Genetic Variants Modulating Ventricular Fibrillation in the Setting of Myocardial Infarction

UvA-DARE (Digital Academic Repository)

Subtle killers and sudden death: Genetic variants modulating ventricular fibrillation in the setting of myocardial infarction

Pazoki, R.

Publication date 2015 Document Version Final published version

Link to publication

Citation for published version (APA): Pazoki, R. (2015). Subtle killers and sudden death: Genetic variants modulating ventricular fibrillation in the setting of myocardial infarction.

General rights It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).

Disclaimer/Complaints regulations If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.

UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl)

Download date:28 Sep 2021 C hapter 5 Multiple genomic approaches for identifi cation of genetic modifi ers of ventricular fi brillation risk in the setting of acute myocardial infarction in the AGNES study

Manuscript in preparation for submission

Raha Pazoki, Lia Crotti, Reza Jabbari, Jonas S.S.G. de Jong, Nienke Bruinsma, Michiel E. Adriaens, Leander Beekman, Lukas R.C. Dekker, Anna M. Di Blasio, Jacob Tfelt-Hansen, Peter J. Schwartz, Gaetano M. De Ferrari, Arthur A.M. Wilde, Michael W.T. Tanck, Connie R. Bezzina 74 Chapter 5

Abstract

Background: Sudden cardiac death (SCD) from ventricular fibrillation (VF) during acute myocardial infarction (MI) is a leading cause of total and cardiovascular mortality. Genetic factors underlying risk for VF in the setting of acute MI remain largely unknown. We previously started exploring the role of common genetic variants in modulation of VF risk in acute MI by a genome-wide association study (GWAS) in the AGNES population, consisting of Dutch individuals with a first acute MI, where cases suffered VF and controls not. The aim of the current study was to carry out GWAS in an extended set of AGNES patients to identify additional susceptibility loci and to overlay this GWAS data with other genomic data in an effort to identify loci modulating risk of VF in this setting. Methods: In total, 1433 AGNES patients with a first acute MI (672 cases, 761 controls) were analyzed by GWAS. We sought to replicate SNPs displaying association P-values < 1 × 10−5 in two additional independent but similar case-control sets, namely PREDES- TINATION and GEVAMI, recruited in Italy and Denmark, respectively. We additionally reviewed literature to find SNPs previously implicated in SCD and investigated their association with VF in the AGNES study. In an effort to identify candidate causal genes at associating loci, the most strongly associating SNPs from the AGNES GWAS were searched in our human heart eQTL (expression quantitative trait locus) database and in publicly available heart and blood eQTL databases in order to investigate whether any of these SNPs could be related to the expression of genes located in cis. In a reverse approach, we selected previously-identified eQTLs from the human heart eQTL databases (representing putatively functional SNPs) and investigated their association with VF in the AGNES sample. We additionally performed pathway analysis on the SNPs identified in the AGNES GWAS in an effort to identify biological mechanisms that may be involved in the occurrence of VF in the setting of MI. Results: rs2824292 remained as most-strongly and only genome-wide significant SNP associated with VF in the AGNES study (OR = 0.65; 95% CI: 0.56, 0.76; P = 2.75 × 10−8). Out of the 10 top independent AGNES SNPs, only rs1750041 (OR = 1.64; 95% CI: 1.09, 2.46; P = 0.02) was nominally significant in the PREDESTINATION study. This SNP was however not associated with VF in the GEVAMI case-control set. Our candidate SNP analysis in AGNES uncovered a nominal association with VF for rs6795970 in the SCN10A gene (OR = 0.85; 95% C.I: 0.73, 0.99; P = 0.02). Among the 10 most-strongly associated SNPs in the AGNES GWAS, rs1750041 and rs10096381 displayed eQTL effects with genes in cis. in peripheral blood. In the reverse approach, we uncovered that rs6982308 (an eQTL SNP that controls the expression of XPR6 in human heart) increased the risk of VF by 1.58 fold (95% CI: 1.28, 1.94; P = 1.8 × 10−5). Pathway analysis showed that VF-associated genes from the current study are enriched in pathways mainly involved in signal transmission. Multiple genomic approaches to identify genetic susceptibility to VF in MI 75

Conclusion: Using several integrated approaches, we highlighted several SNPs (i.e. rs2824292, rs1750041, rs10096381, rs6982308, and rs6795970) that merit further analysis in future studies. We also highlighted several pathways that connected together to possibly form a larger relevant biological mechanisms for risk of VF in the setting of MI. Chapter 5 76 Chapter 5

Introduction

Sudden cardiac death (SCD) accounts for 20% of total mortality and up to 50% of cardiovascular mortality among adults in Western Societies 1,2. Ventricular fibrillation (VF) is the most common cardiac arrhythmia underlying SCD 3. SCD in the general population occurs in individuals with an average age of 65 years 4 who have complex disease stemming from multiple common acquired disorders, especially coronary artery disease (CAD) or associated conditions such as myocardial ischemia, infarction (MI), post-MI myocardial scarring, and ischemic cardiomyopathy 3. While multiple studies have provided evidence for a heritable component in the determination of SCD risk in these settings 5-8, the identification of genetic risk factors has only recently started to be explored and progress has been slow 9-11. Difficulties in gene identification are hindered by a number of factors, amongst which are the paucity of patients that are available for genetic studies and the likely complex underlying genetic architecture of SCD. While as for other complex phenotypes, the genetic architecture of SCD susceptibility is unknown, SCD in the individual patient is likely governed by the cumulative effect of a broad spectrum of inherited genetic variants that occur at different frequencies in the general population and that carry different effect sizes on risk for SCD 1. Molecular mechanism underlying SCD may differ between different cardiac pathologies and gene identification efforts are therefore likely to benefit from a deep characterization of the cardiac condition in which SCD occurs as this allows the identification of different disease sub-groups for genetic studies. In the last years, our group has been focused on the identification of genetic factors modulating risk of VF in the setting of a first acute MI 5,9,12. To this aim we established the Arrhythmia Genetics in the NEtherlandS (AGNES) study, consisting of patients with a first acute ST-segment elevation MI presenting with VF (ECG-documented; cases) and without VF (controls) 5. In this case-control set we established that family history of sudden death is a risk factor for VF 5, providing a strong rationale for the exploration of genetic factors modulating risk of VF. Subsequently, through a genome-wide association study (GWAS) for common genetic variants in this case-control set, we identified the first susceptibility locus at chromosome (chr.) 21q21 9. The aim of the current study was to identify additional VF susceptibility loci by (1) conducting GWAS in an extended set of AGNES, (2) testing the role of expression quantitative trait loci (eQTL), and (3) implementing a pathway-based analysis. Multiple genomic approaches to identify genetic susceptibility to VF in MI 77

Methods

Patient samples The AGNES sample: The AGNES case-control set has been described in detail previously 5,9. In brief, it consists of cases that survived to hospital admission and had ECG- registered VF that occurred within 24 hours after the onset of symptoms and before reperfusion therapy in the setting of an acute and first ST-segment elevation MI. The majority (89%) of the VF events occurred within the first 2 hours after the onset of the symptoms and 99% occurred in the first 12 hours. Controls were patients with a first acute ST-segment elevation MI without VF. Cases and controls (age > 18 and < 80 years 5) were recruited at seven heart centers in the Netherlands between 2001 and 2011. We excluded individuals with an actual non-ST-segment elevation MI, prior MI, congenital heart defects, known structural heart disease, severe co-morbidity, electrolyte disturbances, trauma at presentation, recent surgery, previous coronary artery bypass graft or use of class I and III antiarrhythmic drugs. Individuals who developed VF during or after percutaneous coronary intervention were not eligible. Furthermore, because early reperfusion limits the odds of developing VF, potential control subjects undergoing percutaneous coronary intervention within 2 hours after onset of myocardial ischemia symptoms were not included. This time interval was based on the observation that a majority (> 90%) of cases develop VF within 2 hours after the onset of symptoms. Chapter 5 The study protocol was approved by the Institutional Review Board of the Aca- demic Medical Centre, University of Amsterdam, and was conducted according to the principles of the Declaration of Helsinki. The medical ethics committees of the hospi- tals participating in the study approved the study protocols, and all participants gave written informed consent. AGNES cases and controls that were used in genetic studies were of self-declared Dutch European descent. Thus far, the AGNES study consists of two subsets; a first set (n = 972) collected up until 2008 and a second set (n = 485) collected from 2008 to 2011. The PREDESTINATION sample: The first replication sample for the current study was from the PREDESTINATION (PRimary vEntricular fibrillation and suDden dEath during a firST myocardial INfArcTION) study. Similar to the AGNES study conducted by us in the Netherlands, the PREDESTINATION study is a multicenter study performed prospectively at multiple coronary care units in Italy. The recruiting centers were part of the group ANMCO (Associazione Nazionale Medici Cardiologi Ospedalieri). Patients eligible for the study had at least one ECG-documented VF occurring within 24 hours from the onset of an acute MI. The controls had the same clinical features in the absence of VF. Patients older than 75 and patients with previous MI in medical history, pre-existing significant cardiac disease and/or ejection fraction ≤ 30%, and the 78 Chapter 5

presence of conditions favoring the risk of VT/VF independently from the ischemic episode (e.g. inherited cardiomyopathy or primary electrical disease) were excluded from the study. Patients in the PREDESTINATION study read and signed an informed consent form approved by the Ethics Committee of the University of Pavia. The GEVAMI sample: We used a second replication sample from a Danish sample called GEVAMI (the GEnetic causes to Ventricular Arrhythmia in patients during first ST-elevation Myocardial Infraction). Similar to the AGNES and PREDESTINATION studies, the GEVAMI sample consisted of ECG-documented VF cases that suffered from VF within 12 hours after a first ST-segment elevation MI. The controls had all the same clinical presentations except VF. Cases and controls were collected nationwide in Denmark. Patients older than 18 and younger than 80 were included in the GEVAMI study. Patients with prior MI, previous coronary artery disease, coronary artery bypass graft, percutaneous coronary intervention, and non-ST-segment-elevation MI were not included. Patients or their next-of-kin signed the informed consent. Procedures in the GEVAMI sample are in accordance with the ethical standards of the national ethics committee on human research and with the Helsinki Declaration of 1975, as revised in 1983. Permission from the Danish Data Protection Agency was also obtained before the study was initiated (Jr.nr. 2010-41-5688).

Genotyping Previously, genome-wide genotyping of single nucleotide polymorphisms (SNPs) was performed on 515 AGNES cases and 457 AGNES controls using the Human610- Quad SNP array 9. In the current study, an additional 172 AGNES cases and 313 AGNES controls were genotyped using Illumina HumanOmni2.5 arrays. The analyses were performed using the BeadStudio (Genotyping module version 3.2.33) and the Ge- nomeStudio V2011.1 software (Genotyping module version 1.9.0). For both arrays, the GenCall score cutoff was 0.15 as recommended by Illumina. The average sample call rate was greater than 99%. Genome-wide genotyping of SNPs was performed on 220 Predestination cases and 421 Predestination controls using the Illumina Human OmniExpress‑12 SNP array. The analysis was performed using the Genome Studio software as described above. The GenCall score cutoff was 0.15. The average sample call rate was 99.2%. A subset of SNPs was genotyped in 256 cases and 527 controls from the GEVAMI sample using KASP technology (LGC Genomics, Herts, UK) based on competitive allele-specific PCR.

Quality control and imputation Multiple quality control measures were implemented before imputation. The estimated sex for each individual as determined by genotyping was compared with the Multiple genomic approaches to identify genetic susceptibility to VF in MI 79 phenotypic sex. Exclusion criteria were deviation from Hardy-Weinberg equilibrium at P ≤ 10−4 (estimated in controls), sample call rate < 0.95, and SNP call rate < 0.98. To use the information of typed rare variants when imputing from the reference panels, we selected all typed SNPs without regard to minor allele frequency (MAF). Minimal MAF of the typed SNPs in the current study was 0.001. Genotypes for SNPs from GE- VAMI sample passed quality control thresholds applied in the GEVAMI study (call rate ≥92%, Hardy-Weinberg equilibrium P > 0.01 in control subjects). To correct for possible genetic heterogeneity within the study sample, we performed principal component analysis in the AGNES (n = 1457) and the PREDESTINATION (n = 641) case-control sets using a multi-dimensional scaling technique applied on the Identity-By-State (IBS) matrix implemented in the R package GenABEL 13. Pairwise IBS values were calculated in the AGNES and PREDESTINATION samples separately and Prim’s algorithm implemented in the R package nnclust 14 was used to identify clusters and outliers (threshold: 0.3). The outliers thus identified were excluded from the subsequent analyses. The first two principal components were included as covariates in subsequent analyses. In total, 1433 AGNES and 629 PREDESTINATION cases and controls successfully passed all the quality control steps. The resulting outliers were excluded from the analysis. Genotype imputation was done using the Markov-chain Monte Carlo method implemented in MACH1.0 15,16. Un-typed polymorphisms were imputed using data from HAPMAP separately into the two AGNES datasets and the PREDESTINATION Chapter 5 dataset. The two imputed data sets of the AGNES sample were quality checked for potential swaps in genotype coding and were then combined. After imputation, an r2 threshold of 0.5 was implemented to identify and discard low-quality imputed SNPs.

Genome-wide association analysis Differences in continuous phenotypic variables between cases and controls were tested using an independent t test where data were normally distributed or a Mann- Whitney U test otherwise. Here values are presented as mean ± s.d. or median and inter-quartile range, respectively. Differences in categorical variables were compared using a Fisher exact test and values are presented as number and percentages.

1. Discovery in the AGNES study Genome-wide association analysis was conducted in the combined AGNES case- control set using logistic regression method in ProbABEL 17 while assuming an additive genetic model and including the first two principal components, age and sex as covariates. The P-values were corrected for the genomic control factor. The P-value threshold for genome-wide statistical significance 18,19 was set at P ≤ 5 × 10−8. An arbitrary threshold of P ≤ 1 × 10−5 was also considered. SNPs that passed this arbitrary 80 Chapter 5

threshold and were not in LD with other SNPs were considered independent and were clustered into independent genomic loci. At each genomic locus passing the arbitrary threshold, the SNP with the lowest P-value was considered as the lead-SNP. To account for independent signals, we selected all the SNPs passing this arbitrary threshold at each genomic locus and performed a conditional analysis to account for independent signals within each locus. We did this by running a regression model including all the SNPs at each genomic locus. We also performed a conditional analysis on the lead- SNPs by running a multiple regression model including all lead-SNPs to check their statistical independence.

