Selected Abstracts from Approved DNA Committee Applications
Risk Factors for Age Related Bone Loss
PI: Drs. Douglas Kiel and David Karasik
The Framingham Osteoporosis Study, an ancillary study of the Framingham Heart Study, has contributed substantially to the understanding of risk factors for age-related bone loss and fractures in men and women. For the past fifteen years, this research program has been investigating a variety of risk factors for bone loss and fractures by assessing bone mineral density (BMD) using dual energy x-ray absorptiometry (DXA), and by ascertaining fracture incidence in the Framingham Study Original and Offspring Cohorts. This is an addendum to a previously approved application examining SNPs in candidate genes related to bone mineral density. The investigators wish to increase their sample size from 3300 to 5965 individuals and expand the list of SNPs to be genotyped. The additional individuals are participants with newly obtained bone mineral density data.
While the amount of DNA requested is within the stated limits, it appears to be more than is needed for the proposed Illumina assays. The investigators should clarify with the FGL whether custom DNA plates are really needed, or if standard plates would provide all the samples needed. The investigators list 425 SNPs in the application, but they should send the final list of 384 SNP once it is available.
Type 2 Diabetes Phenotype and Genotype Risk Prediction Models in the Framingham Heart Study
PI: Dr. James Meigs
In this addendum we propose to add nine (9) additional 'proven' diabetes risk SNPs to the 17 proven diabetes risk SNPs that were previously approved for genotyping in order to develop a state-of-the-art phenotypic / genotypic type 2 diabetes risk prediction model. The investigators propose to test 3 aims:1) that risk of incident DM increases in proportion to # of DM risk alleles present, 2) That genetic information increases risk for DM independent of Phenotype Risk Score and 3) Genetic information will reclassify risk status for modest % of people assigned to low, intermediate & high risk. Outstanding application, by an experienced, well published investigative team, addressing a critical question regarding the extent to which genetic risk markers markedly advance the objective of risk personalized medicine by enhancing risk prediction and reclassification.
Genetics of Atrial Fibrillation: Fine mapping the chromosome 4q25 locus for AF and the role of ion channel variants in AF
PI: Drs. Emelia Benjamin and Patrick Ellinor
Although there have been rare reports of familial forms of AF, traditionally AF has not been viewed as a genetic condition. However, in the last five years our understanding of this arrhythmia has evolved and it has become increasingly clear that AF is heritable. Mutations in six different ion channels have been described to cause AF yet the role of these variants remains unclear. A genome wide association study for AF has identified a locus on Chromosome 4q25 and these results have recently been replicated in a large number of cohorts with AF. We propose to 1) fine map this locus for AF in subjects from the MGH and FHS in an attempt to determine if other SNPs in this region are more strongly associated with AF and 2) resequence all of the subjects from FHS to determine the frequency of variants in those with and without AF. This research is significant and well justified. This is a very strong application, except for the lack of discussion regarding analyses of biologically related subjects. However, these investigators are knowledgeable in such methods and will presumably perform appropriate analyses.
A Feasibility Study of Gene Expression and MicroRNA Profiling from 3 Sources of WBC-Derived RNA
PI: Dr. Daniel Levy
The purpose of this DNA Application is to request biological specimens to determine feasibility of RNA expression profiling with a specific goal of assessing the quality of RNA and the acceptability of expression profile data in small numbers of samples from Framingham Offspring and Third Generation participants. Samples from only 50 individuals from each cohort will be studied comparing 3 sources of WBC-derived RNA per individual: a) buffy coats, b) PAXgene tubes, and c) immortalized cell line pellets (non-viable). The technical feasibility study will also be used to determine if existing RNA samples for the Third Generation cohort are suitable for processing on the Affymetrix gene expression platform or if new methods of collection (e.g. PAXgene tubes, or freshly extracted RNA from PBMCs) are needed at the next Third Generation cohort exam (to begin in 2008).
