If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Department of Nephrology and Hypertension, Friedrich‐Alexander Universität Erlangen‐Nürnberg, Erlangen, GermanyDepartment of Nephrology and Medical Intensive Care, Charité—Universitätsmedizin Berlin, Berlin, Germany
Corresponding author. Institute of Genetic Epidemiology, Department of Genetics, Medical University of Innsbruck, Schoepfstrasse 41, A-6020, Innsbruck, Austria.
HDL-C and triglycerides are the main biochemical determinants of cholesterol efflux capacity each explaining >11% of variance.
•
In a GWAS, 2 novel genetic loci (KLKB1 & CLSTN2) and the known APOE/C1 locus are associated with cholesterol efflux capacity.
•
The KLKB1 locus is significant irrespective of adjustment for kidney function, HDL-C, triglycerides and apolipoprotein A-IV.
•
The association with the CLSTN2 locus is suggestive and reaches genome-wide significance with adjustment for triglycerides.
Abstract
Background and aims
HDL-mediated cholesterol efflux capacity (CEC) may protect from cardiovascular disease. Thus, we aimed to identify its genetic and non-genetic determinants.
Methods
We measured CEC to 2% apolipoprotein B-depleted serum using BODIPY-cholesterol and cAMP-stimulated J774A.1 macrophages using serum samples from 4,981 participants in the German Chronic Kidney Disease (GCKD) study. Variance of CEC explained by clinical and biochemical parameters in a multivariable linear regression model was calculated by proportional marginal variance decomposition. A genome-wide association study with 7,746,917 variants was performed based on an additive genetic model. The main model was adjusted for age, sex and principal components 1-10. Further models were selected for sensitivity analysis and to reduce residual variance by known CEC pathways.
Results
Variables that explained 1% and more of the variance of CEC were concentrations of triglycerides (12.9%), HDL-cholesterol (11.8%), LDL-cholesterol (3.0%), apolipoprotein A-IV (2.8%), PCSK9 (1.0%), and eGFR (1.0%). The KLKB1 (chr4) and APOE/C1 (chr19) loci were genome-wide significantly (p < 5x10−8) associated with CEC in our main model (p = 8.8x10−10 and p = 3.3x10−10, respectively). KLKB1 remained significantly associated after additional adjustment for either kidney parameters, HDL-cholesterol, triglycerides or apolipoprotein A-IV concentrations, while the APOE/C1 locus was not significantly associated anymore after adjustment for triglycerides. Adjustment for triglycerides also revealed an association with the CLSTN2 locus (chr3; p = 6.0x10−9).
Conclusions
We identified HDL-cholesterol and triglycerides as the main determinants of CEC. Furthermore, we newly found a significant association of CEC with the KLKB1 and the CLSTN2 locus and confirmed the association with the APOE/C1 locus, likely mediated by triglycerides.
Foam cells are a hallmark of atherosclerotic cardiovascular disease (ASCVD). They develop from macrophages due to accumulation of excess amounts of cholesterol. HDL-mediated cholesterol efflux capacity (CEC) is the ability of HDL particles to induce efflux of cholesterol via transporters such as ABCA1, ABCG1 and SR-BI and via passive diffusion from cells and can be measured in vitro using a cell-based assay [
]. Cholesterol efflux represents the initial step of reverse cholesterol transport (RCT) from macrophages, a process that may prevent and reverse atherosclerosis [
]. Indeed, CEC is inversely associated with prevalent and incident ASCVD independently of HDL-cholesterol (HDL-C) in most cohorts investigated so far [
The biological function of HDL particles is complex. Despite HDL-C concentrations being inversely associated with prevalent and incident ASCVD, Mendelian randomization studies have indicated that this relationship might not be causal [
]. Further, CEC is the initial step of macrophage RCT but it is debated which steps of this process may be necessary to prevent and reverse atherosclerosis [
]. To elucidate the complexity of HDL particles, we need to know the biological processes governing the different HDL functions and steps of the macrophage RCT. Identification of genetic determinants of CEC may help to understand the involved biological processes.
So far, only few aspects of the genetic determinants of CEC are known. A family-based study estimated that CEC is 13% heritable, independently of HDL-C [
]. A genome-wide association study (GWAS) in 5,293 French Canadians using cAMP-stimulated J774 A.1 macrophages found a significant association with CETP and APOE/C1/C2/C4, but this was mediated by HDL-C and triglycerides [
]. This study also used other CEC measures (from unstimulated J774A.1 macrophages, stimulated ABCA1-expressing BHK cells and ABCA1-dependent CEC from J774 A.1 macrophages) that were partly associated with other loci (CETP, LIPC, LPL, APOA1/C3/A4/A5, APOE/C1/C2/C4, PPP1CB/PLB1 and RBFOX3/ENPP7) [
]. Using estimations of cAMP-stimulated CEC by a Bayesian linear regression model of nuclear magnetic resonance data, the association with the CETP locus was replicated and an association with the LIPC locus was also found in 20,372 individuals [
Genetic and observational evidence: No independent role for cholesterol efflux over static high-density lipoprotein concentration measures in coronary heart disease risk assessment.