2. Replication in the PREDESTINATION and GEVAMI samples The lead-SNPs from the discovery stage were then carried forward for replication in PREDESTINATION. Next, we selected a subset of SNPs that displayed the strongest combined association with VF in the AGNES and PREDESTINATION studies and took them forward for replication and meta-analysis in the Danish GEVAMI study. Analyses for the AGNES and PREDESTINATION were performed in R 20. Data analysis for a subset of SNPs in GEVAMI was performed using the Stata software package V.12.0 (StataCorp, College Station, Texas, USA). Similar to the discovery stage, logistic regression was used to analyze the SNPs that were taken forward for replication. A P-value threshold corresponding to 0.05 with a Bonferroni correction for the number of SNPs tested in the replication phase was used. Meta-analysis of the replication SNPs was carried out in R package meta 21 using a random effect model. We also searched articles presenting association results in relevant case-control sets for our previously reported SNP (rs2824292) and meta-analyzed the results with the findings of the current study.

Literature-based candidate SNP analysis We reviewed publications indexed in the National Center for Biotechnology Informa- tion (NCBI) PubMed database to identify genetic variants associated with SCD. We focused on full text publications, written in English, and studies performed among human subjects of European ancestry, published from 1st January 1998 until 3rd De- cember 2013. We used a search query that included: “common missense variant”, “VF”, “single nucleotide polymorphism”, “genome-wide association”, “SNP”, “allele”, “sudden cardiac death”, “sudden death”, “arrhythmic cardiac death”, and “sudden cardiac arrest”. We excluded review articles and case reports. Publications reporting statistically significant findings were reviewed for selection of candidate SNPs 11,22-38. Out of 26 relevant SNPs, 24 were typed or imputed in the AGNES study. The remaining 2 SNPs (rs41312391, and rs16847549) were not available in the AGNES study. rs16847549 is monomorphic in the HapMap sample of European ancestry and rs41312391 is not a recognized HaPMaP SNP and has no available proxy and was therefore not imputed. Multiple genomic approaches to identify genetic susceptibility to VF in MI 81

The 24 typed or imputed SNPs were investigated for association with VF in the AGNES case-control set. Here, the association P-value threshold corresponded to a P-value of 0.05 with a Bonferroni correction for the 24 candidate SNPs in this analysis (P ≤ 0.002).

Overlap with expression quantitative trait locus data We used two different approaches to search for VF-associated SNPs that also display eQTL effects.

1. Forward approach (from VF-SNP to eQTL) We selected SNPs that were associated with VF at P ≤ 1 × 10−5 in the current AGNES GWAS and investigated whether these SNPs displayed effects on the mRNA transcript abundance of genes located within 1Mb region spanning the SNPs (cis-eQTL effects). To this end, we employed three different databases including (1) an in- house eQTL resource obtained in left ventricle from 129 donor human hearts that was generated in our laboratory in the setting of another study 39; (2) The human, left ventricle eQTL resource from the Genotype-Tissue Expression (GTEx) pilot project 40 obtained in 83 post- mortem human heart samples; and (3) the Genenetwork eQTL resource of peripheral blood samples 41. All 3 databases assumed an additive genetic model. The P-value threshold applied in this analysis corresponded to 0.05 with a Bonferroni correction for the total number of transcripts within 1 Mb spanning all the SNP tested. Chapter 5

(1.1) The human donor heart eQTL database: In brief, our in-house human heart genome-wide eQTL data was generated in 129 left ventricular samples that were collected from non-diseased human donor hearts. Genome-wide left ventricular transcript abundance and genotypic data were combined to identify SNPs associating with the abundance of specific transcripts. Each of the 18,402 transcripts and 897,683 SNP genotypes that passed pre-processing and stringent quality control were tested for eQTL effects, identifying 771 eQTLs, regulating 429 unique transcripts at genome- wide statistical significance.

(1.2) GTeX human heart, left ventricle eQTL database: The GTEx project 40 collected human heart tissue samples from post-mortem or organ transplant settings to identify the tissue specific expressions and their correlation with genotypes in human. At present, the GTEx project includes 83 human left ventricule samples.

(1.3) The Genenetwork peripheral blood eQTL database: Genenetwork 41 is a large eQTL database that collected eQTLs in 8086 whole blood and white blood cell samples. Various gene expression platforms were used to measure gene expression values in 82 Chapter 5

the Genenetwork eQTL database. The large sample size of the Genenetwork database provides higher power to detect eQTL effects which was an important reason for us to use this database in our search for eQTL effects of SNPs associating with VF, alongside the relatively small human heart eQTL databases that were selected because of tissue specificity.

2. Backward approach (from eQTL to VF-SNP) In a backward approach, we tested whether independent SNPs that display statistically significant eQTL effects in human heart also display association with VF in the AGNES study. For this we used the 2 human heart databases including (1) our in-house eQTL database and (2) the human heart, left ventricle tissue from the GTEx pilot project (for details see above). The genome-wide significant thresholds as applied in the respective studies were used in our selection of significant SNPs. The two databases together contained 911 independent SNPs (r2 ≤ 0.7) with a MAF > 15% displaying a significant eQTL effect in left ventricular tissue. TheP- value threshold for the association analysis with VF was thus set at P ≤ 5.5 × 10−5, corresponding to a P-value of 0.05 with a Bonferroni correction for 911 independent tests. We also set an arbitrary threshold of P ≤ 0.001.

Pathway analysis We used a gene set enrichment analysis (GSEA) approach to test whether specific biological pathways were enriched with the VF-associated genes from the current AGNES GWAS. We tested pathways annotated in the Ingenuity, BIOCARTA, REAC- TOME, Gene Ontology (GO), Kyoto Encyclopaedia of Genes and Genomes (KEGG), and Protein Analysis THrough Evolutionary Relationships (PANTHER) databases. In total, 3216 gene sets were tested for enrichment of multiple modest associations. Pathway analyses were carried out using MAGENTA 42. In brief, all P-values of SNPs included in the AGNES GWAS were used as an input for MAGENTA. The analysis within MAGENTA consists of three steps: (1) Determine the smallest P-value at each gene. Each SNP was assigned to a gene using extended gene boundaries which correspond to 110kb upstream to the gene’s most extreme transcript start site and 40kb downstream to the gene’s transcript most extreme end site. These extended gene boundaries represent the 99th percentile of the distances of cis-eQTLs from their adjacent gene’s transcript start and end sites which is according to a comprehensive genome-wide analysis of putative functional regulatory elements (cis-eQTLs) using expression data from human lymphoblastoid cell lines 43. (2) Group the genes according to pathway information of each database, i.e. GO, PANTHER, Ingenuity, KEGG, REACTOME, and BIOCARTA. (3) Determine if a pathway is enriched for low-P-value genes by comparing the P-value distribution with the one under the null hypothesis Multiple genomic approaches to identify genetic susceptibility to VF in MI 83 while correcting for confounders, such as gene size, SNP density and LD-related properties using a multivariate linear regression model. In the current study, a GSEA P-value less than 2 × 10−3 (corresponding to false discovery rate (FDR) < 0.1) was considered as statistical significance threshold for pathway analysis.

Results

Genome-wide association analysis

1. Discovery in the AGNES study In total, 1433 AGNES patients with a first acute MI (672 patients with VF and 761 without VF) were included in the analysis. Baseline characteristics of the study populations are shown in Table 1A-C. AGNES cases had a lower prevalence of diabetes mellitus and hypercholesterolemia. The average body mass index (BMI) was lower in

Table 1A | Baseline characteristics of the AGNES case-control set Characteristic Na Total Cases Controls P-valueb (n = 1433) (n = 672) (n = 761) c Sex (male) 1433/672/761 1142(80) 537(80) 605 (79) 0.85 Chapter 5 Age at MI (years) d 1433/672/761 57.5±10.8 56.7±1 58.3±10.6 0.004 ST segment deviation (mm) e 491/251/240 15(16) 18(19) 14(12) < 0.001 MI location c 1366/619/747 0.001f Anterior MI 726(53) 359(58) 367(49) Inferior MI 640 (47) 260(42) 380(51) CK-MB (μg/L) e 1044/415/629 205(277) 226(356) 188(244) < 0.001 Family history of sudden 1416/657/759 400(28) 205(31) 195(26) 0.002 death c Beta blocker usage c 1384/632/752 139(10) 66(10) 73(10) 0.65 Cardiovascular risk factors Current Smoking c 1374/627/747 809(59) 396(63) 413(55) 0.003 Diabetes Mellitus c 1349/613/736 102(8) 29(5) 73(10) 0.0004 Hypertension c 1298/584/714 426(33) 185(32) 241(34) 0.43 Hypercholesterolemia c 1250/563/687 406(32) 164(29) 242(35) 0.02 BMI (kg/m2) d 1342/597/745 26.5±3.86 26.1±3.66 26.8±3.99 0.002 MI, myocardial infarction; CK-MB, creatine kinase-MB; aSample sizes of the total, case and control sets for whom information is available are given. bP-value for comparison of cases and controls for each item; cNumber (%); dmean ± s.d.; eMedian (interquartile range); fP-value for comparison of inferior and anterior MI between cases and controls. 84 Chapter 5

Table 1B | Baseline characteristics of the PREDESTINATION case-control set Characteristic Na Total Cases Controls P-valueb (n = 1433) (n = 672) (n = 761) Sex (male) c 754/316/438 642(85) 266(84) 376(86) < 0.001 Age at MI (years) d 750/316/434 58.3±10.1 58.5±10.3 58.1±10.0 0.9 ST segment deviation (mm) e - - - - MI location c 753/315/438 Anterior MI 374(50) 170(54) 204(47) 0.08 Inferior MI 339(45) 133(42) 206(47) < 0.001 CK-MB (μg/L) e 250/138/112 145.6(240) 183.2(282) 109.5(244) < 0.001 Family history of sudden death c 430/201/229 47(11) 27(13) 20(9) 0.3 Beta blocker usage c 379/134/245 101(25) 44(33) 57(23) 0.2 Cardiovascular risk factors Current Smoking c 716/299/417 387(54) 166(56) 221(53) 0.005 Diabetes Mellitus c 701/288/413 97(14) 35(12) 62(15) 0.006 Hypertension c 712/294/418 390(55) 172(59) 218(52) 0.02 Dyslipidemia c 701/288/413 296(42) 130(45) 166(40) 0.04 BMI (kg/m2) e 417/193/224 102(24) 45(23) 57(25) 0.2 MI, myocardial infarction; CK-MB, creatine kinase-MB; aSample sizes of the total, case and control sets for whom information is available are given. bP-value for comparison of cases and controls for each item; cNumber (%); dmean ± s.d.; eMedian (interquartile range); fP-value for comparison of inferior and anterior MI between cases and controls.

Table 1 C | Baseline characteristics of the GEVAMI case-control set Characteristic Na Total Cases Controls P-valueb (n = 1433) (n = 672) (n = 761) Sex (male) c 852/269/583 678(80) 231(86) 447(77) 0.002 Age at MI (years) d 852/269/583 60±10.4 59.0±9.7 60.5±10.7 0.04 ST segment deviation (mm) e 590/201/389 3(1-5) 4(2-6) 2(1-6) < 0.001 MI location c Anterior MI 396(47) 145 (55) 251(43) 0.002 Non-Anterior MI 453(53) 121(45) 332(57) CK-MB (μg/L) e 654/221/433 165(252) 232(318) 139(221) < 0.001 Family history of sudden death c 833/257/576 241(29) 100(39) 141(25) < 0.001 Beta blocker usage c 832/261/571 67(8) 21(8) 46(8) 1.00 Cardiovascular risk factors Current Smoking c 847/265/582 462(55) 147(55) 315(54) 0.08 Diabetes Mellitus c 852/269/583 81(10) 31(11) 50(9) 0.2 Hypertension c 852/269/583 316(37) 110(41) 206(35) 0.1 Hypercholesterolemia c 852/269/583 291(34) 106(39) 185(32) 0.03 BMI (kg/m2) d 849/269/580 27.2 ±4.3 31.7 ±71.7 27.2 ±4.3 0.4 MI, myocardial infarction; CK-MB, creatine kinase-MB; aSample sizes of the total, case and control sets for whom information is available are given. bP-value for comparison of cases and controls for each item; cNumber (%); dmean ± s.d.; eMedian (interquartile range); fP-value for comparison of inferior and anterior MI between cases and controls. Multiple genomic approaches to identify genetic susceptibility to VF in MI 85 the AGNES cases than in the controls. AGNES cases more often had a family history of sudden cardiac death than the controls. Additionally, AGNES cases were younger at the time of MI, smoked less frequently, and had greater creatine kinase myocardial isoenzyme (CK-MB) levels than AGNES controls (Table 1A). AGNES GWAS showed an excess of SNPs associated with VF compared to the number expected under the null hypothesis of no association (Figure 1). The genomic control factor was small (λ = 1.057), indicating that overall inflation of genome-wide statistical results due to possible residual population stratification was minimal. The distribution of P-values for the association of SNPs with VF is shown in Fig- ure 2. One locus (Table 2) exceeded the genome-wide significance threshold (P ≤ 5 × 10−8). The SNP displaying the most significant association at this locus was rs2824292 at chr. 21q21 (OR = 0.65; 95% CI: 0.56, 0.76; P = 2.75 × 10−8) that was also identified in the first AGNES GWAS 9. An additional 9 independent SNP clusters (Figures 3a & 3b) passed the arbitrary threshold of P ≤ 1 × 10−5 (Table 2). At each genomic locus, the SNP with the smallest P-value was considered as the lead-SNP. Lead-SNPs remained independently associated with VF in a multiple regression model containing all lead-SNPs. In multivariate models containing all SNPs with P ≤ 1 × 10−5 at each genomic locus, 2 SNPs i.e. rs4841508 (OR = 0.47; 95% CI: 0.27, 0.82; P = 0.007) and rs7812879 (OR = 0.72; 95% CI: 0.57, 0.90; P = 0.005) at locus 8p23 remained associated with VF independently. These 2 SNPs were independent of each Chapter 5 other (r2 = 0.05) and of the lead SNP rs10096381 (r2 = 0.4). We didn’t find such independent signals in other 9 genomic loci.

