Available online at www.sciencedirect.

com

ScienceDirect
Journal of Nutritional Biochemistry 48 (2017) 112 – 119

Selenium increases hepatic DNA methylation and modulates one-carbon metabolism

in the liver of mice☆

Bodo Speckmann a , Sarah Schulz a , Franziska Hiller a , Deike Hesse b, c , Fabian Schumacher d, e ,
Burkhard Kleuser d, f , Jürgen Geisel g, Rima Obeid g, h , Tilman Grune a, c, i, Anna P. Kipp j,⁎
a
German Institute of Human Nutrition Potsdam-Rehbruecke, Department of Molecular Toxicology, Nuthetal, Germany
b
German Institute of Human Nutrition Potsdam-Rehbruecke, Department of Experimental Diabetology, Nuthetal, Germany
c
German Center for Diabetes Research (DZD), München-Neuherberg, Germany
d
Institute of Nutritional Science, Department of Nutritional Toxicology, University of Potsdam, Nuthetal, Germany
e
Department of Molecular Biology, University of Duisburg-Essen, Essen, Germany
f
NutriAct – Competence Cluster Nutrition Research Berlin-Potsdam, Germany
g
Department of Clinical Chemistry and Laboratory Medicine, University Hospital of the Saarland, Homburg, Germany
h
Aarhus Institute of Advanced Studies, Aarhus University, Aarhus C, Denmark
i
German Center for Cardiovascular Research (DZHK), Berlin, Germany
j
Institute of Nutrition, Department of Molecular Nutrition Physiology, Friedrich Schiller University Jena, Germany

Received 14 February 2017; received in revised form 7 June 2017; accepted 6 July 2017

Abstract

The average intake of the essential trace element selenium (Se) is below the recommendation in most European countries, possibly causing sub-optimal expression of
selenoproteins. It is still unclear how a suboptimal Se status may affect health. To mimic this situation, mice were fed one of three physiologically relevant amounts of Se. We
focused on the liver, the organ most sensitive to changes in the Se supply indicated by hepatic glutathione peroxidase activity. In addition, liver is the main organ for synthesis
of methyl groups and glutathione via one-carbon metabolism. Accordingly, the impact of Se on global DNA methylation, methylation capacity, and gene expression was
assessed. We observed higher global DNA methylation indicated by LINE1 methylation, and an increase of the methylation potential as indicated by higher S-
adenosylmethionine (SAM)/S-adenosylhomocysteine (SAH) ratio and by elevated mRNA expression of serine hydroxymethyltransferase in both or either of the Se groups.
Furthermore, increasing the Se supply resulted in higher plasma concentrations of triglycerides. Hepatic expression of glycolytic and lipogenic genes revealed consistent Se-
dependent up-regulation of glucokinase. The sterol regulatory element-binding transcription factor 1 (Srebf1) was also up-regulated by Se. Both effects were confirmed in
primary hepatocytes. In contrast to the overall Se-dependent increase of methylation capacity, the up-regulation of Srebf1 expression was paralleled by reduced local
methylation of a specific CpG site within the Srebf1 gene. Thus, we provided evidence that Se-dependent effects on lipogenesis involve epigenetic mechanisms.
© 2017 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: Selenium; DNA methylation; Liver; Lipogenesis; Srebf1

Abbreviations: 5-mdC, 5-methyl-2′-deoxycytidine; Acaca, acetyl-CoA car-

boxylase α; CBS, cystathionine beta synthase; dC, 2′-deoxycytidine; DNMT, DNA
methyltransferase; Fasn, fatty acid synthase; Gck, glucokinase; GPx, glutathione
peroxidase; Hcy, homocysteine; LINE1, long interspersed nuclear elements;
NQO1, NAD(P)H quinone oxidoreductase 1; Pfkfb3, fructose-2,6-biphosphatase
3; SAH, S-adenosylhomocysteine; SAM, S-adenosylmethionine; Scd1, stearoyl-
CoA desaturase 1; Se, selenium; Selh, selenoprotein H; SeMet, selenomethionine;
SePP, selenoprotein P; SHMT, serine hydroxy-methyltransferase; Srebf1, sterol
regulatory element-binding transcription factor 1; T2D, type 2 diabetes; TrxR,
thioredoxin reductase; TS, transsulfuration.
☆ selenometabolites have direct effects independent of the selenopro-
Grants/founding sources: This work was supported by the German
Research Foundation (KI 1590/2-1) and NutriAct – Competence Cluster
Nutrition Research Berlin-Potsdam funded by the Federal Ministry of
Education and Research (FKZ: 01EA1408A-B).
⁎ Corresponding author at: Friedrich Schiller University Jena, Institute of
Nutrition, Department of Molecular Nutrition Physiology, Dornburgerstr. 24,
D-07743 Jena, Germany. Tel.: +49 3641 949609.
E-mail address: anna.kipp@uni-jena.de (A.P. Kipp).
1. Introduction
The essential trace element selenium (Se) has beneficial effects
towards human health, including proper functioning of the immune
response, and prevention of some forms of cancer [1]. Several of the
human selenoproteins, which contain Se in the form of the
proteinogenic amino acid selenocysteine are supposed to mediate
these effects by changing the cellular redox status. In parallel,
selenocompounds such as selenomethionine (SeMet) but also
tein expression, which however has not been fully elucidated for all
compounds, yet.
Up to now, results from large-scale Se intervention trials have been
inconsistent regarding health effects of Se. Differences in participant's
baseline Se status are supposed to be the main confounding factor.
While most Europeans are only suboptimally supplied with Se, the
intake is substantially higher in countries such as the USA. At present,