]. Further GWAS are needed to confirm the observed associations and to identify additional loci.
In addition to understanding CEC under physiologic conditions, the study of CEC may additionally be informative in individuals at high risk of ASCVD such as people with chronic kidney disease (CKD). Individuals with CKD have an altered lipid metabolism, with pronounced changes in HDL composition and decreased CEC in individuals with end-stage kidney disease on dialysis treatment [
]. GWAS in individuals with CKD may thus identify novel modulators of CEC that are especially relevant in CKD. In addition, while HDL-C and triglycerides have been found as main non-genetic determinants of CEC in the general population [
], little is known about determinants of CEC in moderate CKD.
In the present study we therefore i) measured CEC of cAMP-stimulated J774A.1 macrophages in serum of 5,171 participants of the German Chronic Kidney Disease (GCKD) study, ii) determined clinical and biochemical parameters independently associated with CEC and the variance in CEC that they explain and iii) performed a GWAS, identified independent signals at the significant loci and calculated the variance explained by these loci.
2. Materials and methods
2.1 Study description
The German Chronic Kidney Disease (GCKD) study is a multi-center study in 5,217 Caucasian individuals with moderately severe chronic kidney disease. The cohort has been described before in detail [
Disease burden and risk profile in referred patients with moderate chronic kidney disease: composition of the German Chronic Kidney Disease (GCKD) cohort.
]. In brief, patients with chronic kidney disease and an estimated glomerular filtration rate (eGFR) of 30–60 mL/min/1.73 m2 (KDIGO stage G3, A1-3) or an eGFR >60 mL/min/1.73 m2 in the presence of overt proteinuria (KDIGO stage G1-2, A3) were enrolled. Overt proteinuria is defined as an urinary albumin-to-creatinine ratio (UACR) of >300 mg/g or a protein-to-creatinine ratio in 24-h urine of >500 mg/g. Exclusion criteria were active malignancy, New York Heart Association (NYHA) stage IV heart failure, renal or any other transplantation, non-Caucasian ethnicity and legal attendance. Non-fasting serum samples with a median (interquartile range) time since last meal of 2.8 (2.0–4.4) hours (no information for 117 individuals in analysis dataset) were collected and stored at −80 °C. CEC measurements and all other laboratory measurements were made in biosamples from the same baseline specimen collection.
The study has been approved by the review boards of the participating institutions and informed consent was obtained from all participants. The study protocol conforms to the Declaration of Helsinki.
2.2 CEC measurement
The CEC assay was performed on sera of all available GCKD samples (n = 5,171) in duplicates per sample as described before in detail [
] provide a detailed protocol including explanatory comments on stock solutions, volumes, cell handling and quality control. In short, 70,000 J774A.1 macrophages were plated per well in 96-well plates. After overnight incubation (37 °C, 5% CO2), cells were stained with 25 μM BODIPY-cholesterol for 1 h and equilibrated for 16-18 h in the presence of 0.3 mM cAMP and 2 μg/mL ACATi. ApoB-depletion was performed from sera stored at −80 °C and thawed on ice overnight by mixing 4 μL of 20% PEG6000 with 10 μL of serum, incubation (20 min, room temperature) and centrifugation (3,220 g, 30 min, 4 °C) immediately before efflux measurement. Subsequently, 2% of the apoB-depleted supernatant was added in duplicates to the macrophages (2.2 μL apoB-depleted serum per well in a total volume of 110 μL efflux medium) and incubated for 4 h (37 °C, 5% CO2). Afterwards, fluorescence in the supernatant and in the cell lysates (prepared with 1% cholic acid by shaking at room temperature at 1200 rpm for 1 h) was determined at excitation/emission 485 nm/530 nm with a SPARK microplate reader (Tecan Group Ltd., Männedorf, Switzerland). After subtraction of background fluorescence, CEC was calculated as the percentage of fluorescence that effluxed in 4 h to the acceptor (2% apoB-depleted sera) relative to the total fluorescence of the well (). The unit of CEC is defined as percent effluxed BODIPY-cholesterol to total BODIPY-cholesterol content of macrophages and named “unit” in the following to avoid mix-up with other percentages. Efflux to medium without acceptor was subtracted. CEC values were normalized for plate-to-plate differences based on four controls on each plate, and assay performance was monitored with two additional controls on each plate. Liquid handling steps that did not require sterile conditions were performed with an automated workstation Biomek i7 (Beckman Coulter GmbH, Vienna, Austria).
