ANTIOXIDANTS & REDOX SIGNALING Volume 22, Number 15, 2015 ª Mary Ann Liebert, Inc. DOI: 10.1089/ars.2015.6248
NEWS & VIEWS
Do Genes Modify the Association of Selenium and Lipid Levels? Inmaculada Galan-Chilet,1 Eliseo Guallar,2,3 J. Carlos Martin-Escudero,4 Griselda De Marco,1,5 Alejandro Dominguez-Lucas,6 Isabel Gonzalez-Manzano,4 Raul Lopez-Izquierdo,4 Josep Redon,6–8 F. Javier Chaves,1,5 and Maria Tellez-Plaza 6,9
Abstract
The interaction of selenium, a component of antioxidant selenoproteins, with genetic variation in lipid-related pathways has not been evaluated earlier as a potential determinant of blood lipid levels. We aimed at evaluating the effects of gene-environment interactions between plasma levels of selenium and polymorphisms in lipid metabolic pathways on plasma lipid levels in a study population from Spain (N = 1,315). We observed statistically significant associations between plasma selenium and lipid levels (differences in total, low-density lipoprotein [LDL]-cholesterol, and triglycerides comparing the 80th with the 20th percentiles of plasma selenium levels were, respectively, 12.0 (95% confidence interval 6.3, 17.8), 8.9 (3.7, 14.2), and 9.0 (2.9, 15.2) mg/dl). We also found statistically significant interactions at the Bonferroni-corrected significance level ( p = 0.0008) between selenium and rs2290201 in FABP4 for total and LDL cholesterol levels and rs1800774 in CETP for elevated LDL cholesterol. Other polymorphisms showed statistically significant differential associations of plasma selenium levels and lipids biomarkers at the nominal p-value of 0.05. Reported statistical interactions with genes involved in lipid transport and transfer provide biological support to the positive associations of selenium with lipids shown in cross-sectional studies and lead to the hypothesis that selenium and lipid levels share common biological pathways that need to be elucidated in mechanistic studies. Antioxid. Redox Signal. 22, 1352–1362.
Introduction
S
elenium-containing supplements are marketed as antioxidant supplements, although the health benefits of selenium supplementation or of dietary selenium intake are controversial. In cross-sectional epidemiological studies from populations with a broad range of selenium intake, including the United States, United Kingdom, Denmark, France, Italy, Norway, Japan, and Spain, increasing selenium levels were consistently associated with increasing levels of total cholesterol, low-density lipoprotein (LDL)-cholesterol, and triglycerides (TG); whereas the association with high-
density lipoprotein (HDL) was inconsistent (3). The interaction of selenium status with genetic variation in genes involved in lipid-related pathways, however, has not been evaluated as a potential determinant of blood lipid levels even though genome-wide association studies (GWAS) indicate that blood lipids levels are strongly influenced by genetic factors. Altogether, there is a need to assess the joint associations of exposure to selenium and polymorphisms in candidate genes on lipid levels. The objective of this study was thus to evaluate the geneenvironment interactions between selenium levels and genetic variants on lipid levels in a population-based study
1
Genetic Diagnosis and Genotyping Unit, Biomedical Research Institute Hospital Clinic of Valencia (INCLIVA), Valencia, Spain. Departments of 2Epidemiology and 3Medicine, Johns Hopkins Medical Institutions, Baltimore, Maryland. Department of Internal Medicine, University Hospital Rio Hortega, Valladolid, Spain. 5 CIBER of Diabetes and Associated Metabolic Diseases (CIBERDEM), Institute of Health Carlos III, Madrid, Spain. 6 Area of Cardiometabolic and Renal Risk, Biomedical Research Institute Hospital Clinic of Valencia (INCLIVA), Valencia, Spain. 7 Department of Internal Medicine, University Hospital Clinic of Valencia, Valencia, Spain. 8 CIBER Physiopathology of Obesity and Nutrition (CIBEROBN), Institute of Health Carlos III, Madrid, Spain. 9 Department of Environmental Health Sciences, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland. 4
1352
GENE-ENVIRONMENT INTERACTION OF SELENIUM AND LIPIDS Innovation
Cross-sectional epidemiological evidence shows an association between increasing selenium levels and total cholesterol, low-density lipoprotein-cholesterol, and triglycerides levels. However, the interaction of selenium status with genetic variation has not been evaluated as a potential determinant of blood lipid levels. We identified statistically significant interactions of selenium with genes involved in lipid-related pathways, thus providing biological support to the association between selenium and lipids. Experimental research is needed to elucidate the causal pathways underlying these associations, which may lead to novel therapeutic approaches. conducted in Valladolid, Spain. We were particularly interested in interactions between selenium levels and candidate genes associated with lipid metabolic pathways. Plasma selenium levels in Spain are similar to those in other European countries and lower than US levels (1). Results Description of the study population
The mean age of study participants was 50 years, and 49% of them were men. The mean concentrations of plasma selenium, total cholesterol, LDL cholesterol, HDL cholesterol, and TG were 87.0 lg/L, 199.2 mg/dl, 114.5 mg/dl, 52.6 mg/dl, and 160.8 mg/dl, respectively. Men had higher selenium levels compared with women. The association of selenium and lipid profiles
In crude analyses, higher plasma selenium concentrations were associated with increased levels of total cholesterol, LDL cholesterol, and TG (Table 1). The multivariateadjusted differences (mg/dl) in total, LDL, and HDL cholesterol, and TG comparing the 80th with the 20th percentiles of plasma selenium levels were 12.0 (95% confidence interval [CI] 6.3, 17.8), 8.9 (3.7, 14.2), 0.7 ( - 0.4, 1.9), and 9.0 (2.9,
1353
15.2), respectively (Table 2). The associations of lipid biomarkers and plasma selenium were mostly linear, although total and LDL cholesterol showed a leveling of the association with very low selenium levels ( p-values for nonlinearity 0.05 and 0.01, respectively) (Table 2 and Fig. 1). The findings in secondary analyses from logistic regression models using elevated lipid levels as categorical outcomes were similar (not shown). Gene-environment interaction
In gene-environment interaction analyses, we found statistically significant interactions below the Bonferronicorrected significance level between selenium levels and several single nucleotide polymorphisms (SNPs) in candidate genes (Figs. 2 and 3), including rs2290201 in FABP4 for associations with total cholesterol (average difference per an interquintile range increase in plasma selenium [95% CI] 13.8 mg/dl [8.5 to 19.2] and - 14.9 mg/dl [ - 37.1 to 7.2] for CC + CT and TT genotypes, respectively; p-interaction = 0.0005) and LDL cholesterol (corresponding difference 10.3 mg/dl [5.4 to 15.3] and - 17.6 mg/dl [ - 38.2 to 3.0] for CC + CT and TT genotypes, respectively; p-interaction = 0.0006) (data not shown) and rs1800774 in CETP for associations with elevated LDL cholesterol (odds ratio [OR] per an interquintile range increase in plasma selenium [95% CI] 0.97 [0.74 to 1.27], 1.76 [1.38 to 2.25] and 3.20 [1.93 to 5.28] for CC (N cases/non cases = 180/396), CT (160/366) and TT (49/96) genotypes, respectively; p-interaction = 0.0002) (data not shown). Genotypes from other candidate genes showed statistically significant differential associations of plasma selenium levels and lipids biomarkers at the nominal p-value of 0.05, including ABCG8 (rs6720173, rs4148211, rs4953020, and rs6544718), FABP1 (rs2241883 and rs2197076), FABP3 (FABP3.345TC), FABP4 (rs16909225), UCP3 (rs647126, rs1800849 and rs2075577), CETP (rs4783962 and rs5882), ABCG1 (rs1893590, rs1044317, rs4148102, and rs2234719), ABCA1 (rs4149341, rs2230806, and rs4149313), MTTP (rs1800591 and rs3816873), APOA5 (rs619054), LIPG (rs9958947), LIPC (rs6083 and rs2070895),
Table 1. Characteristics of Study Participants by Plasma Selenium Quartiles Quartile of plasma selenium (interval in mg/L)
N (participants) Plasma selenium, (lg/L) Age, mean (SE) Gender,% male (SE) Education < high school,% (SE) Body mass index (kg/m2), mean (SE) Smoking status,% (SE) Former Current Cotinine (ng/mL), mean (SE) Total cholesterol (mg/dl), mean (SE) LDL cholesterol (mg/dl), mean (SE) HDL cholesterol (mg/dl), mean (SE) Triglycerides (mg/dl), mean (SE)
Overall
Quartile 1 ( < 72.16)
1315 87.0 (0.63) 49.7 (0.2) 49.0 (0.0) 22.1 (1.0) 26.1 (0.1)
331 61.4 (0.48) 49.6 (0.8) 45.6 (2.5) 23.6 (2.2) 26.3 (0.2)
28.3 26.9 438.9 199.2 114.5 52.6 160.8
(1.2) (1.3) (30.1) (1.0) (0.9) (0.4) (1.9)
29.1 26.1 420.2 192.4 110.5 51.8 151.3
HDL, high-density lipoprotein; LDL, low-density lipoprotein.
