|Home | About | Journals | Submit | Contact Us | Français|
Left atrial enlargement is associated with cardiovascular disease. Genetic factors contributing to the left atrium (LA) dimension are poorly understood. We sought to map susceptibility genes for LA size in a large Dominican family dataset and an independent population-based cohort from the Northern Manhattan Study (NOMAS).
100 Dominican families consisting of 1350 individuals were used to estimate heritability and map quantitative trait loci for LA size using variance components analysis. LA dimension was measured by transthoracic echocardiography. A polygenic covariate screening was used to identify significant covariates. LA size had a moderate estimate of heritability (h2=0.42), after adjusting for significant covariates. Linkage analysis of 405 microsatellite markers revealed suggestive evidence on chromosome 10p19 (D10S1423, MLOD=2.00) and 17p10 (D17S974, MLOD=2.05). Ordered subset analysis found significantly enhanced (p<0.05 for increase of LOD score) evidence for linkage at 17p10 (MLOD=2.9) in families with lower LDL level. 2233 single nucleotide polymophisms (SNPs) were used to perform a peak-wide association mapping across 17p10 in 825 NOMAS individuals. Strong evidence for association were found in NTN1, MYH10, COX10, and MYOCD genes (p=0.00005 to 0.005).
Using non-biased genome-wide linkage followed by peak-wide association analysis, we identified several possible susceptibility genes affecting LA size. Among them, MYOCD has been shown to serve as a key transducer of hypertrophic signals in cardiomyocytes in vitro. Evidence from our linkage and association study, together with the known function, strongly suggests that polymorphisms in MYOCD gene modify LA size.
The size of the left atrium (LA) has important prognostic implications. LA enlargement has been linked to increased mortality in subsets of high-risk patients with left ventricular dysfunction, 1 or atrial arrhythmias2 but also in the general population.3, 4 An enlarged LA is also associated with development of atrial fibrillation, 5, 6 a condition that by itself increases the risk of ischemic stroke and death. 3, 4, 7. The association between increased risk of ischemic stroke and enlarged LA has been well documented, 8–10 even in subjects without atrial fibrillation. We have previously shown an increased stroke risk in subjects with increased LA size, whether measured directly by echocardiography 10 or inferred by electrical abnormalities detected on EKG.11
Multiple acquired conditions have been associated with LA enlargement.12 Among them are mitral valve disease, arterial hypertension and any condition that increases the left ventricular filling pressures. 12 Understanding the genetic influence on LA size would help in identifying subjects at increased risk for developing an enlarged LA especially at an early stage and in the absence of the hemodynamic predisposing conditions. In addition, this knowledge is essential for understanding cardiac structure and function at the molecular level and identifying therapeutic targets in the management of LA enlargement.
Little is known about the genetic basis of LA size with sparse studies mainly among Caucasian populations.13–17 Herein, we estimated the heritability and mapped quantitative trait loci (QTLs) for LA size in Caribbean Hispanic families from the Family Study of Stroke Risk and Carotid Atherosclerosis. To identify the genes underlying the QTLs, we performed a peak-wide association study in an independent prospective community-based cohort from the Northern Manhattan Study (NOMAS).
Probands in the family study were drawn from the Caribbean Hispanic participants in NOMAS. Previously, we have reported the detailed ascertainment scheme on NOMAS and the family study 18. Briefly, eligible NOMAS participants had never been diagnosed with a stroke, were at leaset 40 years of age, and resided for at least 3 months in a household with a telephone in Northern Manhattan. A total of 3298 subjects were enrolled between 1993 and 2001. In order to maximize the genetic component in the families, we assembled a high-risk Caribbean Hispanic family dataset using the following criteria to define a qualifying proband: (1) reporting a sibling with a history of myocardial infarction or stroke; or (2) having 2 of 3 quantitative risk phenotypes (maximal carotid plaque thickness, left ventricular mass, or homocysteine level above the 75th percentiles in the NOMAS cohort). Eighty percent of the families were recruited based on the first criteria. All subjects provided informed consent to participate in the study and the study was approved by the Institutional Review Boards of Columbia University, University of Miami, and the National Bioethics Committee and the Independent Ethics Committee of Instituto Oncologico Regional del Cibao in the Dominican Republic (DR).
