G. Pilia formulated and directed the project; W.-M. Chen performed the statistical analysis. A. Scuteri, M. Orrú, G. Albai, M. Dei, S. Lai, G. Usala, M. Lai, P. Loi, C. Mameli, L. Vacca, M. Deiana, N. Olla, M. Masala, and A. Cao recruited subjects and made and entered all observations. S. S. Najjar, A. Terracciano, T. Nedorezov, A. Sharov, A. B. Zonderman, G. R. Abecasis, P. Costa, E. Lakatta, and D. Schlessinger provided traits and tests. G. R. Abecasis, W.-M. Chen, and D. Schlessinger wrote the manuscript.
The authors have declared that no competing interests exist.
In family studies, phenotypic similarities between relatives yield information on the overall contribution of genes to trait variation. Large samples are important for these family studies, especially when comparing heritability between subgroups such as young and old, or males and females. We recruited a cohort of 6,148 participants, aged 14–102 y, from four clustered towns in Sardinia. The cohort includes 34,469 relative pairs. To extract genetic information, we implemented software for variance components heritability analysis, designed to handle large pedigrees, analyze multiple traits simultaneously, and model heterogeneity. Here, we report heritability analyses for 98 quantitative traits, focusing on facets of personality and cardiovascular function. We also summarize results of bivariate analyses for all pairs of traits and of heterogeneity analyses for each trait. We found a significant genetic component for every trait. On average, genetic effects explained 40% of the variance for 38 blood tests, 51% for five anthropometric measures, 25% for 20 measures of cardiovascular function, and 19% for 35 personality traits. Four traits showed significant evidence for an X-linked component. Bivariate analyses suggested overlapping genetic determinants for many traits, including multiple personality facets and several traits related to the metabolic syndrome; but we found no evidence for shared genetic determinants that might underlie the reported association of some personality traits and cardiovascular risk factors. Models allowing for heterogeneity suggested that, in this cohort, the genetic variance was typically larger in females and in younger individuals, but interesting exceptions were observed. For example, narrow heritability of blood pressure was approximately 26% in individuals more than 42 y old, but only approximately 8% in younger individuals. Despite the heterogeneity in effect sizes, the same loci appear to contribute to variance in young and old, and in males and females. In summary, we find significant evidence for heritability of many medically important traits, including cardiovascular function and personality. Evidence for heterogeneity by age and sex suggests that models allowing for these differences will be important in mapping quantitative traits.
Genetic analysis of complex traits, which are influenced by many different variables, is difficult because different genes and environmental factors can affect each individual. To simplify analysis, the authors turned to Sardinia, one of the rare, relatively isolated populations. They recruited 6,148 participants, aged 14–102 y, from four neighboring towns. Their sample includes many related individuals, including, for example, approximately 5,000 pairs of brothers and sisters. The authors measured 98 traits in each individual, including different aspects of blood composition and several cardiovascular and personality measures.
Here, the authors evaluate the overall impact of genes and environment on each trait and show that genes can explain many of the differences and similarities between individuals. Genetic influences were typically larger in females and in younger individuals, but interesting exceptions were observed. For example, genetic factors accounted for approximately 26% of the variation in blood pressure for those more than 42 y old, but only for approximately 8% in younger individuals. Their analysis also showed that the same genetic factor could influence multiple traits, for example by affecting multiple features of personality or of cardiovascular function. DNA analyses of this cohort will eventually allow researchers to identify genes that affect each of the traits studied, including important risk factors for cardiovascular disease.
Complex traits, including aging-associated conditions, can be influenced by a multiplicity of genetic and environmental factors. Because each factor is expected to make only a small contribution to trait variability, and this contribution may itself be influenced by interactions with other susceptibility factors, identifying the genetic basis of complex traits is challenging and requires large sample sizes [
Sardinia is the second largest island in the Mediterranean. Its modern population numbers approximately 1.65 million and constitutes a genetically isolated founder population [
Here, we use a large cohort of 6,148 Sardinians to study the heritability of a spectrum of 98 quantitative traits. Studying broad groups of traits, we could assess the generality of any trends, such as changes in heritability with aging. To increase the potential clinical utility of the results, we focused on traits that affect major domains of clinical interest. For example, in addition to anthropometric features, we quantified levels of plasma and serum markers, including total cholesterol, high-density lipoprotein (HDL), and low-density lipoprotein (LDL) levels, and measured subclinical vascular alterations [
Our study uses the full range of phenotypic variation in the population to dissect the genetic contribution and provide a quantitative assessment of the impact of inherited variation on each trait. In addition, we report evidence for heterogeneity in the genetic and environmental contributions to variation, by comparing variances and covariances between males and females and between the younger and older individuals in our cohort. Finally, we examine evidence for an overlap in the genetic determinants of multiple traits, identifying clusters of traits that appear to be influenced by the same genes. The joint study of cardiovascular and personality traits afforded us an opportunity to look for a genetic factor that might contribute to the association of certain personality traits and cardiovascular problems [
We recruited and phenotyped 6,148 individuals, male and female, age 14 y and above (
(A) Shows the number of recruited females (black bars) and males (white bars) from the four clustered towns.