Systems Approach to Biomarker Research in Cardiovascular Disease (SABRe CVD Initiative), Project 2: Gene Expression Signatures of Atherosclerosis and Metabolic Syndrome Risk Factors
PI: Dr. Daniel Levy
The Systems Approach to Biomarker Research in Cardiovascular Disease (SABRe CVD) initiative is a research program to advance personalized medicine through biomarker discovery and validation. The central objective of SABRe CVD is to identify biochemical signatures of disease. The Framingham Heart Study resource can facilitate the identification of a wide variety of new biomarkers, providing significant public health benefits. To that end, the SABRe CVD initiative will generate extensive biomarker data from 7000 study participants using multiple high throughput platforms including immunoassays, proteomics, metabolomics/lipomics, and gene expression and microRNA profiling. Project 2, the subject of this DNA Committee application will generate a set of gene expression signatures from RNA in up to 7000 FHS participants from the Offspring and Third Generation cohorts. The objectives are to characterize the gene expression signatures of disease with a focus on atherosclerosis and metabolic syndrome. We also aim to use individual gene expression signatures to identify common functional genomic polymorphisms (eSNPs) and to identify genomic convergence signals of disease by identifying the intersection of disease/phenotype associations from GWAS and gene expression profiling. We will submit the entire gene expression database into dbGaP (SHARe) to make this resource available to the scientific community.
Genetics of Water Balance
PI: Dr. David Cohen
In the original application, we proposed to perform a genome-wide association study to identify polymorphisms associated with aberrant water balance. Although we have obtained provocative preliminary data through this approach, the 100K Affymetrix data set (and the 500K follow-up data set) offers poor single-nucleotide polymorphism coverage of several candidate genes of greatest interest. Therefore, we propose to augment our analysis with a candidate gene-based association analysis. For this we request access to banked genomic DNA from the Framingham subjects.
Estimation of the Rate and Spectrum of Spontaneous Mutations in Humans
Pl: Dr. Diddahally Govindaraju
The goal of this study is to develop a highly reliable data base on the rate of origin of mutations in the human genome and the spectrum of their molecular effects, free from the biases inherent in previous indirect studies. The ideal approach to identifying mutations in an unbiased manner is to perform a mutation-accumulation experiment; i.e., to maintain a series of initially identical inbred lines through multiple generations, followed by the mutational screening of random segments of the genome by direct sequencing. Although such an experiment is out of the question for humans, a well-established set of multi-generational pedigrees is formally equivalent in nature. In terms of time span and breadth of population coverage, the Framingham Heart Study clearly provides the best opportunity to estimate the basic mutational parameters for the human species. We propose to exploit the materials offered by this unique study to perform high-density sequencing to identify mutations that have accumulated over three generations within independent pedigrees. Nothing will be changed from the original application. All aspects of the project will remain the same. But, the sequencing of the mitochondrial genome will be done in Dr. Church's laboratory, instead of Dr. Lynch's laboratory.
Validation of Polymorphic and Rare Gene Sequence Variants Contributing to Extremes of Blood Pressure in the Framingham Heart Study: Request to use Third Party Genotyping Service
PI: Dr. Christopher Newton-Cheh
Despite extensive efforts by numerous investigators, the identification of genetic variants that alter blood pressure levels in the general population has proved to be an elusive challenge. This project aims to examine the prevalence of rare variants in several key blood pressure regulatory pathway genes in Framingham Heart Study individuals with extreme blood pressure phenotypes. DNA from 144 Framingham participants with extreme high and 138 participants with extreme low blood pressure values will be extensively re-sequenced for missense and nonsense variants. These individuals represent approximately the top and bottom 2% of blood pressure values. Re-sequencing will be performed by the NHLBI Re-sequencing and Genotyping Service, which offers extensive gene coverage. Chi square tests will be used to compare the prevalence of coding variants between the two extreme phenotype groups. We will genotype all polymorphic variants with an asymmetric distribution between the high and the low diastolic blood pressure groups. In the current proposal we propose to use a third-party genotyping service, Illumina, to genotype the 275 variants.