2. Replication in the PREDESTINATION and GEVAMI samples SNP rs2824292 (Figure 3b) and the lead-SNPs from the other 9 loci were then selected for the first replication attempt in the PREDESTINATION study Table( 1B, Table 2). None of the SNPs displayed an association in PREDESTINATION at the pre-set Bonferroni-corrected significance threshold for 10 independent tests (P ≤ 0.005). However, one SNP, rs1750041, showed a nominally significant association with VF in PREDESTINATION (OR = 1.64; 95% CI: 1.09, 2.46; P = 0.017; Table 2). The direction of effect was consistent across the AGNES and PREDESTINATION studies, with the C-allele increasing the risk for VF in both studies. Upon meta-analysis of the two sets, the association P-value for rs1750041 was improved (OR = 1.66; 95% CI: 1.36, 2.02; P = 5.22 × 10−7). After meta-analysis of the 10 SNPs, 2 SNPs (rs1750041 and rs2135787), which had the strongest combined association with VF, were taken forward to the second replication in the Danish GEVAMI study (Table 1C). Neither rs1750041 (OR = 1.00; 95% CI: 0.74 - 1.34; P = 0.99) nor rs2135787 (OR = 1.01; 95% CI: 0.81, 1.26; P = 0.99) 55

86 Chapter 5

FigureFigure 11 || QuantileQuantileFigure- 1Quantile | Quantile-quantile Plot of plotTest of Statisticstest statistics ((P P(P--values)-values)values) for forfor the thethe associations AssociationsAssociations of SNPs of withof SNPs SNPs VF with with VF VF in in Plotted on the x-axis are expected P-values under the null hypothesis thethe AGNESAGNES CaseCasein the-Control AGNES case-control Set. Plotted set. on the xx--axisaxis areare expectedexpected PP--valuesvalues under under the the null null hypothesis hypothesis and and and on the y-axis the observed P-values after genomic control (λ = 1.057) has been applied. onon thethe yy--axisaxis the observed P-values after genomic controlcontrol (λ(λ == 1.057)1.057) hashas beenbeen applied. applied.

Figure 2 | Overview of the genome-wide association scan in the AGNES case-control set. –log10 (P- Figure 2 | OverviewFigure 2 of | Overview the genome of the -genome-widewide association association scan scan in in the the AGNES AGNES case-control case-control set. −log set.10 –log10 (P- values)values) correctedcorrected(P- values)for age, corrected sex, population for age, sex, stratification population stratifi andand cation inflationinflation and infl factorfactor ation are are factor shown shown are shownfor for each foreach each single single nucleotide polymorphismsingle nucleotide (SNP) polymorphism tested (ordered (SNP) tested by physical (ordered by position). physical position). The horizontal Th e horizontal lines indicatelines the nucleotide polymorphism (SNP) tested (ordered by physical-8 position).−8 The horizontal lines indicate-5 the presetpreset thresholdthresholdindicate for genome the preset-wide threshold significance for genome-wide ((PP ≤≤ 55signifi ×× 1010 cance-8)) andand (P ≤ the5the × 10 arbitraryarbitrary) and the thresholdarbitrarythreshold threshold ( (PP ≤ ≤ 1 1 × × 10 10-).5). (P ≤ 1 × 10−5). -8 OneOne locuslocus (Table 2) exceeded the genome--widewide significancesignificance thresholdthreshold ( (PP ≤ ≤ 5 5 × × 10 10-8).). The The

SNPSNP displayingdisplaying the most significant association atat thisthis locuslocus waswas rs2824292rs2824292 at at chr chr. .21q21 21q21 -8 (OR = 0.65; 95% CI: 0.56-0.76; P = 2.75 × 10 -8) that was also identified in the first (OR = 0.65; 95% CI: 0.56-0.76; P = 2.75 × 10 ) that was also identified in the first 9 AGNES GWAS 9. An additional 9 independent SNP clusters (Figure 3a-c) passed the AGNES GWAS . An additional 9 independent SNP clusters (Figure 3a-c) passed the arbitrary threshold of P ≤ 1 × 10-5 (Table 2). arbitrary threshold of P ≤ 1 × 10-5 (Table 2). At each genomic locus, the SNP with the smallest P-value was considered as the lead- At each genomic locus, the SNP with the smallest P-value was considered as the lead- SNP. Lead-SNPs remained independently associated with VF in a multiple regression SNP. Lead-SNPs remained independently associated with VF in a multiple regression

94 | Chapter 5 94 | Chapter 5

Multiple genomic approaches to identify genetic susceptibility to VF in MI 87 a heterogenoty heterogenoty , 2 Genes I In JPH1 In In XKR6 In Closest In SYNJ2 In Near OR2B2 Near Near MAT2B Near Near KCND3 Near Near CXADRNear Near PCSK5-RFKNear Near SYN2-C3orf31 Near Near ANGPT4-RSPO4Near 7 c −7 0.36 0.38 0.17 0.23 0.21 0.66 0.03 0.11 value P- 5.22 × 10 × 5.22 9.26 × 10- × 9.26 b (n = 2062) = (n Combined OR(95%CI) 1.3(0.72,2.35) 0.81(0.52,1.27) 1.26(0.91,1.75) 0.57(0.22,1.44) 0.68(0.37,1.25) 0.72(0.63,0.82) 0.83(0.36,1.89) 1.68(1.04,2.73) 1.28(0.95,1.74) 1.66(1.36,2.02) Odds ratios and of 95% codedCI per Oddscopy allele adjusted ratios b c 0.8 0.7 0.7 0.7 0.9 0.3 0.6 0.07 0.18 0.02 value P- b (n = 629) = (n PREDESTINATION OR(95%CI) -value adjusted for genomic control. adjusted P -value c 1.03(0.81,1.30) 0.96(0.73,1.26) 1.05(0.83,1.34) 0.92(0.56,1.52) 0.97(0.56,1.69) 0.79(0.61,1.02) 1.28(0.89,1.82) 1.27(0.78,2.06) 1.08(0.82,1.41) 1.64(1.09,2.46) Chapter 5 c −8 −7 −6 −6 −6 −6 −6 −6 −6 −5 P -value 2.7 × 10 × 2.7 1.6 × 10 × 1.6 10 × 8.9 3.2 × 10 × 3.2 10 × 4.3 10 × 1.0 3.5 × 10 × 3.5 2.4 × 10 × 2.4 2.3 × 10 × 2.3 2.6 × 10 × 2.6 b AGNES (n = 1433) = AGNES (n OR(95%CI) 0.65(0.56,0.76) 1.48(1.26,1.73) 1.47(1.24,1.75) 0.70(0.60,0.81) 2.09(1.52,2.85) 1.66(1.33,2.09) 0.55(0.43,0.71) 0.35(0.23,0.55) 1.75(1.41,2.16) 0.52(0.39,0.68) Genes within an area of 1 Mb centered at the SNP are listed; the SNP are at Genes centered of 1 withinMb an area a 4/6 CAF 53/51 36/38 28/28 41/36 90/90 87/89 11/13 50/47 85/86 -value; P -value; C/T C/T C/T A/C A/C G/T G/T A/G A/G C/G C/N Chr: location 20:848416 8:75374629 3:11952551 9:78064589 6:27991180 8:10851314 21:17709047 1:112372262 5:163438447 6:158418507 SNP rs2824292 rs6988935 rs17739527 rs2135787 rs13173656 rs1750041 rs1353342 rs1497526 rs497602 rs10096381 Chr, chromosome; C/N, coded allele/ non coded allele; CAF, frequency of coded allele in the cases and controls for AGNES/PREDESTINATION; AGNES/PREDESTINATION; for controls and cases the in allele coded of frequency CAF, allele; coded non allele/ coded C/N, chromosome; Chr, Table 2 | Meta-analysis of Single Nucleotide Polymorphisms (SNPs) in the of Single PolymorphismsAGNES (SNPs) Nucleotide and Predestination 2 | Meta-analysis case-control set Table for age and sex, random eff ratios;ect model is used for the combined odds and sex,for age eff random I-squared; Het.P, heterogenoty heterogenoty Het.P, I-squared; 88 Chapter 5 5 were associated with VF in the GEVAMI sample. Meta-analysis of association sta- tisticsBonferroni from- correctedAGNES, PREDESTINATIONsignificance threshold and for GEVAMI10 independent resulted tests in ( Pa ≤combined 0.005). odds ratio of 1.39 (95% CI: 0.99, 1.96; P = 0.06) per copy of the C-allele for rs1750041, and However, one SNP, rs1750041, showed a nominally significant association with VF in a combined odds ratio of 0.82 (95%CI: 0.65, 1.02; P = 0.08) per copy of the A-allele for rs2135787.PREDESTINATION (OR = 1.64; 95% CI: 1.09, 2.46; P = 0.017; Table 2). The direction Additionally, we meta-analyzed our results for rs2824292 with previously pub- lishedof effect results was consistentfor this SNP across 9,44. We the included AGNES 4and case-control PREDESTINATION studies in thisstudies, meta-analysis: with the C- (1) results from the current AGNES GWAS; (2) results from replication sample of our allele increasing the risk for VF in both studies. Upon meta-analysis of the two sets, the previous GWAS 9; (3) results from the PREDESTINATION study; and (4) results from theassociation study by P -Bugertvalue for et rs1750041al. 44 After wasmeta-analysis improved (OR (Figure = 1.66; 4), 95% the CI:association 1.36, 2.02; was not statistically signifi cant (OR = 1.22; 95% CI: 0.93, 1.60; P = 0.14). P = 5.22 × 10-7).

Figure 3a | Locus-specific association map spanning around rs17739527 (top left), rs2135787 (top right), Figurers13173656 3a | Locus-specifi(bottom left), rs1497526 c association (bottom map right) spanning using thearound SNP annotationrs17739527 and (top proxy left), search rs2135787 (SNAP) (top 36. right),The plot rs13173656 is centered (bottom at the strongest left), rs1497526 SNP at each (bottom locus right)depicted using by redthe diamondSNP annotation at the highest and proxy point; searchthe (SNAP)color legend 36. Th ate theplot upper is centered right of at each the plotsstrongest corresponds SNP at toeach the locus color ofdepicted each SNP by redand diamondillustrates atthe the high- estimated LD (by means of r2) of that SNP with the strongest SNP; Superimposed on the plot are gene estlocations point; (green)the color and legend recombination at the upper rates right(blue). of Chromosome each plots corresponds positions are tobased the oncolor HapMap of each release SNP and22 2 illustratesbuild 36.2 .the estimated LD (by means of r ) of that SNP with the strongest SNP; Superimposed on the plot are gene locations (green) and recombination rates (blue). Chromosome positions are based on HapMap release 22 build 36.2.

96 | Chapter 5

Multiple genomic approaches to identify genetic susceptibility to VF in MI 89 5

5 5

Figure 3b | Locus-specific association map spanning around rs1750041(top left), Chapter 5 rs6988935 (top right), rs10096381(bottom left), rs1353342(bottom right) using the SNP annotation and proxy search (SNAP) 36. The plot is centered at the strongest SNP at each locus depicted by red diamond at the highest point; the color legend at the upper right of each plots corresponds to the color of each SNP and illustrates the estimated LD (by means of r2) of that SNP with the strongest SNP; Superimposed on the plot are gene locations (green) and recombination rates (blue). Chromosome positions are based on HapMap release 22 build 36.2.

FigureFigure 3c | Locus 3c | -Locusspecific-specific association association map spanning map spanning around around rs497602(left), rs497602(left), and rs2824292 and rs2824292 (right) (right)using using theFigure SNPthe annotation 3bSNP | Locus-specifiannotation and proxy and searchproxyc association search(SNAP) (SNAP) 36.map The spanning 36 plot. The is plotcentered around is centered at rs1750041(topthe strongestat the strongest SNP left), at SNP each rs6988935 at l ocuseach l (topocus depictedright),depicted byrs10096381(middle red bydiamond red diamond at the left), highestat the rs1353342(middle highest point; thepoint; color the legend colorright), legend at rs497602(bottom the upperat the rightupper of right left),each of plotsand each rs2824292corresponds plots corresponds (bot- 2 2 totom the color toright) the of color using each of SNPthe each SNP and SNP illustratesannotation and illustrates the and estimated the proxy estimated searchLD (by LD (SNAP)means (by meansof 36 .r Th) of ofe thatrplot) of SNPis thatcentered with SNP the with at strongest the the strongest strongest SNP;SNP SuperimposedSNP; at each Superimposed locus on depicted the onplot the by are plotred gen diamondaree locations gene locations at (green) the highest (green) and recombination point; and recombination the color rates legend (blue). rates at (blue). Chromosome the upper Chromosome right of positionseachpositions plots are basedcorresponds are onbased HapMap on to HapMap the release color release 22 of build each 22 36.2. buildSNP and36.2. illustrates the estimated LD (by means of r2) of that SNP with the strongest SNP; Superimposed on the plot are gene locations (green) and recombi- Afternation Aftermeta rates -metaanalysis (blue).-analysis Chromosome of the of 10 the SNPs, positions 10 SNPs, 2 SNPs are 2 based SNPs (rs1750041 on (rs1750041 HapMap and release rs2135787) and 22rs2135787) build, 36.2.which, which had the had the

strongeststrongest combined combined association association with VFwith, were VF, weretaken taken forwa forward to rdthe to second the second replication replication in in

the Danishthe Danish GEVAMI GEVAMI study study (Table (Table 1C). N1Ceither). Neither rs1750041 rs1750041 (OR =(OR 1.00; = 1.00;95% 95%CI: 0.74 CI: -0.74 -

1.34; 1.34;P = 0.99) P = 0.99)nor rs2135787 nor rs2135787 (OR =(OR 1.01; = 1.01;95% CI:95% 0.81 CI:- 1.26;0.81- 1.26;P = 0.99) P = 0.99)were wereassociated associated

with VFwith in VF the in GEVAMI the GEVAMI sample. sample. Meta -Manalysiseta-analysis of association of association statistics statistics from fromAGNES, AGNES, Multiple genomic approaches to identify genetic susceptibility to VF in MI | 97 PREDESPREDESTINATIONTINATION and GEVAMI and GEVAMI resulted resulted in a combinedin a combined odds oddsratio ratioof 1.39 of (95%1.39 (95% CI: CI:

0.99, 0.99,1.96; 1.96;P = 0.06) P = 0.06)per copy per copyof the of C -theallele C-allele for rs1750041 for rs1750041, and ,a andcombined a combined odds ratioodds ofratio of

0.82 (95%CI0.82 (95%CI: 0.65:- 1.0.0265;- 1.P 02= ;0. P08 =) 0.per08 copy) per copyof the of A the-allele A-allele for rs2135787 for rs2135787. .

