Se is an essential micronutrient that is incorporated specifically as selenocysteine into the active site of selenoproteins (such as the glutathione peroxidases; GPX). Se as an antioxidant helps to maintain intracellular redox balance, promoting a reducing environment and thereby limiting oxidative damage caused by free radicals. It has been suggested that Se intakes above the level necessary to saturate selenoproteins can reduce the risk of cancer. In the Nutritional Prevention of Cancer Trial published in 1996, a daily supplement of 200 μg Se, as Se-enriched yeast, in patients with a history of basal cell or squamous cell carcinoma of the skin showed a significant increase in recurrence of skin cancer, but a significant reduction in overall cancer risk, in risk of cancers with a more severe prognosis, such as prostate, colorectal and lung cancer, and consequently a decrease in total cancer mortality(Reference Clark, Combs and Turnbull1–Reference Duffield-Lillico, Slate and Reid3). The strongest effect was found in individuals with a baseline plasma Se (pSe) level below 106 μg/l, while no protective effect was found in the upper tertile(Reference Duffield-Lillico, Reid, Turnbull, Combs, Slate, Fischbach, Marshall and Clark4). Most of the participants in the Nutritional Prevention of Cancer Trial had initial Se intakes above the level required to saturate selenoproteins, indicating that the observed effects of Se are not limited to antioxidant functions through selenoproteins. Based on results from numerous rodent cancer models, it has also been proposed that alterations in carcinogen metabolism resulting in increased enzymic defence play a role in the cancer-protective effect of Se. But results are far from consistent and may vary with choice of animal model, choice of carcinogen, Se source, Se dosage level and tissue studied(Reference Combs and Gray5–Reference El-Bayoumy and Sinha7).
Se deficiency has been associated with a decrease in GPX activity as well as in mRNA levels(Reference Rao, Puschner and Prolla8). Besides affecting expression and activity of selenoproteins, Se is also involved in the regulation of other gene products. Se supplementation has been reported to increase blood glutathione (GSH) levels, indicating reduced oxidative stress(Reference El-Bayoumy, Richie, Boyiri, Komninou, Prokopczyk, Trushin, Kleinman, Cox, Pittman and Colosimo9). Glutathione reductase (GR) is the principal enzyme maintaining GSH in its reduced form. GSH is also a cofactor for GPX and is necessary for GSH conjugation by glutathione S-transferases (GST). γ-Glutamylcysteine ligase (GCL) composed of a catalytic (GCLC) and a modulatory (GCLM) subunit is the first step in the de novo synthesis of GSH(Reference Gipp, Chang and Mulcahy10, Reference Gipp, Bailey and Mulcahy11). Common for GCLC and GCLM and other genes encoding phase 2 detoxifying enzymes, including NAD(P)H:quinone oxidoreductase (NQO1) and GST, is the presence of electrophile response elements (EpRE) in the promoter regions (TGACNNNGC), which mediate the transcriptional up regulation in response to a variety of inducing agents(Reference Chen and Kong12). The AP-1 transcription factor Fos-related antigen 1 (Fra1) may contain an EpRE in its promoter region(Reference Yoshioka, Deng, Cavigelli and Karin13) and is able to down regulate NQO1 and GCLC gene expression(Reference Venugopal and Jaiswal14, Reference Jardine, MacNee, Donaldson and Rahman15) through embedded AP-1 sites (TGACTCA)(Reference Li and Jaiswal16, Reference Mulcahy and Gipp17). Se could induce or repress these gene responses through EpRE.
Up regulation of genes encoding phase 1 xenobiotic metabolising enzymes in the liver has been reported in Se-deficient mice(Reference Reiter and Wendel18). Some xenobiotics interact with the aryl hydrocarbon receptor (AhR) and transactivate target genes such as cytochrome P450 through the xenobiotic response element (XRE). The aryl hydrocarbon receptor repressor (AhRR) regulates AhR function by a negative feed-back mechanism, thereby down regulating genes that are regulated by AhR(Reference Mimura, Ema, Sogawa and Fujii-Kuriyama19). The promoter region of AhRR contains 3 XRE binding sites, and could potentially be affected by Se supplementation(Reference Cauchi, Stucker, Cenee, Kremers, Beaune and Massaad-Massade20).