A Feasibility Study of Gene Expression and Micro-RNA Profiling from 3 Sources of WBC-Derived RNA
PI: Dr. Daniel Levy
The purpose of this DNA Application is to request biological specimens for determining feasibility of RNA expression with a specific goal of assessing the quality of RNA and the acceptability of expression profiles in small numbers of samples from Framingham Third Generation participants. Samples will be profiled from 50 individuals comparing 4 sources of WBC-derived RNA per individual: a) buffy coats, b) PAXgene tubes, c) immortalized cell lines, and d) PBMCs. Results from this pilot study will be used to guide decisions regarding the optimal RNA source to be used in future gene expression profiling initiatives. Identification of robust, reliable phenotypes, relatively constant across sample types and over time, would be an important step in identification of predictive expression profiles.
Type 2 Diabetes Phenotype and Genotype Risk Prediction Models in the Framingham Heart Study
PI: Dr. James Meigs
Type 2 diabetes is prevalent, predictable, heritable and preventable. We recently developed a FHS Diabetes 'Phenotype' Risk Score with an aROC of 0.85 based on age, sex, family history of diabetes, body mass index, and levels of glucose, insulin, HbA1c, triglycerides, HDL-C and blood pressure. Others have begun to develop 'genetic risk scores' based on the presence of one or more replicated diabetes risk alleles (for instance, TCF7L2, KCNJ11, PPARG). We propose to develop and test a risk prediction model based on a Phenotype Risk Score and a Genetic Risk Score (to be developed) to test three primary hypotheses: 1) risk of incident diabetes increases in proportion to the number of risk alleles present; 2) genetic information increases risk for diabetes independent of Phenotype Risk Score, and increases the model discriminatory capacity beyond that of the Phenotype Risk Score; and 3) genetic information reclassifies risk status for a modest percent of people assigned as low risk by their phenotype. We specify a total of 36 SNPs in 30 genes or chromosomal regions to be tested for the Genetic Risk Score. The study will test at the population level the impact of phenotypes and genotypes on risk for type 2 diabetes.
Improving Inference via Statistical Re-sampling and the Random Forest in Linkage Analysis of Quantitative Trait Loci
PI: Dr. Shelley Bull
This application is a resubmission of approved addendum project 2005.07.01 as the original application was due to expire. The purpose of the research is to develop and evaluate statistical methods designed to reduce the bias in effect size estimates and improve inference obtained in genome-wide linkage scans using re-sampling and random forest techniques. We have implemented these computational methods for the genetic linkage analysis setting, and have conducted simulation studies that suggest these methods can substantially reduce selection bias in effect estimates and improve the detection of multiple linkage regions. To complement the simulation studies, we are applying the re-sampling and random forest methods to a quantitative blood pressure phenotype derived in the GAW13 Framingham Heart Study dataset that includes pedigrees of varying size and structure.
Metabolic Trait Levels and Common Variation in Endocannabinoid Pathway Genes
PI: Dr. James Meigs
Inhibition of the endocannabinoid (EC) receptor CB1 is associated with weight loss, improved insulin sensitivity and glucose and lipid levels, and reduced carotid artery atherosclerosis. We hypothesize that common variants in two genes with major endocannabinoid regulatory roles, CNR1 (encodes CB1) and FAAH (a major EC regulatory enzyme), are associated with variation in metabolic trait levels and risk of type 2 diabetes or CVD. We tested this hypothesis using 21 tag SNPs typed in ~2,400 Framingham Offspring subjects. We found association with several SNPs in regulatory regions around CNR1. We propose to fine map around these high-interest SNPs using information derived from new mapping of dbSNPs in the HapMap CEU samples, select new tag SNPs to capture the majority of variation at a minor allele frequency >0.5%, genotype these novel tag SNPs in the Offspring Unrelated and Family plates, and test this high-resolution SNP map for association with diabetes, CVD, and related traits. Preliminary data suggest a possible role of the EC system in glucose homeostasis and insulin resistance. The proposed studies extend these findings to identify rare or causative variants, to inform functional gene expression studies, and to inform follow-on replication studies in additional independent samples.
