Contribution of Rare and Common Genetic Variants to Plasma Lipid Levels and Carotid Stiffness and GeometryCLINICAL PERSPECTIVE
A Substudy of the Paris Prospective Study 3
Background—We assess the contribution of common and rare putatively functional genetic variants (most of them coding) present on the Illumina exome Beadchip to the variability of plasma lipids and stiffness of the common carotid artery.
Methods and Results—Measurements were obtained from 2283 men and 1398 women, and after filtering and exclusion of monomorphic variants, 32 827 common (minor allele frequency >0.01) and 68 770 rare variants were analyzed. A large fraction of the heritability of plasma lipids is attributable to variants present on the array, especially for triglycerides (fraction of variance attributable to measured genotypes: V(G)/Vp=31.4%, P<3.1×10–11) and high-density lipoprotein cholesterol (V(G)/Vp=26.4%, P<4.2×10–12). Plasma lipids were associated with common variants located in known candidate genes, but no implication of rare variants could be established. Gene sets for plasma lipids, blood pressure, and coronary artery disease were defined on the basis of recent meta-analyses of genome-wide association studies. We observed a strong association between the plasma lipids gene set and plasma lipid variables, but none of the 3 genome-wide association studies gene sets was associated with the carotid parameters. Significant V(G)/Vp ratios were observed for external (14.5%, P<2.7×10–5) and internal diameter (13.4%, P<4.3×10–4), stiffness (12.5%, P<8.0×10–4), intima-media thickness (10.6%, P<7.9×10–4), and wall cross-sectional area (13.2%, P<2.4×10–5). A significant association was observed between the common rs2903692 polymorphism of the CLEC16A gene and the internal diameter (P<4.3×10–7).
Conclusions—These results suggest an involvement of CLEC16A, a gene that has been reported to be associated with immune disorders, in the modulation of carotid vasodilatation.
Increased central arterial stiffening is a hallmark of the aging process and the consequence of many disease states, such as diabetes mellitus, atherosclerosis, and chronic renal disease. Aortic stiffness has independent predictive value for total and cardiovascular mortality, coronary morbidity and mortality, and fatal stroke in patients with essential hypertension, end-stage renal failure, or diabetes mellitus.1–4 Arterial stiffness is defined by a reduction in arterial distensibility. We previously showed that aortic stiffness (carotid-femoral pulse wave velocity) remained a significant predictor of coronary events in hypertensive patients after adjustment for classical cardiovascular risk factors.5 The growing prevalence and associated risk of arterial stiffness provide a major incentive to increase our understanding of the underlying molecular, cellular, and genetic causes and the resulting physiological impact of this condition.
Clinical Perspective on p 636
Since the initial study showing an association between a common polymorphism of the angiotensin II type 1 receptor gene (1166 A/C) and arterial stiffness parameters,6 genome-wide linkage studies have identified chromosomal regions associated with arterial stiffness, without relying on any prior biological hypothesis.7 In addition, gene expression profiling studies have identified novel transcriptional biomarkers of arterial stiffness in humans,8 and variants in genes encoding molecules involved in cytoskeletal organization, vascular smooth muscle cell (VSMC) differentiation, or the contractile state of the cell have been suggested to contribute to cardiovascular stiffness.9
The first genome-wide association study (GWAS) of arterial stiffness was performed using a 100K panel of common single nucleotide polymorphisms (SNPs).10 The most recent GWAS with a larger sample size and using more informative arrays have identified several new candidate genes, some of them being potentially involved in the pathophysiology of arterial stiffness, in particular collagen, type IV (COL4A1),11 the B-cell CLL/lymphoma 11B (BCL11B) gene desert, a transcriptional repressor of various genes that may be relevant to aortic stiffness,12 and adrenomedullin (ADM), a potent vasodilator and angiogenic peptide.13 The recent development of whole-exome arrays targeting mainly coding regions of genes offers great potential for the identification of rare or common variants with important phenotypic effects.