(B) Shows the birthplace distribution of participants, in progressively larger geographic units: Lanusei, L.I.E.A. (Lanusei and the three surrounding towns of Ilbono, Elini, and Arzana), the Lanusei valley, the region of Ogliastra, the province of Nuoro, and all of Sardinia.
(C) Shows the birthplace distribution for grandparents of participants in the same progressively larger geographic units.
Nearly all subjects were born in Sardinia (5,857 [95%]) and, specifically, in the Ogliastra region (5,442 [89%];
To examine the effect of age and sex on each trait, we first generated and reviewed summary plots for each trait. The complete set of plots is available online (
Relative densities are plotted for males (solid lines) and females (dashed lines) for two serum values (cholesterol levels [A] and HDL [B]), two measures of cardiovascular function (IMT of the carotid artery [C] and PWV [D]), and two personality facets (NEO_N3 [E] and NEO_O5 [F]). A complete set of plots, including all traits, is available online (
These are the same traits as in
We next calculated the mean and standard deviation for all traits, both in the entire cohort and after stratifying the sample by sex and age. When stratifying the sample by age, we considered four age bands (14–29, 30–44, 45–59, and 60–102 y of age), each including approximately 25% of sampled individuals. The results are summarized in
We next used quantile normalization to convert each trait to approximate normality, and fitted a simple model with two variance components (a heritable additive polygenic component and an individual specific environmental component) and five covariates (sex, age, age2, and the terms for the interaction of sex with age and age2) [
Heritability of Blood Phenotypes and Anthropometric Measures
Heritability for Measures of Cardiovascular Function and Personality
A wide range of heritabilities was observed for each group of traits (
All heritability estimates are statistically significant (
We next proceeded to examine variance component models that allowed for either genetic dominance or shared sibling environment (
We detected a significant (
Having evaluated the standard variance component models and estimated a heritable component for each trait, we proceeded to evaluate the evidence for heterogeneity in genetic and environmental sources of variance in males and females. We considered a series of models with heterogeneity in variance components (heterogeneity in environmental variance only, heterogeneity in genetic variance only, heterogeneity in both variance components, and a model in which the genetic and environmental variances differed between males and females by a constant factor). We also considered models that included an X-linked or mitochondrial variance component (because those types of non-autosomal inheritance also induce sex-dependent differences in covariances between relatives). We selected the best-fitting model for each trait using the Bayesian information criteria (BIC) criterion (
Model Comparisons between Males and Females
The 40 traits showing significant evidence for heterogeneity of variance components by sex included all five anthropometric traits and many of the blood test results (12 of 34), cardiovascular traits (eight of 20), and personality traits (15 of 35). When heterogeneity was detected, the BIC criterion selected a model with heterogeneity in environmental variances for eight traits (the environmental variance was larger among males in five cases); with heterogeneity in genetic variances for 13 traits (the genetic variance was larger among females in 11 cases); and a model with heterogeneity in the total variance for the remaining 19 traits (in these cases, the environmental and genetic variances differed by a constant factor between males and females, and the total variance was estimated to be larger among females in 15 cases). Interestingly, the biggest differences were observed for body weight (estimated heritability of approximately 50% in females, but approximately 35% in males), hip circumference (heritability of approximately 48% in females, but approximately 27% in males) and γ-glu-transferase levels (heritability of 42% in females, but 24% in males). The A (agreeableness), N (neuroticism), and E (extraversion) personality factors and four facets showed approximately 10% higher heritability in females.