We repeated samples in case of a CV of the duplicates of >15.0%, or in case of technical errors. Finally, we had high-quality CEC measurements available from 5,167 individuals with an average intra-assay CV of 3.64%. Correction for plate-to-plate differences based on four controls reduced the inter-assay CV (calculated with two separate controls) for positive control 1 from 14.73% to 7.95% and for positive control 2 from 10.14% to 6.65%, showing that correction improved assay performance. Genotyping data was available from 5,034 individuals, resulting in 4,981 individuals with CEC measurements, age, sex and genotyping data available.
2.3 Further measurements
Triglycerides, LDL and HDL cholesterol were determined by an enzymatic colorimetric test on a P800 analyzer (Modular, Roche Diagnostics, Rotkreuz, Switzerland). PCSK9 concentrations were quantified with a commercial human PCSK9 ELISA kit (R&D Systems, Minneapolis, USA) [
]. Plasma apoA-IV concentrations were measured with a double-antibody ELISA using an affinity-purified polyclonal rabbit anti-human apoA-IV antibody for coating and the same antibody coupled to horseradish peroxidase for detection [
2.4 Association analysis of CEC with baseline characteristics
All statistical analyses were performed with R version 4.0.5 (R Foundation for Statistical Computing, Vienna, Austria) and RStudio version 1.4.1106 (RStudio Team, Boston, USA). To assess differences between groups, Kruskal-Wallis test was used for continuous variables and Chi-square test for categorical variables. Variables that significantly differed between quartiles of CEC were tested for their association with CEC in a multivariable regression model. Variance inflation factors were used to check for multicollinearity. Relative importance of covariates was assessed with the relaimpo package for R by proportional marginal variance decomposition (pmvd) [
]. This method calculates the R2 of all individual covariates based on weighted averaging over sequential R2 that sum to the total R2 of the model. One thousand bootstrapping runs were used to calculate the confidence intervals (CIs).
2.5 Genotyping and imputation
Samples were genotyped with Illumina HumanOmni2.5-8 v1.2 BeadChip (Illumina, GenomeStudio, Genotyping Module Version 1.9.4). Before imputation, quality control was performed [
]. Samples were excluded if the call rate was <0.97, if there was a sex mismatch or if they failed mean heterozygosity, genetic ancestry and cryptic relatedness checks. SNPs were excluded prior to imputation if the call rate was <0.96, if positions were duplicated or if they deviated from the Hard-Weinberg equilibrium (p <10−5). Genotypes were imputed with the Michigan imputation Server (Available at: https://imputationserver.sph.umich.edu/index.html#) [
] based on an additive genetic model. For data reproducibility, we used the nf-gwas Nextflow pipeline that first performs data preparation steps (validation of files, file format conversion, pruning and quality control of genetic data), then runs the GWAS using REGENIE and finally creates an overview of the GWAS results (summary statistics and plots and annotation with the nearest gene of the most significant variants) (version v0.1.14; available at: https://github.com/genepi/nf-gwas). GWAS was performed on CEC adjusted for age, sex and principal components (PC) 1-10 as the main model (model 1). For sensitivity analyses and to reduce residual variance by known pathways that determine CEC, model 2-5 were additionally adjusted for either eGFR and log-transformed UACR values (model 2), HDL-C (model 3), triglycerides (model 4) or apoA-IV concentrations (model 5). Since CEC is altered depending on sex [
], interaction analyses and stratified GWAS were also performed.
For step 1 of REGENIE (fitting of a whole genome regression model to account for population structure and relatedness) we included all genotyped variants with a minor allele frequency of >1%, minor allele count >100, genotype missingness <10% and Hardy-Weinberg equilibrium test p >10−15 pruned for linkage disequilibrium (1,000 variant window, 100 step size and r2 < 0.9) [
]. For step 2 (single-variant association testing conditional on step 1 predictions according to the leave-one-chromosome-out scheme), all imputed SNPs with a minor allele count of 100 and a minimum imputation score of 0.3 were used. A p <5x10−8 was considered genome-wide significant.
2.7 Post-GWAS analysis
Regional plots were generated with LocalZoom (version v0.14.0-beta.2; available at: https://statgen.github.io/localzoom/) with linkage disequilibrium (LD) information from the European population from the 1000 Genomes Project based on human genome build GRCh37 [
] was run to identify independent signals at the loci that were significantly associated with CEC (lead SNP±500 kb). Conditional analysis was performed with estimated LD data from the GCKD study, a p cutoff = 5x10−8 and a collinearity cutoff = 0.9.
] was performed to calculate the variance explained by the significantly associated loci around the lead SNP (±500 kb).