In Denmark, the average daily dietary Se intake (excluding supplements) is low compared with other countries(Reference Rayman21), being 42 μg/d for men and 33 μg/d for women according to the latest survey(Reference Lyhne, Christensen, Groth, Fagt, Biltoft-Jensen, Hartkopp, Hinsch, Matthiessen, Møller, Saxholt and Trolle22). But Se supplementation through multivitamin and mineral supplements is very common, adding on average 25 μg. We assumed that a population like the Danish, with relatively low Se intakes, would be appropriate for studying the biological effects of Se supplementation.
The aim of the present study was to investigate whether the dietary Se intake of elderly Danes is sufficient to obtain saturation of GPX expression in different blood compartments, and whether increased Se intake will affect gene expression and activity of selected phase 2 enzymes and transcription factors containing EpRE or XRE in their promoter, as has been reported in animal studies. We therefore investigated whether a daily Se supplement of 100, 200 or 300 μg as Se-enriched yeast for 5 years would increase activities of GPX, GR and GST in erythrocytes, plasma or thrombocytes, and modulate gene expression of Fra1, GCLC, GPX1, AhRR and NQO1 in leucocytes. The use of three dosage levels enabled us to look closer at dose–response relationships. We also investigated the overall effect of Se supplementation by pooling the supplemented groups and testing them against the placebo group.
Materials and methods
Subjects and study design
In 1999, a cohort of 500 men and women was established in Denmark. Participants were recruited from the county of Funen, were between the ages of 60 and 74 years, had no previous cancer diagnosed, had no severe or life-threatening diseases and a daily intake of Se supplementation not exceeding 50 μg (a typical multivitamin/mineral supplement sold in Denmark contains 40–50 μg inorganic Se). A 1-month run-in period where subjects were supplemented with placebo tablets ensured exclusion of participants who were not compliant. The study was a randomised, double-blinded, placebo-controlled intervention experiment with a parallel design. At recruitment each participant was randomised to one of four groups, a placebo group and three dosage levels: 100, 200 or 300 μg Se as Se-enriched yeast (SelenoPrecise®; Pharma Nord, Vejle, Denmark). The Se-enriched yeast has previously been reported to consist of about 81 % selenomethionine in the fraction liberated by proteolysis(Reference Larsen, Hansen, Paulin, Moesgaard, Reid and Rayman23). At study entry and at 6, 12, 18, 24, 36 and 60 months the participants visited the centre to donate a blood sample, collect tablets for the following period and answering questions concerning compliance and side effects, such as garlic breath, hair loss and nail brittleness. Based on counting of returned pill packets after years 1, 2 and 3, compliance was >96 %. No serious adverse effects were reported and there was no correlation to Se dose (details will be published elsewhere). The supplementation lasted 5 years. The last blood samples were collected after 5 years of supplementation and the present study was performed on a random sub-sample of 105 participants, selected consecutively among those who attended the centre for the terminal visit within a certain time period. All participants gave oral and written consent according to the second Helsinki declaration. The study was approved by the regional ethical committees of Vejle and Funen (journal no. 19980186).
Blood sampling and storage
After supine rest for 10 min, blood samples were collected from non-fasting participants in 10 ml EDTA-coated tubes (Becton Dickinson, Franklin Lakes, NJ, USA), and centrifuged for 15 min (150 g; room temperature). Thrombocytes were isolated from the thrombocyte-rich plasma by centrifugation (1500 g; room temperature; 5 min). The thrombocyte pellet was washed twice in 0·9 % NaCl and re-suspended in 0·5 ml 0·9 % NaCl. Erythrocytes and buffy coat were isolated by centrifugation (1500 g; room temperature; 10 min). The buffy coat was re-suspended in RLT buffer (QIAgen, Ballerup, Denmark). Erythrocytes were lysed by adding an equal amount of MilliQ water to the tube. Plasma, thrombocytes, buffy coat and erythrocytes were stored at − 80°C.
Plasma selenium status
pSe concentration was determined as previously described using an ELAN 6100 ICP-DRC-MS (PerkinElmer SCIEX, Concord, ON, Canada) equipped with an AS-91 autosampler (PerkinElmer, Norwalk, CT, USA)(Reference Sloth, Larsen, Bugel and Moesgaard24). For every fifteen unknown plasma samples one duplicate sample, one field blank, one certified reference sample and three Se (10 μg/l) standard solutions were included. The latter were used to correct for instrumental drift. Based on the determination of field blanks the limit of detection was estimated at 0·1 μg Se/l and the uncertainty was estimated from the differences between double determinations at 2·1 % relative sd. The accuracy was determined by including a reference sample with a known content of 73 (sd 8) μg/l (Seronorm Serum; Nycomed Pharma Diagnostics, Oslo, Norway). We found an Se concentration of 77 (sd 6) μg/l in this sample.
