Genetic loci on chromosome 5 are associated with circulating levels of interleukin-5 and eosinophil count in a European population with high risk for cardiovascular disease

Highlights • Two loci identified for IL-5 levels on chromosomes and 14.• One locus on chromosome 5 was identified for eosinophil count.• The chromosome 5 loci are in close proximity, but independent and non-overlapping.• IL-5-associated SNPs were not associated with carotid intima-media thickness.• Eosinophil-associated SNPs were not associated with carotid intima-media thickness.

Eosinophil count RAD50 a b s t r a c t IL-5 is a Th2 cytokine which activates eosinophils and is suggested to have an atheroprotective role. Genetic variants in the IL5 locus have been associated with increased risk of CAD and ischemic stroke. In this study we aimed to identify genetic variants associated with IL-5 concentrations and apply a Mendelian randomisation approach to assess IL-5 levels for causal effect on intima-media thickness in a European population at high risk of coronary artery disease. We analysed SNPs within robustly associated candidate loci for immune, inflammatory, metabolic and cardiovascular traits. We identified 2 genetic loci for IL-5 levels (chromosome 5, rs56183820, BETA = 0.11, P = 6.73E À5 and chromosome 14, rs4902762, BETA = 0.12, P = 5.76E À6 ) and one for eosinophil count (rs72797327, BETA = À0.10, P = 1.41E À6 ). Both chromosome 5 loci were in the vicinity of the IL5 gene, however the association with IL-5 levels failed to replicate in a meta-analysis of 2 independent cohorts (rs56183820, BETA = 0.04, P = 0.2763, I 2 = 24, I 2 À P = 0.2516). No significant associations were observed between SNPs associated with IL-5 levels or eosinophil count and IMT measures. Expression quantitative trait analyses indicate effects of the IL-5 and eosinophil-associated SNPs on RAD50 mRNA expression levels (rs12652920 (r2 = 0.93 with rs56183820) BETA = À0.10, P = 8.64E À6 and rs11739623 (r2 = 0.96 with rs72797327) BETA = À0.23, P = 1.74E À29 , respectively). Our data do not support a role for IL-5 levels and eosinophil count in intima-media thickness, however SNPs associated with IL-5 and eosinophils might influence stability of the atherosclerotic plaque via modulation of RAD50 levels.

Introduction
Atherosclerosis is a process leading to coronary artery disease (CAD) and ischemic stroke, and is among the leading causes of mortality and morbidity throughout the world [1]. It is thought that a key component of atherosclerosis is lipid accumulation in the arterial wall, leading to leukocyte recruitment, inflammation http://dx.doi.org/10.1016/j.cyto.2016.01.007 1043-4666/Ó 2016 Elsevier Ltd. All rights reserved.
Abbreviations: IMT mean-max , mean of the maximum IMT measures along the entire carotid tree; IMT-CC mean , mean of IMT measures in the common carotid; IMT-CC max , max of IMT measures in the common carotid; IMT-Bif mean , mean of IMT measures in the carotid bifurcation; IMT-Bif max , max of IMT measures in the carotid bifurcation. and participation of both innate and adaptive immune responses. To support the immune responses hypothesis, a number of immune or inflammatory biomarkers for cardiovascular disease (CVD) have been investigated. Some, such as C-reactive protein (CRP), appear to reflect general disease processes [2,3], whereas a number of cytokines (such as IL-1, IL-2, IL-6, IL-8, IL-10, IL-12, IL-17 and TNFa) are robustly associated with CVD and are thought to be actively involved in the inflammatory processes leading to atherosclerosis [4,5].
Genetic studies have identified single nucleotide polymorphisms (SNPs) that are associated with increased risk of CAD in the interleukin-5 gene (IL5) [6] and increased risk of ischemic stroke in the interleukin-5 receptor alpha subunit gene (IL5RA) [7]. As with most genetic variants associated with CVD, the mechanism by which the IL5 SNP influences disease is unclear. A number of studies (including one from our group) have also proposed an athero-protective role for IL-5 [8][9][10][11], which is supported by the fact that it is produced mainly by anti-inflammatory T helper 2 (Th2) cells.
In addition to activated Th2 cells, IL-5 can be produced by mast cells and to a lesser extent by eosinophils. IL-5 is a growth and differentiation factor for B-cells, an immunoglobulin-A (IgA)enhancing factor and a key cytokine in eosinophil maturation, differentiation, activation and survival [12,13]; thus, it is interesting that the IL5 locus has also been associated with eosinophil count [14]. Cellular responses are dependent upon IL-5 activating a dimeric receptor consisting of an a and a b-subunit. The b-subunit is shared with IL-3 and granulocyte-macrophage colony-stimulating factor (GM-CSF) [15], however the a-subunit is specific to IL-5. Thus, availability of the a-chain on the cell surface controls responsiveness of the cell to IL-5 [16,17].
Thickness of the intima-media layer of the carotid artery can be considered a marker of cardiovascular risk. Differences in carotid intima-media thickness (IMT) over time can be indicative of changes in the structure, and possibly functions, of the vessel wall. The aims of this study were to identify genetic variants associated with IL-5 levels and to use these in a Mendelian randomisation approach to assess IL-5 levels for causal effects on carotid IMT. As eosinophil count and IL-5 levels are inter-related, we also investigated the genetic regulation of eosinophil count. We used the prospective IMPROVE study (the Carotid Intima Media Thickness [IMT] and IMT-Progression as Predictors of Vascular Events in a High Risk European Population study), which is based on 3703 subjects at high risk of CVD with extensive genetic data, biochemical phenotyping and repeat detailed ultra-sound measurements of the carotid artery IMT [18,19]. Genetic markers robustly associated with increased levels of a biomarker or constituting a causal component of increased IMT should be useful for risk stratification and early prevention of CVD.