Adrenergic Genetic Determinants of Hypertension
PI: Dr. Daniel O'Connor
This proposal represents an opportunity to define the genetic basis of common variations in catecholamine biosynthesis for their mechanistic consequences, their role in risk for development of hypertension. Human hypertension is in part genetically determined, but how individual genes contribute to the disease is not well understood. The sympathetic nervous system has been implicated as sympathetic activity is elevated in both of these states. Catecholamines are synthesized by an intricate cascade of steps which include enzymes for synthesis, storage, and metabolism. Here we focus on common genetic variation in the pre-synaptic catecholaminergic pathway as a cause of inter-individual variation in sympathetic activity and blood pressure, which can then contribute to hypertension. Our proposal identifies such adrenergic candidate loci and then utilizes tagging single nucleotide polymorphisms in linkage disequilibrium blocks. We then resequence the locus in systematic polymorphism discovery to isolate the precise genetic differences that may contribute to the phenotypic differences, and also confirm in vitro functional activity of the loci. The relevance is to define the role of catecholamine biosynthetic polymorphisms in hypertensive states and aid in understanding of these contributions to the pathophysiology of this complex disease trait, which may lead to early identification and treatment.
This proposal was postponed until the investigator responded to several reviewers' questions. The investigator satisfactorily proveded additional information and the proposal was approved.
Relationship of loci identified by genome-wide association studies with endothelial progenitor cell populations via the Framingham Heart Study
PI: Dr. Anthony Rosenzweig
Several recent studies have identified a locus on the 9p21 chromosome that is associated with coronary artery disease independently of the usual cardiac risk factors of dyslipidemia and inflammation. The locus lies in a 100kb stretch upstream of the CDKN2A and CKDN2B genes, which encode p16Ink4a, ARF and p15Ink4b proteins, cyclin-dependent kinase inhibitors known to play critical roles in regulating cell proliferation, senescence and apoptosis in a variety of cell types. Endothelial progenitor cells (EPC) represent a potential mechanism for vascular maintenance and repair, and may therefore modulate atherosclerotic disease. We hypothesize that variation at the 9p21 locus is associated with EPC number and function.
We therefore propose to examine the heritability of EPC phenotypes, as well as the relationship between a validated SNP on 9p21 (rs10757278) and the number and function of endothelial progenitor cells in the Framingham Offspring 8th examination cohort. DNA from this cohort will be genotyped for the rs10757278 SNP on the Sequenom hME platform. EPCs have been enumerated in this cohort using flow cytometry analysis and in vitro colony-forming unit assays.
Association of Functionally Enriched Variants with Cardiovascular Disease Phenotypes
PI: Dr. Daniel Levy
We propose to genotype a subset of the 2.5 million HapMap SNPs that is enriched with functional variants and thus may be more likely to be associated with phenotypic variation than the remainder of SNPs. Three types of putative functional SNPs will be studied: a) non-synonymous coding SNPs (nSNPs); b) SNPs associated with altered tissue specific gene expression (eSNPs); and c) SNPs that identify copy number variation (cSNPs). For all nSNPs, eSNPs and cSNPs that we have captured from existing databases, we will examine associations with multiple phenotypes, including traditional risk factors and selected biomarkers, subclinical vascular and ventricular disease measures, and clinically apparent CVD outcomes, and we will compare results with those of using the entire 2.5 million SNPs imputed from HapMap and now being analyzed in conjunction with the Framingham Heart Study SNP Health Association Resource (SHARe). In this way, we will compare the unbiased GWAS approach available via SHARe and dbGaP with a biased, functionally enriched set of SNPs.