Defence enzyme activities (glutathione peroxidase, glutathione reductase and glutathione S-transferase)
Erythrocyte, plasma and thrombocyte GPX and GR activities were determined spectrophotometrically on a Cobas Mira analyser (F. Hoffmann-La Roche Ltd, Basel, Switzerland) according to Wheeler et al. (Reference Wheeler, Salzman, Elsayed, Omaye and Korte25). t-Butyl hydroperoxide was used as substrate for GPX. GST activity was measured in thrombocytes and erythrocytes as previously described(Reference Dragsted, Pedersen, Hermetter, Basu, Hansen, Haren, Kall, Breinholt, Castenmiller, Stagsted, Jacobsen, Skibsted, Rasmussen, Loft and Sandström26). Enzyme activities measured in erythrocyte lysates were related to the amount of Hb in the sample. Enzyme activities determined in thrombocytes and plasma were related to the amount of total protein. Hb and total protein contents were determined using commercially available kits (catalogue no. HG 980, Randox, Crumlin, Co. Antrim, UK; catalogue no. A11A01669, ABX Diagnostics, Montpellier, France, respectively). Control samples (erythrocytes, plasma and thrombocytes) were included for every tenth sample analysed. Inter- and intraday variations for the control samples were < 7 %.
RNA isolation and cDNA synthesis
Total RNA was extracted from the buffy coat using the QIAamp® RNA blood mini kit (QIAgen, Ballerup, Denmark). The concentration and purity of the RNA was determined by measuring the absorbance at 260 and 280 nm using the NanoDrop® ND-1000 UV-Vis spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). A ratio (A260:A280) of >1·8 indicated suitable purity. First-strand cDNA was synthesised using 500 ng total RNA in a final volume of 10 μl with Random Hexamer primer and SuperscriptTM II Rnase H-RT according to the manufacturers (Invitrogen, Taastrup, Denmark).
Primer and probe design
Oligonucleotide primers and TaqMan® probes were designed with Primer Express software v. 1.5 (Applied Biosystems, Stockholm, Sweden), based on sequences from the Genbank database (Table 1). Genomic DNA amplification was excluded whenever possible by designing intron spanning probes.
Oligonucleotides and TaqMan® fluorogenic probes
GPX1, glutathione peroxidase 1; 6-FAM, 6-carboxyfluorescein; MGB, minor groove binding; GCLC, γ-glutamylcysteine ligase catalytic subunit; NQO1, NAD(P)H quinone oxidoreductase 1; Fra1, FOS-like antigen 1; AhRR, aryl-hydrocarbon receptor repressor.
Leucocyte mRNA quantification of Fos-related antigen 1, glutathione peroxidase 1, γ-glutamylcysteine ligase catalytic subunit, NAD(P)H:quinone oxidoreductase and aryl hydrocarbon receptor repressor. Real-time PCR of the five genes was performed on an ABI PRISM® 7900HT Sequence Detection System (Applied Biosystems, Stockholm, Sweden). The PCR reaction was determined in a 20 μl final volume adding 2 ng cDNA, using TaqMan® Universal PCR Master Mix (Applied Biosystems, Stockholm, Sweden). For each target gene, the probe concentration was 0·3 μm and the primer concentrations for the detection of GCLC, AhRR and NQO1 genes were 0·3 μm, 0·5 μm for GPX1, and 0·9 μm for Fra1. The constitutively expressed 18S rRNA primer/probe-set (Applied Biosystems, Stockholm, Sweden) was selected as an endogenous control to correct for potential variation in RNA loading or efficiency of the amplification reaction and used as recommended by the manufacturers.
The PCR amplification was performed in ninety-six-well plates in triplicates using the following cycling protocol: 50°C for 2 min, 95°C for 10 min, forty-five cycles at 95°C for 15 s and 60°C for 1 min. To confirm equal amplification efficiencies, we used the criterion of a regression slope of less than 0·1 for each target gene normalised to 18S rRNA. This confirms that we could use the comparative cycle threshold (Ct) method for the relative quantification of target without running standard curves on the same plate (data not shown). Relative changes in gene expression were calculated by the ΔΔCt method (Applied Biosystems, 2001, User Bulletin no. 2, ABI PRISM 7700 Sequence Detection System, Foster City, CA, USA). Inter-plate variation for a control sample (n 27) was < 2 % for all genes.