http://dx.doi.org/10.1016/j.jnutbio.2017.07.002
0955-2863/© 2017 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 B. Speckmann et al. / Journal of Nutritional Biochemistry 48 (2017) 112–119
selenoprotein P (SePP), the main Se transport protein that shuttles Se
from the liver to peripheral tissues, is supposed to be the best marker
for the Se status [2]. Plasma concentrations of SePP reach a plateau at
plasma Se concentrations of approximately 120 μg/L. Concentrations
above this level appear to have no further beneficial effect but might
even have adverse effects. Two US cancer prevention trials, the NPC [3]
and Sel/Cel trial [4], found a trend for a higher diabetes incidence that
was obvious only in the highest tertile of baseline plasma Se. Several
cross-sectional studies reported that the Se status was positively
associated with the presence of type 2 diabetes (T2D) [5,6] and traits of
the metabolic syndrome such as hypertension [7], elevated fasting
plasma glucose [8], insulin resistance [9], or dyslipidemia [9–14].
While these observational studies cannot prove a putative causal
association, they indicate that the Se intake may be linked to
carbohydrate and lipid metabolism. Also in animal studies, long-
term Se supplementation in different forms and doses caused
hyperinsulinemia, insulin resistance, and weight gain [15–18]. Thus,
Se was hypothesized to interfere with insulin-dependent metabolic
pathways through modulation of intracellular hydrogen peroxide
levels [19,20].
The development of T2D and the metabolic syndrome is influenced
by lifestyle factors including diet. The molecular mechanisms of this
interaction may affect DNA methylation via epigenetic regulation
[21–23]. DNA methyltransferases (DNMTs) catalyze the covalent
addition of a methyl group to the 5-position of desoxycytosine (5-
mdC) using S-adenosyl methionine (SAM) as the methyl donor.
Methylation of CpG dinucleotides is often clustered within gene
promoters or in repetitive DNA elements such as long interspersed
nuclear elements (LINE1). However, the impact of nutrition on
epigenome-disease interactions remains poorly understood which
also includes the intake of micronutrients such as Se. A number of cell
culture and animal studies have shown that Se modifies marks and
editors of epigenetic information (reviewed in [24]). As the Se supply
enhances the flux of homocysteine (Hcy) through the transsulfuration
(TS) pathway, it may also affect one-carbon metabolism and thereby
the methylation potential [25].
A limitation of most Se feeding studies is the use of non-
physiological Se doses ranging from overt deficiency to excess intake.
Based on that, we aimed to identify the effects of physiologically
relevant doses of Se on DNA methylation. To this end, mice received
diets for a six-week feeding period containing Se as SeMet, which is
the most relevant selenocompound in foods. As previously established
[26], the suboptimal diet (0.07 ppm) was about half of the adequate
diet (0.15 ppm) full-filling the recommended amount of Se for mice,
while 0.60 ppm was chosen as supranutritional but achievable dose.
We used the murine liver as a model to test the hypothesis that
physiological ranges of Se intake can affect one-carbon metabolism,
global DNA methylation and the expression of genes involved in
hepatic glucose and lipid metabolism.
2. Materials and methods
2.1. Mouse feeding trial and preparation of primary hepatocytes
C57BL/6 J mice were housed individually in independently ventilated cages under
specific-pathogen-free conditions with a 12-h light/dark cycle and free access to food
and water. Mice were fed a Torula yeast-based chow (C1045, Altromin), which had a
basal Se content of 0.07 mg/kg. For the +Se (0.15 ppm) and the ++Se (0.60 ppm) diets,
the basal chow was enriched with L(+)-selenomethionine (Fisher Scientific). The Se
concentration of the chow was verified by ICP-MS/MS (in collaboration with Prof. Tanja
Schwerdtle, Potsdam). Directly after weaning at an age of 3–4 weeks, mice were
randomized to one of three feeding groups (n=6) for further 7 weeks. Then mice were
anesthetized with isoflurane (Abbott) and blood was withdrawn by heart puncture into
heparinized tubes. Plasma was obtained after centrifugation of the blood for 15 min
(1200×g, 4°C). Anesthetized animals were sacrificed by cervical dislocation and whole
livers were directly snap-frozen. Before further analysis, frozen liver was pulverized
using a tissueLyser II (Qiagen). Plasma and hepatic triglycerides were quantified using
the colorimetric Triglyceride Quantification Assay Kit (ab65336, Abcam).
113
Animal experiments were performed in compliance with the German animal protection
law (Tierschutzgesetz). The mice were housed and handled in accordance with good animal
practice as defined by the Federation of Laboratory Animal Science Associations (FELASA,
www.felasa.eu/) and the national animal welfare body (GV-SOLAS, www.gv-solas.de/). The
animal welfare committees of the German Institute of human Nutrition (DIfE) as well as the
local authorities (Landesamt für Umwelt, Gesundheit und Verbraucherschutz, Brandenburg,
Germany) approved all animal experiments (project number V3–2347–29-2012, date of
approval: 26 October 2012).