To look for heterogeneity in variance components by age, we divided individuals into two groups. The “younger” group included individuals less than 42 y of age (the median age in our sample), whereas the “older” group included individuals 42 y of age and older. We found significant evidence for heterogeneity in variance components by age in 62 of the 98 traits examined (the results are summarized in
Model Comparisons between Young and Old
For 21 traits, a model in which only genetic variance differed between the young and old was selected, and heritability was higher in the young for 15 traits (12 personality traits and three blood test results). It is noteworthy that the six traits more heritable in the old included several blood pressure–related traits (SBP, DBP, mean blood pressure, and pulse pressure). For these cardiovascular traits, heritability increased an average of 18% among older individuals, from approximately 8% for younger individuals to approximately 26% in older individuals. For 15 other traits, a model in which heritabilities between the young and old differed by a constant factor provided the best fit to the data, whereas for one trait (fractionated bilirubin), both environmental and genetic variance components appeared to differ between the young and old.
We calculated genetic correlation coefficients for all pairings of 93 traits (including the 38 blood phenotypes, five anthropometric measures, 20 cardiovascular traits, and 30 facets of personality, but excluding the five factors of personality, which are derived from the 30 facets). This corresponds to a total of 8,556 genetic correlation coefficients, of which 118 coefficients were greater than 0.50. In contrast, only 36 of the overall correlation coefficients were greater than 0.50. A full matrix of pairwise correlation coefficients is available
We identified 18 clusters of traits with a genetic correlation greater than 0.50 (
The 98 quantative traits are classified into clusters inferred from genetic correlations between any two traits, with an “average” distance measure used in the clustering algorithm. Classes of traits are color-coded as personality (red), serum composition (blue), cardiovascular (black), and anthropometric (green). Overlap of the apparent genetic contribution to variance is indicated on the ordinate, with larger overlaps towards the bottom. Eighteen values exceed 50% overlap (see text).
We looked specifically for a genetic link between personality traits and cardiovascular disease [
The cohort of Sardinians described here provided us with a valuable opportunity to investigate the heritability of multiple traits simultaneously. For some traits, the size of our cohort exceeds the total number of individuals examined in all previously published studies of their heritability. The large size of the cohort and the diversity of the relationships sampled enabled us not only to consider the overall heritability of each trait, but also to investigate the possibility of heterogeneity in genetic effects by age or sex, as well as the evidence for shared genetic determinants between different traits. To facilitate downstream studies, complete results of all our analyses (including likelihoods and parameter estimates for each model fitted) are available online (
Overall, we estimated heritabilities of approximately 0.40 on average for individual blood test results, approximately 0.51 for anthropometric measures, approximately 0.25 for measures of cardiovascular function, and approximately 0.19 for personality factors and facets. In general, our results appear to be consistent with previous studies (see, for example, [
Nearly all traits showed highly significant evidence (
Similarly, we found age differences in variance components for 62 traits. In some cases, these differences affected the genetic variance; in other cases, they affected environmental variances; and in still other cases, they affected the total variance. In the majority of cases in which we saw a difference in heritabilities between the young and old, we observed higher heritabilities among younger individuals. The trend likely reflects an expected increase of environmental insults with age [
Although our previous discussion focused on total heritability, our analysis also allowed us to examine the effects of X-linked loci. We found a substantial X-linked component influencing G6PD levels—a result that was expected, because mutations in the G6PD gene
Although the relationship between variance components and the effects of age and sex remains conjectural, and the particular source of the heterogeneity was sometimes hard to distinguish, Tables 3 and 4 clearly show that heterogeneity in variances is too great to be ignored in analyses of many traits: for the traits listed, models without heterogeneity were always rejected when compared to models with heterogeneity. Thus, modeling the variance heterogeneity between different groups or stratifying analysis by age or sex could be valuable in molecular studies. In addition, it may be desirable to focus sample collection and analysis in genetic studies on the most informative individuals (for example, our results show traits such as blood pressure have very low heritability in individuals less than 40 y of age and may be more fruitfully studied in older individuals).
We found that the correlation of genetic variance components across age groups and across sexes did not significantly deviate from 1.0 and thus, despite evidence for heterogeneity, our results do not suggest that different genes determine heritability in males and females, or in the young and old. Instead, we infer that, at any age, the alleles involved consistently increase (or decrease) values of a particular trait in relation to the age-specific population mean. If the cumulative effects of these alleles become functionally severe only at older ages, when reproductive life is generally over, deleterious alleles may still reach substantial frequencies in the population.