Discovery cohort: the IMPROVE cohort
The design features, biobank and dataset of the IMPROVE study have been reported [18,19]. Of 3703 participants with at least three cardiovascular risk factors but without symptoms or diagnosed prevalent disease, 3435 passed phenotypic quality control and were considered for the current study. Ethics committee approval for the study was obtained from each centre and written informed consent was obtained from all participants.

Phenotypes
A wide range of anthropometric and biochemical measurements were acquired, including fasting concentrations of HDL and LDL cholesterol (by Friedewald's formula), triacylglycerol (TG), high sensitivity CRP (hs-CRP), creatinine and plasma glucose as described [19]. Diabetes was defined as diagnosis of diabetes and/or fasting blood glucose P7 mmol/L and/or glucose lowering treatment. Levels of IL-5 were measured in EDTA-plasma samples using the ultra-sensitive kit for human IL-5 (MesoScale Discovery, Gaithersburg, MD, USA) following instructions provided by the manufacturer. Details of the high-resolution ultrasound measurements of carotid IMT at baseline and follow-up have been reported [18,19]. A number of studies suggested that atherosclerotic changes can be site-specific and have distinct clinical implications [20][21][22]. Thus one composite and four segment-specific measurements of carotid IMT by were included in the study: IMT mean-max (mean of the maximum IMT measures along the entire carotid tree), IMT-CC mean and IMT-CC max (mean and maximum of common carotid segment), IMT-BIF mean and IMT-BIF max (mean and maximum of the bifurcation). The ultrasonographic measurements were obtained at baseline and after 15 and 30 months. These three time points were used in linear regression to assess change over time in IMT [18], referred to here as progression of IMT. Characteristics of the IMPROVE cohort with available genetic and phenotypic data (n = 3435) are presented in Table 1.