In the present report, we investigated for the first time the possible influence of variants present on the Illumina exome beadchip on phenotypes related to arterial pathophysiology, that is, plasma lipids and parameters defining the geometry and distensibility of the carotid artery, in a population-based study.
Material and Methods
The design and objectives of the Paris Prospective Study 3 (PPS3) have been previously described.14 It is an ongoing observational prospective study evaluating the possible implication of numerous vascular health parameters in cardiovascular disease in healthy men and women. Briefly, the PPS3 cohort consists of individuals working or in early retirement and their families. The participants are recruited on a voluntary basis in one of the largest French preventive health centers, the Center d’Investigations Preventives et Cliniques (IPC Center) that is subsidized by the National Health Insurance System for wage earners (Securite Sociale–CNAMTS). Female subjects represent ≈1/3 of the participants, and the age of inclusion ranges from 50 to 75 years. The study protocol was approved by the Ethics Committee of the Cochin Hospital (Paris) and, all the participants have signed an informed consent form. Our study is registered in the international trial registry (NCT00741728). Here, we report a substudy in which we have selected 4056 index subjects. After exclusion of samples with >1% missing genotypes (n=96), outliers, and genetically related individuals (n=279, see data preprocessing), 3681 subjects were kept in the analyses.
Measurement of Lipid Variables
The standard health check-up included a complete clinical examination, coupled with standard biological tests performed after overnight fasting. Total cholesterol and triglycerides were measured using standardized enzymatic methods (automat HITACHI 917); high-density lipoprotein cholesterol (HDLc) was measured by direct enzymatic assay with cyclo-dextrin; low-density lipoprotein cholesterol (LDLc) was estimated using the Friedwald formula in individuals with plasma triglycerides below 450 mg/dL.
Assessment of Carotid Geometry and Distensibility
Carotid properties were determined using a high-resolution echotracking system (Artlab, Esaote, Maastricht, the Netherlands). Briefly, a 10 MHz 128 transducer linear array probe was positioned on the carotid area. Measurements were performed on a 4 cm segment of the right common carotid artery, 1 cm proximal to the bifurcation/sinus throughout the cardiac cycle for 6 seconds. A longitudinal section showing clear interfaces for blood/intima and media/adventitia was obtained. The system allows real-time radiofrequency signal analysis with operator-independent determination of external diameter (Dext), internal diameter (Dint), and intima-media thickness (IMT) on 128 lines throughout the cardiac cycle. Distension was measured on 14 lines at high-pulsed radiofrequency (600 Hz). The axial resolution was 34 μm for diameter, 17 μm for IMT, and 1.7 μm for distension.15 Central pressure was estimated from the distension waveform according to van Bortel et al.16 The distensibility coefficient (DC), representing the elastic properties of the artery as a hollow structure, was calculated as dLCSA/(LCSA×PPC), where LCSA is the lumen cross-sectional area and PPC is the central pulse pressure. Carotid stiffness (Cstif) was calculated as DC−0.5 and circumferential wall stress as diastolic blood pressure×Dint/2×IMT. WCSA is wall cross-sectional area.
The Illumina HumanExome-12v1.1 BeadChip targets 242 901 markers and was designed to genotype variants with strong predicted functional impact, such as variants altering protein sequences and reported GWAS hits (see http://genome.sph.umich.edu/wiki/Exome_Chip_Design for a description of the array). We used the standard Infinium HD Assay Ultra protocol according to the manufacturer’s instructions (See Material, Section I.1 in the Data Supplement for details).
Genotype calling was done with Illumina GenomeStudio software, and quality control was performed with the PLINK software17 and in the R environment.18 Markers with genotyping success rates <99% (n=2065) and not in Hardy–Weinberg equilibrium (P<10−5, n=4325) were not investigated further. Outlier samples identified using the neighbor function implemented in PLINK were excluded. We used the genome-wide complex trait analysis software GCTA19 to detect cryptic relatedness among study participants and removed one individual from each pair of related samples (using a genetic relatedness matrix cutoff =0.1 which corresponds to a moderate level of relatedness). After exclusion of monomorphic variants (n=134 914), the final data set included 3681 individuals and 101 597 polymorphic markers.