To identify missense or otherwise functionally relevant variants that are in LD with newly identified conditionally independent GWAS results we used the LDproxy tool of LDlink (version 5.2) [
] with LD data from the European 1000 Genomes cohort (SNP±500 kb). Results were filtered for D’ >0.95, annotated by rs-ID using the Ensembl variant effect predictor [
] and further filtered for missense variants, a CADD-PHRED score of >10 or a delta score of SpliceAI (for acceptor or donor gain or loss, respectively) > 0.5. Protein quantitative trait loci (pQTLs) of these SNPs were looked up with the variant annotation module of SNiPA [
3.1 Association of CEC with HDL-C, triglycerides and other baseline characteristics
The clinical characteristics of the patients are given in Table 1. The mean ± SD of CEC in this cohort was 40.3 ± 4.6 units. Several baseline characteristics such as body mass index, prevalence of hypertension and medication with fibrates and immunosuppressive drugs differed significantly between quartiles of CEC (Table 1). In addition, all lipid parameters and kidney function parameters differed significantly between quartiles (Table 1). To test whether these parameters are associated with CEC independently of other baseline parameters, multivariable linear regression was performed including all significant variables (Table 2). Age, sex, hs-CRP, frequent alcohol consumption, and medication with immunosuppressive drugs were not independently associated with CEC and thus excluded in the multivariable model except for age and sex. Variables which explained 1% and more of the variance of CEC were triglycerides (12.9%), HDL-C (11.8%), LDL-C (3.0%), apoA-IV concentrations (2.8%), eGFR (1.0%) and PCSK9 concentrations (1.0%). The multivariable model including all independently associated variables (Table 2) explained 34% of the variance in CEC (adjusted R2, multiple R2 = 0.34). Together, triglycerides and HDL-C explained almost three quarters of the entire variance of the model. Interaction analysis with sex, diabetes status and eGFR revealed a significant interaction of serum albumin concentrations with diabetes status (p < 0.001) and eGFR (p < 0.001) and a significant interaction of PCSK9 concentrations with eGFR (p = 0.01). However, the main determinants were still triglycerides and HDL-C when the analysis was stratified for sex, diabetes status or high and low eGFR (Supplementary Tables 1-3).
Table 1Baseline characteristics of all individuals in the GCKD cohort with CEC measurements and genotyping data available and stratified by quartiles of CEC.
Continuous variables: mean ± standard deviation plus 25th; 50th; 75th percentile in case of pronounced skewness. Categorical variables: number (n) and percentage (%).
Table 2Relationship of baseline characteristics with cholesterol efflux capacity (CEC). Multivariable regression analysis was performed to identify baseline characteristics independently associated with CEC. Except for sex and age, non-significant variables were excluded from the model. Order descending by relative importance in the multivariable analysis as provided by proportional marginal variance decomposition. For continuous variables β estimates and CIs are given per standard deviation increase. Pairwise Spearman's correlation coefficients of all parameters from the model are given in Supplementary Fig. 1.
Two genetic loci were significantly associated with CEC in a genome-wide association study in our main model adjusted for age, sex and principal components 1-10 (Fig. 1 model 1). The first locus is located on chromosome 4 in the KLKB1 gene region, and the second locus is located on chromosome 19 near the APOE and APOC1 genes. Subsequently, models 2-4 were adjusted for additional biochemical parameters for sensitivity analysis and to reduce residual variance by known pathways that determine CEC. The KLKB1 and APOE/C1 locus remained significantly associated with CEC after additional adjustment for either eGFR and UACR (model 2) or HDL-C (model 3) (Fig. 1). In model 4 additionally adjusted for triglycerides, the APOE/C1 locus was no longer significantly associated with CEC, but the KLKB1 locus remained significantly associated and an additional locus on chromosome 3 in the CLSTN2 gene became significantly associated with CEC (Fig. 1). Since the KLKB1 locus was associated with apoA-IV concentrations in a previous GWAS study [
The role of apolipoprotein A-IV in reverse cholesterol transport studied with cultured cells and liposomes derived from an ether analog of phosphatidylcholine.
], we tested in model 5 whether the association of KLKB1 with CEC is independent from apoA-IV concentrations. Thereby, the association with the KLKB1 locus became much stronger than in model 1 and the APOE/C1 locus remained significantly associated (Fig. 1). There was no indication of inflation in any of the models (Fig. 1 right side).
Fig. 1Manhattan-plot (left) and QQ-plot (right) for five GWAS models on CEC.
The red and gray dashed line in the Manhattan-plots mark p = 5x10−8 (genome-wide significance) and p = 1x10−5, respectively. The closest genes to the genome-wide significant results are highlighted in the Manhattan-plot. p: p-value; PC: principal component. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
All imputed SNPs at the APOE/C1 locus together (9,039 SNPs) explained 1.85% (95%CI 0.08–3.62%, p = 0.009) of the CEC variance. In total, five SNPs were significantly associated with CEC in our main model (Fig. 2A). Conditional stepwise analysis yielded one independent genome-wide significant SNP at the locus (rs7412; chr19:45412079). The other genome-wide significant SNPs are in LD with rs7412 (Fig. 2A). In univariable regression analysis, rs7412 explained 0.87% of the variance in CEC values.
