanism of autophagy and Crohn’s Disease supported by molecular genetic studies

Figure 4, Wellcome Trust Case Control Consortium, 2007

WTCCC SNP in LD with SNP T300ARED—Replicated defined markers/possible genes

GREEN-Novel Markers/possible genes

IL23R ATG16L1 CARD15 5q31

10q21 5q13.1 IRGM BSN/MST1

NKX2-3 PTPN2

Crohn’s Disease Strong Associations

IL23R Interleukin-23 receptor

ATG16L1 Involved in autophagy

5q13.1 Gene desert

5q31 Causative gene in dispute b/c of high LD in region

10q21 Non coding

CARD15 1st confirmed susceptibility gene

IRGM Involved in autophagy

BSN/MST1 Many genes in this region: BSN closest (but brain related); MST1: encodes a protein that induces phagocytosis by resident peritoneal macrophages.

NKX2-3 Lymphoid tissue abnormalities

PTPN2 Negative regulator of inflammatory resp.

Coronary artery disease (CAD)

Plaque buildup in arteries Environmental (diet) and genetic factors Previously associated genes not replicated

here (APOE, p-val:1.7x 10^-1) Found a new region of interest 9p21.3

(1.8x10^-14) and several moderate associations

Rheumatoid arthritis (RA)

Chronic inflammatory disease, destruction of joints, severe disability

Again, environmental and genetic factors Previously associated genes replicated here (HLA-DRB1, p-vals: 10^-27; PTPN22, p-vals: 10^-25)

Found two new regions of interest and several moderate associations

Most interesting is the sex effect (p-val: 3.9 x 10^-7), additive effect in females only

Common Loci for Autoimmune Diseases

CD25 region Encodes IL-2 receptor Association with both RA and T1D (p~10-8 and

p~10-6, respectively) PTPN2

Encodes a key negative regulator of inflammatory responses

Strong association with CD and T1D (p~10-8) and weaker but consistent association with RA (p~10-2)

Discussion/Conclusions GWAS yielded multiple association findings for multiple

diseases, many of them novel Large study; still power issues for OR<1.2 Extensive quality control Used both linear trend and 2 df genotypic test Replication is key

“winner’s curse”; ORs will tend to be overestimated for loci discovered

Several studies have replicated; more work needs to be done Incomplete coverage of Affy chip for some SNPs (T1D INS)

Functional studies needed to make inferences about molecular and physiological mechanisms involved and causal variants

No real gene-gene/gene-environment interactions tested Findings to date only explain a small proportion of the genetic

variation in these diseases Information is publicly available!

http://www.wtccc.org.uk/info/access_to_data_samples.shtml

Thank You!