Analysis of Two SNPs in the TRPV5 Gene and their Association with Bone Mineral Density
PI: Dr. Andreas Pasch
The Transient Receptor Potential Cation Channel, Subfamily V, Member 5 (TRPV5) is a main regulator of calcium excretion by the kidneys and is also expressed in osteoclasts. TRPV5 knock out-mice are osteopenic and hypercalciuric. In humans, renal calcium leak hypercalciuria (with a probable contributory role of TRPV5) has been associated with low bone mineral density. We hypothesize that SNPs in TRPV5 might be related to changed (increased or reduced) bone mineral density. Therefore, we aim to study in the Framingham Osteoporosis Study cohort two abundant polymorphisms in the TRPV5 gene with regard to bone mineral density and hip geometry (DXA). By comparing genotype-phenotype data from two cohorts (Framingham osteoporosis study population and Bern stone former cohort) and by combining these data with functional in vitro analyses (expression of the SNPs in frog oocytes), we aim to identify and functionally substantiate an association between bone mineral density and these polymorphisms.
Develop a Likelihood-based Trait-Model-Free Approach for Linkage Detection of Binary Trait
PI: Dr. Sanoli Basu
Trait-model-free (or 'allele-sharing') approaches to linkage analysis are a popular tool in genetic mapping of complex traits, due to the absence of explicit assumptions about the underlying mode of inheritance of the trait. The likelihood framework introduced by Kong and Cox (1997) allows calculation of accurate p-values and LOD scores to test for linkage between a genomic region and a trait. Their method relies on the specification of a model for the trait-dependent segregation of marker alleles at a genomic region linked to the trait. Here we propose a new such model that is motivated by the desire to extract as much information as possible from extended pedigrees containing data from individuals related over several generations. However, our model is also applicable to smaller pedigrees, and has some attractive features compared with existing models, including the fact that it does not require specification of an appropriate identity by descent (IBD) measure, or pedigree weights.
Follow-up of Monocyte Chemoattractant Protein-1 Genome-wide Association Study Findings
PI: Dr. Emelia Benjamin
MCP1 is associated with cardiovascular risk factors including advancing age, cigarette smoking, triglycerides, body mass index, waist-to-hip, hypertension, and diabetes. Increasing MCP1 concentrations have been associated with prevalent subclinical (IMT) and cardiovascular disease (myocardial infarctions), as well as recurrent events.
The genetic determinants of MCP1 concentrations are not fully understood. MCP1 concentrations are heritable, and have a linkage peak on chromosome 1. In our 550K preliminary findings MCP1 concentrations have a high linkage peak (10-16) and achieve genome-wide significance in a genomic region on chromosome 1 near a genomic region rich in genes coding for chemokines. We had two SNPs that achieved genome-wide significance for SNPs in the DARC genomic region with an additive genetic model LME analysis: rs3027012 had a p=1.36x10-76 (FBAT p=1.53 x10-18)and rs863017 had a p=1.76 x10-67 (8.85 x10-22) in a least squares mean genetic analysis; FBAT p-values for the same SNPs were. We seek to conduct additional genotyping in the DARC region to better understand SNPs associated with MCP1 concentrations.
Validation and Finemapping of Genome-wide Significant Loci Identified in GWAS Meta-analysis of QT Interval from 3 Studies
PI: Dr. Christopher Newtom-Cheh
Electrocardiographic QT interval prolongation is associated with increased risk of sudden cardiac death.1 We have been investigating the genetic determinants of QT interval duration, a heritable intermediate trait,2 in the Framingham Heart Study and other cohorts in the hopes of identifying genetic variants that may influence risk of ventricular arrhythmias in the general population or upon exposure to QT-prolonging medications. We have completed candidate gene studies3-5 and genome-wide association studies6, 7 to identify novel variants and novel genes that influence myocardial repolarization in the general population. In the QTGEN consortium, under the aegis of the CHARGE consortium, we have completed meta-analysis of three QT interval genome-wide association studies in the Framingham Heart Study, the Rotterdam Study and the Cardiovascular Health Study, with a total sample size of 13,109 individuals of European ancestry. We used the fixed genotyping arrays available in each study to impute the genotypes of 2.5 million SNPs with linkage disequilibrium information available in HapMap using the