Additionally,Additionally, we meta we - metaanalyzed-analyzed our results our results for rs2824292 for rs2824292 with with previously previously published published

resultsresults for this for SNP this 9,44 SNP. We 9,44 . includedWe included 4 case 4 -controlcase-control studies studies in this in meta this - metaanalysis:-analysis: (1) (1)

resultsresults from from the current the current AGNES AGNES GWAS GWAS; (2) ; results (2) results from from replication replication sample sample of our of our

previousprevi GWASous GWAS 9; (3) 9results; (3) results from fromthe PREDESTINATION the PREDESTINATION study; study; and (4) and results (4) results from from

the studythe study by Bugert by Bugert et al .et 44 alAfter. 44 After meta - metaanalysis-analysis (Figu (reFigu 4),re the 4) , associationthe association was not was not

statisticallystatistically significant significant (OR =(OR 1.22; = 1.22;95% CI:95% 0.93 CI:- 1.60;0.93- 1.60;P = 0.14) P = .0.14) .

98 | Chapter98 | Chapter 5 5

90 Chapter 5 5

FigureFigure 4 4 | | Forest plot plot for for meta-analysis meta-analysis of rs2824292. of rs2824292. SE, standardSE, standard Error; Error;OR, odds OR, Ratio; odds CI, Ratio; Confi CI,- Confidencedence Interval; Interval; W, W, weight. weight. Odds Odds ratios ratios are are given given per per copy copy of theof the G-allele. G-allele.

Literature-basedLiterature-based candidate candidate SNP SNP analysis analysis We selected 24 SNPs previously reported to be associated with SCD/VF (Table 3) to Wetest selected for association 24 SNPs withpreviously VF in thereported AGNES to becase-control associated set.with None SCD/VF of the (Table tested 3 )SNPs to test reached the Bonferroni-corrected P-value threshold (P ≤ 0.002). At a nominally forsignifi association cant level, with theVF inA-allele the AGNES at rs6795970 case-control was associated set. None with of thea decreased tested SNPs risk reached of VF (OR = 0.85; 95% C.I: 0.73, 0.99; P = 0.04). By testing this association in the PREDESTI- the Bonferroni-corrected P-value threshold (P ≤ 0.002). At a nominally significant level, NATION study, rs6795970 showed a trend for association with VF (OR = 0.80; 95% CI: the0.63, A-allele 1.01; P at = 0.06). rs6795970 After was meta-analysis associated of with the AGNES a decreased and the risk PREDESTINATION of VF (OR = 0.85; results for this SNP, each copy of the A-allele at rs6795970 decreased the risk of VF by 95%0.83 C.I, fold 0.73-0.99; (95% CI: 0.73, P = 0.95; 0.04). P = By 0.006). testing this association in the PREDESTINATION

study, rs6795970 showed a trend for association with VF (OR = 0.80; 95% CI, 0.63-1.01; Overlap with expression quantitative trait locus data P = 0.06). After meta-analysis of the AGNES and the PREDESTINATION results for 1. Forward approach (from VF-SNP to eQTL) thisWe SNP, assessed each thecopy 10 of independent the A-allele VF-associated at rs6795970 locidecreased from the the currentrisk of AGNESVF by 0.83 GWAS fold for cis-eQTL eff ects in three eQTL databases (see material and methods). A total of 235 (95% CI, 0.73-0.95, P = 0.006). genes were present in the 1 Mb regions spanning the 10 loci, resulting in a Bonferroni- corrected eQTL P-value threshold of 0.05/235 = 2 × 10−4). Overlap with expression quantitative trait locus data (1.1) Th e human donor heart eQTL database: In the human donor heart eQTL database 39, none of the 10 SNPs tested showed a Bonferroni-corrected signifi cant associa- 1. Forward approach (from VF-SNP to eQTL) tion; 3 SNPs however showed a nominally signifi cant association with the expression level of genes located in cis (Table 4 & Figure 5). rs2824292 decreased the expression Welevel assessed of the the BTG3 10 independent transcript ( PVF-associated = 0.02; Table loci 4). rs17739527from the current was nominallyAGNES GWAS associ- for ated with a decreased level of the ATP5F1 transcript (P = 0.001), and with increased cislevels-eQTL of effects three transcripts in three eQTL namely databases RHOC (see (P = 0.04), material CEPT1 and methods). (P = 0.047), A and total DDX20 of 235 genes were present in the 1 Mb regions spanning the 10 loci, resulting in a Bonferroni-

corrected eQTL P-value threshold of 0.05/235 = 2 × 10-4).

Multiple genomic approaches to identify genetic susceptibility to VF in MI | 99 Multiple genomic approaches to identify genetic susceptibility to VF in MI 91

Table 3 | Association of candidate SCD SNPs with ventricular fibrillation in AGNES SNP Ref. Chr. Position C/N CAFa AGNES (n = 1433) Gene OR (95% CI) P-value rs6795970 24 3 38,741,679 A/G 39 0.85(0.73,0.99) 0.04 SCN10A rs2383207 34 9 22,105,959 A/G 49 0.86(0.74,1.01) 0.06 CDKN2B-AS1 rs10503929 29 8 32,733,525 C/T 18 0.86(0.71,1.04) 0.12 NRG1 rs7366407 38 1 116,219,426 A/T 27 0.89(0.75,1.05) 0.17 Near HNRNPA1P43 rs3864180 11 13 91,234,489 A/G 63 1.11(0.95,1.3) 0.19 GPC5 rs1042714 26,37 5 148,186,666 C/G 57 1.09(0.94,1.28) 0.26 ADRB2 (Gln27Glu) rs2283222 22 11 2,712,177 C/T 32 1.08(0.92,1.26) 0.34 KCNQ1 rs1805124 32 3 38,620,424 C/T 22 1.08(0.90,1.31) 0.40 SCN5A (Pl(A1/A2)) rs12084280 38 1 160,290,433 C/G 10 0.89(0.69,1.16) 0.40 Near NOS1AP rs17500488 38 1 116,013,286 C/T 08 1.09(0.84,1.43) 0.51 VANGL1 rs360717b 27 11 111,539,935 A/G 28 0.95(0.80,1.12) 0.53 IL‑18 rs9862154 38 3 32,096,706 C/G 81 0.94(0.78,1.14) 0.54 Near GPD1L rs10918859 38 1 160,435,892 A/G 18 1.06(0.87,1.30) 0.56 NOS1AP rs3010396 38 1 116,075,311 A/G 44 0.96(0.83,1.12) 0.62 CASQ2 rs2077316 28 10 63,895,454 A/C 90 0.92(0.64,1.32) 0.64 ZNF365

rs2200733 31 4 111,929,618 C/T 87 0.95(0.76,1.19) 0.67 Near MIR297 Chapter 5 rs6730157 28 2 135,623,558 A/G 67 1.03(0.88,1.2) 0.73 RAB3GAP1 rs12567209 25 1 160,303,103 A/G 08 0.95(0.72,1.26) 0.73 NOS1AP rs710448 36 3 187,935,579 A/G 61 0.98(0.84,1.14) 0.79 KNG1 rs1801253 45 10 115,795,046 C/G 73 1.02(0.86,1.21) 0.83 ADRB1 (Arg389Gly) rs6065 33 17 4,777,161 C/T 92 1.03(0.78,1.37) 0.84 GP1BA (HPA‑2 Met/ VNTR B) rs5918 32 17 42,715,729 C/T 15 1.01(0.82,1.25) 0.90 ITGB3 (Leu33Pro of GPIIIa) rs3766871 35 1 235,844,707 A/G 3 0.98(0.62,1.54) 0.91 RYR2 rs1492099 36 3 149,920,193 C/T 86 0.99(0.8,1.23) 0.94 AGTR1 Chr, chromosome; Ref, reference; C/N, coded allele/ non coded allele; CAF, coded allele frequency; OR, odds ratio per copy of coded allele. aCAF, frequency of coded allele in the cases and controls combined; bProxy for rs187238 (r2 = 1.0). 92 Chapter 5

(P = 0.037). rs1750041 was associated with a decreased level of the ARID1B transcript (Beta = −0.19; SE: 0.03; P = 0.03; Table 4 & Figure 5).

(1.2) GTeX human heart, left ventricle eQTL database: None of the 10 selected SNPs that were searched in the human heart left ventricle GTEx resource displayed eQTLs at the Bonferroni-corrected level of statistical significance, but seven out of ten SNPs showed associations with expression levels at the nominal significance level Table( 4). These were rs10096381 which was associated with FDFT1 (P = 0.03), rs6988935 associated with TMEM70 (P = 0.001), rs497602 with TRIB3 (P = 0.04), rs2135787 with TIMP4 (P = 0.03) and HRH1 (P = 0.01), rs17739527 with CEPT1 (P = 0.02) and ADORA3 (P = 0.004), and finally rs1750041 was associated with SERAC1 (P = 0.02).

(1.3) The Genenetwork peripheral blood eQTL database (n = 5311): Two SNPs (rs1750041 and rs10096381) out of 10 SNPs displayed eQTL effects in peripheral blood passing the Bonferroni-corrected statistical significance level. Each copy of the T-allele at rs1750041 was associated with a decreased level of transcript expression of GTF2H5 (P = 1.66 × 10−46), SERAC1 (P = 7.94 × 10−5) and SYNJ2 (P = 6.34 × 10−4). Each copy of the

Table 4 | The strongest AGNES SNP sought in the human eQTL databases SNP Chr:Positiona Allelesb MAF Gene Probea Human heart 39 Effect Size SE P-value rs2824292 21:18787176 G/A 0.48 BTG3 ILMN_1707339 −0.19 0.08 0.02 rs10096381 8:10813904 ** NA No eQTL NA NA NA rs6988935 8:75212074 C/A 0.37 TMEM70 ILMN_1739032 0.07 0.09 0.40 rs1497526 6:27883201 A/G 0.03 TRIM27 ILMN_1655482 −0.01 0.12 0.90 PGBD1 ILMN_1791949 0.08 0.12 0.51 rs497602 20: 900416 ** ** TRIB3 ** NA NA NA rs2135787 3: 11977551 A/C 0.36 TIMP4 ILMN_1663399 −0.20 0.16 0.22 HRH1 ILMN_1673953 −0.07 0.05 0.19 rs1353342 9: 78874769 A/C 0.08 NA No eQTL NA NA NA rs13173656 5: 163505869 T/C 0.11 NA No eQTL NA NA NA rs17739527 1: 112372262 C/T 0.25 ATP5F1 ILMN_1672191 −0.36 0.13 0.005 ATP5F1 ILMN_1721989 −0.2 0.06 0.001 RHOC ILMN_1673305 0.1 0.05 0.04 CEPT1 ILMN_1676588 0.09 0.05 0.05 DDX20 ILMN_1694983 0.07 0.03 0.04 rs1750041 6: 158498519 A/G 0.17 ARID1B ILMN_1660433 −0.07 0.03 0.03 SYNJ2 ILMN_1660433 0.06 0.08 0.4 Chr, Chromosome; MAF, Minor Allele frequency; SE, Standard Error; NA, not available; aPosition and Probe are based on Koopman et al. bAlleles refers to minor allele/major allele. ** SNP not available in Koopman et al. Multiple genomic approaches to identify genetic susceptibility to VF in MI 93

