Applications of GWAS and Next-Gen Sequencing Technology to the Study of Viral and Drug-Induced Liver Disease
Thomas J. Urban, PharmD, PhD
Center for Human Genome Variation
Duke University Medical Center
3.22.2011
Outline
• Value of common and rare genetic variants in predicting disease and drug response (GWAS)
– Hepatitis C treatment outcomes
– Drug-induced liver injury
• Potential value of rare variation in prediction of serious adverse drug events
Challenges of PegIFN + ribavirin treatment for HCV infection
• Only 40-50% of patients with HCV
genotype 1 show sustained virological response (SVR)
• Serious adverse events, so many drop out prior to completion of 48 week course of therapy
• Treatment is 2 fold more effective in groups of European vs. African ancestry with HCV genotype 1
Study design
N = 871 Caucasians
N =75 Hispanics
N = 191 African Americans
N = 1671
DNA genotyping Nature, 2009
N = 1137
Exclusions: Undetectable HCV RNA at baseline Other/unknown ancestry Compliance < 80% at 12wk QC failure
N = 3070
IDEAL Study NEJM, 2009
Quality control 2 Gender, relatedness, HWE…
Genotypic calls
Correction for population stratification, correcting for
local genomic characteristics
Signal intensity
RAWCNV format
PennCNV
Standard PED/MAP formats; PLINK binary BED/BIM/FAM
formats
*EIGENSOFTplus
*PipeCMD
PLINK/*PipeCMD
Quality control 3
Enrichment of SNP functions and
gene sets
Association with common CNVs
Quality control 4 - Lab re-check Export to WGAViewer for annotation
Rare/large CNV/Pathway
analysis
Association analysis
Logistic regression Linear regression Survival analysis
Genomic Analysis Facility Quality control 1
PLINK/GWSurv
Computer software used:
PipeCMD
PLINK EIGENSOFTplus
WGAViewer
rs12979860 on chromosome 19 is strongly associated with SVR
Ge D. et al, Nature 2009
CC genotype at rs12979860 associates with a 2-3 fold
greater rate of SVR than CT/TT genotype
Ge D. et al, Nature 2009
Ge D. et al, Nature 2009
Marcello T et al., Gastroenterology 2006
Possible mechanisms for cis-acting IL28B genetic variants
IL28B variants: effects on IL28B expression in blood or PBMCs
Tanaka et al., Nat Genet 2009
Suppiah et al., Nat Genet 2009
Urban TJ et al., Hepatology 2010
Differential hepatic gene expression by IL28B genotype:
N = 61 patients with
Chronic HCV infection
Liver mRNA expression
Profiling on Illumina HumanHT-12
Expression microarray
-37G>C
(rs28416813)
Lys70Arg
(rs8103142)
1,340 bp
IVS2+134C>T
(rs11881222) 3’UTR+52T>G
(rs4803217) -312A>G
(rs4803219)
Fine mapping of IL28B gene region
IFN-λ3-Arg70Lys: Potential role in receptor binding affinity/specificity?
Gad HH et al., J Biol Chem 2009
Antiviral potencies of E. coli-expressed IFN-λ3 variants
Drug-Induced Liver Injury (DILI)
• Primary cause of drug attrition from market
• Idiosyncratic, suspected genetic predisposition
• Most represented drugs among DILI cases:
– Amoxicillin/clavulanate
– Valproic Acid (VPA)
– Isoniazid (INH)
– Nitrofurantoin
– Minocycline
• Common variation implicated in risk, but positive predictive value (PPV) fairly low
• Rare variation may contribute significantly to this rare adverse drug event
The Drug-Induced Liver Injury Network (DILIN)
https://dilin.dcri.duke.edu/
Potential to capture 20 Million lives
Coordinated by DCRI
Proposed Mechanisms for DILI
www.liv.ac.uk/drug-safety/Research/dili.htm
Augmentin liver injury study
145 Cases
iSAEC/DILIN Collaboration
201 Genotyped Cases
56 Cases
Augmentin GWAS results 201 cases, 532 POPRES controls
MHC
Lucena MI et al., 2010 (under review)
Augmentin MHC association Top SNP Chr:Mbp P-value OR (95% CI) P-value conditioned on
rs3135388 (DR2)
P-value OR (95% CI)
rs9274407 6:32.74 4.8e-14 3.1 (2.2-4.2) 0.00011 3.2 (1.8-5.8)
DR2 tag
Lucena MI et al., 2010
Augmentin MHC association Conditioned on top class II SNP
