Stroke and cardiovascular diseases are of major public health concern, particularly among the rapidly growing Hispanic population 1. Studies have shown that a positive family history is a strong predictor of stroke, even after adjustment for traditional risk factors, suggesting the existence of a substantial genetic component 2. However, the genetic architecture of this common complex disease remains elusive. Given the extreme complexity of genetic and environmental contributions to stroke, evaluation of the precursor (risk) phenotypes, especially the quantitative ones, may reduce heterogeneity and increase statistical power for susceptibility gene discovery. Our Family Study of Stroke Risk and Carotid Atherosclerosis is uniquely designed to evaluate genetic contribution to cerebrovascular risk phenotypes, such as carotid intima-media thickness and left ventricular mass (LVM).

LVM is an important risk factor for stroke. In the Framingham Heart Study, the highest quartile of LVM was associated with a 2.7-fold increased risk of stroke when compared to the lowest quartile of LVM, after adjusting for age, sex, systolic blood pressure, treatment of hypertension and other established stroke risk factors 3. We have also demonstrated that LVM is significantly associated with an increased risk of stroke and combined vascular events in a multi-ethnic cohort comprised predominately of Hispanics 4. The adjusted hazard ratio for the combined endpoint of myocardial infarction, stroke, or vascular death was 1.34 per one standard deviation (SD) increase in LVM [95% CI 1.10 to 1.63].

As a quantitative trait, LVM is highly heritable. In the Framingham family study the heritability of LVM was estimated to be 0.24–0.32. 5 In the HyperGEN offspring study, the estimates of LVM heritability were 0.46 in African-Americans and 0.47 in Caucasians 6. Heritability of LVM among Japanese subjects living in Hawaii was estimated to be 0.6 7, similar to what we have seen in our sample of Caribbean Hispanics 8. Consistent with a strong genetic influence, heritability estimates for LVM are often higher in twin studies than in sibling or family cohorts; the estimate is as high as 70% for monozygotic twins 910. These studies suggested a substantial genetic contribution to LVM variations and encouraged an effort to map the quantitative trait loci (QTL) for this trait.

Previously, we have published a comprehensive heritability analysis on LVM as well as adjusted LVM measurements: LVM divided by body surface area or height; relative wall thickness; LV end-diastolic diameter; LV end-systolic diameter; interventricular septum thickness; and posterior wall thickness. The estimates of heritability for these LV measurments ranged from 0.23 to 0.65 after different adjustment for covariates, with LVM demonstrating the highest heritability (0.47 to 0.65 in different models) 8. Therefore, we focused our QTL mapping effort on LVM in this study. In addition, we implemented the Ordered-subset Analysis (OSA) based on LVM-related covariates to strengthen linkage, narrow the critical linkage region and identify a more etiologically homogeneous sample that can be used to guide further fine mapping studies.

Overall, 110 families with high vascular disease burden (Dominican Republic 100, Puerto Rico 4, Cuba 2, Ecuador 2, Nicaragua 1, and Colombia 1) were enrolled in our family study. Seventy percent of the subjects were enrolled in Northern Manhattan and 30% in the Dominican Republic. In order to reduce population stratification, we restricted our QTL mapping to the 100 Dominican families, which was composed of 2182 individuals. The mean family size was 22 ± 11 with median of 20, and range of 4–87. LVM measurements and DNA samples were available for 1360 subjects from the 100 Dominican families. Among the 1360 individuals, there were 1207 sib pairs, 362 half-sib pairs, and 1713 avuncular pairs. With our 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. The mean age of the study cohort was 46 years. The mean LVM was 174.85 ± 56.14 (Table 1). Men had significantly greater mean LVM than women (201.42 vs 158.15, p < 0.0001).

The polygenic covariate screening identified age-sex interaction, sex, ever smoking, systolic blood pressure (SBP), and BMI as significant covariates. Heritability and LOD scores were calculated adjusting for these covariates. The significant covariates together explained 0.33 of the total LVM variance. The heritability estimate of LVM is 0.58 (p < 0.0001) for the remaining variance of LVM after adjusting for the significant covariates.

To localize the genetic loci affecting LVM variations, we completed a QTL mapping on LVM. This analysis revealed a significant region on chromosome 12p11 with peak marker at D12S1042 (MLOD = 3.11, empirical p value = 0.0003) (Figure 1). The one-LOD unit down critical interval extends from approximately 25 Megabase (Mb) to 51 Mb on chromosome 12, encompassing 181 annotated genes. No evidence for linkage (LOD > 1.5) was found in any other region through the genome (data not shown). Estimation of the locus specific heritability is 0.25 at D12S1042 and this locus accounted for 43% of the total heritability of LVM in our dataset.