The conditionally independent SNPs are marked in purple and LD information (color coded, legend on left side of each plot) refers to these SNPs. The gray dashed line marks genome-wide significance. Lead SNPs and regions depicted are: (A) rs7412 ± 250 kb, (B) rs2292423 ± 500 kb, (C) rs6791077 ± 500 kb in model 1, (D) rs6791077 ± 500 kb in model 4. PC: principal component. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
The SNP rs7412 is a missense variant located in the APOE gene. The T allele of rs7412 was associated with 0.96 units (95%CI: 0.66–1.26) higher CEC compared to the C allele, and this association remained robust after additional adjustment for eGFR and UACR or HDL-C (Table 3). However, the association was not significant anymore after adjustment for triglycerides but remained still significant after adjustment for apoA-IV concentrations (Table 3).
Table 3GWAS results of the independent SNPs of the APOE/C1 and KLKB1 locus.
APOE/C1 locus (rs7412)
KLKB1 locus (rs2292423)
CLSTN2 locus (rs6791077)
Reference allele/effect allele
C/T
T/A
T/C
Imputation Mach R2
1.03
1.01
1.02
Effect allele frequency
8.52%
40.0%
24.2%
β [95% CI]
p
β [95% CI]
p
β [95% CI]
p
Model 1 (age, sex, PC1-10)
0.96 [0.66; 1.26]
3.3x10−10
0.54 [0.37; 0.71]
8.8x10−10
−0.52 [−0.71; −0.32]
2.0x10−7
Model 2 (as model 1, eGFR, log(UACR))
0.96 [0.66; 1.25]
1.8x10−10
0.52 [0.35; 0.69]
1.7x10−9
−0.50 [−0.69; −0.30]
4.5x10−7
Model 3 (as model 1, HDL-C)
0.93 [0.64; 1.22]
1.1x10−10
0.56 [0.39; 0.72]
6.9x10−11
−0.50 [−0.66; −0.28]
1.4x10−6
Model 4 (as model 1, triglycerides)
0.71 [0.42; 0.99]
1.1x10−6
0.55 [0.38; 0.71]
6.1x10−11
−0.55 [−0.74; −0.37]
6.0x10−9
Model 5 (as model 1, apoA-IV)
0.89 [0.60; 1.18]
1.1x10−9
0.68 [0.52; 0.85]
5.1x10−16
−0.46 [−0.65; −0.26]
1.3x10−6
Estimates are given for the effect allele. Adjustments of the respective models are given in brackets. PC: principal component.
All imputed SNPs at the KLKB1 locus together (9,929 SNPs) explained 2.25% (95%CI: 0.47–4.03, p = 0.0001) of the CEC variance. Conditional stepwise analysis yielded one independent genome-wide significant SNP at the locus (rs2292423; chr4:187175722). The other genome-wide significant SNPs were in LD with rs2292423 (Fig. 2B). In univariable regression analysis this SNP explained 0.72% of the variance in CEC values.
CEC was higher by 0.54 units (95%CI 0.37–0.71) in carriers of the A allele of rs2292423 (Table 3). This association remained robust after additional adjustment for eGFR and UACR, HDL-C or triglyceride concentrations (Table 3). When the analysis was adjusted for apoA-IV concentrations, the association became even stronger (0.68 units, 95%CI 0.52–0.85, p = 5.1x10−16, Table 3). Comparison of effect estimates of rs2292423 between CEC and apoA-IV concentrations revealed an inverse association of the A allele (−0.12 units [−0.16; −0.08], p = 1.9x10−10 for apoA-IV versus +0.54 units [0.37; 0.71], p = 8.8x10−10 for CEC), while apoA-IV concentrations and CEC are positively associated with each other (Table 2).
The SNP rs2292423 is an intronic variant in the KLKB1 gene. It is located 20 bp before the start of exon 11 but according to SpliceAI [
] does not have an effect on splicing. Lookup of possibly functional SNPs in LD with this SNP revealed four variants in KLKB1 and two in CYP4V2 (Supplementary Table 4). All except one have a CADD-PHRED score of >10, indicating that they are in the top 10% of deleterious SNPs in the genome but all of them were predicted as tolerated and benign according to SIFT and PolyPhen.
3.5 The CLSTN2 locus
The lead SNP at the CLSTN2 locus (rs6791077; chr3:140176199) had reached a suggestive significance in model 1 (β −0.52 units, 95%CI [−0.71; −0.32], p = 2.0x10−7) and became significant in model 4 (also adjusted for triglycerides): β −0.55 units, 95%CI [−0.74; −0.37], p = 6.0x10−9) (Table 3 and Fig. 2C + D). The lead SNP was not associated with triglycerides (p = 0.8), and the effect estimate for CEC remained robust irrespective of the adjustments (Table 3).