(2.5) (2.5) (58.5) (2.0) (1.8) (0.8) (3.8)
Quartile 2 Quartile 3 (72.16–84.63) (84.63–99.31) 332 78.2 (0.21) 48.4 (0.8) 46.0 (2.5) 20.4 (2.2) 25.7 (0.2) 26.7 27.2 450.9 195.4 110.9 53.7 154.1
(2.5) (2.6) (60.5) (2.0) (1.9) (0.8) (4.0)
326 91.7 (0.24) 51.5 (0.8) 51.1 (2.5) 23.2 (2.2) 26.4 (0.2) 27.6 27.4 425.4 202.8 116.3 52.4 170.7
(2.5) (2.6) (61.4) (2.0) (1.8) (0.8) (4.4)
Quartile 4 ( > 99.31)
p Linear trend
326 116.5 (0.91) 49.3 (0.8) 53.3 (2.5) 21.3 (2.1) 26.0 (0.2)
0.74 0.06 0.81 0.74
29.7 27.1 459.0 206.1 120.4 52.4 167.2
(2.6) (2.6) (62.4) (2.1) (1.9) (0.8) (4.4)
0.76 0.69 0.67 < 0.001 < 0.001 0.88 < 0.001
1354
54.02 (0.86) 2.0 ( - 0.1, 4.2) 1.3 ( - 0.8, 3.4) 1.3 ( - 0.8, 3.3) 154.68 (4.39) 7.0 ( - 4.1, 18.0) 10.7 (0.1, 21.3) 10.7 (0.1, 21.3)
149.49 (4.1) 0 (Reference) 0 (Reference) 0 (Reference)
110.62 (2.1) 1.9 ( - 3.6, 7.3) 2.2 ( - 3.2, 7.7) 2.0 ( - 3.4, 7.5)
109.85 (1.98) 0 (Reference) 0 (Reference) 0 (Reference) 51.88 (0.79) 0 (Reference) 0 (Reference) 0 (Reference)
195.57 (2.15) 5.3 ( - 0.5, 11.1) 5.7 ( - 0.1, 11.4) 5.4 ( - 0.4, 11.2)
191.62 (2.24) 0 (Reference) 0 (Reference) 0 (Reference)
Quartile 2 (72.16–84.63)
171.81 (4.86) 17.1 (5.3, 28.8) 17.2 (5.8, 28.6) 17.2 (5.8, 28.5)
52.26 (0.88) 1.3 ( - 0.8, 3.4) 1.2 ( - 0.8, 3.2) 1.1 ( - 0.9, 3.2)
115.9 (1.93) 5.2 ( - 0.1, 10.4) 5.6 (0.4, 10.8) 5.4 (0.2, 10.6)
202.53 (2.11) 9.9 (4.1, 15.7) 10.2 (4.5, 15.9) 10.0 (4.2, 15.7)
Quartile 3 (84.63–99.31)
168.3 (4.86) 15.2 (3.6, 26.8) 17.2 (6.1, 28.3) 17.2 (6.1, 28.4)
52.17 (0.85) 1.2 ( - 0.9, 3.3) 0.8 ( - 1.2, 2.8) 0.9 ( - 1.1, 2.8)
120.77 (2.09) 11.5 (6.0, 16.9) 11.8 (6.4, 17.3) 12.2 (6.7, 17.6)
206.6 (2.31) 15.7 (9.7, 21.7) 16.1 (10.1, 22.1) 16.4 (10.5, 22.4)
Quartile 4 ( > 99.31)
8.6 (2.2, 15.1) 9.0 (2.9, 15.2) 9.0 (2.9, 15.2)
0.8 ( - 0.4, 2.0) 0.7 ( - 0.4, 1.8) 0.7 ( - 0.4, 1.9)
8.3 (3.0, 13.6) 8.6 (3.4, 13.8) 8.9 (3.7, 14.2)
11.4 (5.6, 17.2) 11.6 (5.9, 17.4) 12.0 (6.3, 17.8)
80th to 20th percentilea
< 0.001 < 0.001 < 0.001
0.18 0.22 0.21
< 0.001 < 0.001 < 0.001
< 0.001 < 0.001 < 0.001
p linear trend
0.53 0.37 0.37
0.53 0.77 0.82
0.01 0.01 < 0.001
0.07 0.07 0.05
p nonlinear trend
Model 1: Adjusted by age (years), sex (male, female), education ( < high school, ‡ high school), and fasting time (log hours). Model 2: Further adjusted by body mass index, smoking status (current, former, never), and categorized urine cotinine levels ( < 34, 34–500, and > 500 ng/mL). Model 3: Further adjusted by anti-hypercholesterolemia treatment. The 80th and 20th percentiles of blood selenium distributions are 68.69 and 103.37 lg/L, respectively. a Models comparing the 80th with the 20th percentiles of plasma selenium distributions and associated test for trend were obtained from linear regression models with log-transformed selenium as a continuous variable, except for total cholesterol and LDL where selenium was introduced as restricted quadratic splines with knots at percentiles 10th, 50th, and 90th because of the nonlinear relationships.
Total Cholesterol (mg/dl) Mean (SD) Model 1 Diff. (95% CI) Model 2 Diff. (95% CI) Model 3 Diff. (95% CI) LDL Cholesterol (mg/dl) Mean (SD) Model 1 Diff. (95% CI) Model 2 Diff. (95% CI) Model 3 Diff. (95% CI) HDL Cholesterol (mg/dl) Mean (SD) Model 1 Diff. (95% CI) Model 2 Diff. (95% CI) Model 3 Diff. (95% CI) Triglycerides (mg/dl) Mean (SD) Model 1 Diff. (95% CI) Model 2 Diff. (95% CI) Model 3 Diff. (95% CI)
Quartile 1 ( < 72.16)
Table 2. Change of Lipid Biomarkers Levels by Plasma Selenium Levels
GENE-ENVIRONMENT INTERACTION OF SELENIUM AND LIPIDS
1355
Total Cholesterol
HDL−Cholesterol
20
15
5
15
10
10
0
10
0
5
−5
5
−10
0
−10
0
50
75 125 Plasma Selenium (mg/L)
Differences (mg/dL)
20
Differences (mg/dL)
10
200
50
75 125 Plasma Selenium (mg/L)
200
Triglycerides 30
20
20
15
20
15
10
10
10
10
0
5
0
5
−10
0
−10
0
50
75
125
Plasma Selenium (mg/L)
200
Differences (mg/dL)
20
50
75
125
Weighted Percentage
30
Weighted Percentage
Differences (mg/dL)
LDL−Cholesterol
Weighted Percentage
20
Weighted Percentage
30
200
Plasma Selenium (mg/L)
FIG. 1. Adjusted differences (95% confidence interval [CI]) of lipid biomarkers by plasma selenium levels. The curve represents adjusted mean differences of lipid biomarkers based on restricted quadratic splines with knots at the 10th, 50th, and 90th percentiles (61.5, 84.4, and 114.5 lg/L respectively) of the plasma selenium distribution. The reference value (Difference = 0) was set at the 10th percentile of the plasma selenium distribution (61.5 lg/L). Models were adjusted for age, sex, education, fasting time (log hours), body mass index (BMI), smoking status (never, former, and current), and urine cotinine levels ( < 34, 34–500, and > 500 ng/mL) and anti-hypercholesterolemia treatment. The histogram represents the frequency distribution of plasma selenium in the study sample.
‰ FIG. 2. Candidate genes-selenium interaction–log10 p-values for adjusted differences of lipid biomarker concentrations. p-Values for the interaction of plasma selenium and 59 and 38 single nucleotide polymorphisms (SNPs) derived from linear and nonlinear (for total and LDL cholesterol) regression models, respectively, (dominant, recessive, and additive model) adjusted by age, sex, education, fasting time (log hours), BMI, smoking status (never, former, and current), urine cotinine levels ( < 34, 34–500, and > 500 ng/mL), and anti-hypercholesterolemia treatment are presented on the left y axis on the logarithmic scale according to the position of the SNPs on the chromosome (x axis). The horizontal solid line corresponds to a nominal significance level of 0.05. Horizontal dashed line corresponds to the Bonferroni-corrected significance level of 0.0008. To see this illustration in color, the reader is referred to the web version of this article at www.liebertpub.com/ars
- log10(P interaction) 7
6
5
4
3
2
1
7
6
5
4
3
2
1 2
2 4
4 7
7
7 8
LIPC: rs2070895
LDL Cholesterol
9
1
0 9
8
Chromosome
8 11 15
Chromosome
9
Triglycerides
8
1
0 9 11
9
11
ABCG1: rs1893590
CETP: rs4783962
15 16
16
ABCG1: rs1044317
4 ABCA1: rs2230806
Chromosome
11
15 16 18 21
15
16
18
21
ABCG1: rs2234719
2 FABP4: rs2290201
8 9
ABCG1: rs4148102
1 8
LIPG: rs9958947
2 7
LIPG: rs9958947
3 4
CETP: rs5882
4
ABCG8: rs4148211
ABCG1: rs1893590
CETP: rs4783962
UCP3: rs647126
FABP4: rs2290201
FABP1: rs2241883
8
APOA5: rs619054
5
UCP3: rs1800849
6 2
LPL: rs320
1
UCP1: rs12502572
2
FABP1: rs2241883
3
ABCG8: rs4953020
4
ABCG8: rs4148211
5
ABCA1: rs2230806
7
ABCG8: rs4953020
- log10(P interaction) 6
ABCG8: rs4953020 ABCG8: rs4148211
- log10(P interaction) 7
ABCG8: rs6544718
- log10(P interaction)
1356 GALAN-CHILET ET AL.
9
Total Cholesterol additive dominant recessive
1
0 21
Chromosome 21
9
HDL Cholesterol
8
1
0
GENE-ENVIRONMENT INTERACTION OF SELENIUM AND LIPIDS
LPL (rs10099160, rs11570892, rs256 and rs17410577), UCP1 (rs12502572), and FABP2 (rs1511025) (Figs. 2 and 3). Discussion
Increasing plasma selenium levels were associated with increasing levels of total, LDL, and TG cholesterol, but not HDL cholesterol. We also found an interaction of selenium levels with genetic variation in several genes related to lipid metabolic pathways, suggesting potentially inter-related biological pathways in selenium and lipid metabolism. This is the first study evaluating the interaction of selenium levels with lipid-related candidate genes on lipid levels using an epidemiologic sample representative of a population with a low burden of disease. The observed positive associations of selenium with total and LDL cholesterol were largely consistent with findings from cross-sectional studies (3). While few studies have evaluated the association between selenium and TG, they were also consistent with the positive association observed in our data. In the United States, participants in the highest quartile of serum selenium had 10% higher TG concentrations than participants in the lowest quartile (3). The associations between selenium and HDL cholesterol, however, were inconsistent across studies. The biological mechanisms underlying the association of plasma selenium with lipid concentrations in cross-sectional studies are not well understood. The liver has a key role in the regulation of both selenium and lipid levels. The liver accumulates small-molecule forms of selenium and hepatocytes secrete selenoprotein P, which binds selenium for plasma transport to various extrahepatic tissues (2). Selenium excess induced hepatic dysfunction in experimental studies, including alterations in lipid metabolism (8) and there is experimental and epidemiological evidence supporting the fact that high selenium levels may increase oxidative stress. Selenium was associated with redox changes in isolated hepatocytes and high levels of selenium damaged the liver of rats and increased the liver concentration 8-hydroxy-2¢-deoxyguanosine (2). Furthermore, plasma selenium levels above 110 lg/L were positively associated with urine 8-hydroxy-2¢-deoxyguanosine in our study sample (1). It is thus biologically possible that selenium excess can induce liver damage and disrupt lipid metabolism, although it is also possible that hepatic damage may simultaneously affect selenium and lipid metabolism. While selenium deficiency has also been related to liver damage in experimental animals (2), selenium and lipids were not associated with the lower range of selenium exposure in our study. For total and LDL cholesterol, we found Bonferroni-corrected statistically significant interactions of selenium with SNPs in FABP4 (rs2290201) and CETP (rs1800774), genes which are highly expressed in the hepatic and adipose tissue.