Statistical analysis
Treatment effects were analysed with the general linear model, with sex, smoking and their interactions with treatment included in the model. Weight was significantly higher among men than women, and was therefore included in the model as a covariate. Since a large proportion of subjects reported taking multivitamin/mineral supplements, use of additional supplements was included in the statistical analyses as a covariate. For comparison of two groups, t tests were used (placebo against pooled Se-supplemented groups). Data that could not meet the criteria of variance homogeneity and normal distribution after log transformation were analysed by non-parametric tests (Kruskal–Wallis H test or Mann–Whitney U test). Group differences for categorical data were analysed using the χ2 test. For all tests a P value less than 0·05 was considered statistically significant. Pearson correlations were used to identify correlating variables. All statistical analyses were performed using SAS (SAS Institute, Cary, NC, USA).
Results
Study population
Baseline characteristics for the study participants are presented in Table 2 as mean values and standard deviations (age, weight) or as percentages (smoking status, sex). The participants were predominantly non-smokers and in the selected sub-sample for the present study women assigned to the highest dosage level were older compared with women in the lowest and medium dosage groups (Table 2). Women had significantly lower body weight compared with men (P < 0·001). Otherwise, there were no statistically significant differences in baseline characteristics between the groups.
(Mean values and standard deviations or percentages)
* Compared with men, women had significantly lower body weight (P < 0·001; ANOVA).
† Highest dosage group was significantly different from lowest dosage group (100 μg/d) and medium dosage group (200 μg/d) (P < 0·05; ANOVA).
Records of the use of multivitamin/mineral supplements among the participants (registered at time of final blood sampling, year 5) indicate that on average 36 % took additional supplements (as ordinary multivitamin/mineral tablets); however, the percentage of supplement users was about twice as high among participants from the highest dosage group compared with the rest (Table 2). No participants reported use of specific Se supplements.
Compliance
pSe concentration was measured to test for compliance. Se status increased linearly (R 2 0·98; P < 0·0001) with increasing Se intake (Table 3), indicating a high degree of compliance. Smoking status did not influence pSe (P>0·05) (data not shown).
(Medians and 10th and 90th percentiles and means and standard deviations)
* Significantly different from placebo group (P < 0·05; ANOVA).
Enzyme activities
There was a statistically significant effect on thrombocyte GPX activity (tGPX) (P = 0·003), with a statistically significant treatment × sex interaction (P = 0·008). When subjects were stratified according to sex, we found an effect of Se supplementation on tGPX activity in women only (P women = 0·006) (Table 3). tGPX activity was significantly increased in groups supplemented with 200 or 300 μg/d (P placebo − 200 μg = 0·001; P placebo − 300 μg = 0·008), while there was a tendency towards increased tGPX activity at the lowest dosage level (P placebo − 100 μg = 0·084) compared with placebo. We found no statistically significant difference in tGPX between the two highest dosage levels (P 200 μg–300 μg = 0·688). No effect of Se supplementation on tGPX was observed in men (P men = 0·401). We found no effect of Se supplementation on plasma GR or GPX, on erythrocyte GR, GPX or GST, or on GR or GST measured in thrombocytes (Table 3). We found no interaction with smoking on any of the activities.
When all supplemented groups were pooled and tested against the placebo group, only tGPX was significantly increased with Se supplementation (P = 0·027) and, again, only women responded (P women = 0·004; P men = 0·596). pSe correlated with plasma GPX activity (P = 0·046; r 0·212) but not tGPX (P = 0·106) or erythrocyte GPX (P = 0·162).
Gene expression
Se supplementation did not affect gene expression of GPX1, Fra1, GCLC, NQO1 or AhRR in leucocytes (Table 4), when all groups were tested against each other. There was a large inter-individual variation in the expression of the transcription factors Fra1 and AhRR, especially in the placebo group (Table 4).
(Medians and 10th and 90th percentiles and means and standard deviations)
When the Se-supplemented groups were pooled and tested against placebo, we found a statistically significant decrease in expression of Fra1 and GCLC (P = 0·019 and P = 0·042, respectively) and a tendency towards a decreased expression of NQO1 (P = 0·067) and AhRR (P = 0·085). There was no effect of Se on GPX1 (P = 0·281). AhRR was the only gene expression marker influenced by smoking status. Smoking increased the expression of AhRR (P = 0·0001) (Fig. 1), an effect that was independent of Se supplementation.