Table 4 | The strongest AGNES SNP sought in the human eQTL databases (continued) SNP Genes within 1 Mba rs2824292 NS: C21orf34, BTG3, C21orf91, PRSS7, C21orf37, NCRNA00157; BP: C21orf114, C21orf91-OT1, CXADR, U03241, CHODL, TBX1 rs10096381 NS: MSRA, T-SP1, GATA4, XKR6, SOX7, NEIL2, C8orf74; BP: MTMR9, C8orf49, AMAC1L2, FDFT1, BX105370, C8orf16, CTSB, PINX1, BG181200, FAM167A, TDH, BLK, C8orf13 rs6988935 NS: LY96, TMEM70, STAU2, GDAP1, STAU2, GDAP1, CRISPLD1, PI15, UBE2W, LOC100128126, JPH1, TCEB1; BP: AK125786, RDH10, TYW3, DB201663, PRKRIR, BM674958, DR033690 rs1497526 NS: GPX6, GUSBL1, NKAPL, PP14762, FKSG83, ZNF165, HIST1H2AI, HIST1H2BK, TRIM27, ZNF391, ZSCAN16, HIST1H4I, ZNF192, ZKSCAN4, RFP, ZKSCAN3, HIST1H4L, ZNF323, PGBD1, SCAND3, HIST1H1B, ZNF187, ZSCAN12L, HIST1H3J, GPX5, ZNF193; BP: AK311627, ZSCAN12, HIST1H2AL, OR2B2, HIST1H2AG, HIST1H2AM, HIST1H2BJ, HIST1H4J, PRSS16, HIST1H4K, BC043177, AX747641, HIST1H2BM, HIST1H2BN, HIST1H3I, RPSA, AK123552, HIST1H2BO, HIST1H2AH, BI915787, OR2B6, ZSCAN23, HIST1H2AJ, HIST1H3H, HIST1H2BL, POM121L2, ZNF204, ZNF184, HIST1H2AK rs497602 NS: DEFB128, SNPH, PSMF1, SDCBP2, SIRPA, NSFL1C, NRSN2, FKBP1A, C20orf55, DEFB127, TCF15, SRXN1, TBC1D20, ZCCHC3, SIRPD, C20orf54, C20orf46; BP: TRIB3, DEFB125, SIRPB1, SIRPG, AK000809, DEFB126, FAM110A, SCRT2, SOX12, C20orf96, RBCK1, AI948563, DEFB129, DEFB32, ANGPT4, CSNK2A1, DEFB132, RSPO4 rs2135787 NS: RPL32, TIMP4, SYN2, TSEN2, VGLL4, SLC6A11, C3orf31, SLC6A1, Chapter 5 LOC100129480, IQSEC1, HRH1, RAF1; BP: LOC100128644, BX111027, LOC100506990, MKRN2, ATG7, BG754745, CD240199, AI769555, GSTM1L, SNORA7A, TMEM40, PPARG, CAND2 rs1353342 NS: VPS13A, RFK, PRUNE2, PCSK5, LAMR1P15, KIAA0367; BP: BP432430, AF070595, FOXB2, LOC100130426, GCNT1, PCA3 rs13173656 NS: NUDCD2, HMMR, CCNG1, MAT2B; BP: AW836649 rs17739527 NS: WDR77, RAP1A, FAM19A3, CEPT1, LOC149620, DDX20, KCND3, RHOC, ADORA3, ST7L, MOV10, FAM212B, C1orf183, CHIA, WNT2B, C1orf88, SLC16A1, C1orf162, CTTNBP2NL, ATP5F1, LOC100129269; BP: AK123703, AX747733, PPM1J, CAPZA1, BC063548, OVGP1, TMEM77, DENND2D, FLJ36116, CR738423, CHI3L2, AFARP1 rs1750041 NS: SYNJ2, RSPH3, C6orf35, TAGAP, TULP4, TMEM181, ARID1B, ZDHHC14, SNX9, VIL2, DYNLT1 BP: BX117128, BX116720, BX119561, GTF2H5, SYTL3, EZR, AK125637, CU680699, SERAC1, OSTCP1 NS, not significant (i.e. probe passed quality control and background expression threshold, but no significant eQTL effect); BP, bad probe (i.e. probe did not pass quality control: bad probe design resulting in ambiguous mapping or potential spurious eQTL effects) or not detected (i.e. probe passed quality control, but is not expressed above background levels). a List of non-significant genes within 1Mb region spanning the SNP 94 Chapter 5 5

Figure 5 | Plot of mRNA intensities for strongest AGNES SNPs with eQTL eff ects in the human heart39 . Figure 5 | Plot of mRNA intensities for strongest AGNES SNPs with eQTL effects in the human heart 39.

(1.3) The Genenetwork peripheral blood eQTL database (n = 5311): Two SNPs

(rs1750041 and rs10096381) out of 10 SNPs displayed eQTL effects in peripheral blood passing the Bonferroni-corrected statistical significance level. Each copy of the T-allele at rs1750041 was associated with a decreased level of transcript expression of GTF2H5

Multiple genomic approaches to identify genetic susceptibility to VF in MI | 103 Multiple genomic approaches to identify genetic susceptibility to VF in MI 95 gene XKR6 Cognate 40 Ref. a 0.91 P -value OR PREDESTINATION (95% CI) 0.98(0.75,1.29) a −5 10 × AGNES P -value 1.80 Chapter 5 AGNES OR(95% CI) 1.58(1.28,1.94) 51 CAF AGNES C/G C/N AGNES 10,231,182 Position 8 Chr SNP rs6982308 value adjusted for age and sex for age study. in the AGNES adjusted value P- Table 5 | Association VF in a significant of MI the displaying eQTL effect SNPs (i.e. heart) with in human of eQTL SNPs setting Table a C/N, coded allele/ non coded allele; CAF, coded allele frequency; OR, odds ratio per copy of coded allele; Ref, reference. of coded allele; Ref, per copy coded allele frequency; OR, odds ratio coded allele/ non coded allele; CAF, C/N, 96 Chapter 5

G-allele at rs10096381 was associated with a decreased level of expression of XKR6 (P ≤ 9.12 × 10−5). No significant eQTLs were found for rs1353342 and rs13173656 in any of the databases we investigated.

2. Backward approach (from eQTL to VF-SNP) We selected 911 previously identified eQTLs in human heart 39,40 and tested their association with VF in the AGNES study (Table 5) using a Bonferroni-corrected significance threshold of P < 5.5 × 10−5 (0.05/911). In this analysis, one SNP, rs6982308 regulating XKR6 gene in the human heart left ventricle tissue of the GTEx database, displayed a Bonferroni-corrected significant effect on VF. At nominal statistical significant level 199 SNPs (45 independent SNPs at r2 < 0.7) were additionally associated with VF. Each copy of the C-allele at rs6982308 increased the risk of VF by 1.58 fold (95% CI: 1.28, 1.94; P = 1.8 × 10−5; Table 5). After meta-analysis of test statistics from the AGNES and the PREDESTINATION for these SNPs, each copy of the C-allele at rs6982308 increased the risk of VF by 1.31 fold (95% CI: 1.10, 1.55; P = 0.002).

Pathway analysis We used a gene set enrichment analysis (GSEA) approach to test whether previously identified biological pathways were enriched for VF-associated genetic loci in the current study (Table 6). A GSEA P-value of less than 2 × 10−3 corresponding to FDR < 0.1 was considered for the selection of significant pathways (see methods for detail). The associations in the current study were mainly enriched in pathways involved in (i) Pathways related to signal transmission to CREB phosphorylation (Figure 6) such as Neurotransmitter receptor binding and downstream transmission in the postsynaptic cell (P = 1.3 × 10−3), Trafficking of Amino-3-hydroxy-5-methyl-4-isoxazolepropionate (AMPA) receptors (P = 1.0 × 10−4), Post N-methyl-D-aspartate (NMDA) receptor activation events (P = 3.5 × 10−3), phosphorylation of cyclic AMP response element binding protein (CREB) through the activation of CAMKII (P = 8.0 × 10−4), CACAM pathway (P = 3.2 × 10−3); RAS activation upon Ca2+ influx through NMDA receptor (P = 9.0 × 10−4), CREB phosphorylation through the activation of RAS (P = 1.2 × 10−3), and PGC1A pathway (P = 3.0 × 10−4); (ii) Cell cycle pathways such as Packaging of telomere ends (P = 4.4 × 10−5), and Telomere maintenance (P = 3.0 × 10−4); and (iii) Other Pathways such as RNA polymerase I promoter opening (P = 1.0 × 10−4) and Glucuronidation (P = 7.4 × 10−3). Multiple genomic approaches to identify genetic susceptibility to VF in MI 97

Table 6 | Pathway analysis of SNPs tested in the AGNES case control set Database Pathway Genes Nominal GSEA P-value CREB phosphorylation RELATED pathways: REACTOME Trafficking of AKAP5, AP2A1, AP2A2, AP2B1, AP2M1, AP2S1, CACNG2, 1.0 × 10−4 AMPA receptors CACNG3, CACNG4, CACNG8, CAMK2A, CAMK2B, CAMK2D, DLG1, EPB41L1, GRIA1, GRIA2, GRIA3, GRIA4, GRIP2, LOC651907, MDM2, MYO6, NSF, PICK1, PRKCA, PRKCB, PRKCG BIOCARTA PGC1A pathway CALM1, CALM2, CALM3, CAMK1, CAMK1G, CAMK2A, 3.0 × 10−4 CAMK2B, CAMK2D, CAMK2G, CAMK4, CYCSP35, ESRRA, HDAC5, LOC124827, LOC147908, MEF2A, MEF2BNB-MEF2B, MEF2C, MEF2D, PPARA, PPARGC1A, PPP3CA, PPP3CB, PPP3CC, SLC2A4, YWHAH REACTOME CREB phosphory- AKAP9, CREB1, GRIN1, GRIN2A, GRIN2B, GRIN2C, GRIN2D, 8.0 × 10−4 lation through NEFL, CALM1, CALM2, CALM3, CAMK2A, CAMK2B, CAMK2D, ACTN2 the activation of CAMKII REACTOME RAS activation AKAP9, GRIN1, GRIN2A, GRIN2B, GRIN2C, GRIN2D, HRAS, 9.0 × 10−4 upon Ca2+ influx NEFL, RASGRF1, RASGRF2, CALM1, CALM2, CALM3, CAMK2A, CAMK2B, CAMK2D, ACTN2 through NMDA receptor REACTOME Post NMDA ACTN2, ADCY1, ADCY3, ADCY8, AKAP9, BRAF, CALM1, 3.5 × 10−3 receptor activation CALM2, CALM3, CAMK2A, CAMK2B, CAMK2D, CAMK4, CAMKK1, CREB1, GRIN1, GRIN2A, GRIN2B, GRIN2C,

events Chapter 5 GRIN2D, HRAS, MAPK1, NEFL, PDPK1, PRKACB, RAF1, RASGRF1, RASGRF2, RPS6KA1, RPS6KA2, RPS6KA3, RPS6KA6, RRAS BIOCARTA CACAM pathway CALM1, CALM2, CALM3, CAMK1, CAMK1G, CAMK2A, 3.2 × 10−3 CAMK2B, CAMK2D, CAMK2G, CAMK4, CAMKK1, CAMKK2, CREB1, CYCSP35, LOC124827, LOC147908 REACTOME Neurotransmitter ACTN2, ADCY1, ADCY2, ADCY3, ADCY4, ADCY5, ADCY6, 1.3 × 10−3 receptor binding ADCY7, ADCY8, ADCY9, AKAP5, AKAP9, AP2A1, AP2A2, AP2B1, AP2M1, AP2S1, ARHGEF9, BRAF, CACNG2, CACNG3, and downstream CACNG4, CACNG8, CALM1, CALM2, CALM3, CAMK2A, transmission in the CAMK2B, CAMK2D, CAMK4, CAMKK1, CHRFAM7A, postsynaptic cell CHRNA1, CHRNA2, CHRNA3, CHRNA4, CHRNA5, CHRNA6, CHRNA7, CHRNA9, CHRNB2, CHRNB3, CHRNB4, CHRND, CHRNE, CHRNG, CREB1, DLG1, DLG3, EPB41L1, GABBR1, GABBR2, GABRA1, GABRA2, GABRA3, GABRA4, GABRA5, GABRA6, GABRB1, GABRB2, GABRB3, GABRG2, GABRG3, GABRR1, GABRR2, GNAI1, GNAI2, GNAI3, GNAL, GNB1, GNB2, GNB3, GNB4, GNB5, GNG10, GNG11, GNG12, GNG13, GNG2, GNG3, GNG4, GNG5, GNG7, GNG8, GNGT1, GNGT2, GRIA1, GRIA2, GRIA3, GRIA4, GRIK1, GRIK2, GRIK3, GRIK4, GRIK5, GRIN1, GRIN2A, GRIN2B, GRIN2C, GRIN2D, GRIP2, HRAS, KCNJ10, KCNJ12, KCNJ15, KCNJ16, KCNJ2, KCNJ3, KCNJ4, KCNJ5, KCNJ6, KCNJ9, LOC651907, LOC732445, MAPK1, MDM2, MYO6, NCALD, NEFL, NSF, PDPK1, PICK1, PLCB1, PLCB2, PLCB3, PRKACB, PRKCA, PRKCB, PRKCG, RAF1, RASGRF1, RASGRF2, RPS6KA1, RPS6KA2, RPS6KA3, RPS6KA6, RRAS 98 Chapter 5

Table 6 | Pathway analysis of SNPs tested in the AGNES case control set (continued) Database Pathway Genes Nominal GSEA P-value REACTOME CREB phosphory- ACTN2, AKAP9, BRAF, CALM1, CALM2, CALM3, CAMK2A, 1.2 × 10−3 lation through the CAMK2B, CAMK2D, CREB1, GRIN1, GRIN2A, GRIN2B, GRIN2C, GRIN2D, HRAS, MAPK1, NEFL, PDPK1, RAF1, activation of RAS RASGRF1, RASGRF2, RPS6KA1, RPS6KA2, RPS6KA3, RPS6KA6, RRAS - Cell Cycle pathways: REACTOME Packaging of telo- HIST4H4, HIST3H2BB, HIST1H2BA, POT1, TINF2, 4.4 × 10−5 mere ends HIST1H2AE, HIST1H2AD, H2AFX, H2AFZ, HIST1H2BD, HIST1H2BB, TERF2IP, HIST2H4B, ACD, TERF1, TERF2, HIST2H2AA4, HIST3H3, HIST1H4I, HIST1H2AJ, HIST1H2AC, HIST1H2AB, HIST2H2AA3, HIST2H2AC, HIST1H2BG, HIST1H2BL, HIST1H2BN, HIST1H2BM, HIST1H2BF, HIST1H2BE, HIST1H2BH, HIST1H2BI, HIST1H2BC, HIST1H2BO, HIST2H2BE, HIST1H4A, HIST1H4D, HIST1H4F, HIST1H4K, HIST1H4J, HIST1H4C, HIST1H4H, HIST1H4B, HIST1H4E, HIST1H4L, HIST2H4A, HIST1H2BK, HIST1H2BJ REACTOME Telomere mainte- ACD, DKC1, DNA2, FEN1, H2AFX, H2AFZ, HIST1H2AB, 3.0 × 10−4 nance HIST1H2AC, HIST1H2AD, HIST1H2AE, HIST1H2AJ, HIST1H2BA, HIST1H2BB, HIST1H2BC, HIST1H2BD, HIST1H2BE, HIST1H2BF, HIST1H2BG, HIST1H2BH, HIST1H2BI, HIST1H2BJ, HIST1H2BK, HIST1H2BL, HIST1H2BM, HIST1H2BN, HIST1H2BO, HIST1H4A, HIST1H4B, HIST1H4C, HIST1H4D, HIST1H4E, HIST1H4F, HIST1H4H, HIST1H4I, HIST1H4J, HIST1H4K, HIST1H4L, HIST2H2AA3, HIST2H2AA4, HIST2H2AC, HIST2H2BE, HIST2H4A, HIST2H4B, HIST3H2BB, HIST3H3, HIST4H4, LIG1, NHP2, PCNA, POLA1, POLA2, POLD1, POLD2, POLD3, POLD4, POLE, POLE2, POT1, PRIM1, PRIM2, RFC2, RFC3, RFC4, RFC5, RPA1, RPA2, RPA3, RUVBL1, RUVBL2, TERF1, TERF2, TERF2IP, TERT, TINF2, WRAP53 - Other pathways: REACTOME RNA polymerase I H2AFX, H2AFZ, H3F3A, H3F3AP5, H3F3AP6, H3F3B, 1.0 × 10−4 promoter opening HIST1H2AB, HIST1H2AC, HIST1H2AD, HIST1H2AE, HIST1H2AJ, HIST1H2BA, HIST1H2BB, HIST1H2BC, HIST1H2BD, HIST1H2BE, HIST1H2BF, HIST1H2BG, HIST1H2BH, HIST1H2BI, HIST1H2BJ, HIST1H2BK, HIST1H2BL, HIST1H2BM, HIST1H2BN, HIST1H2BO, HIST1H3A, HIST1H3B, HIST1H3C, HIST1H3D, HIST1H3E, HIST1H3F, HIST1H3G, HIST1H3H, HIST1H3I, HIST1H3J, HIST1H4A, HIST1H4B, HIST1H4C, HIST1H4D, HIST1H4E, HIST1H4F, HIST1H4H, HIST1H4I, HIST1H4J, HIST1H4K, HIST1H4L, HIST2H2AA3, HIST2H2AA4, HIST2H2AC, HIST2H2BE, HIST2H3A, HIST2H3C, HIST2H3D, HIST2H4A, HIST2H4B, HIST3H2BB, HIST4H4, LOC731820, MAPK3, MBD2, UBTF REACTOME Glucuronidation SLC35D1, UGDH, UGP2, UGT1A1, UGT1A3, UGT1A4, 7.4 × 10−3 (Metabolism) UGT1A5, UGT1A6, UGT1A7, UGT1A8, UGT1A9, UGT2A1, UGT2B10, UGT2B11, UGT2B17, UGT2B28, UGT2B4, UGT2B7 Multiple genomic approaches to identify genetic susceptibility to VF in MI 99