Top SNP Chr:Mbp P-value OR (95% CI)
rs2523822 6:29.94 1.3e-9 2.3 (1.7-2.9)
Lucena MI et al., 2010
Augmentin MHC association Conditioned on top class I and II SNPs
Lucena MI et al., 2010
Augmentin top SNPs and HLA alleles
HLA-DRB1 HLA-DRA1
rs3135388 HLA-DRB1*1501 (and DR2) rs9274407
r2 ~ 1
r2 ~ 0.77
HLA-DQB1 HLA-A
rs2523822
SNP Tagged
Alleles Notes
rs9274407 DR2 In LD with DR2
rs3135388 DR2 Reported to be associated by two previous
studies;
rs2523822 HLA-A*0201 Not reported in previous studies
HLA-A*0201
r2 ~ 0.9
DR2 = DRB1*1501–DQB1*0602
Augmentin conclusions
29
• DR2 association is confirmed (DRB1*1501)
• Additional association in region of HLA-A
• Significant interaction between HLA-A and HLA-DRB1
• Cannot determine likely causal variants
– Investigation in additional populations likely required to sort out
Genetic Influence on DILI Risk: Across drugs, across phenotypes…
Drug Reaction
Details Prev Risk Allele Freq.
Rel
Risk PPV 1 - NPV
Ximelagatran 0.08 HLA-DRB1*0701 0.08 4 0.22 0.055
Augmentin <0.001 HLA-DRB1*1501
A*0201/B*1801
0.15 4 5.7e-4 5.7e-5
Isoniazid 0.15 CYP2E1*1 & 2 0.13 7 0.59 0.084
Lapatinib 0.09 HLA-DQA1*0201
(HLA-DRB1*0701)
0.08 9 0.17 0.03
Lumiracoxib 0.013 HLA-DRB1*1501 0.15 13 0.039 0.0030
Ticlopidine <0.001 HLA-A*3303 0.07 36 1.2e-3 3.5e-5
Tranilast 0.12 UGT1A1*28 0.30 48 0.23 0.0048
Flucloxacillin <0.001 HLA-B*5701 0.04 81 0.0022 2.8e-5
Irinotecan Neutropenia 0.20 UGT1A1*28 0.30 28 0.36 0.013
Mercaptopurine Neutropenia 0.12 TPMT*2/3A/3B/3C 0.05 9.0 0.77 0.086
Abacavir Hypersensitivity 0.04 HLA-B*5701 0.04 >1000 0.50 5.0e-4
Carbamazepine SJS/TEN 0.003 HLA-B*1502 0.04 >1000 0.038 3.8e-5
Early Conclusions from GWAS
• Top-associated SNPs in for amoxicillin-clavulanate (and flucloxacillin, SAEC) revealed secure associations with common SNPs in the HLA region
– Evidence suggesting possibility of shared HLA risk alleles across different drugs, ADRs
• Other common genetic risk factors for DILI in general, or factors specific to certain clinical presentations (injury type, delay of onset, etc.) were not appreciable
• Thus, a combined analysis of all genotyped cases from DILIN and SAEC provides the best chance of discovering additional common DILI risk alleles (HLA or otherwise)
DILIN-SAEC collaborative GWAS
32
403 Cases
iSAEC/DILIN Collaboration
968 Genotyped Cases (783 European)
655 POPRES Controls
2,345 WTCCC2 Controls
565 Cases
DILIN-SAEC: Total Enrollment
• After genotype and ancestry pruning:
• N=783 DILI cases
– 401 DILIN, 382 SAEC
• N=3,000 controls
– 2,345 WTCCC2, 655 POPRES
• Genotyped on Illumina 1M or 1Mduo array
• 800,769 SNPs overlapping after QC
• Amoxicillin/clavulanate and flucloxacillin highly represented (n=296); stratified analyses performed with and without these cases
All DILI Cases (n = 783)
• Association signal(s) in the MHC region reflect previously-described Augmentin and flucloxacillin risk variants
• No genome-wide significant effects outside of MHC
All Non-Augmentin/Flucloxacillin DILI Cases (n = 487)
• No deviation from expected p-value distribution
• No SNPs in MHC approaching genome-wide significance
Power to detect common genetic
risk factors for DILI
Searching for Rare DILI Risk Variants by Whole-Genome/Exome Sequencing
• Genome-wide association studies have revealed no common genetic risk factors for DILI overall, or to DILI due to individual drugs (sans Augmentin, flucloxacillin), perhaps owing to limited sample sizes
• Common variation cannot be highly predictive of DILI, by definition
• Whole-genome/exome sequencing allows essentially complete survey of both common and rare functional variation across the known functional regions of the genome
• Goal: Identify genetic risk factors causal for DILI, with the aim of developing a diagnostic test with clinically useful predictive power