Effect of smoking on gene expression of aryl hydrocarbon receptor repressor (AhRR) relative to the mean expression in the placebo group. Values are means, with standard deviations represented by vertical bars. Participants are sub-divided according to number of cigarettes smoked per d: 0 (n 77); 1–5 (n 8); 6–19 (n 11); +20 (n 6). Pipe and cigar smokers are placed in the group with smokers of 1–5 cigarettes per d. * Mean value is significantly different from that of non-smokers (P < 0·05; Kruskal–Wallis).
Discussion
The anti-carcinogenic effects of Se have been linked to increased expression of protective enzymes. In the present study, we investigated the effect of 5 years of Se supplementation on the activity and gene expression of selected selenoproteins and genes involved in carcinogen metabolism. The pSe concentration was determined to test for compliance and we found a statistically significant increase in pSe with increasing Se supplement intake. It has previously been shown in numerous studies that pGPX and eGPX activities increase with Se intake, and that the increase reaches a plateau at intakes corresponding to pSe concentrations of about 70–90 μg/l and higher(Reference Alfthan, Aro, Arvilommi and Huttunen27, Reference Thomson, Robinson, Butler and Whanger28). In the present study pSe concentrations in the placebo group were between 71 and 111 μg/l with a mean of 92 μg/l. If we assume that the pSe in the control group is representative of the mean Se status in the participants before entry into the study, the lack of effect on pGPX and eGPX activities could be explained by these high baseline values, indicating that GPX is saturated in erythrocytes and plasma. We found a statistically significant increase in tGPX activity with Se supplementation, using Se-enriched yeast which consists primarily of the organic Se compound selenomethionine(Reference Larsen, Hansen, Paulin, Moesgaard, Reid and Rayman23). A sub-group analysis showed that the effect of Se was sex specific. Se supplementation only affected tGPX activity in women, and only at the two highest dosage levels. The responses at these two levels were not statistically different, suggesting that tGPX activity is saturated at pSe concentrations between 165 and 221 μg/l (means of the two groups). This is a bit higher than previously reported where tGPX activity plateaued at pSe concentrations < 150 μg/l(Reference Alfthan, Aro, Arvilommi and Huttunen27–Reference Nève30). tGPX activity has been reported to saturate at higher pSe concentrations compared with eGPX and pGPX(Reference Nève, Vertongen and Capel31), which could explain the observed positive effect on GPX activity in thrombocytes. We are not aware of other studies reporting this sex-specific difference in tGPX activity in response to Se supplementation, and we cannot exclude that a similar effect would have been observed in men, had the number of participants been higher. Although not statistically significant, we note that GPX activity in all blood fractions increased with Se supplementation. In the present study, five years of Se supplementation did not affect GR activity in plasma, erythrocytes or thrombocytes, or GST activity in erythrocytes or thrombocytes.
Large inter-individual differences in response to Se supplementation have previously been reported(Reference Brown, Pickard, Nicol, Beckett, Duthie and Arthur32). Brown et al. found that the lower the baseline activity, the greater was the increase in response to Se supplementation. Thus it is possible that we would have seen effects of Se supplementation on GPX activity in erythrocytes and plasma if we had had baseline measurements available, thereby allowing us to investigate the individual responses, at least in individuals having low baseline levels.
GPX activity was measured in three different blood compartments, and only pGPX correlated weakly with pSe concentration. Nève et al. have previously reported significant correlations between pSe and pGPX and tGPX activities(Reference Nève, Vertongen and Capel31) but we were not able to show any correlations between pSe and eGPX or tGPX activity. We did not measure Se status in erythrocytes or thrombocytes and it is possible that Se is channelled differently into various blood compartments as GPX synthesis is requested.
Smoking has been reported to decrease Se status and GPX activity(Reference Bukkens, de Vos, Kok, Schouten, de Bruijn and Hofman33, Reference Ravn-Haren, Olsen, Tjonneland, Dragsted, Nexo, Wallin, Overvad, Raaschou-Nielsen and Vogel34). We found no interactions with smoking. One possible explanation could be that the group of smokers in the present study was much smaller than the group of non-smokers, and individual smokers were very different regarding smoking habits. Some participants smoked very rarely while others smoked every day.