1357
Adipocyte fatty acid–binding protein-4 (FABP4/adipocyte P2) is involved in lipid transport and is positively associated with plasma lipid levels. In European Americans from the Cardiovascular Health Study, the minor alleles of six polymorphisms in FABP4 (including allele A for rs2290201) were associated with lower FABP4 levels (4). In our geneinteraction analysis, homozygotes for the minor allele in rs2290201 (T in our study population) presented lower total and LDL cholesterol with increasing selenium levels compared with carriers of CC and CT genotypes. CETP codifies for the cholesteryl ester transfer protein and is involved in the transfer of cholesteryl esters from HDL to other lipoproteins in the blood. CETP inhibition increases HDL cholesterol levels and decreases non-HDL cholesterol levels. In our study population, carriers of genotype CC in rs1800774 showed lower total and LDL-cholesterol levels with higher plasma selenium levels, although the interaction was not statistically significant for HDL cholesterol. In the Atorvastatin Comparative Cholesterol Efficacy and Safety Study (ACCESS) trial, four polymorphisms in the 5¢ end of CETP (VNTR, rs1800775, rs708272, and rs15723185) were significantly associated with higher concentrations of HDL cholesterol (7). Consistent with our results, rs1800774 was not associated with HDL cholesterol levels. The associations with total and LDL cholesterol, however, were not reported. In addition to gene-selenium interactions with FABP4 and CETP, we found suggestive evidence for interactions of selenium levels with other candidate genes associated with lipid levels in GWAS and metanalysis, including ABCG8, LPL, LIPC, APOA5, ABCA1, or LIPG. Large epidemiologic studies with sufficient power to evaluate gene-environment interactions are needed to replicate our findings, but our geneinteraction analyses provide biological support for the observed cross-sectional associations between selenium and lipids and support the fact that selenium and lipid metabolic pathways may be inter-related. While our findings support a biological connection between selenium and lipid levels, the cross-sectional nature of our study limits our ability to determine the direction of this association or whether common factors (such as liver dysfunction or dietary factors) are responsible for simultaneous changes in selenium and lipid levels. The only two prospective studies that have evaluated the association between selenium and lipids using longitudinal data found that baseline selenium levels did not predict future changes in lipid levels. Similarly, in the UK PRECISE Pilot trial, selenium yeast had modestly beneficial effects on plasma lipid levels at 6 months after randomization compared with placebo (6). Additional data are also available from two trials in which selenium supplementation was combined with other vitamins and minerals (9). In a population from rural China, supplementation with selenium combined with vitamins C and E
‰ FIG. 3. Candidate genes-selenium interaction–log10 p-values for adjusted odds ratio of altered lipid biomarkers levels. p-Values for the interaction of plasma selenium and 59 SNPs derived from logistic regression models (dominant, recessive, and additive model) adjusted by age, sex, education, fasting time (log hours), BMI, smoking status (never, former, and current), urine cotinine levels ( < 34, 34–500, and > 500 ng/mL), and anti-hypercholesterolemia treatment are presented on the left y axis on the logarithmic scale according to the position of the SNPs on the chromosome (x axis). The horizontal solid line corresponds to a nominal significance level of 0.05. Horizontal dashed line corresponds to the Bonferronicorrected significance level of 0.0008. To see this illustration in color, the reader is referred to the web version of this article at www.liebertpub.com/ars
- log10(P interaction) 6
5
4
3
2
1
6
5
4
3
2
1
2
2
4 7
4 7
7 8 9
8
Chromosome
8
1358 9
9 11
0
Chromosome
11
15
8
Low HDL Cholesterol
1
0
Chromosome
High Triglycerides
8
7
1
0
8
CETP: rs1800774
UCP3: rs647126 UCP3: rs2075577
FABP4: rs16909225 ABCA1: rs4149341
15 16 18 21
16 18 21
15 16 18 21
15
16
18
21
ABCG1: rs1044317
ABCG1: rs4148102
1 LIPG: rs9958947
11 LIPC: rs6083
4
11
ABCG1: rs4148107 ABCG1: rs2234719
7 2
APOA5: rs619054
1 9
UCP3: rs1800849
2 8
ABCA1: rs2230806
3 7
ABCA1: rs4149313
4 4
LPL: rs11570892
5 2
FABP4: rs2290201
6
LPL: rs17410577 LPL: rs256 LPL: rs10099160 LPL: rs11570892
1
UCP1: rs3811787
7
ABCG1: rs1893590
CETP: rs1800774
UCP3: rs647126
ABCA1: rs4149341
FABP1: rs2241883
ABCG8: rs6720173
2
UCP1: rs3811787
3
FABP1: rs2241883
4
FABP3: FABP3.345TC
5
FABP3: FABP3.345TC FABP3: rs12401792 ABCG8: rs6720173
- log10(P interaction) 6
ABCG8: rs4953020 ABCG8: rs4148211 FABP1: rs2197076 FABP1: rs2241883 MTTP: rs1800591 MTTP: rs3816873
- log10(P interaction)
7
FABP2: rs1511025
- log10(P interaction)
8
High Total Cholesterol additive dominant recessive
1
0
Chromosome
High LDL Cholesterol
GENE-ENVIRONMENT INTERACTION OF SELENIUM AND LIPIDS
resulted in small but significant increases in total and LDLcholesterol concentrations, but not in HDL-cholesterol (9). In the SU.VI.MAX study, the group assigned to selenium, b-carotene, zinc, and vitamins C and E had a borderline significantly higher frequency of hypercholesterolemia and elevated TG levels 7.5 years after randomization (9). Additional data from prospective studies, clinical trials, and mechanistic studies are needed to understand the causal effects underlying the associations between selenium and lipids observed in cross-sectional studies. In addition to the cross-sectional design, our study has other limitations. First, we only had access to nonfasting plasma samples. While total and HDL cholesterol levels can be measured in nonfasting samples, LDL cholesterol and TG levels are usually measured in fasting samples. There is recent evidence, however, supporting the fact that nonfasting LDL and TG levels are also associated with increased cardiometabolic risk (5). Second, we had only a single measurement of plasma selenium and lipids, which may be subject to within-person variability. Nondifferential measurement error, however, would induce an attenuation of the observed associations. Finally, total plasma selenium levels do not provide information of individual selenium compounds. More detailed analyses of different selenoprotein levels and activity are needed to better understand the association of selenium with lipid biomarkers. Strengths of our study include the sampling design that allows our findings to be generalized to a general population framework, the availability of detailed information for adjusting for multiple determinants of lipid levels, and the availability of genotyped SNPs for 19 candidate genes associated with lipid pathways. In conclusion, we observed approximately linear associations between plasma selenium and total, LDL-cholesterol, and TG levels and found that these associations were modified by genetic polymorphisms involved in lipid metabolic pathways, including lipid transport and transfer. As prospective epidemiologic studies and clinical trials with measurements of selenium and genome-wide genetic data become increasingly available, collaborative meta-analyses will allow us to establish the biological and clinical implications of our findings under different levels of selenium exposure. Notes Study population
We conducted a population-based survey among adults 15–85 years old residing in the catchment area of the Rio Hortega University Hospital in Valladolid (Spain) in 1997– 2003. In Spain, tertiary hospitals have assigned specific geographic areas for patient referral, and they integrate the network of primary care centers in each area. The complex study design and data collection methods have been described in detail elsewhere (1). The study sample consisted of 1502 participants with stored plasma samples for selenium and lipid determinations. We excluded 12 participants missing plasma selenium measurements, 2 participants missing lipid measurements, 53 participants with TG levels > 400 mg/dl, and 120 participants missing other relevant covariates, leaving 1315 participants for analysis. The research protocol was approved by Ethics Committee of the Rio Hortega University Hospital, and all participants provided written informed consent.
1359
Plasma selenium levels
Plasma selenium levels were measured by inductively coupled-plasma mass spectrometry (ICPMS) on an Agilent 7500CEx ICPMS (Agilent Technologies) following a standardized protocol. The lower and upper detection limits for plasma selenium levels were 29.85 and 205.30 lg/L, respectively. No sample had levels below or above these limits. The inter-assay coefficient of variation for plasma selenium levels was 5.6%. Plasma lipid levels
Plasma total cholesterol, HDL-cholesterol, and TG concentrations in nonfasting samples were determined using a Hitachi 917 analyzer. LDL cholesterol was calculated using the Friedewald formula, which is highly correlated with directly measured LDL-cholesterol in both fasting and nonfasting samples (5). Elevated total cholesterol was defined as > 200 mg/dl, elevated LDL cholesterol was defined as > 130 mg/dl, low HDL cholesterol was defined as < 40 mg/dl or < 50 mg/dl for men and women, respectively, and elevated TG was defined as > 150 mg/dl. Fasting time was defined as the number of hours from the last food or drink (excluding water) intake to blood extraction. DNA isolation, SNP selection, and genotyping
DNA was isolated from peripheral blood cells using Chemagic System (Chemagen), and its quality was assessed with PicoGreen dsDNA Quantification Reagent (Invitrogen). DNA was diluted to a final concentration of 100 ng/ll. We used bibliography searches and the SYSNPS program, based on public data sources, including Ensembl and HapMap, to identify 96 SNPs from 19 candidate genes implicated in lipid pathways, including fatty acid uptake, transport and metabolism, regulation of plasma TG levels, and lipoprotein metabolism and assembly. We included SNPs previously reported to be related to health outcomes in humans or to have functional implications. The SNPs were genotyped using an oligo-ligation-assay (SNPlex; Applied Biosystems) following the manufacturer’s guidelines. The mean genotyping coverage across all genotyped SNPs was 94%. Twenty-four SNPs were excluded because of genotyping errors or because they did not meet Hardy–Weinberg equilibrium ( p-value < 0.01). Thirteen SNPs were further excluded, because they had an MAF < 0.05 or a minor genotype frequency < 20 individuals in the study population. The final number of SNPs included in linear geneenvironment interaction analyses was 59. The complete list of finally selected SNPs is available in Table 3. In nonlinear SNPselenium interaction analysis, we additionally excluded 21 SNPs that had a minor genotype frequency lower than 5% above or below the median of selenium distribution to have the genotypes sufficiently represented across the range of selenium levels and obtain robust estimates of the nonlinear associations. Other variables
Information on age, sex, education, smoking status, and lipid-lowering medication was based on self-report. Body mass index (BMI) was calculated using measured height and weight. Urine cotinine was measured by enzyme-linked immunosorbent assay (ELISA) (Kit ‘‘Ana´lisis DRI Cotinina,’’
Table 3. Single Nucleotide Polymorphisms Finally Selected Gene name
Gene symbol
ATP-binding cassette, sub-family A (ABC1), member 1
ABCA1
ATP-binding cassette, sub-family G (WHITE), member 1
ABCG1
ATP-binding cassette, sub-family G (WHITE), member 8
ABCG8
Apolipoprotein A-V
APOA5
Cholesteryl ester transfer protein, plasma
CETP
Fatty acid binding protein 1, liver
FABP1
Fatty acid binding protein 2, liver
FABP2
Fatty acid binding protein 3, liver
FABP3
Fatty acid binding protein 4, liver
FABP4
Lecithin-cholesterol acyltransferase Lipase, hepatic
LCAT LIPC
Lipase, endothelial Lipoprotein lipase
LIPG LPL
Microsomal triglyceride transfer protein
MTTP
Paraoxonase 1 Mitochondrial uncoupling protein 1
PON1 UCP1
Mitochondrial uncoupling protein 2 Mitochondrial uncoupling protein 3
UCP2 UCP3
SNP rs2230808 rs4149313 rs7031748 rs2230806 rs1800977 rs2246293 rs2422493 rs4149341 rs1893590 rs1378577 rs4148102 rs4148107 rs2234719 rs1044317 rs6720173 rs4953020 rs4148211 rs6544718 rs619054 rs662799 rs10750097 rs4783962 rs1800774 rs5882 rs2197076 rs2241883 rs1511025 rs6857641 rs2282688 rs10034579 rs2271072 fabp3.345tc rs12401792 rs16909225 rs2305319 rs2290201 rs1109166 rs1077834 rs1800588 rs2070895 rs6083 rs9958947 rs17410577 rs256 rs320 rs10099160 rs11570892 rs3811800 rs1800591 rs3816873 rs2306985 rs854571 rs12502572 rs3811787 rs1800592 rs660339 rs647126 rs2075577 rs1800849
SNP, single nucleotide polymorphism.