Study Design Overview
Whole Genome Sequencing
CHGV Genomic Analysis Facility: -12 Illumina Genome Analyzers
-4 Illumina HiSeq 2000s
“Extreme Phenotype” Sequencing
1. Identify subjects (carefully!)
2. Sequence to 30x coverage
3. Alignment of reads to reference genome
4. Identify SNPs, indels and CNVs
– Indels=small insertions/ deletions <50bp
> 3 million SNPs and
> ½ million indels
What to do with it all???
SequenceVariantAnalyzer, software to manage,
annotate, and analyze the large number of
unique variants detected in whole genome
sequencing projects.
http://humangenome.duke.edu/software
Quality filtering
Analysis module:
Analysis module output: • Gene prioritization
• Fisher’s exact test
• Search for deviations from Hardy-
Weinberg Equilibrium
• Search for compound heterozygotes
• Simple listings
• … Whole genome, whole exome, genomic region, or gene specific:
• Variant tables (all variants)
• Variant listings (cases vs. controls)
• Coverage summarizations
Quality Analysis: Coverage
Coverage (Captured
Regions)
% of Captured Region
with >5X Coverage % Pairs Aligned
Average 73.1 0.956 0.953
Standard Deviation 9.62 0.013 0.018
0
1
2
3
4
5
6
7
8
9
60 62.5 65 67.5 70 72.5 75 77.5 80 82.5 85 More
Read Depth
0
2
4
6
8
10
12
14
16
% Bases with RD>5
QA: Concordance with GWAS Data
Based on 20,347 SNVs overlapping exome and
1Mduo data after SNV quality filtering
0
5
10
15
20
25
30
35
Example of a Trait Mapped by Whole Exome/Genome Sequencing: Metachondromatosis
Sobreira NLM et al.,
PLoS Biology 2010
Sobreira NLM et al.,
PLoS Biology 2010
Example of a Trait Mapped by Whole Exome/Genome Sequencing: Metachondromatosis
PTPN11: Protein tyrosine phosphatase SHP-2
• Gain-of-function mutations in multiple Mendelian disorders
• Loss-of-function mutations in metachondromatosis
Example of a Trait Mapped by Whole Exome/Genome Sequencing: Metachondromatosis
Sampling of current sequencing
projects
Neuropsychiatric/Cognition – 100 Schizophrenics (enriched in families)
– 50 individuals with epilepsy (family based)
– Focus on extremes in treatment response/resistance
Drug induced adverse events – Severe Adverse Event Consortium (SAEC) partnership
– Drug-Induced Liver Injury (part of DILIN)
– Clozapine-Related Agranulocytosis
– Drug-Induced Prolonged QT Syndrome, TdP
Other – Children with peanut and other food allergies
– Multiple Mendelian diseases
Conclusions • GWAS have revealed several
common variants of strong effect on some drug response phenotypes
– IL28B and HCV treatment response
– ITPA and RBV-induced anemia
• GWAS have revealed common variants of moderate effect on DILI risk
– Tendency for drug-specific effects
• Comprehensive evaluation of rare functional variation may soon allow for identification of high-impact genetic determinants of DILI and other drug response phenotypes
Acknowledgements • David Goldstein • Paul Watkins • DILIN Clinical
Investigators • Duke CHGV
• Kevin Shianna • Yuki Hitomi • Dongliang Ge • Kristen Linney
• DCRI • John
McHutchison • James Rochon • Thomas Phillips • Katherine Galan
• iSAEC
• Arthur Holden
• Matt Nelson
• Yufeng Shen
• Ann Daly
• NIDDK
• Jose Serrano
• Jay Hoofnagle
• Schering Plough Research Institute (Merck)