It has been hypothesised that Se at pharmacological doses would influence expression of phase 1 and phase 2 genes containing XRE or EpRE in their promoters(Reference El-Bayoumy and Sinha7). We found no dose–response relationship on expression of AhRR, GPX1, GCLC, NQO1 or Fra1 in leucocytes. Pooling the Se groups and testing them against the placebo group showed decreased expression of Fra1, GCLC and NQO1, although the latter did not reach statistical significance. Fra1 has been reported to down regulate GCLC and NQO1 (Reference Venugopal and Jaiswal14, Reference Jardine, MacNee, Donaldson and Rahman15). If this were the case in the present study, we would have expected to see an up regulation of Fra1. Instead we found the opposite. This may indicate that Fra1 is not involved in the observed down regulation of GCLC and NQO1. We found increased AhRR gene expression with smoking. Cigarette smoke contains carcinogens such as polycyclic aromatic hydrocarbons that are metabolised by cytochrome P450 enzymes. Therefore AhRR is co-induced with cytochrome P450 enzymes via XRE(Reference Tsuchiya, Nakajima, Itoh, Iwanari and Yokoi35). Whether increased AhRR expression in leucocytes is a marker for smoking needs further investigation. None of the other genes were transcribed differently as a function of smoking and there were no significant interactions between smoking and pSe on gene transcription.
Gene expression of phase 2 enzymes has been shown to vary with the studied organ, and results on whether blood expression profiles can be used as surrogate markers of organ expression are conflicting(Reference Carlsen, Myhrstad, Thoresen, Moskaug and Blomhoff36, Reference Whitney, Diehn, Popper, Alizadeh, Boldrick, Relman and Brown37). Finnstrom et al. found no correlation between blood and liver mRNA expression patterns of cytochrome P450 genes, but they noted that genes that were highly expressed in the liver at the same time could be expressed at lowered levels in the blood(Reference Finnstrom, Thorn, Loof and Rane38). Rauchy et al. on the other hand found good correlation(Reference Raucy, Schultz, Wester, Arora, Johnston, Omdahl and Carpenter39). Results are inconclusive and further studies as to whether a blood sample can be used to determine the expression profile in target organs are needed. Furthermore, leucocytes are a mixture of many different cell types, and it has been shown that the profile of these cells vary from subject to subject based on cell type counts(Reference Whitney, Diehn, Popper, Alizadeh, Boldrick, Relman and Brown37). There might therefore be different expression profiles among different leucocyte cell types. Decreased gene expression of phase 2 enzymes in the blood compartment reported here may thus reflect changed composition of the leucocyte cell pool. It may also reflect altered expression in other tissues. In a study in transgenic mice containing a GCLC-luc construct, luciferase activity increased in brain and muscle but decreased in liver following treatments with berry extracts or ellagic acid(Reference Carlsen, Myhrstad, Thoresen, Moskaug and Blomhoff36), indicating that complex up and down regulation may take place in different tissues as a response to redox active dietary components. We speculate that Se compounds may elicit similar complex actions in different tissues leading to an overall down regulation of EpRE-regulated genes in blood leucocytes.
We conclude that Se supplementation affects tGPX activity differently in men and women. Se status in healthy elderly Danish women is insufficient to obtain saturation of tGPX activity and higher Se intake might increase oxidative defence in this blood compartment. The present results indicate that Se supplementation can down regulate leucocyte genes having EpRE in their promoter and that Fra1 does not seem to be involved in down regulation of GCLC and NQO1 with Se. Decreased expression of EpRE-regulated genes could potentially increase the risk of cancer. However, further studies are needed to establish whether this down regulation in leucocytes reflects a similar expression pattern in other tissues.
Acknowledgements
We thank all those who volunteered for the study, Vibeke Kegel, Marianne Hansen, Geert Eisensee and her staff for excellent technical support, and Asta Christiansen and Hanne Hansen for their dedicated handling of the participants in the study. The present study was financially supported by a grant from the Danish Ministry of the Interior and Health, Research Centre for Environmental Health Fund.
The authors' responsibilities were as follows: S. C., K. O. and S. M. designed the original trial; L. O. D. designed the present sub-study; L. O. D. and E. H. L. obtained funding; S. C. supervised blood sample collection; B. N. K. was responsible for gene expression analyses; E. H. L. was responsible for Se analyses; G. R. H. was responsible for enzyme assays, undertook the statistical analyses and wrote the first draft of the manuscript; K. O. and L. O. D. provided statistical support. All authors approved the final manuscript. All authors declared that they had no conflict of interest.