Conditional stepwise analysis yielded one independent genome-wide significant SNP at the locus (rs6791077; chr3:140176199). The other two genome-wide significant hits in this region in model 4 were in LD with the lead SNP (Fig. 2D). The lead SNP is an intronic variant 2,244 bp upstream of exon 7 of CLSTN2. Lookup of possibly functional SNPs in LD with this SNP revealed two missense and one intronic variant in CLSTN2 (Supplementary Table 5). All have a CADD-PHRED score of >10, indicating that they are in the top 10% of deleterious SNPs in the genome, but all of them were predicted as tolerated and benign according to SIFT and PolyPhen.
3.6 Interaction with and stratification for sex, diabetes status and eGFR at the top genetic loci
There was no significant interaction with sex, diabetes and continuous eGFR in our main model of the GWAS at the genome-wide significant loci (Supplementary Table 6). However, stratification revealed that the following associations may be slightly attenuated: APOE/C1 locus in women vs. men (β[95% CI] 0.59 [0.15; 1.03] vs. 0.98 [0.60; 1.35]), KLKB1 locus in individuals with low eGFR vs. high eGFR (0.27 [0.03; 0.52] vs. 0.66 [0.43; 0.89]) and CLSTN2 locus in individuals with diabetes vs. no diabetes (−0.13 [−0.44; 0.18] vs. −0.65 [−0.89; −0.42];Supplementary Table 6).
4. Discussion
In the present study we aimed to identify genetic and non-genetic factors that influence CEC. We identified HDL-C and triglycerides as the main determinants of CEC in a cohort with moderate CKD. As genetic determinants we observed two novel loci in the KLKB1 and CLSTN2 genes and additionally replicated the association with the APOE/C1 locus.
4.1 Association of CEC with HDL-C, triglycerides and other parameters
Several parameters were independently associated with CEC. The most important parameters were triglycerides (12.9%), HDL-C (11.8%), LDL-C (3.0%), apoA-IV concentrations (2.8%), eGFR (1.0%) and PCSK9 concentrations (1.0%), with triglycerides and HDL-C already explaining almost three quarters of the entire explained variance. This is in line with analyses in the GRAPHIC cohort, a general population from the United Kingdom [
Cholesterol efflux capacity is an independent predictor of all-cause and cardiovascular mortality in patients with coronary artery disease: a prospective cohort study.
]. The strength of this association, as indicated by the pairwise Spearman's correlation coefficient, was higher in the GCKD study (r = 0.23,Supplementary Fig. 1) than what has been observed in the Dallas Heart Study (population-based cohort free from ASCVD) (phase 1: r = 0.07 and phase 2: r = 0.14) [
Cholesterol efflux capacity is an independent predictor of all-cause and cardiovascular mortality in patients with coronary artery disease: a prospective cohort study.
]. However, it is important to note that the strength of this association may be influenced by the assay system used: Sankaranarayanan and colleagues showed that the CEC assay using BODIPY-cholesterol is more dependent on ABCA1 than radioactively labelled cholesterol, and thus depends more strongly on preβ-HDL, which is poorly reflected by HDL-C [
Cholesterol efflux capacity is an independent predictor of all-cause and cardiovascular mortality in patients with coronary artery disease: a prospective cohort study.
]. In line, in a large direct comparison of the two assay systems in the Dallas Heart Study phase 2 correlation was also lower for BODIPY-cholesterol compared to H3-cholesterol (0.14 versus 0.36) [
HDL cholesterol efflux capacity is inversely associated with subclinical cardiovascular risk markers in young adults: the cardiovascular risk in Young Finns study.
Cholesterol efflux capacity is an independent predictor of all-cause and cardiovascular mortality in patients with coronary artery disease: a prospective cohort study.
] previous studies. Reasons for these divergent results might include the varying composition of the cohorts and differences between the CEC assay systems, as we discussed recently [
] often have higher CEC values. Likely this is due to increased numbers of preβ HDL particles in hypertriglyceridemia, which are very efficient in mediating efflux via ABCA1 [
]. This may also explain the inconsistent results since the strength may thus depend on the number and functionality of preβ HDL particles and not directly on triglyceride concentrations.
4.2 Genetic loci associated with CEC
Most importantly, in the GWAS on CEC we identified two genome-wide significant hits in our main model. The first significant locus in our main model is in the APOE/C1 locus. The identified SNP rs7412 has been found in more than 100 GWASs, especially on LDL-C, total cholesterol and HDL-C but also triglycerides [
]. However, additional adjustment for triglyceride concentrations indicated, that this association might be mediated by triglyceride concentrations. The SNP rs7412 is a known missense variant that defines the E2 isoform of apo E [
Plasma levels of apolipoprotein E, APOE genotype, and all-cause and cause-specific mortality in 105 949 individuals from a white general population cohort.
The C-terminal lipid-binding domain of apolipoprotein E is a highly efficient mediator of ABCA1-dependent cholesterol efflux that promotes the assembly of high-density lipoproteins.