Discussion

Using a multi-tiered approach in three independent patient sets (AGNES, PREDESTI- NATION, GEVAMI) with a first acute MI and complimentary genetic approaches in- volving eQTL analysis, we identified 5 SNPs with varying evidence of association with VF that merit further investigation in future studies (rs2824292, rs1750041, rs10096381, rs6982308, and rs6795970). We also attempted to identify putatively causal candidate genes at these loci by overlaying our association data with eQTL data. Furthermore, by assessing SNPs associated with VF in AGNES we identified a number of enriched biological pathways that could be involved in predisposition to VF.

1. rs2824292 In our extended GWAS analysis in the AGNES population, the locus at chr. 21q21 (rs2824292), which we identified in the initial AGNES GWAS 9, remained associated with VF at a genome-wide statistical significance level. In the initial AGNES GWAS 9, the association of rs2824292 with VF was replicated in an independent Dutch case- control set (chapter 4 of this thesis) consisting of patients presenting with VF in the setting of MI (cases) drawn from the ARREST study 46 and MI-survivors (controls) drawn from the GENDER study 47. In spite of this initial replication, and the fact that the signal remained at genome-wide statistical significance in the extended GWAS presented herein, in the current study the effect of rs2824292 on VF was not replicated Chapter 5 in the PREDESTINATION case-control set which has inclusion criteria similar to AG- NES (Table 2). Similarly, another study 44 conducted in a small case-control set from Germany did not detect an association of this locus with VF. After meta-analysis of the test statistics for rs2824292 in these 4 case-control sets, the locus was not associated with VF. This meta-analysis also demonstrated a remarkable heterogeneity in the effect size and direction between the studies (Figure 4) which may be due to differences in phenotype definition and study design, statistical power or pure chance 10. The only genes within 1Mb of rs2824292 are CXADR (encoding the coxsackie and adenovirus receptor) and BTG3 (encoding B-cell translocation gene 3). The coxsackie and adenovirus receptor (CAR) is a transmembrane protein located at the intercalated disks of cardiomyocytes and has a role in conduction of the cardiac electrical impulse at the level of the atrio-ventricular node. We 48 investigated CXADR as a candidate gene for the effect on VF at this locus and demonstrated that decreased expression of CXADR slows conduction in the ventricles and modulates susceptibility to arrhythmia in the setting of myocardial ischemia. These observations make CXADR a likely causal gene for the effect on VF observed in AGNES at this locus. This is further supported by our observation 48 that the risk allele (G) at rs2824292 is associated with a lower expression of CXADR in human heart and has no effect on BTG3 transcript abundance. This 100 Chapter 5

data, generated by qRT-PCR, is at variance with the observation made by Koopmann et al. who using an array-based approach for determination of transcript abundance showed that rs2824292 was an eQTL for BTG3 albeit at a nominally-significant level. It is here imperative to realize that these two studies were conducted on the same human heart samples and therefore the differences in association observed across these studies is very likely due to the different approaches used in the measurement of mRNA abundance, with the qRT-PCR assessment of Marsman et al. being the most accurate. It is clear that the effects of rs2824292 on the expression ofCXADR and BTG3 require further assessment in additional heart samples which are unfortunately scarce. Thus far, in aggregate, data from functional and gene expression studies suggests that rs2824292 may increase the risk of VF by decreasing the level of expression of CXADR.

2. rs1750041 Besides rs2824292, 9 other independent loci with association below our arbitrary threshold were taken forward for replication in the PREDESTINATION and GEVAMI studies. Of these, one SNP (rs1750041 at chr. 6p25) was nominally associated with VF in PREDESTINATION. This SNP however did not display association with VF in GEVAMI. rs1750041 is located in an intron of the SYNJ2 gene encoding synaptojanin 2. This gene has not been previously implicated in arrhythmia susceptibility. SYNJ2 is widely expressed and is also expressed in human heart and cardiac myocytes and mostly localized in cytosol and plasma membrane. Synaptojanin 2 is a member of the inositol-polyphosphate 5-phosphatase family. This family is involved in synaptic transmitter release and interacts with ion channels 49 such as inward rectifier potassium (Kir) channels 50 and calcium channels 51,52. Phosphatidylinositol-4,5-bisphosphate (PIP2) malfunction has been implicated in arrhythmia 53 (reviewed in ref. 54), and in QT-interval prolongation in the Long-QT syndrome 55,56. A mutation in the Phospha- tidylinositol-4,5-bisphosphate binding site of the potassium channel encoded by KCNJ2 has been linked to a familial disorder featuring ventricular arrhythmias 57. In spite of these suggestive observations, we found no Bonferroni-corrected eQTL effect of rs1750041 on SYNJ2 expression in any of the eQTL databases that we investigated. However, we found that rs1750041 modulates the expression of GTF2H5 (in blood), and SERAC1 (in blood and left ventricle). The strongest eQTL effects were reported in blood. None of these have been previously implicated in arrhythmia susceptibility.

3. rs6795970 The current study builds further on our earlier analysis 12,24 that provided the first evidence that rs6795970 (a non-synonymous SNP in SCN10A) is associated with VF in AGNES. The statistical significance of this association improved in the larger Multiple genomic approaches to identify genetic susceptibility to VF in MI 101

AGNES case-control set studied here. Furthermore, this SNP also showed a trend for association in PREDESTINATION with a similar direction of effect and as expected, the association P-value was lower when the AGNES and PREDESTINATION sets were meta-analyzed. In recent years a large body of literature has accumulated on rs6795970 and the haplotype on which it occurs. This SNP was first linked to cardiac electrical function in GWAS on the electrocardiographic PR-interval 24,58,59 and subsequently QRS-duration 60. In all these studies, the A-allele of rs6795970 was associated with slower cardiac conduction (i.e. prolongation of the PR-interval and QRS duration). These same studies 58 and subsequent studies also linked this SNP to pacemaker implantation as a result of atrioventricular block 58, and atrial fibrillation (AF) 61-64. Surprisingly, the conduction-slowing A-allele is protective of VF in AGNES and PRE- DESTINATION and is also protective of AF in the respective studies 59,61,62. Several studies have attempted to identify the genetic and electrophysiological mechanism by which genetic variation in the haplotype tagged by rs6795970 modulates cardiac conduction. rs6795970 causes a Valine to Alanine substitution at residue 1073 in the Nav1.8 sodium channel encoded by SCN10A. Strong evidence exists linking another variant in this haplotype with the effects observed on conduction. Studies working on SCN10A identified a common variant (rs6801957) that resides in an enhancer element located in an intron of SCN10A. This variant is in high LD with rs6795970 (r2 = 0.93). rs6801957 contains a TBX3/TBX5 (T-box transcription factors) binding site and regulates the expression of the nearby SCN5A gene (which encodes Chapter 5 Nav1.5, the mediator of cardiomyocyte depolarization and a critical mediator of conduction in the heart). A series of studies have provided convincing evidence that the A‑allele at rs6801957 disrupts TBX3/TBX5 binding to the enhancer, thereby reducing the expression of SCN5A 65,66. In support of this, the A‑allele at SCN10A rs6801957 has been shown to be associated with reduced SCN5A mRNA in human heart 66.

4. rs6982308 and rs10096381 The second most significant SNP in the AGNES GWAS (rs10096381) is located in an intron of the XKR6 gene (encoding XK, Kell blood group complex subunit-related family, member 6). This SNP displayed an eQTL effect onXKR6 in peripheral blood 41. Of note, in our reverse approach (eQTL to VF-SNP) we uncovered that another SNP, rs6982308, which displays an eQTL effect onXKR6 in human heart 40, is also associated with VF in the AGNES study. rs6982308 and rs10096381 are distant SNPs separated by multiple genes and are not in LD with each other. In fact, SNP rs6982308 is located in the methionine sulfoxide reductase (MSRA) gene, which is located 467 kb far from XKR6 (Figure 3b). MSRA (the protein expressed by MSRA gene) is a regulator of oxi- dative stress and lifespan 67, protects cardiomyocytes against hypoxia/reoxygenation, stress and cell death 68 and protects the heart from ischemia-reperfusion injury 69. 102 Chapter 5

Currently, little is known about the XKR6 gene. Common genetic variants in XKR6 (e.g. rs7819412) have been previously linked to triglyceride levels 70 hypertri- glyceridemia 71 and systemic lupus erythematous (rs6985109, rs4240671, rs11783247, rs6984496) 72-75. The underlying pathology linking XKR6 to all these apparently different phenotypes is still unknown. rs10096381 that is identified in the current study is in high LD with rs6984496 and rs11783247 (r2 > 0.9; mentioned above) which have been previously identified for systemic lupus erythematous.

Pathway analysis In the current study, we demonstrated that SNPs associating with VF in the AGNES study were enriched in 12 biological pathways (Table 6). Few of these pathways such as cell cycle, RNA polymerase I promoter opening, PGC1a, and Glucuronidation could not be grouped into clear SCD-related molecular mechanisms. However, with extensive literature review we grouped the rest of the pathways (i.e. AMPA, NMDA, CaM, CACAM, CAMKII, Ras, CREB) to an important arrhythmia-related process that starts from signal transmission and eventually leads to CREB phosphorylation in a calcium-dependent manner (Figure 6). When signal arrives to the cell membrane of post synaptic cell, AMPA and NMDA receptors transmit the signal from the pre synaptic cell into the post-synaptic cell. These ligand-gated ion channels 76 regulate intracellular calcium 77. Chronic NMDA activation increases susceptibility to ventricular arrhythmias 78 whereas antagonists of NMDA or AMPA reduce calcium accumulation in cardiac mitochondria 79 and thus prevent ventricular arrhythmias 80,81. Calcium ions that are transferred inside the cell, bind to a calcium-regulating protein called

Figure 6 | Schematic figure of pathways starting from activation of AMPA and NMDA receptors to CREB phosphorylation (adapted from Figure 5 of ref. 92). Multiple genomic approaches to identify genetic susceptibility to VF in MI 103

Calmodulin (CaM) 82-84. CaM also regulates cardiac potassium 85 and sodium 86,87 channels. CaM decreases conduction in gap junction of ventricular myocytes at moderately elevated calcium level 88. Mutations in CaM gene cause various types of ventricular arrhythmias 89,90. Calcium and CaM form a calcium-calmodulin (CACAM) complex that activates enzymes called calcium-calmoduline kinases (CaMKs) 91 that mediate cardiac calcium current in ventricular myocytes and have the capacity to phosphorylate CREB in vitro and in vivo (reviewed in ref. 92 and 93). CAMKII is the most important member of CAMKs family in cardiomyocytes 94. CAMKII prevents CREB from binding to DNA and exerting its effects as a transcription factor 95. CAMKII modulates potassium current in cardiac sinoatrial (SA) node cells in an intracellular calcium dependent manner 96. Over-activation of CaMKII initiates ventricular arrhythmias 97 and inhibitors of CaMKII prevent arrhythmias 98 by reducing cardiac contractility 99. Another parallel pathway to CaMKII-mediated CREB phosphorylation is trough calcium-dependent activation of Ras (small G-protein that is involved in signal transduction) 100 which once activated can phosphorylate CREB. Ras signaling pathway is important in cardiac hypertrophy 101-103. Ras inhibitors improve cardiac performance, 104 reduce ischemia-reperfusion injury, 104-106 and decrease size of scar tissue 104.