1360
Allele T/C T/C G/T C/T G/A C/G G/A T/C A/C T/G G/A T/C C/T A/G G/C G/T A/G T/C G/A G/A G/A T/C C/T G/A G/A T/C T/C T/C T/C A/C C/G C/T G/T A/G T/C G/A T/C T/C C/T G/A A/G C/T G/C C/T T/G T/G A/G T/C G/T T/C C/G T/C G/A T/G T/C G/A A/G G/A G/A
Consequence Nonsynonymous Nonsynonymous Synonymous Nonsynonymous 5¢UTR 5¢UTR 5¢UTR 3¢UTR Intergenic Intergenic Intronic Intronic Intronic 3¢UTR Nonsynonymous Intronic Nonsynonymous Nonsynonymous 3¢UTR 5¢UTR 5¢UTR 5¢UTR Intronic Nonsynonymous Intronic Nonsynonymous Synonymous 5¢UTR 5¢UTR 5¢UTR Intronic Intronic 5¢UTR 5¢UTR Intronic Intronic Intronic 5¢UTR 5¢UTR 5¢UTR Nonsynonymous 5¢UTR Intronic Intronic Intronic Intronic 3¢UTR 5¢UTR 5¢UTR Nonsynonymous Nonsynonymous 5¢UTR Intronic 5¢UTR 5¢UTR Nonsynonymous 3¢UTR Synonymous 5¢UTR
GENE-ENVIRONMENT INTERACTION OF SELENIUM AND LIPIDS
Ref. 0395 Microgenics laboratories). One thousand sixty five individuals had cotinine concentrations below the limit of detection (34 ng/ml) (1). Statistical methods
We used the survey package in R software (version 3.0.3; R Development Core Team 2014) to account for the sampling design and survey weights. Plasma selenium concentrations were log transformed for analyses. Cut-offs for plasma selenium quartiles were based on weighted distributions in the study sample. We estimated the average differences in lipid concentrations (with 95% CIs) by selenium levels using multivariable linear regression models. As secondary analyses, we estimated the odd ratios for elevated total cholesterol, LDL cholesterol, and TG, or for low HDL cholesterol (with 95% CIs) using multivariable logistic regression models. Plasma selenium concentrations were introduced in the models as quartiles, comparing each of the three highest quartiles of plasma selenium with the lowest quartile, or as a log-transformed (continuous) variable to compare lipid levels in the 80th versus the 20th percentiles of the selenium distribution (interquintile range). Statistical models were initially adjusted for sex, age, education ( < high school, ‡ high school), and fasting time (logtransformed number of hours) (Model 1). We additionally adjusted for BMI, smoking status (never, former and current smoker), and urine cotinine levels ( < 34, 34–500 and > 500 ng/mL) (Model 2). Finally, we further adjusted for use of lipid-lowering medications (Model 3). p-values for linear trend were obtained from Wald tests for log-transformed plasma selenium in regression models. To further explore the shape of the relationship between plasma selenium and lipid biomarkers, we modeled plasma selenium levels as restricted quadratic splines with knots at the 10th, 50th, and 90th percentiles of the plasma selenium distribution (61.5, 84.6, and 114.6 lg/L respectively). p-values for nonlinearity were obtained from Wald tests for the nonlinear spline terms. We conducted sensitivity analyses using not-transformed plasma selenium levels (instead of log-transformed), with essentially identical results. Gene-environment interaction analyses were based on interaction terms for log-transformed plasma selenium levels with indicator variables for genotypes. For each SNP, we estimated three models for the interaction with selenium in separate models assuming dominant, recessive, and additive inheritance, and calculated the p-values for interaction in each model by using Ftests for global significance by comparing nested models with and without the corresponding interaction terms. Since candidate genes for the gene-environment interaction study were selected based on a priori knowledge of their role in lipid metabolic pathways, the selected alpha level for statistical significance was 0.05. For descriptive purposes, however, we calculated a Bonferroni-corrected p-value of 0.0008 (estimated as 0.05 divided by 59 finally selected SNPs). For the Bonferroni-corrected significant SNPs, in cases where only 1 inheritance model showed statistically significant SNP-selenium interactions, we reported the associations of selenium and lipid levels among subgroups of participants with the genotypes of interest in this specific model. If more than 1 inheritance model showed statistically significant SNP-selenium interactions, we reported the best fitting inheritance model selected by comparing a general model that included separate dummy variables for the heterozygote and
1361
minor allele homozygote (reference major allele homozygote) with the models assuming dominant (minor allele homozygote and heterozygote versus major allele homozygote), recessive (minor allele homozygote versus heterozygote and major allele homozygote), or additive inheritance (0, 1, or 2, minor allele dosage). Acknowledgments
This work was supported by the Strategic Action for Research in Health sciences [CP12/03080, PI10/0082, PI13/ 01848, and PI11/00726], GRUPOS 03/101; PROMETEO/ 2009/029 and ACOMP/2013/039 from the Valencia Government, GRS/279/A/08 from Castilla-Leon Government, and European Network of Excellence Ingenious Hypercare (EPSS037093) from the European Commission; CIBER Fisiopatologı´a Obesidad y Nutricio´n (CIBERobn) [CIBER-02-08-2009, CB06/03 and CB12/03/30016]; and CIBER de Diabetes y Enfermedades Metabo´licas Relacionadas (CIBERDEM). The Strategic Action for Research in Health Sciences, CIBEROBN and CIBERDEM are initiatives from Carlos III Health Institute Madrid and the Spanish Ministry of Economy and Competitiveness and are co-funded with European Funds for Regional Development (FEDER). References
1. Galan-Chilet I, Tellez-Plaza M, Guallar E, De Marco G, Lopez-Izquierdo R, Gonzalez-Manzano I, Carmen Tormos M, Martin-Nunez GM, Rojo-Martinez G, Saez GT, MartinEscudero JC, Redon J, and Javier Chaves F. Plasma selenium levels and oxidative stress biomarkers: a gene-environment interaction population-based study. Free Radic Biol Med 74: 229–236, 2014. 2. Ho¨gberg J and Alexander J. Selenium. In: Handbook on the Toxicology of Metals, edited by Nordberg GF, Fowler BA, Nordberg M, Friberg LT. London, Waltham, San Diego: Academic Press, 2007, pp. 783–807. 3. Laclaustra M, Stranges S, Navas-Acien A, Ordovas JM, and Guallar E. Serum selenium and serum lipids in US adults: national Health and Nutrition Examination Survey (NHANES) 2003–2004. Atherosclerosis 210: 643–648, 2010. 4. Mukamal KJ, Wilk JB, Biggs ML, Jensen MK, Ix JH, Kizer JR, Tracy RP, Zieman SJ, Mozaffarian D, Psaty BM, Siscovick DS, and Djousse L. Common FABP4 genetic variants and plasma levels of fatty acid binding protein 4 in older adults. Lipids 48: 1169–1175, 2013. 5. Nordestgaard BG and Benn M. Fasting and nonfasting LDL cholesterol: to measure or calculate? Clin Chem 55: 845– 847, 2009. 6. Rayman MP, Stranges S, Griffin BA, Pastor-Barriuso R, and Guallar E. Effect of supplementation with high-selenium yeast on plasma lipids: a randomized trial. Ann Intern Med 154: 656–665, 2011. 7. Thompson JF, Durham LK, Lira ME, Shear C, and Milos PM. CETP polymorphisms associated with HDL cholesterol may differ from those associated with cardiovascular disease. Atherosclerosis 181: 45–53, 2005. 8. Wang X, Zhang W, Chen H, Liao N, Wang Z, Zhang X, and Hai C. High selenium impairs hepatic insulin sensitivity through opposite regulation of ROS. Toxicol Lett 224: 16–23, 2014. 9. Zhang L, Gail MH, Wang YQ, Brown LM, Pan KF, Ma JL, Amagase H, You WC, and Moslehi R. A randomized factorial study of the effects of long-term garlic and micro-
1362
GALAN-CHILET ET AL.
nutrient supplementation and of 2-wk antibiotic treatment for Helicobacter pylori infection on serum cholesterol and lipoproteins. Am J Clin Nutr 84: 912–919, 2006.
Address correspondence to: Dr. Maria Tellez-Plaza Area of Cardiometabolic and Renal Risk Biomedical Research Institute Hospital Clinic of Valencia (INCLIVA) Av. Menendez Pidal 4 accesorio Valencia 46010 Spain E-mail: [email protected] [email protected]
Date of first submission to ARS Central, January 5, 2015; date of acceptance, January 20, 2015.