The C-terminal lipid-binding domain of apolipoprotein E is a highly efficient mediator of ABCA1-dependent cholesterol efflux that promotes the assembly of high-density lipoproteins.
Plasma levels of apolipoprotein E, APOE genotype, and all-cause and cause-specific mortality in 105 949 individuals from a white general population cohort.
]. This data indicates that there is either no direct effect of rs7412 on CEC but its effect is rather mediated via elevated triglyceride concentrations, or the concomitantly increased triglycerides mask the direct effect of the elevated apo E levels on CEC. In support of our finding, the same locus has been identified in a previous GWAS on CEC, which also observed that the genome-wide significance was lost after additional adjustment for HDL-C and triglyceride concentrations [
The second significant locus is the KLKB1 locus. Lookup of possibly functional SNPs in LD with the identified intronic variant rs2292423 indicated variants in CYP4V2 and KLKB1. KLKB1 encodes for the precursors of the serine proteases plasma kallikrein that cleaves factor XII and releases bradykinin and renin (UniProtKB P03952) [
]. Indeed, in vitro studies have shown that plasma kallikrein directly lowers CEC from macrophage foam cells by degrading apo A-I in preβ-migrating HDL particles [
]. As noted above, plasma kallikrein is involved in the intrinsic coagulation pathway. If the association is mediated by a direct effect of plasma kallikrein on CEC, we cannot exclude that this was due to the usage of serum in our study without addition of protease inhibitors directly after blood drawing [
]. In addition, the association with this locus could also be due to changes in composition and number of HDL subclasses caused by an altered function of the renin-angiotensin or kallikrein-kinin system or of fatty acid metabolism. Importantly, this locus was also significantly associated with apoA-IV concentrations in a GWAS [
]. However, additional adjustment for apoA-IV concentrations showed that its association with CEC is not mediated by apoA-IV concentrations. In fact, the A allele of rs2292423 (or other variants that are correlated with this genotype) seems to have opposing effects on CEC and apoA-IV concentrations. Although opposing effects can indicate the presence of collider bias (false association due to controlling for a variable that is independently caused by exposure and outcome) [
], this seems unlikely since the locus was also significantly associated with CEC without apoA-IV adjustment. Replication and more knowledge about the mechanisms behind these associations is needed to disentangle the possibly direct and indirect effects of this locus on CEC.
Finally, CEC was significantly associated with the CLSTN2 locus after adjustment for triglycerides. The lead SNP reached a suggestive significance of < 1x10−5 in our main model, indicating that we may have had insufficient power to detect this locus without triglyceride adjustment, an important determinant of CEC variance. CLSTN2 encodes for calsyntenin 2 and is mainly described for its function in the brain where it may modulate calcium-mediated postsynaptic signals [
]. Replication and larger studies are needed to investigate the effect of this locus on CEC.
4.3 Strengths and limitations of the study
The main strength is the measurement of cell-based CEC in the largest CKD cohort to date using an extensively optimized protocol with stringent quality control using a plate-to-plate correction factor based on four different control samples on each plate that markedly reduced variation and two separate controls to monitor assay performance [
]. This resulted in a low intra- and inter-assay CV of 3.64% and 7.3%, respectively, indicating excellent reproducibility.
A major challenge of the investigation of CEC is that the protocols in use are not standardized, making comparison between studies and hence replication of associations difficult. However, since BODIPY-cholesterol is more dependent on ABCA1 than protocols with radiolabelled cholesterol [
], results may be comparable but BODIPY-cholesterol may have more power to detect ABCA1-mediated pathways. Indeed, we replicated results from radiolabelled CEC studies [
] but also observed with KLKB1 a novel genetic locus that likely acts via ABCA1 by degrading apo A-I in preβ-migrating HDL particles. As a side note, we did not expect to identify variants in CEC transporters since the assay uses J774A.1 macrophages and not patient-derived macrophages.
Cell-based CEC measurements are laborious and expensive. Thus, methods estimating instead of directly measuring cell-based CEC using for example a Bayesian linear regression model of nuclear magnetic resonance data are enticing for large cohorts [
Genetic and observational evidence: No independent role for cholesterol efflux over static high-density lipoprotein concentration measures in coronary heart disease risk assessment.
] further studies are needed to test the relevance of the novel genetic loci in non-CKD cohorts. However, considering that results remained robust with adjustment for CKD parameters, that we replicated the association with the APOE/C1 locus of a previous GWAS [
] and the biological plausibility of the association with the newly identified KLKB1 locus, the novel genetic loci may point towards relevant mechanisms modifying CEC. Characterising CEC for the first time in a large population of patients with moderate CKD provides the basis for a better understanding of the role of CEC in this high risk population. The stratification and interaction analyses with sex, diabetes status and eGFR, however, revealed results that should be verified and further explored in larger cohorts. Especially interesting are the findings on albumin which can be structurally modified in diabetes and CKD by e.g. glycation and oxidation [
A final limitation is that apoA-I and apoB levels and additional HDL-related parameters such as preβ-HDL are not available in the GCKD study.