Statistical analysis was performed on common and rare variants separately, and a threshold of 0.01 for the minor allele frequency (MAF) was adopted to separate rare from common variants.
All phenotypes were log-transformed before analysis, and association analyses were adjusted for sex, age, body mass index (BMI), and body surface area (BSA). Adjustment on both BMI and BSA was performed because associations with the 2 body size measurements were dissimilar for some arterial parameters. The correlation between BMI and BSA was 0.625, and the correlations of both measurements with sex were 0.212 and 0.681, respectively. The correlation between BMI and BSA was unlikely to introduce a problem of collinearity that might affect the results because (1) our focus was on the relationship between the genetic factors and the lipid/vascular parameters, and BMI, BSA, and the other covariables were used for adjustment only; (2) the correlation between the genetic factors and the covariables was weak; (3) using the set of covariables, the variance inflation factors of BMI and BSA were 1.89 and 3.29, respectively, which is not strong; and (4) when adjustment was done by keeping only BMI or BSA and not both, the results were almost unaffected. We also tested polynomial adjustments, but the results were almost unchanged (not reported).
Associations between phenotypes and common variants (variant-level analysis) assuming an additive mode of inheritance were explored using a mixed linear model implemented in the GCTA software,19 accounting for the remaining genetic relatedness among participants and covariates (age, sex, BMI, and BSA). For each phenotype investigated, we estimated the fraction of the phenotypic variance attributable to measured genotypes (V(G)/Vp) using the mixed linear model approach implemented in GCTA20 and also tested a possible heterogeneity of these associations according to sex. Note that in our analysis, we used the leave-one-chromosome-out analysis implemented in GCTA software to avoid adjusting on variants that are in linkage disequilibrium (LD) with the variant tested.
At the gene and gene set levels, associations with the lipid and vascular traits were investigated using the Sequence Kernel Association Tests (SKAT) implemented in the R package SKAT (R/SKAT) for both rare and common variants.21 All tests were adjusted on age, sex, BMI, BSA, and first 10 principal components to account for population structure by entering these covariates in the SKAT null model. The SKAT test allows for both effect increasing and effect decreasing variants, and this seems appropriate in the context of this study in which tested variants were collected independently of the phenotypes investigated; nevertheless, we also conducted a one-sided collapsing test to account for the possibility that most variants in a gene or a gene set affect a phenotype in the same direction (see Material, Section I.2. in the Data Supplement for details).
For each trait, association P values were summarized by a Manhattan plot (R/qqman package), and the fit of observed to expected association χ2 was plotted as QQ plots (R/snpStats package).
To account for multiple testing, we applied a Bonferroni correction for the number of variants tested in the variant-level analysis of common variants (n=32 827) and for the number of genes tested in the gene-level analysis of rare and common variants (n=13 453 genes harboring at least 2 variants). These corrections are conservative because they do not account for the LD existing among variants in the same gene and nearby genes. To further account for the interdependent vascular phenotypes, we applied a Bonferroni correction for the 3 main phenotype categories: plasma lipids, carotid stiffness, and carotid geometry. After these corrections, the significance thresholds for common variants and rare variants were set to 5×10–7 and to 1×10–6, respectively.
Power of the Study
The power of the variant-level analysis for various MAF (>0.01) and allele effect size are reported in Table I in the Data Supplement. We note that given our study size, for an MAF of 0.01, effect size must be at least equal to 0.7 standard deviation (SD) unit to reach a power of 0.8. For an MAF of 0.1, a similar power is attained for an effect size of 0.23 SD unit. For sets of variants, power estimation requires modeling the genetic architecture (number, respective frequencies and effects of variants, LD between variants). The R/SKAT package provides an analytic method to compute power for SKAT (See Methods I.3 in the Data Supplement). Given the relatively small number of rare variants in our study, the power of the gene-level analysis was low for most genes. On the other hand, the power of the gene set–level analysis was more appropriate (Table II in the Data Supplement).