Transthoracic echocardiography was performed according to the guidelines of the American Society of Echocardiography 19. LA size was measured in parasternal long-axis view at end-systole. To minimize variability, measurements were made in triplicate and averaged. Echocardiographic studies were interpreted by researchers blinded to the clinical characteristics. Interobserver variability ranged between 8% and 10%.
DNA from the family study was sent to the Center for Inherited Disease Research (CIDR) for genotyping a set of 405 microsatellite markers at an average interval of 10 centimorgan (cM) across genome. Autosomal microsatellite genotypes were used to verify and adjust family structure using the programs PREST. 20 Briefly, Maximized Log-Likelihood Ratio (MLLR) test statistics were computed to compare the putative relationship between pairs of individuals to those constructed based on the autosomal genotypes. Relationships with a p-value <0.000001 in a consistent manner across the family were considered an error. Estimated kinship coefficients and identity by descent (IBD) estimates were used to rearrange family structure as needed. Mendelian error checking was performed on the final family structure using Pedcheck.21
DNA from the NOMAS cohort was genotyped using the Genome-Wide Human SNP Array 6.0 chip (AffyMetrix) at the Genotyping Core of the John P Hussman Institute for Human Genomics (HIHG) at the University of Miami. DNA samples were processed according to Affymetrix procedures. The arrays were scanned on the GeneChip Scanner 3000 7G. Image data were analyzed using the Genotyping Console™. Vigorous quality control was applied to both samples and SNPs. Samples were removed from further analysis if they had call rates below 95%, relatedness, gender discrepancies, or were outliers beyond 6 SD from the mean based on Eigenstrat analysis. SNPs with severe deviation from Hardy-Weinberg equilibrium (p < 1E-06) or a genotyping call rate less than 95% were also removed using PLINK 1.0522.
A polygenic covariate screening as implemented in Sequential Oligogenic Linkage Analysis Routines (SOLAR) was run to screen age, sex, smoking, diabetes, dyslipidemia, hypertension, and body mass index (BMI) to determine significant covariates. An interaction between age and sex was automatically included by SOLAR. A permissive threshold of p<0.1 was used to allow for inclusion of any potentially significant covariates. Hypertension was defined as reported history of high blood pressure, systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, or use of antihypertensive medication. Smoking was defined as never versus ever. Dyslipidemia was defined as a history of hyperlipidemia or total cholesterol greater than 240 mg/dL. Diabetes was defined as a history of diabetes or fasting blood sugar greater than 126 mg/dL. Coronary artery disease (CAD) was defined as having a history of bypass surgery, angioplasty, or myocardial infarction.
Variance components methodology as implemented in SOLAR was used to estimate heritability and calculate two-point and multipoint LOD scores 23–25. A mixed-effects model that incorporates fixed covariate effects, additive genetic effects, and residual error was used. Heritability is calculated as the proportion of phenotypic variance explained by additive genetic effects while accounting for covariates. Heritability of LA diameter was assessed by itself or after correction for the indices of body size most commonly used in the literature: body surface area (BSA), and height. For QTL mapping, marker-specific IBDs were computed using the David and Weeks Monte Carlo algorithm. Then, marker-specific IBDs were merged together into one single file to calculate multipoint IBDs using a 1cM grid. Empirical p-values for LOD scores were calculated based on 10,000 replicates in which a fully-informative marker, unlinked to LA diameter, was simulated and used to compute possible LOD scores. Since SOLAR requires that quantitative traits be normally distributed and properly scaled, LA measurements were natural-log transformed, multiplied by 10.
Ordered subset analysis (OSA) was used to identify a more homogeneous subset of families for linkage analysis. 26 First, families were ranked by trait-related quantitative covariates. Then, the family-specific LOD scores were added in order of the increasing or decreasing covariate values of families until the maximum evidence for linkage was achieved. 10,000 random family orderings were permuted to generate empirical p-values for the significance of increase in the LOD score from the overall dataset to the OSA identified subset.