Genetic analysis
Genotyping was performed using Illumina 200 K Cardio-Metabochip [23] and Immunochip [24] genotyping arrays (hence referred to as the combined chip). Whilst the combined chips do not have genome-wide coverage, they do have good coverage of candidate or robustly associated loci for immune, inflammatory, metabolic and cardiovascular traits, which allows for a less conservative threshold for significance compared to genome-wide platforms. Genotyping and calling were performed at the SNP Technology Platform (Uppsala University, Sweden Uppsala). Standard genetic quality control procedures were followed with exclusions for SNPs failing call rate (<95%) and Hardy-Weinberg (p < 5E À7 ) checks or with minor allele frequency (MAF) <1% and for subjects with cryptic relatedness or sex check failures. After quality control, 3433 individuals with genetic and phenotypic data were included in the analysis.
The PROCARDIS cohort consists of cases and controls from the UK, Sweden, Italy and Germany. All cases were diagnosed with CAD before the age of 66 years and most (80%) had a sibling with CAD before 66 years. Controls were recruited from the same populations and did not have a diagnosis of CAD, or a sibling with diagnosis of CAD, before 66 years of age. Plasma levels of IL-5 were measured in duplicate on the same plate by a sandwich immunoassay with electrochemiluminescence detection using a MesoScale Diagnostics platform (Gaithersburg, Maryland, USA). Genomewide genotyping of PROCARDIS was conducting using the Illumina 1 M or 610 K platforms. Genotype calling was performed at the Centre National du Genotypage, Paris or the Uppsala SNP Genotyping platform, Sweden. Standard quality control was conducted as detailed above.
The Malmö Diet and Cancer (MDC) study is a prospective population-based cohort study including 28,449 subjects recruited during 1991-1996. Subjects aged 45-69 years, living in Malmö, Sweden were eligible for participation. Between November 1991 and February 1994, every second enrolled subject was also invited to take part in a substudy of the epidemiology of carotid artery disease. This MDC-CardioVascular cohort consists of 6103 subjects (60% women), 5540 of whom also agreed to have blood collected under standardized fasting conditions. MDC was genotyped using the Illumina OmniExpressExome chip and standard quality control was conducted as detailed above. The methodology used for measuring IL-5 levels has previously been described [28]. Briefly, levels of cytokines released by mononuclear cells after stimulation (with CD3/CD28 beads) were analysed using a multiplex cytokine immunoassay (MesoScale Discovery, Gaithersburg, MD, USA).
Whilst the genotyping platforms utilised by the discovery and replication cohorts are not the same, all 3 platforms are produced by the same company and based upon the same bead chip technology. In addition, the same quality control parameters are used for the genotype calling, as well as for the downstream analyses (as detailed above).

Statistical methods
Differences between two groups according to sex were assessed by Wilcoxon Mann-Whitney ranksum test for continuous variables and Fisher's exact test for binary variables (Table 1). Since a sex difference in IL-5 levels has been observed [11], we adjusted for sex. Risk factors and biomarkers significantly associated with plasma levels of IL-5 (by Spearman rank correlation) were considered for inclusion in regression models. Variables with skewed distributions were log-transformed prior to regression analysis. In both the analyses of IL-5 and of eosinophil count, the basic model was adjusted for population structure (as assessed by the first 3 multi-dimensional scaling components), age and sex. For IL-5 analyses, the extended model also included LDL and hsCRP. For eosinophil count, the extended model included BMI, fasting proinsulin concentration and white blood count. Variables for inclusion in the extended model were determined by backward stepwise selection in a multiple regression model. Linear regression analysis was used to assess associations between SNP genotypes and plasma IL-5 concentrations or eosinophil count, assuming an additive genetic model. The threshold for replication was set at p < 1.00E À5 , whereas array-wide significance was set at p < 1.92E À7 (based on the calculation of the number of uncorrelated SNPs on the combined Metabo-Immuno chips as previously described [29]). Linear regression assuming an additive genetic model was also used for analysis of IMT measures, with adjustment for age, sex and population structure. Further adjustment for established CVD risk markers included age, sex, population structure, BMI, SBP, HDL, TG, hsCRP, current smoking and type 2 diabetes.
Replication analyses were conducted in the same manner as the discovery analyses, but the models were amended for PROCARDIS, where country was used instead of MDS components and casecontrol status was included as a covariate. Statistical analyses were performed in STATA (Statacorp, Houston, TX, USA) except chipwide genetic analyses which were conducted using PLINK [30]. Because of high correlation between IMT measurements of carotid segments [19], correction for testing multiple phenotypes was not applied.

Gene expression analysis
Differential gene expression analysis for SNPs robustly associated with plasma IL-5 concentrations was performed. Firstly, in the Advanced Study of Aortic Pathology (ASAP) [31], where aortic adventitia (n = 132), aortic media (n = 137) and heart (n = 127) tissue samples were analysed. In brief, tissue samples were collected from patients undergoing elective surgery for aortic aneurysm and/ or aortic valve repair. RNA was extracted from biopsy material and hybridised to Affymetrix ST 1.0 exon arrays (Santa Clara, CA, USA). DNA was extracted from whole blood and genotyped on the Illumina 610 W-Quad Bead Array (San Diego, CA, USA). SNPs with call rate >95% were used for imputation, and imputed SNPs with quality scores of MACH <0.3 were excluded. The ASAP study was approved by the ethics committee at Karolinska Institutet. Associations between SNP genotypes and gene expression were assessed assuming an additive genetic model. Significance was set at p < 0.0056 (Bonferroni-correction for 3 SNPs in 3 tissues). Secondly the publicly available MuTHER gene expression data was also interrogated [32]. Of the IL-5 or eosinophil-associated SNPs identified in IMPROVE, 5 were available in the MuTHER dataset, thus p < 0.01 (Bonferroni-correction for 5 SNPs in one tissue) was set for the level of significance. Finally, the publically available rSNPBase [33] was searched effects of SNPs on gene expression. As the data in this resource is qualitative, no threshold for significance was employed.