Primary hepatocytes were freshly isolated from liver tissue of three male 12–14 week
old C57BL/6 J mice [27] fed a standard mouse chow (V153x R/M-H, Ssniff) with a selenium
content of approximately 0.30 ppm to perform three independent cell culture
experiments. Hepatocytes were cultured for additional 48 h at 37°C in a 5% CO2
atmosphere in RPMI-1640 medium (Sigma) with 10% fetal calf serum, 100 U/mL penicillin,
100 mg/mL streptomycin and with or without 200 nM sodium selenite or 1 μM SeMet.
2.2. Isolation of RNA and real-time qPCR analysis
Total RNA was extracted from murine snap-frozen, pulverized liver by using the
InviTrap Spin Universal RNA kit (Stratec Molecular). For primary mouse hepatocytes, the
RNeasy Mini Kit (Qiagen) was used. Both protocols comprised an efficient elimination of
DNA from the sample. From each sample, 1 μg of RNA was transcribed into cDNA with
SuperScript II reverse transcriptase (Invitrogen) using poly (dT)15 primers in a total
volume of 10 μl. cDNA samples were diluted 1:5 in water before expression of mRNA was
analyzed by real-time qPCR using the Mx3005P system (Agilent). 2 μl of diluted cDNA, iQ
SybrGreen Supermix (Biorad) and 0.5 μM of specific primers were combined in a total volume
of 25 μl. Each cycle started at 95°C for 30 s, followed by the annealing phase at 60°C for 30 s
and the elongation phase at 72°C for 30 s. The whole procedure was repeated 40 times. The
specificity of the reaction was confirmed by including a melting curve analysis at the end of
each qPCR assay. PCR products were quantified by using standard curves in the range
between 104 and 109 copies of each amplicon (Table S1) and the MxPro qPCR software.
The RT2 Diabetes PCR Array (Qiagen) contains pre-validated primer pairs for the
detection of 84 mRNA transcripts from diabetes-associated genes, which were
normalized to 5 reference genes (Table S2). It was used together with the RT2 first
strand kit and the RT2 SybrGreen qPCR master mix on the Mx3005P system following
the manufacturer's instructions. Primers were otherwise designed using the Universal
Probe-Library Assay Design Center (Roche) and checked for specificity using NCBI
BlastN (Table 1). For these custom-designed qPCRs, mouse hypoxanthine phosphor-
ibosyltransferase 1 (Hprt1) and ribosomal protein L13a (Rpl13a) were used as reference
genes based on the following criteria described in [28]: least variation in microarray
data between all samples, mean expression level ratio of −Se over +Se groups close to
1, and an expression level of at least ten times the background. The geometric mean of
reference genes was used for normalization as mentioned elsewhere [29].
2.3. Enzyme activity assays
20 mg of liver powder (frozen tissue ground under liquid nitrogen) were
homogenized in a tissue lyzer (Qiagen) for 2 × 2 min at 30 Hz in 500 μl 100 mM Tris/
HCl, 300 mM KCl, 0.1% Triton X-100, pH 7.6, containing 4 μl of protease inhibitor cocktail
(Calbiochem). Cellular debris was removed at 20,000×g, for 15 min, at 4°C. Protein
content was estimated according to Bradford and enzyme activities were expressed as
mU/mg protein. Measurements of NAD(P)H quinone oxidoreductase 1 (NQO1),
thioredoxin reductase (TrxR), and GPx activities were performed as described
previously [26,30,31]. Briefly, NQO1 activity was examined by menadione-mediated
reduction of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT).
TrxR activity was measured by the NADPH-dependent reduction of 5, 5′-dithiobis(2-
nitrobenzoic acid). GPx activity was determined in a NADPH-consuming glutathione
reductase coupled assay. All measurements were carried out in 96-well format using a
microplate reader (Synergy2, BioTek).
2.4. Quantification of 5-mdC/dC percentage by isotope-dilution LC–MS/MS
DNA was extracted from mouse liver using the Invisorb Genomic DNA Kit II (Stratec
Molecular). 10 μg of genomic DNA were hydrolyzed to 2′-deoxynucleosides using
micrococcal nuclease from Staphylococcus aureus, bovine spleen phosphodiesterase,
and calf intestinal alkaline phosphatase (all from Sigma-Aldrich) as described
previously [32] but the incubation time of the two-step digestion approach was 1 h
each. DNA hydrolysates were centrifuged (5 min, 16,000×g) and 10 μl of the
supernatants were diluted 1:10 (v:v) with water. Afterwards, 15 μl of this dilution
were mixed with 15 μl of an aqueous internal standard solution containing 1 μM
[15 N2,13C1]dC and 50 nM 5-mdC-d3 (Toronto Research Chemicals). Samples were
vortexed, briefly centrifuged and the supernatants were subjected to a LC–MS/MS
method [33]. Analyses were conducted with an Agilent 1260 Infinity LC system coupled to
an Agilent 6490 triple quadrupole-mass spectrometer interfaced with an electrospray ion
source operating in the positive ion mode (ESI+). Chromatographic separation was
carried out using an Agilent Poroshell 120 EC-C18 column (2.7 μm, 3.0 × 50 mm). Water
and acetonitrile (VWR), both acidified with 0.0075% formic acid, were used as eluents.
Samples (1 μl) were injected into a mobile phase consisting of 100% water. 2′-
Deoxynucleosides were eluted from the column at 30°C with the following gradient at
a flow rate of 0.4 ml/min: 0–5 min (100–96.5% water), 5–6.9 min (96.5–95% water),
114 B. Speckmann et al. / Journal of Nutritional Biochemistry 48 (2017) 112–119