MACH software as described in Saxena et al (Science 2007).8 We observed association of common variants in genes known to be involved in myocardial repolarization including NOS1AP (p=2x10-45) and Long QT Syndrome genes KCNE1 (p=2x10-8), SCN5A (p=8x10-8) and KCNH2 (p=8x10-9), with 2 signals in KCNQ1 (p=3x10-16, p=6x10-11). In addition, we have identified novel associations with four loci: on 16q21 (p=2x10-14), on 6q22 (p=4x10-11), on 1p36 (p=5x10-9), on 16p13 (p=2x10-8) and on 17q12 (p=1x10-7). Importantly, several of the associations identified are in imputed SNPs which have varying degrees of imputation quality. Imputation quality can be assessed using the observed-to-expected variance ratio (1 = perfect, 0=poorest quality) in several of the constituent cohorts. For example, SNP rs2074238 in intron 1 of KCNQ1 has a minor allele frequency of 0.05 but is associated with a 0.47 standard deviation reduction in QT interval duration (p=3x10-16) equal to the change in QT in response to drugs that cause lethal arrhythmias). However, the SNP is poorly imputed with variance ratios of 0.10 and 0.45 in 2 cohorts and could not be imputed in the third cohort. In the current application, we request to directly genotype in the Framingham Family, Unrelated Offspring (Gen 2A) plates a total of 33 SNPs at genome-wide significant loci only that will allow us to technically validate and partially finemap (typing imputed SNPs with r2>0.3 and low p-values) the 12 genome-wide significant SNPs at these 10 loci (p=10-7). We have in preparation a manuscript that we intend to submit to Nature Genetics in June 2008 and seek expedited review of our application. We will complete genotyping at the Broad Institute where we have existing approved DNA Committee applications (2004-08, 2004-18, 2006-12.4). No new DNA is requested. All analyses will occur at Framingham using existing access to the QT interval data. This application is solely to request expedited review for permission to genotype in these existing samples so that data can be included in the upcoming Nature Genetics submission.
Follow-up and replication of a chromosome 11 region linked and associated with body mass index in Framingham
PI: Dr. Larry Atwood
In 2002 we published strong evidence for linkage to regions on chromosome 6, 10, and 11 (Atwood et al., AJHG 71:1044) on 1114 Framingham participants in 330 families. All three regions had lodscores over 3.0 and were replicated in the literature. In 2004, NIDDK funded a grant to pursue these linked regions. An initial round of fine mapping showed the evidence for linkage was strongest on chromosome 11. We then used the Illumina custom genotyping service to genotype 3072 SNPs in this chromosome 11 region. After Mendelian cleaning and eliminating failed, monomorphic, and rare minor allele SNPs, we performed association analysis on 2668 SNPs in this region.
We have a strong prior hypothesis based on linkage studies (our own and multiple replications in the literature) that there are loci in this region associated with BMI. We do not need to apply a genome wide significance level. The major linkage peak and the major association peak occur at the same locus (124-126MB). There is also a cluster of association results at the minor linkage peak (119.7-120.2MB).
A literature search indicates the presence of at least three candidate genes for obesity in this region. The first, FEZ1, is directly under the major linkage-association peak. FEZ1 is associated with Prader-Willi syndrome which has obesity as a symptom. The second, ARHGEF12, has been associated with insulin action in PIMA Indians. The third, ASAM, is the adipocyte specific adhesion molecule, which, by it's very name, is an obvious candidate for obesity. All three of these genes are directly under association clusters.
The concordance of linkage, association, and bioinformatic evidence is strong, even compelling, evidence that there are multiple BMI variants in this region.
However, this evidence is only in the original 330 Framingham families. Replication is needed as soon as possible. We propose to expand this custom genotyping to all Framingham participants. When combined with the Affy 550K genotypes we will have sufficient power and coverage to replicate this result.
Follow-up of a SNP for Chronic Kidney Disease
PI: Dr. Caroline Fox
Chronic kidney disease (CKD) affects 19 million adults in the United States, and is associated with cardiovascular disease, stroke, peripheral arterial disease, and all-cause mortality. Genetic factors play a role in the progression of renal disease. In the Framingham Heart Study, we have shown that kidney function is heritable, suggesting a role for genetic mechanisms in its etiology.