Abbreviations Used BMI ¼ body mass index GWAS ¼ Genome Whole Association Studies HDL ¼ high density lipoprotein ICPMS ¼ inductively coupled-plasma mass spectrometry LDL ¼ low density lipoprotein OR ¼ odds ratio SNP ¼ single nucleotide polymorphism TG ¼ triglycerides
NEWS & VIEWS
Do Genes Modify the Association of Selenium and Lipid Levels? Inmaculada Galan-Chilet,1 Eliseo Guallar,2,3 J. Carlos Martin-Escudero,4 Griselda De Marco,1,5 Alejandro Dominguez-Lucas,6 Isabel Gonzalez-Manzano,4 Raul Lopez-Izquierdo,4 Josep Redon,6–8 F. Javier Chaves,1,5 and Maria Tellez-Plaza 6,9
Abstract
The interaction of selenium, a component of antioxidant selenoproteins, with genetic variation in lipid-related pathways has not been evaluated earlier as a potential determinant of blood lipid levels. We aimed at evaluating the effects of gene-environment interactions between plasma levels of selenium and polymorphisms in lipid metabolic pathways on plasma lipid levels in a study population from Spain (N = 1,315). We observed statistically significant associations between plasma selenium and lipid levels (differences in total, low-density lipoprotein [LDL]-cholesterol, and triglycerides comparing the 80th with the 20th percentiles of plasma selenium levels were, respectively, 12.0 (95% confidence interval 6.3, 17.8), 8.9 (3.7, 14.2), and 9.0 (2.9, 15.2) mg/dl). We also found statistically significant interactions at the Bonferroni-corrected significance level ( p = 0.0008) between selenium and rs2290201 in FABP4 for total and LDL cholesterol levels and rs1800774 in CETP for elevated LDL cholesterol. Other polymorphisms showed statistically significant differential associations of plasma selenium levels and lipids biomarkers at the nominal p-value of 0.05. Reported statistical interactions with genes involved in lipid transport and transfer provide biological support to the positive associations of selenium with lipids shown in cross-sectional studies and lead to the hypothesis that selenium and lipid levels share common biological pathways that need to be elucidated in mechanistic studies. Antioxid. Redox Signal. 22, 1352–1362.
Introduction
S
elenium-containing supplements are marketed as antioxidant supplements, although the health benefits of selenium supplementation or of dietary selenium intake are controversial. In cross-sectional epidemiological studies from populations with a broad range of selenium intake, including the United States, United Kingdom, Denmark, France, Italy, Norway, Japan, and Spain, increasing selenium levels were consistently associated with increasing levels of total cholesterol, low-density lipoprotein (LDL)-cholesterol, and triglycerides (TG); whereas the association with high-
density lipoprotein (HDL) was inconsistent (3). The interaction of selenium status with genetic variation in genes involved in lipid-related pathways, however, has not been evaluated as a potential determinant of blood lipid levels even though genome-wide association studies (GWAS) indicate that blood lipids levels are strongly influenced by genetic factors. Altogether, there is a need to assess the joint associations of exposure to selenium and polymorphisms in candidate genes on lipid levels. The objective of this study was thus to evaluate the geneenvironment interactions between selenium levels and genetic variants on lipid levels in a population-based study
1
Genetic Diagnosis and Genotyping Unit, Biomedical Research Institute Hospital Clinic of Valencia (INCLIVA), Valencia, Spain. Departments of 2Epidemiology and 3Medicine, Johns Hopkins Medical Institutions, Baltimore, Maryland. Department of Internal Medicine, University Hospital Rio Hortega, Valladolid, Spain. 5 CIBER of Diabetes and Associated Metabolic Diseases (CIBERDEM), Institute of Health Carlos III, Madrid, Spain. 6 Area of Cardiometabolic and Renal Risk, Biomedical Research Institute Hospital Clinic of Valencia (INCLIVA), Valencia, Spain. 7 Department of Internal Medicine, University Hospital Clinic of Valencia, Valencia, Spain. 8 CIBER Physiopathology of Obesity and Nutrition (CIBEROBN), Institute of Health Carlos III, Madrid, Spain. 9 Department of Environmental Health Sciences, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland. 4
1352
GENE-ENVIRONMENT INTERACTION OF SELENIUM AND LIPIDS Innovation
Cross-sectional epidemiological evidence shows an association between increasing selenium levels and total cholesterol, low-density lipoprotein-cholesterol, and triglycerides levels. However, the interaction of selenium status with genetic variation has not been evaluated as a potential determinant of blood lipid levels. We identified statistically significant interactions of selenium with genes involved in lipid-related pathways, thus providing biological support to the association between selenium and lipids. Experimental research is needed to elucidate the causal pathways underlying these associations, which may lead to novel therapeutic approaches. conducted in Valladolid, Spain. We were particularly interested in interactions between selenium levels and candidate genes associated with lipid metabolic pathways. Plasma selenium levels in Spain are similar to those in other European countries and lower than US levels (1). Results Description of the study population
The mean age of study participants was 50 years, and 49% of them were men. The mean concentrations of plasma selenium, total cholesterol, LDL cholesterol, HDL cholesterol, and TG were 87.0 lg/L, 199.2 mg/dl, 114.5 mg/dl, 52.6 mg/dl, and 160.8 mg/dl, respectively. Men had higher selenium levels compared with women. The association of selenium and lipid profiles
In crude analyses, higher plasma selenium concentrations were associated with increased levels of total cholesterol, LDL cholesterol, and TG (Table 1). The multivariateadjusted differences (mg/dl) in total, LDL, and HDL cholesterol, and TG comparing the 80th with the 20th percentiles of plasma selenium levels were 12.0 (95% confidence interval [CI] 6.3, 17.8), 8.9 (3.7, 14.2), 0.7 ( - 0.4, 1.9), and 9.0 (2.9,
1353
15.2), respectively (Table 2). The associations of lipid biomarkers and plasma selenium were mostly linear, although total and LDL cholesterol showed a leveling of the association with very low selenium levels ( p-values for nonlinearity 0.05 and 0.01, respectively) (Table 2 and Fig. 1). The findings in secondary analyses from logistic regression models using elevated lipid levels as categorical outcomes were similar (not shown). Gene-environment interaction
In gene-environment interaction analyses, we found statistically significant interactions below the Bonferronicorrected significance level between selenium levels and several single nucleotide polymorphisms (SNPs) in candidate genes (Figs. 2 and 3), including rs2290201 in FABP4 for associations with total cholesterol (average difference per an interquintile range increase in plasma selenium [95% CI] 13.8 mg/dl [8.5 to 19.2] and - 14.9 mg/dl [ - 37.1 to 7.2] for CC + CT and TT genotypes, respectively; p-interaction = 0.0005) and LDL cholesterol (corresponding difference 10.3 mg/dl [5.4 to 15.3] and - 17.6 mg/dl [ - 38.2 to 3.0] for CC + CT and TT genotypes, respectively; p-interaction = 0.0006) (data not shown) and rs1800774 in CETP for associations with elevated LDL cholesterol (odds ratio [OR] per an interquintile range increase in plasma selenium [95% CI] 0.97 [0.74 to 1.27], 1.76 [1.38 to 2.25] and 3.20 [1.93 to 5.28] for CC (N cases/non cases = 180/396), CT (160/366) and TT (49/96) genotypes, respectively; p-interaction = 0.0002) (data not shown). Genotypes from other candidate genes showed statistically significant differential associations of plasma selenium levels and lipids biomarkers at the nominal p-value of 0.05, including ABCG8 (rs6720173, rs4148211, rs4953020, and rs6544718), FABP1 (rs2241883 and rs2197076), FABP3 (FABP3.345TC), FABP4 (rs16909225), UCP3 (rs647126, rs1800849 and rs2075577), CETP (rs4783962 and rs5882), ABCG1 (rs1893590, rs1044317, rs4148102, and rs2234719), ABCA1 (rs4149341, rs2230806, and rs4149313), MTTP (rs1800591 and rs3816873), APOA5 (rs619054), LIPG (rs9958947), LIPC (rs6083 and rs2070895),
Table 1. Characteristics of Study Participants by Plasma Selenium Quartiles Quartile of plasma selenium (interval in mg/L)
N (participants) Plasma selenium, (lg/L) Age, mean (SE) Gender,% male (SE) Education < high school,% (SE) Body mass index (kg/m2), mean (SE) Smoking status,% (SE) Former Current Cotinine (ng/mL), mean (SE) Total cholesterol (mg/dl), mean (SE) LDL cholesterol (mg/dl), mean (SE) HDL cholesterol (mg/dl), mean (SE) Triglycerides (mg/dl), mean (SE)
Overall
Quartile 1 ( < 72.16)
1315 87.0 (0.63) 49.7 (0.2) 49.0 (0.0) 22.1 (1.0) 26.1 (0.1)
331 61.4 (0.48) 49.6 (0.8) 45.6 (2.5) 23.6 (2.2) 26.3 (0.2)
28.3 26.9 438.9 199.2 114.5 52.6 160.8
(1.2) (1.3) (30.1) (1.0) (0.9) (0.4) (1.9)
29.1 26.1 420.2 192.4 110.5 51.8 151.3
HDL, high-density lipoprotein; LDL, low-density lipoprotein.