Strengths and limitations

As mechanisms of VF likely differ across different cardiac pathologies, the fact that Chapter 5 this study was conducted exclusively in patients that presented with VF in the setting of a first MI, represents an important strength of the study. The presence of such a specific phenotype in the discovery and validation patient sets increases the likeli- hood of shared molecular mechanisms and consequently shared genetic risk factors among the cases; this strategy is thus expected to increase the statistical power for the discovery of genetic variants associated with VF. The replication sets used in this study (PREDESTINATION and GEVAMI) are highly suitable replication sets for associations uncovered in AGNES as their inclusion criteria are very similar to those of AGNES. However, although the AGNES set is large when one considers the challenges encountered in collecting clinical data and DNA from victims of VF/SCD, it is modest in comparison to those used in contemporary GWA studies on other phenotypes. The relatively small size of AGNES limits our ability to identify risk-SNPs with small effects. This underscores the necessity of continued recruitment of highly character- ized patient sets with VF to further our understanding of the underlying genetic risk factors. Due to ethical considerations, the inclusion of VF-cases in AGNES is limited to those who survive the event; individuals who do not survive at least until hospital admission are not included in the study. This implies that findings in AGNES might not be generalizable to non-survivors of VF. 104 Chapter 5

In an original approach in the current study we tested SNPs displaying eQTL effects in human heart tissue for a role in modulation of risk for VF. Such SNPs modulate the level of expression of genes in human heart and are thus highly relevant candidates for a cardiac phenotype like VF. This approach increases the power to identify biologically true associations. We acknowledge that the eQTL databases we used had a small sample size. The sample sizes for the most relevant tissues included 129 human heart samples in our earlier study 39 and 83 human heart left ventricle in the study by Lonsdale et al. 40. Hence, potential eQTLs with small effects may not be present in these eQTL databases. Furthermore, in the generation of the human heart eQTL database by Koopmann et al. not all transcripts were tested for eQTL effect due to sub-optimal probe design. Addi- tionally, quality control criteria such as MAF > 15% and optimal probe design limited the number of SNPs and transcripts included in generation of this eQTL database. Thus, SNPs with eQTL effects might have been missed here. In the study by Westra et al. 41 only eQTLs with FDR < 0.5 (corresponded to a P-value < 0.003) were accessible to us. Therefore, SNPs with an eQTLs effect below this threshold are also missed.

Conclusion We here integrated several approaches in an effort to identify genetic loci and molecular mechanisms modulating risk for VF in the setting of a first acute MI and possible causal genes therein. While the identification of robust and highly reproducible signals for this phenotype remains challenging, a number of SNPs from this study merit further investigation in future studies. These include rs2824292, which although remained highly-significantly associated with VF in an extended sample of AGNES, did not display association in case-control sets with a presumably similar phenotype. Another SNP that deserves investigation in the future is rs1750041 since it has been replicated in the PREDESTINATION study. rs6795970 stands out in our candidate SNP analysis. Although the direction of effect is unexpected (conduction slowing allele is protective for VF) this SNP is emerging as the most reproducible SNP thus far in a vari- ety of arrhythmias. Finally, rs6982308 stands out of our candidate eQTL analysis. This was also important since beside the association with VF, this SNP is an eQTL for XKR6 gene at which the second AGNES SNP rs10096381 occurs. All associations identified in the current study require further research in future studies.

Acknowledgement This study was supported by research grants from the Netherlands Heart Foundation (grants 2001D019, 2003T302 and 2007B202), the Leducq Foundation (grant 05-CVD) and the Interuniversity Cardiology Institute of the Netherlands. Multiple genomic approaches to identify genetic susceptibility to VF in MI 105

References

1. Marsman, R. F., Tan, H. L. & Bezzina, C. R. Genetics of sudden cardiac death caused by ventricular arrhythmias. Nature reviews. Cardiology 11, 96-111, doi:10.1038/nrcardio.2013.186 (2014). 2. Kong, M. H. et al. Systematic review of the incidence of sudden cardiac death in the United States. Journal of the American College of Cardiology 57, 794-801, doi:10.1016/j. jacc.2010.09.064 (2011). 3. Huikuri, H. V., Castellanos, A. & Myerburg, R. J. Sudden death due to cardiac arrhythmias. The New England journal of medicine 345, 1473-1482, doi:10.1056/NEJMra000650 (2001). 4. Bardai, A. et al. Epilepsy is a risk factor for sudden cardiac arrest in the general population. PloS one 7, e42749, doi:10.1371/journal.pone.0042749 (2012). 5. Dekker, L. R. et al. Familial sudden death is an important risk factor for primary ventricular fibrillation: a case-control study in acute myocardial infarction patients. Circulation 114, 1140-1145, doi:10.1161/circulationaha.105.606145 (2006). 6. Kaikkonen, K. S., Kortelainen, M. L., Linna, E. & Huikuri, H. V. Family history and the risk of sudden cardiac death as a manifestation of an acute coronary event. Circulation 114, 1462-1467, doi:10.1161/circulationaha.106.624593 (2006). 7. Friedlander, Y. et al. Sudden death and myocardial infarction in first degree relatives as predictors of primary cardiac arrest. Atherosclerosis 162, 211-216 (2002). 8. Jouven, X., Desnos, M., Guerot, C. & Ducimetiere, P. Predicting sudden death in the population: the Paris Prospective Study I. Circulation 99, 1978-1983 (1999). 9. Bezzina, C. R. et al. Genome-wide association study identifies a susceptibility locus at 21q21 for ventricular fibrillation in acute myocardial infarction. Nature genetics 42, 688-691, doi:

10.1038/ng.623 (2010). Chapter 5 10. Pazoki, R., Tanck, M. W., Wilde, A. A. & Bezzina, C. R. Complex inheritance for susceptibility to sudden cardiac death. Current pharmaceutical design 19, 6864-6872 (2013). 11. Arking, D. E. et al. Genome-wide association study identifies GPC5 as a novel genetic locus protective against sudden cardiac arrest. PloS one 5, e9879, doi:10.1371/journal. pone.0009879 (2010). 12. Pazoki, R. et al. SNPs identified as modulators of ECG traits in the general population do not markedly affect ECG traits during acute myocardial infarction nor ventricular fibrillation risk in this condition. PloS one 8, e57216, doi:10.1371/journal.pone.0057216 (2013). 13. Aulchenko, Y. S., Ripke, S., Isaacs, A. & van Duijn, C. M. GenABEL: an R library for genome- wide association analysis. Bioinformatics (Oxford, England) 23, 1294-1296, doi:10.1093/ bioinformatics/btm108 (2007). 14. nnclust: Nearest-neighbour tools for clustering v. R package version 2.2 (2010). 15. Li, Y., Willer, C. J., Ding, J., Scheet, P. & Abecasis, G. R. MaCH: using sequence and genotype data to estimate haplotypes and unobserved genotypes. Genet Epidemiol 34, 816-834, doi: 10.1002/gepi.20533 (2010). 16. Li, Y., Willer, C., Sanna, S. & Abecasis, G. Genotype imputation. Annual review of genomics and human genetics 10, 387-406, doi:10.1146/annurev.genom.9.081307.164242 (2009). 17. Aulchenko, Y. S., Struchalin, M. V. & van Duijn, C. M. ProbABEL package for genome-wide association analysis of imputed data. BMC bioinformatics 11, 134, doi:10.1186/1471-2105- 11-134 (2010). 106 Chapter 5

18. Dudbridge, F. & Gusnanto, A. Estimation of significance thresholds for genomewide- as sociation scans. Genetic epidemiology 32, 227-234, doi:18300295 (2008). 19. Pe’er, I., Yelensky, R., Altshuler, D. & Daly, M. J. Estimation of the multiple testing burden for genomewide association studies of nearly all common variants. Genetic epidemiology 32, 381-385, doi:18348202 (2008). 20. R: A Language and Environment for Statistical Computing (Vienna, Austria, 2014). 21. meta: Meta-Analysis with R. R package version 3.7-1. (2014). 22. Albert, C. M. et al. Common variants in cardiac ion channel genes are associated with sudden cardiac death. Circulation. Arrhythmia and electrophysiology 3, 222-229, doi:10.1161/ circep.110.944934 (2010). 23. Arking, D. E. et al. Identification of a sudden cardiac death susceptibility locus at 2q24.2 through genome-wide association in European ancestry individuals. PLoS genetics 7, e1002158, doi:10.1371/journal.pgen.1002158 (2011). 24. Chambers, J. C. et al. Genetic variation in SCN10A influences cardiac conduction. Nature genetics 42, 149-152, doi:10.1038/ng.516 (2010). 25. Eijgelsheim, M. et al. Genetic variation in NOS1AP is associated with sudden cardiac death: evidence from the Rotterdam Study. Human molecular genetics 18, 4213-4218, doi:10.1093/ hmg/ddp356 (2009). 26. Gavin, M. C. et al. A common variant in the beta2-adrenergic receptor and risk of sudden cardiac death. Heart rhythm: the official journal of the Heart Rhythm Society 8, 704-710, doi: 10.1016/j.hrthm.2011.01.003 (2011). 27. Hernesniemi, J. A. et al. Interleukin‑18 promoter polymorphism associates with the occurrence of sudden cardiac death among Caucasian males: the Helsinki Sudden Death Study. Atherosclerosis 196, 643-649, doi:10.1016/j.atherosclerosis.2007.07.018 (2008). 28. Huertas-Vazquez, A. et al. Novel loci associated with increased risk of sudden cardiac death in the context of coronary artery disease. PloS one 8, e59905, doi:10.1371/journal. pone.0059905 (2013). 29. Huertas-Vazquez, A. et al. A common missense variant in the neuregulin 1 gene is associated with both schizophrenia and sudden cardiac death. Heart rhythm: the official journal of the Heart Rhythm Society 10, 994-998, doi:10.1016/j.hrthm.2013.03.020 (2013). 30. Kao, W. H. et al. Genetic variations in nitric oxide synthase 1 adaptor protein are associated with sudden cardiac death in US white community-based populations. Circulation 119, 940-951, doi:10.1161/circulationaha.108.791723 (2009). 31. Lahtinen, A. M. et al. Common genetic variants associated with sudden cardiac death: the FinSCDgen study. PloS one 7, e41675, doi:10.1371/journal.pone.0041675 (2012). 32. Mikkelsson, J., Perola, M., Laippala, P., Penttila, A. & Karhunen, P. J. Glycoprotein IIIa Pl(A1/ A2) polymorphism and sudden cardiac death. Journal of the American College of Cardiology 36, 1317-1323 (2000). 33. Mikkelsson, J., Perola, M., Penttila, A. & Karhunen, P. J. Platelet glycoprotein Ibalpha HPA‑2 Met/VNTR B haplotype as a genetic predictor of myocardial infarction and sudden cardiac death. Circulation 104, 876-880 (2001). 34. Newton-Cheh, C. et al. A common variant at 9p21 is associated with sudden and arrhythmic cardiac death. Circulation 120, 2062-2068, doi:10.1161/circulationaha.109.879049 (2009). 35. Ran, Y. et al. Common RyR2 variants associate with ventricular arrhythmias and sudden cardiac death in chronic heart failure. Clinical science (London, England: 1979) 119, 215- 223, doi:10.1042/cs20090656 (2010). Multiple genomic approaches to identify genetic susceptibility to VF in MI 107

36. Sotoodehnia, N. et al. Genetic variation in angiotensin-converting enzyme-related pathways associated with sudden cardiac arrest risk. Heart rhythm: the official journal of the Heart Rhythm Society 6, 1306-1314, doi:10.1016/j.hrthm.2009.06.013 (2009). 37. Sotoodehnia, N. et al. Beta2-adrenergic receptor genetic variants and risk of sudden cardiac death. Circulation 113, 1842-1848, doi:10.1161/circulationaha.105.582833 (2006). 38. Westaway, S. K. et al. Common variants in CASQ2, GPD1L, and NOS1AP are significantly associated with risk of sudden death in patients with coronary artery disease. Circulation. Cardiovascular genetics 4, 397-402, doi:10.1161/circgenetics.111.959916 (2011). 39. Koopmann, T. T. et al. Genome-wide identification of expression quantitative trait loci (eQTLs) in human heart. PloS one 9, e97380, doi:10.1371/journal.pone.0097380 (2014). 40. The Genotype-Tissue Expression (GTEx) project. Nature genetics 45, 580-585, doi:10.1038/ ng.2653 (2013). 41. Westra, H. J. et al. Systematic identification of trans eQTLs as putative drivers of known disease associations. Nature genetics 45, 1238-1243, doi:10.1038/ng.2756 (2013). 42. Segre, A. V., Groop, L., Mootha, V. K., Daly, M. J. & Altshuler, D. Common inherited variation in mitochondrial genes is not enriched for associations with type 2 diabetes or related glycemic traits. PLoS genetics 6, doi:10.1371/journal.pgen.1001058 (2010). 43. Veyrieras, J. B. et al. High-resolution mapping of expression-QTLs yields insight into human gene regulation. PLoS genetics 4, e1000214, doi:10.1371/journal.pgen.1000214 (2008). 44. Bugert, P. et al. No evidence for an association between the rs2824292 variant at chromosome 21q21 and ventricular fibrillation during acute myocardial infarction in a German population. Clinical chemistry and laboratory medicine: CCLM / FESCC 49, 1237-1239, doi: 10.1515/cclm.2011.190 (2011). 45. Tseng, Z. H. et al. Common beta-adrenergic receptor polymorphisms are not associated

with risk of sudden cardiac death in patients with coronary artery disease. Heart rhythm: Chapter 5 the official journal of the Heart Rhythm Society 5, 814-821, doi:10.1016/j.hrthm.2008.03.016 (2008). 46. Blom, M. T. et al. Reduced in-hospital survival rates of out-of-hospital cardiac arrest victims with obstructive pulmonary disease. Resuscitation 84, 569-574, doi:10.1016/j.resuscitation.2012.10.009 (2013). 47. Agema, W. R. et al. Current PTCA practice and clinical outcomes in The Netherlands: the real world in the pre-drug-eluting stent era. European heart journal 25, 1163-1170, doi: 10.1016/j.ehj.2004.05.006 (2004). 48. Marsman, R. F. et al. Coxsackie and adenovirus receptor is a modifier of cardiac conduction and arrhythmia vulnerability in the setting of myocardial ischemia. J Am Coll Cardiol 63, 549-559, doi:10.1016/j.jacc.2013.10.062 (2014). 49. Gamper, N. & Shapiro, M. S. Regulation of ion transport proteins by membrane phosphoinositides. Nature reviews. Neuroscience 8, 921-934, doi:10.1038/nrn2257 (2007). 50. Furst, O., Mondou, B. & D’Avanzo, N. Phosphoinositide regulation of inward rectifier potassium (Kir) channels. Frontiers in physiology 4, 404, doi:10.3389/fphys.2013.00404 (2014). 51. Shen, D., Wang, X. & Xu, H. Pairing phosphoinositides with calcium ions in endolysosomal dynamics: phosphoinositides control the direction and specificity of membrane trafficking by regulating the activity of calcium channels in the endolysosomes. BioEssays: news and reviews in molecular, cellular and developmental biology 33, 448-457, doi:10.1002/ bies.201000152 (2011). 108 Chapter 5