Further analyses can also benefit from the apparent overlap in the genetic determinants of multiple traits. For example, our observations can guide downstream multivariate analysis as well as the construction of composite traits. Combining traits with a shared genetic component can result in composite traits with higher heritability than their component phenotypes [
In ongoing studies, we plan to refine heritability estimates for traits sensitive to major environmental factors, extending the analyses by taking into account recorded information about blood pressure medicines, smoking, and alcohol consumption. We are also attempting to assess possible ascertainment bias at older ages, resulting from the differential death rates among individuals with constitutions associated with premature death (for example, individuals with high blood pressure). In one preliminary approach, we limited ascertainment bias by excluding those over 60 y of age from the analysis (also eliminating most individuals taking blood pressure medication or affected by cardiovascular disease). Nevertheless, even with this truncated sample, estimated heritabilities in young and old individuals follow similar trends to those reported here, and our qualitative conclusions are not affected (work in progress). Ultimately, decisive analyses should be possible based on further longitudinal study of the current cohort.
Overall, the cohort provides a high-yield setting for identifying traits controlling variation in medically important quantitative traits in humans. We originally planned to focus genetic analyses on individuals with extreme values for a few cardiovascular and personality traits. However, technological advances have greatly facilitated large-scale genotyping, and we now expect that a single genome scan can be completed including all the individuals in our cohort. This will allow analysis for multiple traits, and these heritability analyses results suggest that the cohort will provide a valuable resource for gene mapping. Standard power calculations suggest that a linkage scan of these samples should yield expected LOD (logarithm of the odds ratio) scores of 3 or greater for loci explaining more than 8% of the variation in 96 of 98 traits (all but PSA and TSH; W.-M. Chen, unpublished data), whereas a genome-wide association scan could identify common alleles explaining as little as 1% of the variation. Simultaneous genetic analysis of multiple traits in a single cohort will necessarily involve a substantial amount of multiple testing, but careful evaluation of false discovery rates [
Recruitment focused on Lanusei—the largest town in Ogliastra, the location of its only hospital, and site of the local bishopric—and the neighboring towns of Ilbono, Arzana, and Elini. To achieve our goal of recruiting more than 6,000 individuals from the region, the project was advertised through provincial, religious, and municipal authorities, in local television, newspaper, and radio messages, through local physicians, and by mailings and phone calls. Only individuals more than 13 y of age were eligible to participate in the study.
A clinic was established in a quiet but easily accessible sector of Lanusei. Each subject came to the clinic before breakfast, signed consent forms, and gave a sample of fasting blood. Later in the day, each subject returned for a full 2-h evaluation, including blood pressure and anthropometric measurements, cardiovascular assessments and personality testing, and a medical history interview. Two teams of six staff, all Sardinian, worked in parallel so that up to 60 subjects could be examined each week. Each team included: a physician, responsible for the medical history and physical examination; another physician and a technician, responsible for measurements of arterial stiffness and thickness; a tester for the psychological inventory; and, finally, a phlebotomist and a technician responsible for collecting and fractionating blood. Additional backup staff helped in data handling and transfer.
The study, including the protocols for subject recruitment and assessment; the Informed Consent for participants (and Assent Forms for those 14–18 y); and the overall analysis plan was reviewed and approved by institutional review boards for the Istituto di Neurogenetica e Neurofarmacologia (INN; Cagliari, Italy), for the MedStar Research Institute (responsible for intramural research at the National Institutes of Aging, Baltimore, Maryland, United States) and for the University of Michigan (Ann Arbor, Michigan, United States).
Personality phenotypes were assessed with the Revised NEO Personality Inventory (NEO-PI-R [
From each participant screened, 25 ml of blood was drawn and fractionated to provide serum, EDTA-plasma, heparin-plasma, white blood cells, and red blood cells. Clinical laboratories in Sardinia provided blood cell counts and applied a standard battery of blood tests for the measurement of electrolytes, renal function, liver function, thyroid function, and iron metabolism. Given our interest in cardiovascular risk factors, fasting lipid profiles, markers of insulin resistance (glucose, insulin, and hemoglobin A1C), and ESR were also measured. C-reactive protein (CRP) was assessed using the standard low-sensitivity assay [
Blood pressure was measured with a mercury sphygmomanometer. Measurements were taken in the morning, after a light breakfast, and after a 5-min quiet resting period, with subjects in the seated position. The SBP and DBP used here are the average of the second and third measurements from the right arm. Pulse pressure (PP) was calculated as (PP = SBP − DBP) and mean blood pressure (MBP) as (MBP = DBP + PP/3). Standard 12-lead electrocardiography was performed on all participants, from which the PR interval and the QT interval corrected for heart rate (QTC) were measured.