3681 individuals (2283 men and 1398 women) contributed to the analysis. The mean values of the relevant characteristics are reported in Table 1 for men and women separately. The results of association analyses were similar in men and women, and as a consequence, we report results for both genders combined. Correlations among vascular parameters are shown in Table 2 before (above the diagonal) and after (below the diagonal) adjustment on covariables (age, sex, BMI, and BSA).
Heritability of Investigated Traits Attributable to Variants on the Array
The ratio of the genetic variance to the total variance (V(G)/Vp) of lipid and vascular phenotypes are reported in Table 3; these ratios correspond to the heritability attributable to the 101 597 variants available. The ratios for LDLc, HDLc, and triglycerides were 10.5% (P<0.0081), 26.4% (P<4.2×10–12), and 31.4% (P<3.1×10–11), respectively. The V(G)/Vp ratios for blood pressure–related variables (mean blood pressure and PPC) were not significant; on the other hand, the ratios for Cstif (12.5%, P<0.0008), Dext (14.5%, P<2.7×10–5), Dint (13.4%, P<0.00043), IMT (10.6%, P<0.00078), and WCSA (13.2%, P<2.4×10–5) were significantly different from zero; however, the large standard errors (Table 3) imply that the heritability estimates are rather imprecise. To better characterize the source of this heritability, genetic effects were investigated at 3 levels: (1) the variant level, for common variants (MAF >0.01), (2) the gene level, for common and rare variants (MAF ≤0.01), and (3) the gene set level, with 3 gene sets, each combining candidate genes identified in GWAS of plasma lipids, coronary artery disease, and blood pressure, respectively.
The variant-level analysis was conducted on the 32 827 common variants. Results of the association analysis with the mixed linear model adjusted for age, sex, BMI, and BSA are reported for each phenotype as a Manhattan and QQ plot (see Material, Section II and Figures I–III in the Data Supplement); the most significant associations at each associated locus are shown in Table 4, and a complete report of all associated SNPs is provided in Table III in the Data Supplement. The highest estimated lambda statistics for all phenotypes was <1.05 (see QQ plots in Figures I–III in the Data Supplement), suggesting that population stratification if any was well controlled in the mixed linear model and was unlikely to affect the association results in a significant way.
The QQ and Manhattan plots (Figure I in the Data Supplement) reveal the presence of numerous associations with lipid variables. Significant associations were observed for LDLc with variants at the APOE/TOMM40/APOC1, PCSK9, and SNAPC2 loci, for HDLc with variants at the CETP, LPL, ABCA1, SNAPC2, and LIPC loci, and for triglycerides with variants at the GCKR, APOA5, LPL, and TRIB1 loci. These associations are consistent with results reported in large-scale GWAS of plasma lipids.22 Note that the exm1417699 variant (rs116635738) in the SNAPC2 gene is located 445 798 bp from the rs7255436 intronic SNP in the ANGPTL4 gene, which is a GWAS-identified locus for HDLc.22 However, rs7255436 was available in our data set but was not associated with any of the lipid variables (P>0.05).
For arterial phenotypes, a significant association (P<4.2×10–7) was observed between carotid Dint and rs2903692 at the CLEC16A locus. The minor allele (MAF=0.39) of this variant was associated with a reduced Dint value. rs2903692 was also associated with Dext (P<4.5×10–6) but not with WCSA (P<0.04) and IMT (P<0.9), suggesting that it is related to vasodilation rather than to wall hypertrophy. Two other common variants of CLEC16A, rs12708716 and rs12924729, in tight LD with rs2903692 exhibited slightly weaker associations with Dint, P<2.2×10–6 and 3.4×10–6, respectively. No significant association was found with stiffness parameters (Cstif, CWS, and WCSA).