Cardiovascular, metabolic, inflammatory or immune-associated SNPs vs IL-5 levels or eosinophil count
In order to assess whether there are SNPs in loci previously associated with immune, inflammatory, metabolic or CVD traits which influence plasma levels of IL-5 or eosinophil count, the combined chip was analysed. When considering IL-5 levels, there were no genetic variants that demonstrated array-wide significant associations (i.e. p < 1.92E À7 ), however 2 loci reached the threshold set for replication (p < 1.00E À5 , Fig. 1, Upper panel and Table 2). Further adjustment using the extended model had a negligible effect on the results ( Table 2). The chromosome 14 locus consists of a single SNP (Supplementary Fig. 1), whilst the chromosome 5 locus (defined by the positions of the first and last SNPs reaching p < 1.00E À5 in the region) encompasses 79 SNPs extending across and upstream of the IL5 gene. When considering associations with eosinophil count, one locus, including 14 SNPs, reached the threshold for replication but not array-wide significance (Fig. 1, Lower panel and Table 2) in both the basic and extended models ( Table 2). This locus is also on chromosome 5 and extends across the IL5 gene.

Genetic architecture of IL5 and eosinophil-associated loci
Given that only a single SNP was genotyped in the chromosome 14 locus, it is hard to say anything about the genetic architecture of this locus. More genotyping in this region would be needed to further explore this locus.
The 2 loci on chromosome 5 discovered here are located in physical proximity to each other, however the SNPs in the 2 loci are distinct and appear to be independent from each other by linkage disequilibrium (LD, Fig. 2, Upper and Lower panels, Supplementary Fig. 2). The 3 genetic markers in the vicinity of the IL5 gene reported in earlier studies were independent from each other and the loci identified in the current study as assessed by LD ( Supplementary Fig. 2), however the eosinophil-associated SNP (rs4143832) and the CVD-associated SNP (rs2706399) fell within our eosinophil-associated locus. To further test the independence of the 2 loci, conditional analysis was conducted, whereby the IL-5 analysis was adjusted for the lead eosinophilassociated SNP and vice versa. As demonstrated in Table 3, there is negligible attenuation of the effect of the IL-5-associated SNPs on IL-5 levels when including the lead eosinophil-associated SNP as a covariate. The same is observed if the reverse is attempted, thus we conclude that the loci are independent.

Replication of IL5-associated loci
Ideally we would have liked to replicate both the IL-5 and eosinophil findings but due to lack of applicable cohorts it was only possible to attempt replication for the IL-5-assocaited SNPs. Characteristics of the replication cohorts are presented in Supplementary Table 1. Replication was attempted in 2 independent European cohorts for the IL-5-associated loci, namely rs56183820 on chromosome 5 and rs4902762 on chromosome 14. The size of effect in the replication meta-analysis (Table 4) was similar to that seen in the discovery analysis, however the direction was inconsistent and the associations were solidly non-significant.

IL-5 and eosinophil-associated SNPs in relation to IMT
Our aim was to use a Mendelian randomization approach to assess whether the IL-5 or eosinophil-associated loci influenced measures of IMT. As we were not able to replicate the association of the chr 5 locus with IL-5 levels, conducting a Mendelian randomization experiment was unjustified. However, we did assess the effect of these SNPs on IMT measures. Considering 6 independent loci (3 previously reported SNPs, 1 eosinophil-associated locus and 2 IL-5-associated loci), the Bonferroni-corrected significance threshold was set at p < 0.0083. No associations were observed between the SNPs and baseline measures of IMT (Supplementary Table 2). Eosinophil-associated SNPs demonstrated nominally significant associations with progression of IMT-CC max , but only rs2706399 (the G allele of which was previously associated with increased CVD risk [6]) met the Bonferonni-corrected p for significance (G allele, Beta À0.008, 95%CI (À0.012, À0.004), p = 0.0017, Supplementary Table 3). Further adjustment for emerging or established CVD risk markers had negligible effect on the associations (Supplementary Table 3). Inclusion of eosinophil count in the model rendered the eosinophil-associated SNPs nonsignificant, but did not have an effect on the association of rs2706399 (CVD-associated) with IMT-CC max progression (Supplementary Table 3).