To date, there are very few known genes associated with CKD. Using genome-wide association, we hope to uncover novel loci for CKD. To do this, we have formed a collaboration with the ARIC, Rotterdam, and CHS studies, resulting in 22,000 individuals and over 3,000 cases of CKD. We have found evidence for genome-wide association in the UMOD gene (rs4293393, p-value=1.1x10-8). This SNP is just upstream from the UMOD gene, which encodes uromodulin, one of the most abdundant proteins to be found in urine. Defects in this gene are associated with hyperuricemia, gout, and renal failure in children, making this a very compelling and exciting candidate gene for CKD.
The results of this SNP show robust intra-study results in all but the Framingham Heart Study (p=0.70). When we have investigated the results more closely, it appears that the one SNP on the 550K Affymetrix chip in this region was not well genotyped, as it contains several Mendelian inconsistencies. Unfortunately, this is the only genotyped SNP in this genic region, and all other SNPs are imputed based on this SNP. Therefore, in this expedited DNA proposal, we wish to obtain approval to genotype the top 4 SNPs in this region, based on our results in our consortium.
Follow-Up of Meta-Analysis of Four Genome-wide Association Studies of Incident Stroke and Ischemic Stroke in the CHARGE Consortium
PI: Dr. Sudha Seshadri
We sought to identify genetic variants underlying stroke by performing a prospective meta-analysis of genome-wide association data in white subjects from four large cohort studies (the Framingham Heart Study, the Atherosclerosis Risk in Communities Study, the Cardiovascular Health Study, the Rotterdam Study) comprising the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) consortium, and replicated our finding in an African-American cohort, part of the Atherosclerosis Risk in Communities Study. Each discovery cohort used genotype information to impute to HapMap's CEU panel and age- and sex-adjusted Cox models to relate 2.2 million SNPs to incident stroke. Study-specific findings were combined in a fixed-effects meta-analysis including 19,602 stroke-free individuals (mean age 63) who developed 1,544 initial incident strokes (1,164 ischemic) over an average follow-up of 11 stroke (191 ischemic) over 15 years. Two intergenic SNPs at one locus were significantly associated with incident total and ischemic stroke (p<5x10-8): rs11833579 and rs12425791, at chromosome 12p13 within 11kb of the gene NINJ2 encoding ninjurin2, an adhesion molecule expressed in adult glia and induced by injury, and 90kb of the gene lysine-deficient protein kinase 1 (WNK1). For rs12425791, the risk per allele in the discovery cohort was 31% (95% CI: 19-44%) higher for total stroke, and 39% (95% CI: 27-46%) higher for ischemic stroke yielding population attributable risks of 11 and 14%, respectively. In African-Americans, corresponding risks were 39% (95%CI: 5-84%) and 43% (95%CI: 6-92%). This finding has not been previously reported. Thus, our community-based GWAS uncovered a novel association for incident stroke. Whereas rs12425791 was genotyped in the Framingham Heart Study (FHS) sample using the Affymetrix GeneChip® Human Mapping 500K Array Set and 50K Human Gene Focused Panel® rs11833579 which showed the strongest association was with stroke was not genotyped but was imputed in the FHS sample using the Markov Chain Haplotyping (MaCH) package (http://www.sph.umich.edu/csg/abecasis/MACH) version 1.0.15 software (imputed to plus strand of NCBI build 36, HapMap release #22). The observed to expected (O/E) ratio was 0.943.
We had submitted these data for publication and received criticism from the reviewer regarding the uncertainty associated with imputation may have resulted in false positive results.
Despite the good quality imputation, technical validation of the imputed results by genotyping using a Taqman assay will help us to confirm this important finding and satisfy the reviewer. Also, as 'noise' due to incorrect genotype assignment in reduced we may find that the observed association between rs11833579 and the outcome measures of total and ischemic stroke becomes stronger. Genotyping will also permit us to explore the epidemiological, molecular and clinical correlates of genetic variation at this locus.