The gene-level analysis was conducted with SKAT for genes harboring at least 2 variants (rare or common) present in our data set (n=13 453 genes). QQ plots and tables reporting gene–phenotype associations with P values <10–4 are provided in the Material, Section III, Table IV, and Figures IV–VI in the Data Supplement. Although this analysis was more specifically focused on the analysis of rare variants, the results of the joint analysis of all and common variants are also reported in Table IV in the Data Supplement. The gene-level analysis of common variants revealed significant associations (P<10–6) implicating lipid variables, but all associated loci (APOE, TOMM40, LPL, APOA5, GCKR, CETP, LPL, and SNAPC2) had already been identified in the variant-level analysis. No significant gene-level effect of common variants on vascular variables was observed. The results of the SKAT and collapsing tests for rare variants are reported in Table IV in the Data Supplement. Overall for both lipid and cardiovascular phenotypes, the gene-level analysis did not reveal any association reaching the significance threshold of 10–6. It must be considered, however, that because of the large number of monomorphic variants, relatively few rare variants were available for most investigated genes (see Table IV in the Data Supplement), and as a consequence, the study had limited power to detect an effect of rare variants at the gene level. To try overcome the lack of statistical power resulting from the separate investigation of each gene, we extended the analysis to sets of genes whose genetic variability is known to affect lipid and cardiovascular phenotypes.
Gene Set–Level Analysis
We defined candidate gene sets based on the most recent GWAS reports for plasma lipids,22 coronary artery disease,23 and blood pressure/hypertension.24,25 All variants available in our study and located within or close to the sequence of the genes in each gene set were identified and analyzed with SKAT to assess their contribution to the variability of the traits investigated. The genes included in the sets and the number of variants are reported in the Material, Section IV in the Data Supplement, and the results of the SKAT analysis are shown in Tables V–VII in the Data Supplement for all, common, and rare variants, respectively. Analysis of the blood pressure gene set did not reveal any significant association with the lipid or the vascular phenotypes (Table 5). As expected, the plasma lipid gene set was associated with the plasma lipid variables, but this association was only attributable to the common variants. No association of this gene set with the vascular phenotypes was observed. Associations with the coronary artery disease gene set were significant for LDLc and triglycerides, but again these associations were observed for the common variants only.
To assess whether common variants significantly associated with plasma lipids in the variant-level analysis could account for the associations observed with the lipid set, we excluded 38 common variants associated with plasma lipids in the variant-level analysis (P<0.0001) and performed the SKAT analysis again using the reduced plasma lipid gene set. Associations of common variants in this gene set with LDLc (P<9.5×10–5), HDLc (P<2.3×10–5), and triglycerides (P<1.4×10–4) were reduced but remained significant, indicating that common variants with weak effects present in the lipid gene set but nonidentified in the variant-level analysis contribute to plasma lipid levels. In Tables V–VII in the Data Supplement, we also report the results of the gene set–level analysis limited to variants with a high putative severity score as defined by the Combined Annotation–Dependent Depletion (CADD) score. However, this analysis did not reveal any association with the rare variants in the 3 gene sets.
The exome array with its enrichment in rare variants provides a unique tool to investigate the contribution of genetic variants affecting protein structure to various quantitative traits and diseases in the human population. Compared with current genome-wide arrays, it provides a much reduced genome coverage of <10% of variants with a MAF >1% (based on a r2>0.8). The great advantage of the exome array is therefore not related to its genome-wide tagging ability but to its enrichment in putatively functional variants and, therefore, to the increased possibility that an associated variant is causal in comparison to a random marker.
Rare Versus Common Variants
To define rare variants, it has been proposed to adopt a threshold of 1/sqrt(2n), where n is the study sample size. In our study, this value was 0.0116, so we adopted a threshold for rare variants of 0.01. For the gene- and gene set–level analyses, we used the default parameters in R/SKAT, which up-weight rare variants relative to common ones. However, we also performed the analyses without differentially weighting rare and common variants (see Methods I.2 in the Data Supplement), and the results were similar.