(2.5) (2.5) (58.5) (2.0) (1.8) (0.8) (3.8)
Quartile 2 Quartile 3 (72.16–84.63) (84.63–99.31) 332 78.2 (0.21) 48.4 (0.8) 46.0 (2.5) 20.4 (2.2) 25.7 (0.2) 26.7 27.2 450.9 195.4 110.9 53.7 154.1
(2.5) (2.6) (60.5) (2.0) (1.9) (0.8) (4.0)
326 91.7 (0.24) 51.5 (0.8) 51.1 (2.5) 23.2 (2.2) 26.4 (0.2) 27.6 27.4 425.4 202.8 116.3 52.4 170.7
(2.5) (2.6) (61.4) (2.0) (1.8) (0.8) (4.4)
Quartile 4 ( > 99.31)
p Linear trend
326 116.5 (0.91) 49.3 (0.8) 53.3 (2.5) 21.3 (2.1) 26.0 (0.2)
0.74 0.06 0.81 0.74
29.7 27.1 459.0 206.1 120.4 52.4 167.2
(2.6) (2.6) (62.4) (2.1) (1.9) (0.8) (4.4)
0.76 0.69 0.67 < 0.001 < 0.001 0.88 < 0.001
1354
54.02 (0.86) 2.0 ( - 0.1, 4.2) 1.3 ( - 0.8, 3.4) 1.3 ( - 0.8, 3.3) 154.68 (4.39) 7.0 ( - 4.1, 18.0) 10.7 (0.1, 21.3) 10.7 (0.1, 21.3)
149.49 (4.1) 0 (Reference) 0 (Reference) 0 (Reference)
110.62 (2.1) 1.9 ( - 3.6, 7.3) 2.2 ( - 3.2, 7.7) 2.0 ( - 3.4, 7.5)
109.85 (1.98) 0 (Reference) 0 (Reference) 0 (Reference) 51.88 (0.79) 0 (Reference) 0 (Reference) 0 (Reference)
195.57 (2.15) 5.3 ( - 0.5, 11.1) 5.7 ( - 0.1, 11.4) 5.4 ( - 0.4, 11.2)
191.62 (2.24) 0 (Reference) 0 (Reference) 0 (Reference)
Quartile 2 (72.16–84.63)
171.81 (4.86) 17.1 (5.3, 28.8) 17.2 (5.8, 28.6) 17.2 (5.8, 28.5)
52.26 (0.88) 1.3 ( - 0.8, 3.4) 1.2 ( - 0.8, 3.2) 1.1 ( - 0.9, 3.2)
115.9 (1.93) 5.2 ( - 0.1, 10.4) 5.6 (0.4, 10.8) 5.4 (0.2, 10.6)
202.53 (2.11) 9.9 (4.1, 15.7) 10.2 (4.5, 15.9) 10.0 (4.2, 15.7)
Quartile 3 (84.63–99.31)
168.3 (4.86) 15.2 (3.6, 26.8) 17.2 (6.1, 28.3) 17.2 (6.1, 28.4)
52.17 (0.85) 1.2 ( - 0.9, 3.3) 0.8 ( - 1.2, 2.8) 0.9 ( - 1.1, 2.8)
120.77 (2.09) 11.5 (6.0, 16.9) 11.8 (6.4, 17.3) 12.2 (6.7, 17.6)
206.6 (2.31) 15.7 (9.7, 21.7) 16.1 (10.1, 22.1) 16.4 (10.5, 22.4)
Quartile 4 ( > 99.31)
8.6 (2.2, 15.1) 9.0 (2.9, 15.2) 9.0 (2.9, 15.2)
0.8 ( - 0.4, 2.0) 0.7 ( - 0.4, 1.8) 0.7 ( - 0.4, 1.9)
8.3 (3.0, 13.6) 8.6 (3.4, 13.8) 8.9 (3.7, 14.2)
11.4 (5.6, 17.2) 11.6 (5.9, 17.4) 12.0 (6.3, 17.8)
80th to 20th percentilea
< 0.001 < 0.001 < 0.001
0.18 0.22 0.21
< 0.001 < 0.001 < 0.001
< 0.001 < 0.001 < 0.001
p linear trend
0.53 0.37 0.37
0.53 0.77 0.82
0.01 0.01 < 0.001
0.07 0.07 0.05
p nonlinear trend
Model 1: Adjusted by age (years), sex (male, female), education ( < high school, ‡ high school), and fasting time (log hours). Model 2: Further adjusted by body mass index, smoking status (current, former, never), and categorized urine cotinine levels ( < 34, 34–500, and > 500 ng/mL). Model 3: Further adjusted by anti-hypercholesterolemia treatment. The 80th and 20th percentiles of blood selenium distributions are 68.69 and 103.37 lg/L, respectively. a Models comparing the 80th with the 20th percentiles of plasma selenium distributions and associated test for trend were obtained from linear regression models with log-transformed selenium as a continuous variable, except for total cholesterol and LDL where selenium was introduced as restricted quadratic splines with knots at percentiles 10th, 50th, and 90th because of the nonlinear relationships.
Total Cholesterol (mg/dl) Mean (SD) Model 1 Diff. (95% CI) Model 2 Diff. (95% CI) Model 3 Diff. (95% CI) LDL Cholesterol (mg/dl) Mean (SD) Model 1 Diff. (95% CI) Model 2 Diff. (95% CI) Model 3 Diff. (95% CI) HDL Cholesterol (mg/dl) Mean (SD) Model 1 Diff. (95% CI) Model 2 Diff. (95% CI) Model 3 Diff. (95% CI) Triglycerides (mg/dl) Mean (SD) Model 1 Diff. (95% CI) Model 2 Diff. (95% CI) Model 3 Diff. (95% CI)
Quartile 1 ( < 72.16)
Table 2. Change of Lipid Biomarkers Levels by Plasma Selenium Levels
GENE-ENVIRONMENT INTERACTION OF SELENIUM AND LIPIDS
1355
Total Cholesterol
HDL−Cholesterol
20
15
5
15
10
10
0
10
0
5
−5
5
−10
0
−10
0
50
75 125 Plasma Selenium (mg/L)
Differences (mg/dL)
20
Differences (mg/dL)
10
200
50
75 125 Plasma Selenium (mg/L)
200
Triglycerides 30
20
20
15
20
15
10
10
10
10
0
5
0
5
−10
0
−10
0
50
75
125
Plasma Selenium (mg/L)
200
Differences (mg/dL)
20
50
75
125
Weighted Percentage
30
Weighted Percentage
Differences (mg/dL)
LDL−Cholesterol
Weighted Percentage
20
Weighted Percentage
30
200
Plasma Selenium (mg/L)
FIG. 1. Adjusted differences (95% confidence interval [CI]) of lipid biomarkers by plasma selenium levels. The curve represents adjusted mean differences of lipid biomarkers based on restricted quadratic splines with knots at the 10th, 50th, and 90th percentiles (61.5, 84.4, and 114.5 lg/L respectively) of the plasma selenium distribution. The reference value (Difference = 0) was set at the 10th percentile of the plasma selenium distribution (61.5 lg/L). Models were adjusted for age, sex, education, fasting time (log hours), body mass index (BMI), smoking status (never, former, and current), and urine cotinine levels ( < 34, 34–500, and > 500 ng/mL) and anti-hypercholesterolemia treatment. The histogram represents the frequency distribution of plasma selenium in the study sample.
‰ FIG. 2. Candidate genes-selenium interaction–log10 p-values for adjusted differences of lipid biomarker concentrations. p-Values for the interaction of plasma selenium and 59 and 38 single nucleotide polymorphisms (SNPs) derived from linear and nonlinear (for total and LDL cholesterol) regression models, respectively, (dominant, recessive, and additive model) adjusted by age, sex, education, fasting time (log hours), BMI, smoking status (never, former, and current), urine cotinine levels ( < 34, 34–500, and > 500 ng/mL), and anti-hypercholesterolemia treatment are presented on the left y axis on the logarithmic scale according to the position of the SNPs on the chromosome (x axis). The horizontal solid line corresponds to a nominal significance level of 0.05. Horizontal dashed line corresponds to the Bonferroni-corrected significance level of 0.0008. To see this illustration in color, the reader is referred to the web version of this article at www.liebertpub.com/ars
- log10(P interaction) 7
6
5
4
3
2
1
7
6
5
4
3
2
1 2
2 4
4 7
7
7 8
LIPC: rs2070895
LDL Cholesterol
9
1
0 9
8
Chromosome
8 11 15
Chromosome
9
Triglycerides
8
1
0 9 11
9
11
ABCG1: rs1893590
CETP: rs4783962
15 16
16
ABCG1: rs1044317
4 ABCA1: rs2230806
Chromosome
11
15 16 18 21
15
16
18
21
ABCG1: rs2234719
2 FABP4: rs2290201
8 9
ABCG1: rs4148102
1 8
LIPG: rs9958947
2 7
LIPG: rs9958947
3 4
CETP: rs5882
4
ABCG8: rs4148211
ABCG1: rs1893590
CETP: rs4783962
UCP3: rs647126
FABP4: rs2290201
FABP1: rs2241883
8
APOA5: rs619054
5
UCP3: rs1800849
6 2
LPL: rs320
1
UCP1: rs12502572
2
FABP1: rs2241883
3
ABCG8: rs4953020
4
ABCG8: rs4148211
5
ABCA1: rs2230806
7
ABCG8: rs4953020
- log10(P interaction) 6
ABCG8: rs4953020 ABCG8: rs4148211
- log10(P interaction) 7
ABCG8: rs6544718
- log10(P interaction)
1356 GALAN-CHILET ET AL.
9
Total Cholesterol additive dominant recessive
1
0 21
Chromosome 21
9
HDL Cholesterol
8
1
0
GENE-ENVIRONMENT INTERACTION OF SELENIUM AND LIPIDS
LPL (rs10099160, rs11570892, rs256 and rs17410577), UCP1 (rs12502572), and FABP2 (rs1511025) (Figs. 2 and 3). Discussion
Increasing plasma selenium levels were associated with increasing levels of total, LDL, and TG cholesterol, but not HDL cholesterol. We also found an interaction of selenium levels with genetic variation in several genes related to lipid metabolic pathways, suggesting potentially inter-related biological pathways in selenium and lipid metabolism. This is the first study evaluating the interaction of selenium levels with lipid-related candidate genes on lipid levels using an epidemiologic sample representative of a population with a low burden of disease. The observed positive associations of selenium with total and LDL cholesterol were largely consistent with findings from cross-sectional studies (3). While few studies have evaluated the association between selenium and TG, they were also consistent with the positive association observed in our data. In the United States, participants in the highest quartile of serum selenium had 10% higher TG concentrations than participants in the lowest quartile (3). The associations between selenium and HDL cholesterol, however, were inconsistent across studies. The biological mechanisms underlying the association of plasma selenium with lipid concentrations in cross-sectional studies are not well understood. The liver has a key role in the regulation of both selenium and lipid levels. The liver accumulates small-molecule forms of selenium and hepatocytes secrete selenoprotein P, which binds selenium for plasma transport to various extrahepatic tissues (2). Selenium excess induced hepatic dysfunction in experimental studies, including alterations in lipid metabolism (8) and there is experimental and epidemiological evidence supporting the fact that high selenium levels may increase oxidative stress. Selenium was associated with redox changes in isolated hepatocytes and high levels of selenium damaged the liver of rats and increased the liver concentration 8-hydroxy-2¢-deoxyguanosine (2). Furthermore, plasma selenium levels above 110 lg/L were positively associated with urine 8-hydroxy-2¢-deoxyguanosine in our study sample (1). It is thus biologically possible that selenium excess can induce liver damage and disrupt lipid metabolism, although it is also possible that hepatic damage may simultaneously affect selenium and lipid metabolism. While selenium deficiency has also been related to liver damage in experimental animals (2), selenium and lipids were not associated with the lower range of selenium exposure in our study. For total and LDL cholesterol, we found Bonferroni-corrected statistically significant interactions of selenium with SNPs in FABP4 (rs2290201) and CETP (rs1800774), genes which are highly expressed in the hepatic and adipose tissue.