4.4 Conclusions
We found that HDL-C and triglyceride concentrations are associated with CEC in a cohort of patients with moderately severe chronic kidney disease, which is in line with a study in individuals from a general population [
], likely mediated by triglyceride levels. Moreover, we identified KLKB1 as a novel locus associated with CEC, independently of HDL-C and triglyceride concentrations. We also found CLSTN2 as an additional novel locus, which had a suggestive significance in our main model and reached genome-wide significance after adjustment for triglycerides.
Financial support
This study was supported by grants of the Austrian Research Fund (FWF, W-1253 DK HOROS) to Florian Kronenberg and the Dr. Legerlotz Foundation to Johanna F. Schachtl-Riess. The GCKD study was and is supported by the German Ministry of Education and Research (Bundesministerium für Bildung und Forschung, FKZ 01 ER 0804, 01 ER 0818, 01 ER 0819, 01 ER 0820, and 01 ER 0821) and the KfH Foundation for Preventive Medicine (Kuratorium für Heimdialyse und Nierentransplantation e.V. –Stiftung Präventivmedizin) and corporate sponsors (www.gckd.org).
The funding source had no involvement in study design, data collection, analysis and interpretation of data, or preparation of the manuscript.
Current GCKD Investigators and Collaborators with the GCKD Study are: University of Erlangen-Nürnberg: Kai-Uwe Eckardt, Heike Meiselbach, Markus P. Schneider, Mario Schiffer, Hans-Ulrich Prokosch, Barbara Bärthlein, Andreas Beck, André Reis, Arif B. Ekici, Susanne Becker, Ulrike Alberth-Schmidt, Anke Weigel, Sabine Marschall, Eugenia Schefler; University of Freiburg: Gerd Walz, Anna Köttgen, Ulla T. Schultheiß, Fruzsina Kotsis, Simone Meder, Erna Mitsch, Ursula Reinhard; RWTH Aachen University: Jürgen Floege, Turgay Saritas, Alice Gross; Charité, University Medicine Berlin: Elke Schaeffner, Seema Baid-Agrawal, Kerstin Theisen; Hannover Medical School: Hermann Haller; University of Heidelberg: Martin Zeier, Claudia Sommerer, Mehtap Aykac; University of Jena: Gunter Wolf, Martin Busch, Rainer Paul; Ludwig-Maximilians University of München: Thomas Sitter; University of Würzburg: Christoph Wanner, Vera Krane, Antje Börner-Klein, Britta Bauer; Medical University of Innsbruck, Institute of Genetic Epidemiology: Florian Kronenberg, Barbara Kollerits, Lukas Forer, Sebastian Schönherr, Hansi Weissensteiner; University of Regensburg, Institute of Functional Genomics: Peter Oefner, Wolfram Gronwald; University of Bonn, Institute of Medical Biometry, Informatics and Epidemiology, Medical Faculty: Matthias Schmid, Jennifer Nadal.
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
The work described has been part of the academic thesis of Johanna F. Schachtl-Riess to obtain a PhD degree at the Medical University of Innsbruck.
We are grateful for the willingness of the patients to participate in the GCKD study. The enormous effort of the study personnel of the various regional centers is highly appreciated. We thank the nephrologists who provide routine care for the participants and collaborate with the GCKD study (the list of nephrologists currently collaborating with the GCKD study is available at www.gckd.org).
Appendix A. Supplementary data
The following is the Supplementary data to this article.
Genetic and observational evidence: No independent role for cholesterol efflux over static high-density lipoprotein concentration measures in coronary heart disease risk assessment.
Disease burden and risk profile in referred patients with moderate chronic kidney disease: composition of the German Chronic Kidney Disease (GCKD) cohort.
The role of apolipoprotein A-IV in reverse cholesterol transport studied with cultured cells and liposomes derived from an ether analog of phosphatidylcholine.
Cholesterol efflux capacity is an independent predictor of all-cause and cardiovascular mortality in patients with coronary artery disease: a prospective cohort study.
HDL cholesterol efflux capacity is inversely associated with subclinical cardiovascular risk markers in young adults: the cardiovascular risk in Young Finns study.
Plasma levels of apolipoprotein E, APOE genotype, and all-cause and cause-specific mortality in 105 949 individuals from a white general population cohort.
The C-terminal lipid-binding domain of apolipoprotein E is a highly efficient mediator of ABCA1-dependent cholesterol efflux that promotes the assembly of high-density lipoproteins.
00049-7&id=ga1.jpg){kind=link}
00049-7&id=gr1.jpg){kind=link}
00049-7&id=gr2.jpg){kind=link}