The heritability of plasma lipids is close to 50%. The last meta-analysis of GWAS of plasma lipids conducted in 188 578 European ancestry individuals and 7898 non-European ancestry individuals has confirmed 97 previously reported loci and identified 62 new ones.22 In our much smaller study mainly focused on coding variants, we identified only a small subset of these loci, but the V(G)/Vp ratios for the lipid phenotypes indicate that the variants present on the exome array may contribute to >50% of the heritability of HDLc and triglycerides.
In the Framingham Heart Study, heritability of mean arterial pressure and pulse pressure were, respectively, 0.33 and 0.50.26 In sharp contrast in our study, the fraction of phenotypic variance of mean blood pressure and PPC attributable to variants on the genome array was nonsignificant. In addition, variants extracted from the blood pressure gene set genotyped in our study were not associated with mean blood pressure or PPC. Most variants on the exome array were chosen because they are coding and affect protein sequence. We may therefore hypothesize that variants contributing to mean blood pressure and PPC do not belong to this category and are more likely to be regulatory.
In the Framingham Heart Study, heritability of carotid-femoral pulse wave velocity (a measure of arterial stiffness) was 0.40.26 In the PPS3, the V(G)/Vp ratio for Cstif was 12.5% (P<0.0008). However, the association analysis of Cstif at the variant, gene, or gene set level did not reveal any significant association, suggesting that the variants affecting heritability of carotid stiffness may exert a too weak individual effect to be detectable in our study. It must also be considered that PWV is directly measured, whereas Cstif is calculated from measures of diameter, distension, and pulse pressure, raising the possibility that cumulated measurement error may explain our negative result.
In the PPS3, carotid geometry was assessed via IMT, Dint, Dext, and WCSA. All 4 parameters were characterized by a V(G)/Vp ratios differing from zero. In the Framingham Heart Study,27 adjusted heritability was 0.38 for the common carotid artery IMT (P<0.001) and 0.35 for the internal carotid artery IMT (P<0.001). In our study, the heritability of the common carotid artery IMT attributable to the variants on the exome array was 10.6% (P<0.0008). In a meta-analysis of GWAS conducted by the CHARGE consortium in over 40 000 participants of European ancestry, common variants associated with carotid IMT were identified in 4 regions of the genome, near ZHX2, APOC1, PINX1, and SLC17A4.28 The strongest association was with the rs445925, located at position 19:45415640 near the APOC1 gene. This variant was present in our data set, but no association with carotid IMT could be detected (P>0.05). The other lead SNPs identified in the meta-analysis of GWAS of carotid IMT were not genotyped in our study; however, no variant at the respective loci was associated with the trait.
In a genome-wide linkage analysis in 3300 American Indian participants in the Strong Heart Family Study, heritability estimates for carotid artery lumen diameters ranged from 0.29 to 0.45 across field centers. In the PPS3 study, Dext was the carotid parameter showing the largest V(G)/Vp ratio (14.5%, P<2.7×10–5). In the Strong Heart Family Study, significant evidence for a locus influencing carotid artery lumen diameter on chromosome 7q was reported29; the authors proposed KCND2 as a possible candidate at this locus. In the PPS3, an intronic variant located at position 7:120323727 within the KCND2 gene sequence was not associated with carotid diameter, and there was no evidence for an association of Dint with variants located within 5 MB on both sides of KCND2 (>100 variants, minimum P value =0.014).