1357
Adipocyte fatty acid–binding protein-4 (FABP4/adipocyte P2) is involved in lipid transport and is positively associated with plasma lipid levels. In European Americans from the Cardiovascular Health Study, the minor alleles of six polymorphisms in FABP4 (including allele A for rs2290201) were associated with lower FABP4 levels (4). In our geneinteraction analysis, homozygotes for the minor allele in rs2290201 (T in our study population) presented lower total and LDL cholesterol with increasing selenium levels compared with carriers of CC and CT genotypes. CETP codifies for the cholesteryl ester transfer protein and is involved in the transfer of cholesteryl esters from HDL to other lipoproteins in the blood. CETP inhibition increases HDL cholesterol levels and decreases non-HDL cholesterol levels. In our study population, carriers of genotype CC in rs1800774 showed lower total and LDL-cholesterol levels with higher plasma selenium levels, although the interaction was not statistically significant for HDL cholesterol. In the Atorvastatin Comparative Cholesterol Efficacy and Safety Study (ACCESS) trial, four polymorphisms in the 5¢ end of CETP (VNTR, rs1800775, rs708272, and rs15723185) were significantly associated with higher concentrations of HDL cholesterol (7). Consistent with our results, rs1800774 was not associated with HDL cholesterol levels. The associations with total and LDL cholesterol, however, were not reported. In addition to gene-selenium interactions with FABP4 and CETP, we found suggestive evidence for interactions of selenium levels with other candidate genes associated with lipid levels in GWAS and metanalysis, including ABCG8, LPL, LIPC, APOA5, ABCA1, or LIPG. Large epidemiologic studies with sufficient power to evaluate gene-environment interactions are needed to replicate our findings, but our geneinteraction analyses provide biological support for the observed cross-sectional associations between selenium and lipids and support the fact that selenium and lipid metabolic pathways may be inter-related. While our findings support a biological connection between selenium and lipid levels, the cross-sectional nature of our study limits our ability to determine the direction of this association or whether common factors (such as liver dysfunction or dietary factors) are responsible for simultaneous changes in selenium and lipid levels. The only two prospective studies that have evaluated the association between selenium and lipids using longitudinal data found that baseline selenium levels did not predict future changes in lipid levels. Similarly, in the UK PRECISE Pilot trial, selenium yeast had modestly beneficial effects on plasma lipid levels at 6 months after randomization compared with placebo (6). Additional data are also available from two trials in which selenium supplementation was combined with other vitamins and minerals (9). In a population from rural China, supplementation with selenium combined with vitamins C and E
‰ FIG. 3. Candidate genes-selenium interaction–log10 p-values for adjusted odds ratio of altered lipid biomarkers levels. p-Values for the interaction of plasma selenium and 59 SNPs derived from logistic regression models (dominant, recessive, and additive model) adjusted by age, sex, education, fasting time (log hours), BMI, smoking status (never, former, and current), urine cotinine levels ( < 34, 34–500, and > 500 ng/mL), and anti-hypercholesterolemia treatment are presented on the left y axis on the logarithmic scale according to the position of the SNPs on the chromosome (x axis). The horizontal solid line corresponds to a nominal significance level of 0.05. Horizontal dashed line corresponds to the Bonferronicorrected significance level of 0.0008. To see this illustration in color, the reader is referred to the web version of this article at www.liebertpub.com/ars
- log10(P interaction) 6
5
4
3
2
1
6
5
4
3
2
1
2
2
4 7
4 7
7 8 9
8
Chromosome
8
1358 9
9 11
0
Chromosome
11
15
8
Low HDL Cholesterol
1
0
Chromosome
High Triglycerides
8
7
1
0
8
CETP: rs1800774
UCP3: rs647126 UCP3: rs2075577
FABP4: rs16909225 ABCA1: rs4149341
15 16 18 21
16 18 21
15 16 18 21
15
16
18
21
ABCG1: rs1044317
ABCG1: rs4148102
1 LIPG: rs9958947
11 LIPC: rs6083
4
11
ABCG1: rs4148107 ABCG1: rs2234719
7 2
APOA5: rs619054
1 9
UCP3: rs1800849
2 8
ABCA1: rs2230806
3 7
ABCA1: rs4149313
4 4
LPL: rs11570892
5 2
FABP4: rs2290201
6
LPL: rs17410577 LPL: rs256 LPL: rs10099160 LPL: rs11570892
1
UCP1: rs3811787
7
ABCG1: rs1893590
CETP: rs1800774
UCP3: rs647126
ABCA1: rs4149341
FABP1: rs2241883
ABCG8: rs6720173
2
UCP1: rs3811787
3
FABP1: rs2241883
4
FABP3: FABP3.345TC
5
FABP3: FABP3.345TC FABP3: rs12401792 ABCG8: rs6720173
- log10(P interaction) 6
ABCG8: rs4953020 ABCG8: rs4148211 FABP1: rs2197076 FABP1: rs2241883 MTTP: rs1800591 MTTP: rs3816873
- log10(P interaction)
7
FABP2: rs1511025
- log10(P interaction)
8
High Total Cholesterol additive dominant recessive
1
0
Chromosome
High LDL Cholesterol
GENE-ENVIRONMENT INTERACTION OF SELENIUM AND LIPIDS
resulted in small but significant increases in total and LDLcholesterol concentrations, but not in HDL-cholesterol (9). In the SU.VI.MAX study, the group assigned to selenium, b-carotene, zinc, and vitamins C and E had a borderline significantly higher frequency of hypercholesterolemia and elevated TG levels 7.5 years after randomization (9). Additional data from prospective studies, clinical trials, and mechanistic studies are needed to understand the causal effects underlying the associations between selenium and lipids observed in cross-sectional studies. In addition to the cross-sectional design, our study has other limitations. First, we only had access to nonfasting plasma samples. While total and HDL cholesterol levels can be measured in nonfasting samples, LDL cholesterol and TG levels are usually measured in fasting samples. There is recent evidence, however, supporting the fact that nonfasting LDL and TG levels are also associated with increased cardiometabolic risk (5). Second, we had only a single measurement of plasma selenium and lipids, which may be subject to within-person variability. Nondifferential measurement error, however, would induce an attenuation of the observed associations. Finally, total plasma selenium levels do not provide information of individual selenium compounds. More detailed analyses of different selenoprotein levels and activity are needed to better understand the association of selenium with lipid biomarkers. Strengths of our study include the sampling design that allows our findings to be generalized to a general population framework, the availability of detailed information for adjusting for multiple determinants of lipid levels, and the availability of genotyped SNPs for 19 candidate genes associated with lipid pathways. In conclusion, we observed approximately linear associations between plasma selenium and total, LDL-cholesterol, and TG levels and found that these associations were modified by genetic polymorphisms involved in lipid metabolic pathways, including lipid transport and transfer. As prospective epidemiologic studies and clinical trials with measurements of selenium and genome-wide genetic data become increasingly available, collaborative meta-analyses will allow us to establish the biological and clinical implications of our findings under different levels of selenium exposure. Notes Study population
We conducted a population-based survey among adults 15–85 years old residing in the catchment area of the Rio Hortega University Hospital in Valladolid (Spain) in 1997– 2003. In Spain, tertiary hospitals have assigned specific geographic areas for patient referral, and they integrate the network of primary care centers in each area. The complex study design and data collection methods have been described in detail elsewhere (1). The study sample consisted of 1502 participants with stored plasma samples for selenium and lipid determinations. We excluded 12 participants missing plasma selenium measurements, 2 participants missing lipid measurements, 53 participants with TG levels > 400 mg/dl, and 120 participants missing other relevant covariates, leaving 1315 participants for analysis. The research protocol was approved by Ethics Committee of the Rio Hortega University Hospital, and all participants provided written informed consent.
1359
Plasma selenium levels
Plasma selenium levels were measured by inductively coupled-plasma mass spectrometry (ICPMS) on an Agilent 7500CEx ICPMS (Agilent Technologies) following a standardized protocol. The lower and upper detection limits for plasma selenium levels were 29.85 and 205.30 lg/L, respectively. No sample had levels below or above these limits. The inter-assay coefficient of variation for plasma selenium levels was 5.6%. Plasma lipid levels
Plasma total cholesterol, HDL-cholesterol, and TG concentrations in nonfasting samples were determined using a Hitachi 917 analyzer. LDL cholesterol was calculated using the Friedewald formula, which is highly correlated with directly measured LDL-cholesterol in both fasting and nonfasting samples (5). Elevated total cholesterol was defined as > 200 mg/dl, elevated LDL cholesterol was defined as > 130 mg/dl, low HDL cholesterol was defined as < 40 mg/dl or < 50 mg/dl for men and women, respectively, and elevated TG was defined as > 150 mg/dl. Fasting time was defined as the number of hours from the last food or drink (excluding water) intake to blood extraction. DNA isolation, SNP selection, and genotyping
DNA was isolated from peripheral blood cells using Chemagic System (Chemagen), and its quality was assessed with PicoGreen dsDNA Quantification Reagent (Invitrogen). DNA was diluted to a final concentration of 100 ng/ll. We used bibliography searches and the SYSNPS program, based on public data sources, including Ensembl and HapMap, to identify 96 SNPs from 19 candidate genes implicated in lipid pathways, including fatty acid uptake, transport and metabolism, regulation of plasma TG levels, and lipoprotein metabolism and assembly. We included SNPs previously reported to be related to health outcomes in humans or to have functional implications. The SNPs were genotyped using an oligo-ligation-assay (SNPlex; Applied Biosystems) following the manufacturer’s guidelines. The mean genotyping coverage across all genotyped SNPs was 94%. Twenty-four SNPs were excluded because of genotyping errors or because they did not meet Hardy–Weinberg equilibrium ( p-value < 0.01). Thirteen SNPs were further excluded, because they had an MAF < 0.05 or a minor genotype frequency < 20 individuals in the study population. The final number of SNPs included in linear geneenvironment interaction analyses was 59. The complete list of finally selected SNPs is available in Table 3. In nonlinear SNPselenium interaction analysis, we additionally excluded 21 SNPs that had a minor genotype frequency lower than 5% above or below the median of selenium distribution to have the genotypes sufficiently represented across the range of selenium levels and obtain robust estimates of the nonlinear associations. Other variables
Information on age, sex, education, smoking status, and lipid-lowering medication was based on self-report. Body mass index (BMI) was calculated using measured height and weight. Urine cotinine was measured by enzyme-linked immunosorbent assay (ELISA) (Kit ‘‘Ana´lisis DRI Cotinina,’’
Table 3. Single Nucleotide Polymorphisms Finally Selected Gene name
Gene symbol
ATP-binding cassette, sub-family A (ABC1), member 1
ABCA1
ATP-binding cassette, sub-family G (WHITE), member 1
ABCG1
ATP-binding cassette, sub-family G (WHITE), member 8
ABCG8
Apolipoprotein A-V
APOA5
Cholesteryl ester transfer protein, plasma
CETP
Fatty acid binding protein 1, liver
FABP1
Fatty acid binding protein 2, liver
FABP2
Fatty acid binding protein 3, liver
FABP3
Fatty acid binding protein 4, liver
FABP4
Lecithin-cholesterol acyltransferase Lipase, hepatic
LCAT LIPC
Lipase, endothelial Lipoprotein lipase
LIPG LPL
Microsomal triglyceride transfer protein
MTTP
Paraoxonase 1 Mitochondrial uncoupling protein 1
PON1 UCP1
Mitochondrial uncoupling protein 2 Mitochondrial uncoupling protein 3
UCP2 UCP3
SNP rs2230808 rs4149313 rs7031748 rs2230806 rs1800977 rs2246293 rs2422493 rs4149341 rs1893590 rs1378577 rs4148102 rs4148107 rs2234719 rs1044317 rs6720173 rs4953020 rs4148211 rs6544718 rs619054 rs662799 rs10750097 rs4783962 rs1800774 rs5882 rs2197076 rs2241883 rs1511025 rs6857641 rs2282688 rs10034579 rs2271072 fabp3.345tc rs12401792 rs16909225 rs2305319 rs2290201 rs1109166 rs1077834 rs1800588 rs2070895 rs6083 rs9958947 rs17410577 rs256 rs320 rs10099160 rs11570892 rs3811800 rs1800591 rs3816873 rs2306985 rs854571 rs12502572 rs3811787 rs1800592 rs660339 rs647126 rs2075577 rs1800849
SNP, single nucleotide polymorphism.