To better characterize the prominent linkage region on 12p11, we performed OSA to further reduce phenotypic heterogeneity. In this analysis, trait-related quantitative covariates were used to rank families for a more homogeneous subset. We chose SBP, BMI, and waist circumference (WC) as the covariates to order families because high blood pressure and obesity, especially elevated waist circumference (WC), are major risk factors that lead to an increase of LVM in general population 222324. We ordered families by family average SBP, BMI, and WC. OSA results based on high to low (H-L) or low to high (L-H) ranking of the three covariates are listed in Table 2. The OSA maximum LOD score was increased in all H-L rankings (LOD = 3.81 to 4.45) but was not changed in the L-H rankings (LOD = 3.14 to 3.40), when compared to the overall analysis. Permutation tests were performed to evaluate the significance of increases in LOD scores. Among all the OSA ranking strategies that led to a higher LOD score, ordering families by decreasing average WC significantly increased the evidence for linkage (LOD = 4.45) (p = 0.0045 for increase in LOD score). In addition to significantly strengthening the linkage, this ranking strategy also greatly reduces the one-LOD unit down supporting interval size from 26 Mb to 14 Mb (Figure 1). The narrowed critical linkage region extends from 25 Mb to 39 Mb on chromosome 12 harboring 55 annotated genes. The OSA defined subset included 54 families with high WC. The average WC was 38.4 inches for the OSA subset families and 34.4 inches for the rest of the families (Table 2).

Using well-characterized, extended Dominican Republic families, we mapped a novel QTL for LVM on chromosome 12p11. This QTL influences LVM independent of other risk factors such as gender, age, hypertension, BMI, and smoking. By implementing OSA to reduce phenotype heterogeneity, we have greatly strengthened the evidence for linkage, narrowed the linkage region and identified a sub-phenotype that has concentrated genetic effects from chromosome 12p11. Our findings provide a well-defined chromosome region and phenotype for future validation and fine mapping studies to localize genes controlling LVM.

The majority of genetic studies on LVM phenotype to date have focused on candidate genes derived from biological pathways that have impact on the LVM phenotype, such as genes implicated in myocardial cell growth/remodeling, calcium homeostasis, extracellular matrix, hemodynamic load, and blood pressure regulation. Polymorphisms in several genes have been associated with LVM, such as the 825C/T polymorphism in the G protein beta subunit (GNB3) gene, 252627 polymorphism in the beta-1 adrenergic receptor (beta-1 AR) gene, 28, the -344C/T polymorphism and a gene conversion in intron 2 in the aldosterone synthase gene, 2930 and the polymorphism in the angiotensin-converting-enzyme (ACE) gene. 31 However, the association results have been conflicting 1032. Recently, a meta-analysis of 485 reports concluded that the -344C/T polymorphism in the aldosterone synthase gene is not significantly associated with LVM 33.

Other approaches to identify the underpinning genes for a disease-associated trait are unbiased genomic approaches, such as genome-wide linkage and association studies. These studies do not require a priori knowledge on the molecular basis of phenotypic variations and have advantages in identifying unknown/unexpected genes and pathways for complex traits. Previously, two genome-wide studies of LVM as a quantitative trait have been reported. Both of them were conducted in Caucasian populations 3435. One was completed in white British families with hypertension and another one was conducted in families from the Framingham Heart Study. In the British study, suggestive evidence for linkage was found on chromosome 5p (MLOD = 1.6) for echocardiography-derived LVM (echo-LVM), and on chromosome 7q (MLOD = 1.7) and 12q (MLOD = 2.2) for electrocardiography-derived LVM (ECG-LVM) 34. In the Framingham Heart Study, 100,000 of single nucleotide polymorphisms (SNPs) were tested for association with LVM. None met the genome-wide significance level. Three SNPs with a nominal p-value less than 10^-5 were found on chromosomes 2, 6, and 11. One SNP is located within a heat shock protein gene HSPA8; the other two SNPs are not located in or near known genes. Linkage analysis revealed strong evidence for linkage on chromosome 5q (MLOD = 4.4) and suggestive evidence for linkage on chromosomes 1q and 8q (MLOD = 2.4).

There is no overlap among the QTLs delimited in the British, the Framingham Heart, and the current studies. One possible explanation is that the unique ethnic population in our study may have a different genetic basis for LVM than Caucasians analyzed in the other two studies. However, the lack of replication even between the British and Framingham studies as well as the conflicting candidate gene studies, underscore the challenges in genetic analysis of complex traits: the genetic basis is likely to be complex and heterogeneous, involving multiple genes each with moderate effect and interaction with environmental factors. Within a given dataset, only one or two loci with the strongest effect in that population can be detected. In our study, the 12p11 QTL only accounts for less half of the total heritability of LVM. Other loci that have smaller effect sizes are still at large due to the limited power of the current dataset. Future studies in a similar population, including meta analyses are needed to catalogue all the genetic causes for the inter-individual LVM variations.

To reduce heterogeneity, stratification analysis based on a trait-related variable has been proposed in genetic studies of complex traits. The OSA ranks families by a trait-related variable and sequentially introduces the families to the linkage analysis until the maximum evidence of linkage is observed 36. The subset defined by OSA is likely to represent a more etiologically homogeneous sample and thus increases the statistical power. Furthermore, the more homogeneous phenotype defined in the OSA subset provides insight to guide further analysis to pinpoint the genetic variants contributing to the trait. For example, in the GENECARD family study of coronary artery disease (CAD), OSA has identified a subset of families with very early age-of-onset (mean age-of-onset = 37.8 years) with strong evidence for linkage on chromosome 7p (LOD = 4.2). This finding was used to guide further association studies in subjects with early onset CAD and led to the discovery of the Neuropeptide Y gene polymorphism as genetic risk factor for atherosclerosis 37.