Expression QTL
To identify functional effects of the IL-5 or eosinophilassociated loci, SNPs which met the threshold for p < 1.00E À5 were examined for influence on expression of genes at the locus. In the ASAP study, aorta (adventitia and media) and heart tissues were examined. A region 250 kb up and downstream of the lead SNP was defined for each locus, and genes within this region were assessed for genotype-specific expression. Bonferroni correction was made for 3 tissues and 4 loci (one on chromosome 14, 3 on chromosome 5, p for significance <0.0042, ( Supplementary  Fig. 2)). The chromosome 14 region included 2 genes, CTAGE5 and FBXO33 and rs4902762 was associated with CTAGE5 expression levels in heart tissue (p = 0.0073), but this did not reach statistical significance. The chromosome 5 eosinophil-and IL-5associated loci overlapped to a large degree, and included many genes (P4HA2 and PDLIM4 (eosinophil count locus), SLC22A4, SLC22A5, IRF1, IL5, RAD50, IL13, IL4, KIF3A and SEPT8 (both loci), SHROOM1, GDF9, UQCRQ, LEAP2 and AFF4 (IL-5 locus). Rs12652920 was associated with RAD50 expression in aortic media (p = 0.0093), but this did not reach the level of statistical significance. The MuTHER lymphocyte data [32] supports a positive effect of the rs4705959 T allele (associated with increased eosinophil count) on mRNA expression of RAD50 and SLC22A5, but a negative effect on expression of IL5 (Table 5). The C allele of rs12652920 (associated with increased IL-5 levels) was also associated with increased mRNA expression of RAD50 (Table 5). The rSNPBase was queried to identify proxies (r2 > 0.8) for the lead SNPs and whether they have been reported to influence expression of genes. For the IL-5-associated locus (rs56183420) 63 proxies were identified, of which 47 reported eQTL effects on RAD50 but only 24 for IL5. Of note, 35 of these SNPs were available in the IMPROVE cohort, all showed associations (p < 1.00E À4 ) with IL-5 levels, but not eosinophil count. For the eosinophil-associated locus (rs72797327) 4 proxies were identified, of which 2 each were reported for effects on IL-5 and RAD50 expression, but all 4 reported effects on SEPT8 expression. Of these SNPS, all 3 that were available in IMPROVE showed strong (p < 1.00E À5 ) effects on eosinophil but not IL-5 levels. Considering the chromosome 14 locus, 23 proxies were identified with 5 influencing CTAGE5 expression and 7 influencing FBXO33 levels. Only one proxy was available in IMPROVE and was associated with IL-5 levels but not eosinophil count.