1360
Allele T/C T/C G/T C/T G/A C/G G/A T/C A/C T/G G/A T/C C/T A/G G/C G/T A/G T/C G/A G/A G/A T/C C/T G/A G/A T/C T/C T/C T/C A/C C/G C/T G/T A/G T/C G/A T/C T/C C/T G/A A/G C/T G/C C/T T/G T/G A/G T/C G/T T/C C/G T/C G/A T/G T/C G/A A/G G/A G/A
Consequence Nonsynonymous Nonsynonymous Synonymous Nonsynonymous 5¢UTR 5¢UTR 5¢UTR 3¢UTR Intergenic Intergenic Intronic Intronic Intronic 3¢UTR Nonsynonymous Intronic Nonsynonymous Nonsynonymous 3¢UTR 5¢UTR 5¢UTR 5¢UTR Intronic Nonsynonymous Intronic Nonsynonymous Synonymous 5¢UTR 5¢UTR 5¢UTR Intronic Intronic 5¢UTR 5¢UTR Intronic Intronic Intronic 5¢UTR 5¢UTR 5¢UTR Nonsynonymous 5¢UTR Intronic Intronic Intronic Intronic 3¢UTR 5¢UTR 5¢UTR Nonsynonymous Nonsynonymous 5¢UTR Intronic 5¢UTR 5¢UTR Nonsynonymous 3¢UTR Synonymous 5¢UTR
GENE-ENVIRONMENT INTERACTION OF SELENIUM AND LIPIDS
Ref. 0395 Microgenics laboratories). One thousand sixty five individuals had cotinine concentrations below the limit of detection (34 ng/ml) (1). Statistical methods
We used the survey package in R software (version 3.0.3; R Development Core Team 2014) to account for the sampling design and survey weights. Plasma selenium concentrations were log transformed for analyses. Cut-offs for plasma selenium quartiles were based on weighted distributions in the study sample. We estimated the average differences in lipid concentrations (with 95% CIs) by selenium levels using multivariable linear regression models. As secondary analyses, we estimated the odd ratios for elevated total cholesterol, LDL cholesterol, and TG, or for low HDL cholesterol (with 95% CIs) using multivariable logistic regression models. Plasma selenium concentrations were introduced in the models as quartiles, comparing each of the three highest quartiles of plasma selenium with the lowest quartile, or as a log-transformed (continuous) variable to compare lipid levels in the 80th versus the 20th percentiles of the selenium distribution (interquintile range). Statistical models were initially adjusted for sex, age, education ( < high school, ‡ high school), and fasting time (logtransformed number of hours) (Model 1). We additionally adjusted for BMI, smoking status (never, former and current smoker), and urine cotinine levels ( < 34, 34–500 and > 500 ng/mL) (Model 2). Finally, we further adjusted for use of lipid-lowering medications (Model 3). p-values for linear trend were obtained from Wald tests for log-transformed plasma selenium in regression models. To further explore the shape of the relationship between plasma selenium and lipid biomarkers, we modeled plasma selenium levels as restricted quadratic splines with knots at the 10th, 50th, and 90th percentiles of the plasma selenium distribution (61.5, 84.6, and 114.6 lg/L respectively). p-values for nonlinearity were obtained from Wald tests for the nonlinear spline terms. We conducted sensitivity analyses using not-transformed plasma selenium levels (instead of log-transformed), with essentially identical results. Gene-environment interaction analyses were based on interaction terms for log-transformed plasma selenium levels with indicator variables for genotypes. For each SNP, we estimated three models for the interaction with selenium in separate models assuming dominant, recessive, and additive inheritance, and calculated the p-values for interaction in each model by using Ftests for global significance by comparing nested models with and without the corresponding interaction terms. Since candidate genes for the gene-environment interaction study were selected based on a priori knowledge of their role in lipid metabolic pathways, the selected alpha level for statistical significance was 0.05. For descriptive purposes, however, we calculated a Bonferroni-corrected p-value of 0.0008 (estimated as 0.05 divided by 59 finally selected SNPs). For the Bonferroni-corrected significant SNPs, in cases where only 1 inheritance model showed statistically significant SNP-selenium interactions, we reported the associations of selenium and lipid levels among subgroups of participants with the genotypes of interest in this specific model. If more than 1 inheritance model showed statistically significant SNP-selenium interactions, we reported the best fitting inheritance model selected by comparing a general model that included separate dummy variables for the heterozygote and
1361
minor allele homozygote (reference major allele homozygote) with the models assuming dominant (minor allele homozygote and heterozygote versus major allele homozygote), recessive (minor allele homozygote versus heterozygote and major allele homozygote), or additive inheritance (0, 1, or 2, minor allele dosage). Acknowledgments
This work was supported by the Strategic Action for Research in Health sciences [CP12/03080, PI10/0082, PI13/ 01848, and PI11/00726], GRUPOS 03/101; PROMETEO/ 2009/029 and ACOMP/2013/039 from the Valencia Government, GRS/279/A/08 from Castilla-Leon Government, and European Network of Excellence Ingenious Hypercare (EPSS037093) from the European Commission; CIBER Fisiopatologı´a Obesidad y Nutricio´n (CIBERobn) [CIBER-02-08-2009, CB06/03 and CB12/03/30016]; and CIBER de Diabetes y Enfermedades Metabo´licas Relacionadas (CIBERDEM). The Strategic Action for Research in Health Sciences, CIBEROBN and CIBERDEM are initiatives from Carlos III Health Institute Madrid and the Spanish Ministry of Economy and Competitiveness and are co-funded with European Funds for Regional Development (FEDER). References
1. Galan-Chilet I, Tellez-Plaza M, Guallar E, De Marco G, Lopez-Izquierdo R, Gonzalez-Manzano I, Carmen Tormos M, Martin-Nunez GM, Rojo-Martinez G, Saez GT, MartinEscudero JC, Redon J, and Javier Chaves F. Plasma selenium levels and oxidative stress biomarkers: a gene-environment interaction population-based study. Free Radic Biol Med 74: 229–236, 2014. 2. Ho¨gberg J and Alexander J. Selenium. In: Handbook on the Toxicology of Metals, edited by Nordberg GF, Fowler BA, Nordberg M, Friberg LT. London, Waltham, San Diego: Academic Press, 2007, pp. 783–807. 3. Laclaustra M, Stranges S, Navas-Acien A, Ordovas JM, and Guallar E. Serum selenium and serum lipids in US adults: national Health and Nutrition Examination Survey (NHANES) 2003–2004. Atherosclerosis 210: 643–648, 2010. 4. Mukamal KJ, Wilk JB, Biggs ML, Jensen MK, Ix JH, Kizer JR, Tracy RP, Zieman SJ, Mozaffarian D, Psaty BM, Siscovick DS, and Djousse L. Common FABP4 genetic variants and plasma levels of fatty acid binding protein 4 in older adults. Lipids 48: 1169–1175, 2013. 5. Nordestgaard BG and Benn M. Fasting and nonfasting LDL cholesterol: to measure or calculate? Clin Chem 55: 845– 847, 2009. 6. Rayman MP, Stranges S, Griffin BA, Pastor-Barriuso R, and Guallar E. Effect of supplementation with high-selenium yeast on plasma lipids: a randomized trial. Ann Intern Med 154: 656–665, 2011. 7. Thompson JF, Durham LK, Lira ME, Shear C, and Milos PM. CETP polymorphisms associated with HDL cholesterol may differ from those associated with cardiovascular disease. Atherosclerosis 181: 45–53, 2005. 8. Wang X, Zhang W, Chen H, Liao N, Wang Z, Zhang X, and Hai C. High selenium impairs hepatic insulin sensitivity through opposite regulation of ROS. Toxicol Lett 224: 16–23, 2014. 9. Zhang L, Gail MH, Wang YQ, Brown LM, Pan KF, Ma JL, Amagase H, You WC, and Moslehi R. A randomized factorial study of the effects of long-term garlic and micro-
1362
GALAN-CHILET ET AL.
nutrient supplementation and of 2-wk antibiotic treatment for Helicobacter pylori infection on serum cholesterol and lipoproteins. Am J Clin Nutr 84: 912–919, 2006.
Address correspondence to: Dr. Maria Tellez-Plaza Area of Cardiometabolic and Renal Risk Biomedical Research Institute Hospital Clinic of Valencia (INCLIVA) Av. Menendez Pidal 4 accesorio Valencia 46010 Spain E-mail: [email protected] [email protected]
Date of first submission to ARS Central, January 5, 2015; date of acceptance, January 20, 2015.
Abbreviations Used BMI ¼ body mass index GWAS ¼ Genome Whole Association Studies HDL ¼ high density lipoprotein ICPMS ¼ inductively coupled-plasma mass spectrometry LDL ¼ low density lipoprotein OR ¼ odds ratio SNP ¼ single nucleotide polymorphism TG ¼ triglycerides