Dint was associated with common variants on the CLEC16A sequence, and the strongest association was observed for rs2903692 (P<4.21×10–7). Similar but slightly less significant associations were observed for Dext. rs2903692 is located in intron 22 of CLEC16A at position 16:11144926. The 3 intronic variants on the CLEC16A sequence were added to the exome array because they have been reported to be associated with several immunity-related diseases,30 particularly type 1 diabetes mellitus and multiple sclerosis. rs2903692 is located at a distance of 58 910 bp from rs12708716, the lead SNP associated with type 1 diabetes mellitus, and both SNPs are in tight LD (R2=0.96). The mechanism of implication of CLEC16A in immune diseases is not clearly established,30,31 and it is possible that other genes in the region are responsible for the association of this locus with immune diseases. It is interesting that Chromosome Conformation Capture studies have demonstrated physical proximity of the promoter of the nearby DEXI gene and intron 19 of CLEC16A. This intronic region, where several of the disease-associated variants are located, is highly enriched in transcription binding sites.32 In the same report, it was shown that alleles of CLEC16A that confer protection from type 1 diabetes mellitus and multiple sclerosis were associated with increased expression of DEXI in 2 human monocyte data sets, leading to the conclusion that DEXI may be an unappreciated autoimmune candidate gene. In addition to DEXI, 2 nearby genes SOCS1 and CIITA are other possible immunity and inflammation-related candidates. In our study, we observed an association with the diameters but not with the other carotid parameters, including IMT. The internal and external diameters of the common carotid artery are stable and are usually considered as structural parameters, which mostly depend on the extracellular matrix content of the vascular wall. However, they may be affected by physiological or pharmacological changes in vasomotor tone via endothelium-related mechanisms. If the influence of CLEC16A on arterial diameter was confirmed, the possible implication of a vasomotor mechanism would require investigation. It should be noted that in the Strong Heart Family Study, no linkage of carotid diameter with markers located in the region encompassing CLEC16A on chromosome 16 was observed.29
Our results concur with those of other reports33,34 in showing that rare coding variants, identified in a healthy population and included on the exome array, exert little if any effect on the traits investigated. This may be related to an insufficient power because the published studies in which this array was used are of limited sample size compared with the most recent meta-analysis of common variants. A much larger sample size may be needed35 to assess the impact of rare variants on vascular phenotypes.
In this study, we show that coding and GWAS-identified variants present on the Illumina exome array contribute in a significant way to the heritability of plasma lipids and of parameters assessing the geometry of carotid arteries. On the other hand, they exert no influence on blood pressure and carotid stiffness. Our results replicate several previously reported associations with plasma lipids involving common variants, but rare variants analyzed at the gene or gene set levels did not significantly influence the traits studied here. The only significant association involving vascular parameters concerned carotid internal diameter. The candidate region implicated on chromosome 16 encompasses the CIITA/DEXI/CLEC16A/SOCS1 genes and is well-known for its associations with several immune disorders. If this association was confirmed, it would suggest an implication of immune-related genes in carotid vasomotricity.
We thank Erwan Bozec for providing the technicians with echotracking training, Dr M.F. Eprinchard, Dr J.M. Kirzin, and all the medical and technical staff of the IPC Center, the Centre de Ressources Biologiques de l’Hôpital Européen Georges Pompidou staff (C. de Toma and B. Vedie), and the Platform for Biological Resources (PRB) of the Hôpital Européen Georges Pompidou for the management of the biobank. The Paris Prospective Study 3 is organized under an agreement between INSERM and the IPC Center and between INSERM and the Biological Research Center at the Européen Georges Pompidou Hospital, Paris, France. We thank the Caisse Nationale d’Assurance Maladie des Travailleurs Salariés (CNAM-TS, France) and the Caisse Primaire d’Assurance Maladie de Paris (CPAM-P, France) for helping make this study possible. We thank Mary Osborne-Pellegrin for help in article editing.
Sources of Funding
The Paris Prospective Study 3 (PPS3) Study was funded by grants from The National Research Agency (ANR), the Research Foundation for Hypertension (FRHTA), the Research Institute in Public Health (IRESP), and the Region Ile de France (Domaine d’Intérêt Majeur). This PPS3 substudy was supported by the Agence Nationale de la Recherche (ANR-09-GENO-010), the Region Lorraine, and the Communauté Urbaine du Grand Nancy.