The NOMAS cohort included samples from a broader population and population stratification was assessed using Eigenstrat 27. Linear regression analysis was done in PLINK using an additive genetic model and adjusting for the significant covariates identified by a stepwise selection procedure in SAS. As in the family study, the same set of covariates were screened and any covariate with p<0.10 was kept in the model. Additionally, the top three principal components (PCA1, PCA2, and PCA3) as identified by Eigenstrat as well as the number of years between baseline (when risk factor information was collected) and when the echocardiographic measurement was taken were included in the covariate screening. LA measurements were natural log transformed to ensure normality. No additional outliers were eliminated (kurtosis = 0.15).
To correct for multiple testing of SNPs in the peak-wide association mapping, we applied SimpleM 28. In large scale genetics studies many SNPs were tested simultaneously but they are often not independent to each other due to linkage disequilibrium (LD). This correlation violates the independence assumption used in the Bonferroni correction. SimpleM infer independent tests such that a standard Bonferroni correction can be applied. First, a correlation matrix for the SNPs was constructed using the composite linkage disequilibrium. Then, eigenvalues were computed based on pairwise SNP correlation. Finally, PCA was used to estimate the effective number of SNPs used in the study.
In total, we enrolled 100 Dominican families with 2182 individuals in the family study. The mean family size was 22±11 members with a range of 4–87. LA measurement and genotype were available for 1372 subjects. Due to the significant contribution of mitral regurgitation and low left ventricular ejection fraction to LA size, individuals (N=13) with severe mitral regurgitation or ejection fraction less than 30% were excluded. Six individuals with outlier LA measurements (beyond 3 SD from the mean) were further removed: five of them had moderate mitral valve disease, left ventricular hypertrophy, and/or low ejection fraction but did not meet the criteria for removal. They all have high systolic blood pressure (>145mmHg) and were not on medication except one. One patient was excluded because of small LA size for unknown reason. With our final dataset, we had over 80% power to detect QTLs for traits with heritability estimates greater than 0.18 at a LOD score threshold of 2.0.
Within the NOMAS cohort, 1137 individuals were genotyped. Among them, 825 individuals passed genotyping quality control, did not have severe mitral regurgitation, low left ventricular ejection, were independent of the family study, and had phenotype and covariate measurements available. Table 1 summarizes the sociodemographic, vascular risk factors, and LA size measurements in the family dataset and the NOMAS cohort used in the final analysis.
Table 2 summarizes the estimates of heritability for LA diameter and LA diameter corrected for BSA (LA/BSA) and height (LA/HT). Age, dyslipidemia and BMI were significant covariates for all three LA size measurements (p<0.1). Covariates explained 32% of the LA diameter variance (sex and smoking as additional significant covariates), 21% of LA/BSA variance (sex and age-sex interaction as additional significant covariates), and 34% of LA/HT variance (age-sex interaction as additional significant covariates) variance. After adjusting for the significant covariates, the heritability of LA diameter, LA/BSA, and LA/HT were 0.42, 0.35, and 0.34, respectively.
As the non-corrected LA diameter measurement had the highest estimate of heritability, we focused our QTL mapping on non-corrected LA diameter. Two-point linkage analysis found suggestive evidence (LOD scores ≥2.0) on chromosomes 10 and 17 (Figure 1). Multipoint linkage analysis confirmed evidence for linkage in the two regions: 10p19 (MLOD=2.00) and 17p10 (MLOD=2.05) (Table 3). However, neither of them met the criteria for genome-wide significance.
To reduce phenotypic heterogeneity and therefore strengthen the linkage signal, we performed OSA for the two promising regions. Quantitative covariates related to LA size variance were used to rank families from high to low (H-L) or low to high (L-H) order. Five covariates, SBP, HDL, LDL, triglyeride, and waist circumference, were used in our OSA. HDL, LDL and triglyeride were chosen because dyslipidemia significantly contributed to the LA diameter in our Dominican families (Table 2), even though the lipid levels are not traditionally considered as important factors affecting LA size. Among all ranking strategies, significantly enhanced linkage evidence was observed while ordering families by average LDL from L-H on chromosome 17p10 (OSA subset MLOD=2.9, p=0.0314 for increase of linkage) (Figure 2). The LDL-defined subset included 83 families with lower LDL. The average LDL was 105.2 mg/dL for the OSA subset of families and 130.3 mg/dL for the rest of families (data not shown).