Participants also underwent non-invasive assessments of arterial structure and function that are increasingly recognized as potent predictors of adverse cardiovascular outcomes [
During the sonographic evaluation of the common carotid artery, Doppler studies allowed the measurement of PSV, end-diastolic velocity (EDV), pulsatility index (IP), systolic–diastolic ratio (SD_ratio), and acceleration time (AT).
In addition to personality traits, cardiovascular measures, and blood composition, we also considered four anthropometric traits recorded during physical examination of each subject (height, weight, and waist and hip circumference) and one derived quantity (the BMI, kg/m2). For conciseness, additional details of how individual phenotypes, including cardiovascular measures, were collected are supplied as
Quality assessment of the data was carried out using PEDSTATS [
To utilize fully the information in our cohort, and to accommodate covariate effects, we estimated heritabilities using a variance components model [
Variance components analyses are sensitive to outliers, kurtosis, and skewness in the trait distribution. Quantile normalization provides a practical way to deal with these problems in the context of gene mapping and, specifically, variance component analyses [
We considered a base model in which variance is partitioned into a polygenic component σg2 and an environmental component σe2. As usual, the environmental component is unique to each individual, whereas the polygenic component is shared between individuals in proportion to their kinship coefficient. Thus, if
After fitting this base model, we considered refined models including additional variance components to model genetic dominance, σd2, or the effects of shared sibling environment, σs2. To model genetic dominance, we let
To model heterogeneity, we evaluated models in which separate variance components were fitted for males, σg,male2 and σe,male2, and females, σg,female2 and σe,female2. The variances for each trait measurement and the covariances for trait-measurements between individuals of the same sex follow naturally from the formulae given in the section describing the Base Polygenic Model (above). When individuals
Setting
We also considered the possibility of X-linked or mitochondrial inheritance. X-linked inheritance can produce differences in the total variance between males and females, and either of these phenomena can generate sex-dependent covariances between relatives (for example, they can lead to differences between mother–daughter correlations and father–son correlations). We compared models with heterogeneity to models with an X-linked variance component, σx2, and to models with a mitochondrial genetic variance component, σm2. In models with an X-linked variance component,
Similarly to our analysis with heterogeneity by sex, we defined separate variance components for individuals whose age was greater or equal than the sample median (42 y of age), σg,old2 and σe,old2, and for individuals whose age was below the sample median, σg,young2 and σe,young2. When there was evidence for heterogeneity, we considered intermediate models in which heterogeneity was allowed only for environmental effects (i.e., where σg,young2 = σg,old2), or only for genetic effects (i.e., where σe,young2 = σe,old2), or in which the total variance for both genetic and environmental factors changed by a shared factor (σg,young2 =
To investigate the origin of correlations between each pair of traits
All models were fitted by maximizing the standard multivariate normal likelihood [
Recall that the kinship coefficient is the probability that two identical alleles will be sampled from a pair of individuals when we select one allele at random from each. The self-kinship coefficient is the probability that two alleles sampled from one individual, with replacement, are identical. We used a recursive formulation to estimate kinship coefficients for X-linked genes, analogous to the conventional approach described in Lange [
Then, we defined the kinship coefficient for X-linked genes,
Although this definition only covers the situation in which
This section provides a detailed protocol for the assessment of cardiovascular traits.
(18 KB PDF)
This table includes trait means and variances. Trait means are stratified by sex and into four age bands.
(37 KB PDF)
Highlights subsets of traits identified in the clustering analysis, for which the genetic correlation exceeds 0.5.
(7 KB PDF)
The table presents Procrustes-rotated principal components from the genetic correlations among the 30 facets of the NEO-PI-R, targeted to the American normative factor structure.
(11 KB PDF)
We thank warmly Monsignore Piseddu, Bishop of Ogliastra; Mayor Enrico Lai and his administration in Lanusei for providing and furnishing the clinic site; and the mayors of Ilbono, Arzana, and Elini, the head of the local Public Health Unit Ar1, and the residents of the towns, especially Silvana Bacchidu, for their volunteerism and cooperation. We thank Serena Sanna for technical help with manuscript preparation, including identification of monozygotic twins. We also thank Harold Spurgeon and Paul Pullen for invaluable help with equipment and readings; and Michele Evans and Dan Longo for helpful discussions.
Bayesian information criteria
body mass index
diastolic blood pressure
diastolic diameter
systolic diameter
erythrocyte sedimentation rate
high-density lipoprotein
intimal–medial thickness
low-density lipoprotein
prostate-specific antigen
peak systolic velocity
pulse wave velocity
systolic blood pressure
thyroid stimulating hormone