↵* C. Proust and Drs Empana and Boutouyrie contributed equally to this work as first co-authors.
↵† Drs Jouven, Cambien, and Lacolley contributed equally to this work.
The Data Supplement is available at http://circgenetics.ahajournals.org/lookup/suppl/doi:10.1161/CIRCGENETICS.114.000979/-/DC1.
- Received December 1, 2014.
- Accepted June 24, 2015.
- © 2015 American Heart Association, Inc.
- Benetos A,
- Adamopoulos C,
- Bureau JM,
- Temmar M,
- Labat C,
- Bean K,
- et al
- Laurent S,
- Cockcroft J,
- Van Bortel L,
- Boutouyrie P,
- Giannattasio C,
- Hayoz D,
- et al
- Zieman SJ,
- Melenovsky V,
- Kass DA.
- Lacolley P,
- Challande P,
- Osborne-Pellegrin M,
- Regnault V.
- Galmiche G,
- Labat C,
- Mericskay M,
- Aissa KA,
- Blanc J,
- Retailleau K,
- et al
- Tarasov KV,
- Sanna S,
- Scuteri A,
- Strait JB,
- Orrù M,
- Parsa A,
- et al
- Mitchell GF,
- Verwoert GC,
- Tarasov KV,
- Isaacs A,
- Smith AV,
- Beygui F,
- Wild PS,
- Zeller T,
- Germain M,
- Castagné R,
- Lackner KJ,
- et al
- 18.↵Team TRC. R: A Language and Environment for Statistical Computing [Internet]. R Foundation for Statistical Comptuing. 2013; Available from: http://www.R-project.org/.
- Munroe PB,
- Barnes MR,
- Caulfield MJ.
- Mitchell GF,
- DeStefano AL,
- Larson MG,
- Benjamin EJ,
- Chen MH,
- Vasan RS,
- et al
- Fox CS,
- Polak JF,
- Chazaro I,
- Cupples A,
- Wolf PA,
- D’Agostino RA,
- et al
- Bis JC,
- Kavousi M,
- Franceschini N,
- Isaacs A,
- Abecasis GR,
- Schminke U,
- et al
- Davison LJ,
- Wallace C,
- Cooper JD,
- Cope NF,
- Wilson NK,
- Smyth DJ,
- et al
- Holmen OL,
- Zhang H,
- Zhou W,
- Schmidt E,
- Hovelson DH,
- Langhammer A,
- et al
Increased arterial stiffening is a hallmark of the aging process and the consequence of many disease states, such as obesity, diabetes mellitus, atherosclerosis, and chronic renal disease. Aortic stiffness has independent predictive value for total and cardiovascular mortality. Arterial stiffness is defined by a reduction in arterial distensibility and changes in vasomotor tone. In a population-based study including 2283 men and 1398 women, we assess the contribution of common and rare putatively functional genetic variants (most of them coding) present on the Illumina exome Beadchip to the variability of plasma lipids and parameters of arterial stiffness. A large fraction of the heritability of plasma lipids (triglycerides and high-density lipoprotein cholesterol) and vascular parameters (external and internal diameter, carotid stiffness, intima-media thickness, and wall cross sectional area) is attributable to variants present on the array. By testing 32 827 common (minor allele frequency >0.01) and 68 770 rare variants, we found that plasma lipids were associated with common variants located in known candidate genes, but no implication of rare variants could be established. The only significant association involving vascular parameters concerned carotid internal diameter. A significant association was observed between the common rs2903692 polymorphism of the CLEC16A gene and the internal diameter. The region with the CLEC16A gene on chromosome 16 encompasses the CIITA/DEXI/CLEC16A/SOCS1 genes, and this region has been associated with several immune disorders, stressing the implication of immunity mechanisms in cardiovascular diseases. If this association is confirmed in clinical studies, it would suggest an implication of immune-related genes in carotid stiffness.