To fine map our most prominent peak, we conducted a peak-wide association study (PWAS) using SNP data from a recent GWAS completed in the NOMAS cohort. 723,979 SNPs across the whole genome passed quality control. Among them, 2233 SNPs were located within the one-LOD unit down region (6.8 megabase to 14.4 megabase) of the 17p10 linkage peak. Significant covariates based on the stepwise selection procedure were used in all analyses and included PCA1, PCA3, age, sex, BMI, hypertension, and years between baseline and echocardiographic measurement. The effective number of independent tests is 763 for the 2233 SNPs surveyed. Using a conservative Bonferroni correction, the peak-wide significance threshold is 0.05/763=0.00007. The top two associated SNPs met the peak-wide significance criterion: rs1029659 (p=0.00004, beta=0.024) in an intergenic region near Cytochrome C Oxidase Protein 10 (COX10) and rs4791774 (p=0.00005, beta=-0.021) in Netrin 1 (NTN1) (p<0.00005, Figure 3 and Table 4).
Several SNPs with nominal p ≤ 0.005 were found in LOC100128006, Nonmuscle Myosin Heavy Chain 10 (MYH10), NTN1 and intergenic regions (Table 4 and Figure 3). Some of them are located within genes that have been implicated in cardiac hypertrophy. For example, LOC100128006 is next to Myocardin (MYOCD) and codes an antisense RNA for MYOCD. As a co-activator of serum response factor (SRF), MYOCD regulates cardiac gene expression and smooth muscle cell differentiation 29–31. It has been shown that MYOCD is a key regulator of cardiac myocytes hypertrophy in vitro. 32 MYH10 encodes a nonmuscle myosin heavy chain and is expressed in heart. Mice that have heart-specific ablation of MYH10 were born with enlarged cardiac myocytes and developed cardiomyopathy. 33 In addition to the peak-wide significant SNP, two other NTN1 SNPs (rs4791331 and rs11870124) showed evidence of association (p<0.005).
Using well-characterized, extended Dominican Republic families and an independent community-based prospective cohort, we demonstrated that genetic factors explain a moderate proportion of the variance in LA size and mapped a few candidate genes that warrant further study. This is the first comprehensive genetic study on LA size among Caribbean Hispanics.
The most significant finding is for rs1029659 close to COX10. The protein product of COX10 is essential for the cytochrome C oxidase (COX), which is the terminal component of the mitochondrial respiratory chain. COX is essential for energy homeostasis and defects in COX would likely lead to cardiac hypertrophy to compensate for the insufficient energy supply in the cardiac smooth muscle cells. A patient with missense mutations in both alleles of COX10 gene has been reported. One of the clinical symptoms for the patient is severe hypertrophic cardiomyopathy 34. It is worth noting that four additional SNPs within COX10 had nominally significant p values (p=0.006 to 0.008, Figure 3), providing additional support for the association between COX10 and LA size. Another peak-wide significant finding is in NTN1. The protein product the gene belongs to a family of laminin-related secreted proteins and it is thought to be involved in axon guidance and cell migration during development and angiogenesis. NTN1 might contribute to the enlargement of left atrium through its angiogenic activity.
The susceptibility genes in 17p10 seem to be implicated in heart structure in general. However, we did not find evidence for linkage in this region for left ventricular mass (LVM).35 We evaluated the association between the significant SNPs in Table 4 and LVM in the NOMAS cohort (data not shown). Except the SNP (rs7212848, p=0.003) near MYOCD, none of them are significantly associated with LVM, which suggest that these genes have different effects on LA size and LVM. The significant association between MYOCD and LVM, however, further supports that the gene is a master regulator of cardiac smooth muscle differentiation and lineage. Given the strong known functions of this cardiac-specific factor, we believe that genetic variations that modifying MYOCD level or function are likely to influence LA size and LVM.
Literature on the heritability of LA size is sparse, with only two published studies. In the Tecumseh Offspring Study 13, parent-children correlation for LA size was 0.19 (p=0.007). This correlation is approximately equal to a heritability of 38%, which is comparable to our results. In the Framingham Heart Study, the estimate of heritability for LA diameter was 25% 16. Unlike these previous studies, which were conducted in Caucasian cohorts, our families were entirely comprised of Caribbean Hispanics, for whom no data on the heritability of LA size exists.