Using OSA, we found that the linkage signal on chromosome 12p11 is mainly from families with high WC. To exclude the possibility that we mapped a locus for WC instead of LVM, we performed a linkage analysis on WC and found no evidence suggesting that the signal on 12p11 is driven by WC (LOD = 0 at D12S1042 for WC). Therefore, the 12p11 linkage is a bona fide QTL for LVM and genes under this QTL interact with visceral obesity to influence inter-individual LVM variations. Adipose tissue is an active organ that secretes a variety of bioactive peptides. Several proteins of the renin angiotensin system, an essential regulatory axis for blood pressure, are expressed in adipose tissue. 38 Among them, angiotensinogen (AGT), angiotensin receptors type I (AT1), and ACE have higher expression in visceral than subcutaneous adipose tissue. Indeed, plasma AGT and plasma ACE activity are higher in obese individuals. 38 One possible scenario is that the candidate gene(s) in 12p11 QTL interact(s) with the bioactive peptides produced in visceral adipose tissue, such as ACE, to affect LVM. This interaction could confound phenotype-genotype correlations and should be taken into account in future fine-mapping studies.

The OSA based on BMI did not significantly enhance the linkage evidence despite that BMI and WC are correlated measurements, suggesting that WC provided additional information to BMI in the analysis of LVM. Consistently, recent reports from Rodrigues et al. 39 and Tsioufis CP et al. 40 showed that WC is an important predicator of LV hypertrophy and LV diastolic dysfunction. Furthermore, the impact of WC on LV structure and function seemed to be especially strong in women in those two studies. This gender specific effect should be considered in future studies in population-based cohort. The OSA based on SBP did not identify any subset that was significantly different from the randomly ordered families. However, our study is limited by the single clinic measurement of blood pressure and antihypertensive treatment. Blood pressure variation throughout the day is better predictor of left ventricular hypertrophy than a single clinic measurement, and antihypertensive treatment may reduce left ventricular hypertrophy depending on how long the patient has been treated. These could obscure the correlation between blood pressure and LVM.

It is worth noting that in the HyperGEN study suggestive evidence for linkage (LOD = 1.99) for LV geometry was found at the same peak marker (D12S1042) for LV mass in our study 41. LV geometry is a combination of LV mass and relative wall thickness (RWT) of LV, with LV mass contributing to forming the different LV geometric patterns and RWT being an index of concentric geometry. Although the LV measurement is not exactly the same examined in our study, the encouraging LOD score at the same chromosomal location as found in the HyperGEN study corroborates our findings of chromosome 12p11 as a potential locus for LVM due to similarity between LV mass and LV geometry. Nonetheless, further validation studies on LVM are desired to confirm the QTL.

Our success in mapping a strong QTL relies on the strength of our study design. We focused on one ethnic group and used a family study design to minimize the effects of population substructure. All the families have high vascular disease burden and detailed systematic measurements of the quantitative phenotype. The extended family structure (average family size of 20) in our dataset provides unprecedented information content and statistical power. The QTL mapping approach also offers additional statistical advantages over discrete traits. A major limitation of our study is that we did not pinpoint the causal genetic variations that lead to increased LVM. This study is only the first stage to elucidate the genetic basis for LVM. Looking through all the genes under the 12p11 linkage peak did not find any gene that would have an obvious effect on LVM based on the current knowledge. However, the molecular basis of increased LVM is not clear and many genes' functions are not fully understood. We plan to perform fine mapping study across the linkage peak to narrow the critical region and identify the candidate gene in future studies.

Most genetic studies of cardiovascular disease have focused on non-Hispanic populations 4243444546 with a few exceptions such as the San Antonio Family Heart Study and the Multi-ethnic Study of Atherosclerosis (MESA) 4748. Our family study provides essential data to fill the gaps in our knowledge of the genetics of cerebrovascular risk phenotypes in minority populations.

With the unbiased genome-wide approach, we mapped a novel QTL for LVM, a significant risk phenotype for stroke and vascular events. The strong evidence for linkage on chromosome 12p11 warrants further validation and fine mapping efforts. In addition, our family study provides unique data on the genetics of cerebrovascular risk phenotypes in an understudied minority population.

The authors declare that they have no competing interests.

LW: planned the analysis, interpreted data, draft the manuscript. AB: analyzed the data and participated in the interpretation of data. MRD: ascertained and phenotyped the subjects, assisted in study design, revised the manuscript for important intellectual content. SS: designed the analysis, analyzed the data, and participated in the interpretation of data. SHB: designed the analysis, interpreted data, and critically revised the manuscript. TR: ascertained and phenotyped the subjects, assisted in study design, and critically revised the manuscript. RS: conceived the overall study, secured funding, and critically revised the manuscript. All authors have read and approved the content of the manuscript.