52. Michailidis, I. E., Zhang, Y. & Yang, J. The lipid connection-regulation of voltage-gated Ca(2+) channels by phosphoinositides. Pflugers Archiv: European journal of physiology 455, 147-155, doi:10.1007/s00424-007-0272-9 (2007). 53. Dvir, M., Peretz, A., Haitin, Y. & Attali, B. Recent molecular insights from mutated IKS channels in cardiac arrhythmia. Current opinion in pharmacology 15, 74-82, doi:10.1016/j. coph.2013.12.004 (2014). 54. Woodcock, E. A., Kistler, P. M. & Ju, Y. K. Phosphoinositide signalling and cardiac arrhythmias. Cardiovascular research 82, 286-295, doi:10.1093/cvr/cvn283 (2009). 55. Park, K. H. et al. Impaired KCNQ1-KCNE1 and phosphatidylinositol-4,5-bisphosphate in- teraction underlies the long QT syndrome. Circulation research 96, 730-739, doi:10.1161/01. RES.0000161451.04649.a8 (2005). 56. Matavel, A., Medei, E. & Lopes, C. M. PKA and PKC partially rescue long QT type 1 phenotype by restoring channel-PIP2 interactions. Channels (Austin, Tex.) 4, 3-11 (2010). 57. Ai, T. et al. Novel KCNJ2 mutation in familial periodic paralysis with ventricular dysrhyth- mia. Circulation 105, 2592-2594 (2002). 58. Holm, H. et al. Several common variants modulate heart rate, PR interval and QRS duration. Nature genetics 42, 117-122, doi:10.1038/ng.511 (2010). 59. Pfeufer, A. et al. Genome-wide association study of PR interval. Nature genetics 42, 153-159, doi:10.1038/ng.517 (2010). 60. Sotoodehnia, N. et al. Common variants in 22 loci are associated with QRS duration and cardiac ventricular conduction. Nature genetics 42, 1068-1076, doi:10.1038/ng.716 (2010). 61. Andreasen, L. et al. Brugada syndrome risk loci seem protective against atrial fibrillation. European journal of human genetics: EJHG, doi:10.1038/ejhg.2014.46 (2014). 62. Ritchie, M. D. et al. Genome- and phenome-wide analyses of cardiac conduction identifies markers of arrhythmia risk. Circulation 127, 1377-1385, doi:10.1161/circulationaha.112.000604 (2013). 63. Jabbari, J. et al. Common and Rare Variants in SCN10A Modulate the Risk of Atrial Fibrilla- tion. Circulation. Cardiovascular genetics, doi:10.1161/circgenetics.113.000442 (2014). 64. Ilkhanoff, L. et al. A Common SCN5A Variant Is Associated with PR Interval and Atrial Fibrillation Among African Americans. Journal of cardiovascular electrophysiology, doi: 10.1111/jce.12483 (2014). 65. van den Boogaard, M. et al. Genetic variation in T-box binding element functionally affects SCN5A/SCN10A enhancer. The Journal of clinical investigation 122, 2519-2530, doi:10.1172/ jci62613 (2012). 66. van den Boogaard, M. et al. A common genetic variant within SCN10A modulates cardiac SCN5A expression. The Journal of clinical investigation 124, 1844-1852, doi:10.1172/jci73140 (2014). 67. Moskovitz, J. et al. Methionine sulfoxide reductase (MsrA) is a regulator of antioxidant defense and lifespan in mammals. Proceedings of the National Academy of Sciences of the United States of America 98, 12920-12925, doi:10.1073/pnas.231472998 (2001). 68. Prentice, H. M. et al. MsrA protects cardiac myocytes against hypoxia/reoxygenation induced cell death. Biochemical and biophysical research communications 366, 775-778, doi: 10.1016/j.bbrc.2007.12.043 (2008). 69. Zhao, H. et al. Myristoylated methionine sulfoxide reductase A protects the heart from ischemia-reperfusion injury. American journal of physiology. Heart and circulatory physiology 301, H1513-1518, doi:10.1152/ajpheart.00441.2011 (2011). Multiple genomic approaches to identify genetic susceptibility to VF in MI 109

70. Kathiresan, S. et al. Common variants at 30 loci contribute to polygenic dyslipidemia. Nature genetics 41, 56-65, doi:10.1038/ng.291 (2009). 71. Weissglas-Volkov, D. et al. Investigation of variants identified in caucasian genome-wide association studies for plasma high-density lipoprotein cholesterol and triglycerides levels in Mexican dyslipidemic study samples. Circulation. Cardiovascular genetics 3, 31-38, doi: 10.1161/circgenetics.109.908004 (2010). 72. Harley, J. B. et al. Genome-wide association scan in women with systemic lupus erythematosus identifies susceptibility variants in ITGAM, PXK, KIAA1542 and other loci. Nature genetics 40, 204-210, doi:10.1038/ng.81 (2008). 73. Budarf, M. L. et al. A targeted association study in systemic lupus erythematosus identifies multiple susceptibility alleles. Genes and immunity 12, 51-58, doi:10.1038/gene.2010.47 (2011). 74. Alonso-Perez, E. et al. Further evidence of subphenotype association with systemic lupus erythematosus susceptibility loci: a European cases only study. PloS one 7, e45356, doi: 10.1371/journal.pone.0045356 (2012). 75. Jarvinen, T. M. et al. Replication of GWAS-identified systemic lupus erythematosus susceptibility genes affirms B-cell receptor pathway signalling and strengthens the role of IRF5 in disease susceptibility in a Northern European population. Rheumatology (Oxford, England) 51, 87-92, doi:10.1093/rheumatology/ker263 (2012). 76. Dingledine, R., Borges, K., Bowie, D. & Traynelis, S. F. The glutamate receptor ion channels. Pharmacological reviews 51, 7-61 (1999). 77. Stern, J. E. & Potapenko, E. S. Enhanced NMDA receptor-mediated intracellular calcium signaling in magnocellular neurosecretory neurons in heart failure rats. American journal of physiology. Regulatory, integrative and comparative physiology 305, R414-422, doi:10.1152/

ajpregu.00160.2013 (2013). Chapter 5 78. Shi, S. et al. Chronic N-Methyl-d-Aspartate Receptor Activation Induces Cardiac Electrical Remodeling and Increases Susceptibility to Ventricular Arrhythmias. Pacing and clinical electrophysiology: PACE, doi:10.1111/pace.12430 (2014). 79. Sun, X. et al. Increasing glutamate promotes ischemia-reperfusion-induced ventricular arrhythmias in rats in vivo. Pharmacology 93, 4-9, doi:10.1159/000356311 (2014). 80. Lu, J. et al. Nerve sprouting contributes to increased severity of ventricular tachyarrhythmias by upregulating iGluRs in rats with healed myocardial necrotic injury. Journal of molecular neuroscience: MN 48, 448-455, doi:10.1007/s12031-012-9720-x (2012). 81. Liu, Y. et al. A preliminary experimental study on the cardiac toxicity of glutamate and the role of alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor in rats. Chinese medical journal 126, 1323-1332 (2013). 82. Yang, Y. et al. Cardiac myocyte Z-line calmodulin is mainly RyR2-bound, and reduction is arrhythmogenic and occurs in heart failure. Circulation research 114, 295-306, doi:10.1161/ circresaha.114.302857 (2014). 83. Xu, X. et al. Defective calmodulin binding to the cardiac ryanodine receptor plays a key role in CPVT-associated channel dysfunction. Biochemical and biophysical research communications 394, 660-666, doi:10.1016/j.bbrc.2010.03.046 (2010). 84. Anderson, M. E., Braun, A. P., Schulman, H. & Premack, B. A. Multifunctional Ca2+/ calmodulin-dependent protein kinase mediates Ca(2+)-induced enhancement of the L- type Ca2+ current in rabbit ventricular myocytes. Circulation research 75, 854-861 (1994). 110 Chapter 5

85. Bai, C. X. et al. Role of nitric oxide in Ca2+ sensitivity of the slowly activating delayed rectifier K+ current in cardiac myocytes. Circulation research 96, 64-72, doi:10.1161/01. res.0000151846.19788.e0 (2005). 86. Kim, J. et al. Calmodulin mediates Ca2+ sensitivity of sodium channels. The Journal of biological chemistry 279, 45004-45012, doi:10.1074/jbc.M407286200 (2004). 87. Wagner, S. & Maier, L. S. Modulation of cardiac Na(+) and Ca(2+) currents by CaM and CaMKII. Journal of cardiovascular electrophysiology 17 Suppl 1, S26-s33, doi:10.1111/j.1540- 8167.2006.00382.x (2006). 88. Toyama, J. et al. Ca(2+)-calmodulin mediated modulation of the electrical coupling of ventricular myocytes isolated from guinea pig heart. Journal of molecular and cellular cardiology 26, 1007-1015, doi:10.1006/jmcc.1994.1121 (1994). 89. Marsman, R. F. et al. A mutation in CALM1 encoding calmodulin in familial idiopathic ventricular fibrillation in childhood and adolescence. Journal of the American College of Cardiology 63, 259-266, doi:10.1016/j.jacc.2013.07.091 (2014). 90. Limpitikul, W. B. et al. Calmodulin mutations associated with long QT syndrome prevent inactivation of cardiac L-type Ca(2+) currents and promote proarrhythmic behavior in ventricular myocytes. Journal of molecular and cellular cardiology 74, 115-124, doi:10.1016/j. yjmcc.2014.04.022 (2014). 91. Yuan, W. & Bers, D. M. Ca-dependent facilitation of cardiac Ca current is due to Ca-calmodulin-dependent protein kinase. The American journal of physiology 267, H982-993 (1994). 92. Lonze, B. E. & Ginty, D. D. Function and regulation of CREB family transcription factors in the nervous system. Neuron 35, 605-623 (2002). 93. Kreusser, M. M. & Backs, J. Integrated mechanisms of CaMKII-dependent ventricular remodeling. Frontiers in pharmacology 5, 36, doi:10.3389/fphar.2014.00036 (2014). 94. Kobrinsky, E., Duong, S. Q., Sheydina, A. & Soldatov, N. M. Microdomain organization and frequency-dependence of CREB-dependent transcriptional signaling in heart cells. FASEB journal: official publication of the Federation of American Societies for Experimental Biology 25, 1544-1555, doi:10.1096/fj.10-176198 (2011). 95. Wu, X. & McMurray, C. T. Calmodulin kinase II attenuation of gene transcription by pre- venting cAMP response element-binding protein (CREB) dimerization and binding of the CREB-binding protein. The Journal of biological chemistry 276, 1735-1741, doi:10.1074/jbc. M006727200 (2001). 96. Xie, Y., Ding, W. G. & Matsuura, H. Ca/calmodulin potentiates I in sinoatrial node cells by activating Ca /calmodulin-dependent protein kinase II. Pflugers Archiv: European journal of physiology, doi:10.1007/s00424-014-1507-1 (2014). 97. Bourgonje, V. J. et al. Relevance of calmodulin/CaMKII activation for arrhythmogenesis in the AV block dog. Heart rhythm: the official journal of the Heart Rhythm Society 9, 1875- 1883, doi:10.1016/j.hrthm.2012.07.023 (2012). 98. Yang, L. Z. et al. cAMP- and Ca(2)(+) /calmodulin-dependent protein kinases mediate ino- tropic, lusitropic and arrhythmogenic effects of urocortin 2 in mouse ventricular myocytes. British journal of pharmacology 162, 544-556, doi:10.1111/j.1476-5381.2010.01067.x (2011). 99. Bell, J. R., Curl, C. L., Ip, W. T. & Delbridge, L. M. Ca2+/calmodulin-dependent protein kinase inhibition suppresses post-ischemic arrhythmogenesis and mediates sinus bradycardic recovery in reperfusion. International journal of cardiology 159, 112-118, doi:10.1016/j. ijcard.2011.02.038 (2012). Multiple genomic approaches to identify genetic susceptibility to VF in MI 111

100. Schmid, R. S. et al. NCAM stimulates the Ras-MAPK pathway and CREB phosphorylation in neuronal cells. Journal of neurobiology 38, 542-558 (1999). 101. LaMorte, V. J. et al. Gq- and ras-dependent pathways mediate hypertrophy of neonatal rat ventricular myocytes following alpha 1-adrenergic stimulation. The Journal of biological chemistry 269, 13490-13496 (1994). 102. Thorburn, J. & Thorburn, A. The tyrosine kinase inhibitor, genistein, prevents alpha- adrenergic-induced cardiac muscle cell hypertrophy by inhibiting activation of the Ras- MAP kinase signaling pathway. Biochemical and biophysical research communications 202, 1586-1591, doi:10.1006/bbrc.1994.2113 (1994). 103. Thorburn, A. Ras activity is required for phenylephrine-induced activation of mitogen- activated protein kinase in cardiac muscle cells. Biochemical and biophysical research communications 205, 1417-1422, doi:10.1006/bbrc.1994.2823 (1994). 104. Pando, R. et al. Ras inhibition attenuates myocardial ischemia-reperfusion injury. Bio- chemical pharmacology 77, 1593-1601, doi:10.1016/j.bcp.2009.02.010 (2009). 105. Al-Rashdan, I. et al. Cardioprotection from ischemia-reperfusion injury due to Ras-GTPase inhibition is attenuated by glibenclamide in the globally ischemic heart. Cell biochemistry and function 25, 455-461, doi:10.1002/cbf.1353 (2007). 106. Hines, W. A. & Thorburn, A. Ras and rho are required for galphaq-induced hypertrophic gene expression in neonatal rat cardiac myocytes. Journal of molecular and cellular cardiology 30, 485-494, doi:10.1006/jmcc.1997.0613 (1998). Chapter 5