A few studies have investigated association between LA size and polymorphisms in candidate genes. Most of the examined genes are associated with the renin-angiotensin system and results have been conflicting. The LA size was larger in individuals who were homozygous for the insertion polymorphism of the angiotensin converting enzyme in Turkish hypertensive patients 15 but not in Chinese patients14. No association was found between LA size and the −344T/C polymorphism in the aldosterone synthase (CYP11B2) gene in patients with heart failure 17.
In contrast to candidate gene studies, a genome-wide approach, such asthe one used in this study, has the potential to expand our knowledge of the biological basis of LA enlargement by identifying new genes and new pathways. Using 100, 000 SNPs the Framingham Heart Study reported suggestive evidence for linkage on chromosome 13q31.1 (MLOD=2.55) for LA diameter 16. Using 402 microsatellite markers, we did not detect any evidence for linkage in this region, which might be related to the relative lower resolution in our study. Another possible explanation is that our study involves a unique ethnic population, which might have a different genetic basis for LA size.
Recently, a meta-analysis of GWAS on cardiac structure measurements was conducted by the EchoGen consortium consisting of five community-based cohorts, including the Cardiovascular Health Study, the Rotterdam study, the KORA, the Framingham Heart Study, and the Gutenbery Heart Study. 36 No SNP was associated with LA size at the genome-wide significant level. The strongest association was found in SMG-6 (p=9 × 10−7), about 4.3 megabases away from our 17p10 peak. The affymetrix 6.0 chip used in our NOMAS cohort includes 60 SNPs in SMG-6. Only rs12451892 was marginally (p=0.02, data not shown) associated with LA diameter but it would not survive multiple testing correction. Further studies are needed to evaluate the significance of SMG-6 to LA size in different populations.
Arterial hypertension is a well described risk factor for enlarged LA size. However, we found that hypertension did not contribute to LA size significantly (p=0.2) when we evaluated all traditional covariates, including age, sex, smoking, diabetes, dyslipidemia, hypertension, and BMI, together in a polygenic screening model in our family dataset. One possible explanation is that age served as surrogate for hypertension in our model. This notion was supported by the observation that hypertension became significant when age was not included in the covariate screening model.
The present study is the first to evaluate genetic basis to LA size in Hispanic population. There are several strengths of our study. First, the extended Dominican families and large sample size provided substantial statistical power. Second, the echocardiographic assessment in probands and their family members was performed by the same investigators adopting a common protocol assuring consistent phenotyping of the quantitative trait. Third, the genome-wide approach followed by fine-mapping in an independent cohort allowed us to evaluate the genetic contribution to LA size through the whole genome instead of taking a snapshot of one or a few genes at a time. We also acknowledge several limitations. We used LA diameter instead of volume as a measure of the LA size. Volume is considered a more accurate measure of the atrial size, especially in the case of asymmetric enlargement 12. However, the measurement of a single LA diameter has been shown to have good intraobserver and interobserver concordance 37. Another weakness is the possibility that other covariates and some shared environmental factors affecting LA size variance may not have been accounted for in our heritability estimates. This is an inherent pitfall of the heritability estimation, which we addressed by including the most known risk factors associated with LA size. Finally, to reduce heterogeneity and increase our power to map the QTLs, we restricted our study to Dominican families with strong genetic burden for vascular diseases. As a result, our findings should not be directly generalized to other populations.
In conclusion, we have demonstrated a moderate genetic component of LA size variance and mapped several potential genes that influence LA size. Further studies are warranted to evaluate the contribution of these genetic variants to LA size.
We would like to thank all staff of the Northern Manhattan Family Study for their efforts, and in particular Edison Sabala, Carlos Garcia-Lithgow, Kathiurca Almonte, and Suh-Huang Juo.
This work was supported by grants R01 NS NS40807 (RLS, TR, MRDT, SH), and R01 NS047655 (TR, RLS) from the National Institute of Neurologic Disorders and Stroke (NINDS). Dr. Di Tullio was supported by a NINDS Mid-Career Investigator Award in Patient-Oriented Research (K24 NS02241).
Authors have no conflicts of interest to disclose.