Discussion
This study identified genetic regulators of plasma IL-5 levels and eosinophil count, however the lack of replication of the genetic associations precludes the ability to formally assess the causality of IL-5 levels on IMT. To the best of our knowledge this is the largest study with measurement of IL-5 levels and the first to investigate genetic regulation of IL-5 levels.
The genetic loci associated with IL-5 levels and eosinophil count are in close proximity to each other and in the vicinity of the IL5 gene on chromosome 5, but act independently. Earlier studies identified SNPs in the region of the IL5 and IL5a genes as being associated with inflammatory bowel disease (rs2188962 [34]), eosinophil count (rs4143832 [14]) and coronary artery disease (rs2706399 [6]). These variants were present on the combined chip and thus were also analysed. No association was observed between rs2706399 and either IL-5 levels or eosinophil count (Supplementary Table 5). In contrast, rs2188962 demonstrated nominal associations and rs4143832 demonstrated convincing associations with IL-5 levels and eosinophil count (Supplementary Table 5). In agreement with Gudbjartsson et al., the minor allele of rs4143832 was associated with eosinophil count [14] as well as IL-5 levels. Of note, whilst these SNPs demonstrate some significance in relation to levels of IL-5 or eosinophil count, they do not reach the preset level of significance for being forwarded to replication analysis. Rs4143832 and rs2706399 are located within the eosinophilassociated locus reported here but are independent from the SNPs studied here (by LD, Supplementary Table 2).
Whilst SNPs at the loci identified here do show some associations with IL5 mRNA levels, expression of the RAD50 gene demonstrates stronger and more consistent associations across the loci. The results from rSNPBase are less clear, and whilst they do support RAD50 as a possible candidate, they suggest that SEPT8 (chromosome 5) and FBXO33 (chromosome 14) might also be relevant. Recently, up-regulation of RAD50 and other double-strand repair proteins was reported in vascular smooth muscle cells from atherosclerotic plaque [35]. The authors suggest that these proteins are associated with plaque stability rather than atherosclerosis. Thus it is plausible that the IL5 locus influences plaque stability via RAD50-associated mechanisms rather than IL-5-associated mechanisms. This speculative mechanism fits with the IL5 locus being associated with MI [6] and unstable angina [36], typically resulting from plaque rupture, but not with IMT, which is indicative of vessel wall remodelling.
Eosinophil count and IL-5 levels are closely inter-connected: IL-5 is essential for eosinophil maturation and differentiation [37]. Conversely, eosinophil-specific granules store IL-5 [38], indicating that these cells might also contribute to increasing circulating IL-5 levels. In IMPROVE, there is a strong correlation between IL-5 concentrations and eosinophil count (Spearmans rho = 0.560, p < 0.001) [11]. Thus it is interesting that these 2 variables are regulated by independent loci in the same genetic region. One possibility is that a common factor, potentially RAD50, regulates both IL-5 levels and eosinophil count. An alternative scenario is that, given the roles of IL-5 and eosinophils in inflammation, upregulation of RAD50 might be a DNA-protective stress response mechanism. This would explain the puzzling finding that the allele associated with increased eosinophil count shows increased RAD50 and the inverse effect on IL5 gene expression, whilst the allele associated with increased IL-5 levels also shows increased RAD50 gene expression. It is of note that 2 of the top eQTL SNPs for RAD50 [39] are those with the strongest associations with eosinophils (rs11739623 and rs4705959) in IMPROVE. Previous reports have demonstrated the use of studying specific immune cell types to explore immune or inflammation-related diseases [40,41]. Whilst the relationship between the SNPs, RAD50 expression, IL-5 levels and eosinophil counts requires clarification, these findings are of potential relevance for eosinophil-associated disorders.
Whilst this is the largest study with detailed IMT measures, plasma IL-5 concentrations and extensive genetic data for all subjects, there are limitations to the study. This study was conducted in subjects with at least 3 classical CVD risk factors, thus findings might not pertain to other populations. A sex difference in IL-5 levels has been reported [11], thus sex-stratified analyses would be of interest. However, this approach has not been used here, for two reasons: the distribution of outliers (P1.5 interquartile range above the upper quartile) between men and women might explain the sex difference in levels. Because of this and the loss of power when stratifying the analyses, we believe that the chance of detecting false positive signals is too high for this analysis to be robust. The lack of Th2 count or measurements of Th2 cytokines mean that we cannot conclusively determine the origin of IL-5 and whether Th2 responses are activated in these subjects. However, it has previously been shown that the percentage of Th2 cells has only a very weak (non significant) correlation with IL-5 levels released by activated leukocytes [28]. The genotyping chips employed in this study are based on a large scale candidate gene/loci experiment, which allows a less stringent significance threshold for the discovery analysis, however the whole genome is not represented. Thus we cannot comprehensibly describe the genetic architecture of the IL-5-associated locus on chromosome 14, nor can we exclude the possibility that true IL-5-influencing genetic variants exist which lie within areas of the genome not examined in this study.

Conclusions
This study identified genetic loci on chromosome 5 associated with IL-5 levels and eosinophil count. Whilst these loci are close together in the vicinity of the IL5 gene, they are independent. Furthermore, expression analysis suggests RAD50, rather than IL5, as a prime candidate gene for these associations. An additional locus associated with IL-5 levels was observed on chromosome 14. This report suggests that these loci have negligible influence on IMT measures.   Where: EA, Effect allele; EAF, effect allele frequency; The rs4705959 T allele was associated with increased eosinophil levels; The rs12652920 C allele was associated with increased IL-5 levels.
the Swedish Heart-Lung Foundation and Stiftelsen för Gamla