Genetics of atrial fibrillation: Fine mapping the chromosome 4q25 locus for AF and the role of ion channel variants in AF
Although there have been rare reports of familial forms of AF, traditionally AF has not been viewed as a genetic condition. However, in the last five years our understanding of this arrhythmia has evolved and it has become increasingly clear that AF is heritable. Mutations in six different ion channels have been described to cause AF yet the role of these variants remains unclear. A genome wide association study for AF has identified a locus on Chromosome 4q25 and these results have recently been replicated in a large number of cohorts with AF. We propose to 1) fine map this locus for AF in subjects from the MGH and FHS in an attempt to determine if other SNPs in this region are more strongly associated with AF and 2) re-sequence all of the subjects from FHS to determine the frequency of variants in those with and without AF.
Comparison of Markov chain Monte Carlo Methods for Linkage Analysis on General Pedigrees
PI: Shili Lin, PhD
Linkage analysis is most effective when extended pedigrees and multiple, linked markers are analyzed. However, calculation of exact multipoint likelihoods on large pedigrees with many markers remains computationally infeasible. The Markov chain Monte Carlo (MCMC) approach provides a means, by sampling from complex probability distributions, to perform multipoint linkage analyses on arbitrarily large pedigrees. In this study, we wish to provide practical guidelines for efficient use of several competing programs, and an understanding of their limitations. We propose to compare four MCMC programs-the Bayesian oligogenic approach implemented in Loki, the LOD-score programs lm_markers and lm_bayes from MORGAN, and the LOD-score method SimWalk2-through analyses of the GAW13 Problem 1 data comprising 330 extended pedigrees from the Framingham Study, phenotypic data on traits related to the metabolic syndrome, and a genomewide scan of microsatellite markers. An initial analysis on the original data will provide, in addition to linkage profiles, data on program run times and sensitivity of results to the specified mode of inheritance. We will subsequently compare the power of the methods, and their precision in localizing trait loci, using simulated trait and genotype data from pedigrees of the same structure as the GAW13 families.
Feasibility Study of Gene Expression and MicroRNA Profiling from 3 Sources of WBC-Derived RNA
PI: Daniel Levy, MD
The purpose of this DNA Application is to request biological specimens for determining feasibility of RNA expression with a specific goal of assessing the quality of RNA and the acceptability of expression profiles in small numbers of samples from Framingham Offspring participants. Samples will be profilied from 50 individuals comparing 3 existing sources of WBC-derived RNA per individual: a) buffy coats, b) PAXgene tubes, and c) immortalized cell lines. Results from this pilot study will be used to guide decisions regarding the optimal RNA source to be used in future gene expression profiling initiatives. Identification of robust, reliable phenotypes, relatively constant across sample types and over time, would be an important step in identification of predictive expression profiles.
Type 2 Diabetes Phenotype and Genotype Risk Prediction Models in the Framingham Heart Study
Type 2 diabetes is prevalent, predictable and heritable. We recently developed a FHS Diabetes 'Phenotype' Risk Score with an aROC of 0.85 based on age, sex, family history of diabetes, body mass index, and levels of glucose, insulin, HbA1c, triglycerides, HDL-C and blood pressure. Others have begun to develop 'genetic risk scores' based on the presence of one or more replicated diabetes risk alleles (for instance, TCF7L2, KCNJ11, PPARG). We propose to develop and test a risk prediction model based on a Phenotype Risk Score and a Genetic Risk Score (to be developed) to test three primary hypotheses: 1) risk of incident diabetes increases in proportion to the number of risk alleles present; 2) genetic information increases risk for diabetes independent of Phenotype Risk Score, and increases the model discriminatory capacity beyond that of the Phenotype Risk Score; and 3) genetic information reclassifies risk status for a modest percent of people assigned as low risk by their phenotype. We specify a total of 26 SNPs in 20 genes or chromosomal regions to be tested for the Genetic Risk Score. The study will test at the population level the impact of phenotypes and genotypes on risk for type 2 diabetes.