Table 1 2.8. Statistics

Primer sequences and amplicon lengths for custom qPCR assays (5′ → 3′)
Gene Primer fwd Primer rev Amplicon Statistical analyses were conducted by using GraphPad Prism Version 6 (San Diego, CA, US).
Name length (bp) Data are shown as the mean + SD (standard deviation). The distribution of the data was tested
with the Shapiro–Wilk's Test. The one-way analysis of variance (ANOVA) test and Bonferroni's
Acaca GGCTCAAACTGCAGGTATCC TTGCCAATCCACTCGAAGA 63 post-hoc test were used to investigate the differences between the three Se groups. A p-value
Dnmt1 CGCCCAGTTTGTAGTGAGCCA TTGTTGCTCGCCTCTGTCCC 136 b0.05 was considered as statistically significant, and that between 0.05 and b0.10 was
Fasn GCTGCTGTTGGAAGTCAGC AGTGTTCGTTCCTCGGAGTG 75 considered as showing a trend.
Gck GGAGCCCAGTCGTTGACTC ATCCGGCTCATCACCTTCTT 85
Hprt GCAGTCCCAGCGTCGTG GGCCTCCCATCTCCTTCAT 168
Rpl13a GTTCGGCTGAAGCCTACCAG TTCCGTAACCTCAAGATCTGCT 158 3. Results
Scd1 TTCCCTCCTGCAAGCTCTAC CAGAGCGCTGGTCATGTAGT 61
Selh CCTTATTCCACCAACGCGCCA GCGTCAGCTCGTACAATGCTC 154 3.1. Selenium increases serum triglyceride levels and hepatic Se biomarkers
Sepw1 ATGCCTGGACATTTGTGGCGA GCAGCTTTGATGGCGGTCAC 152
Shmt1 GAGGGATATCCAGGCCAAA CCTGTAATGCACGCTTCTGA 84
Srebf1 GGTTTTGAACGACATCGAAGA CGGGAAGTCACTGTCTTGGT 60 C57BL6/J mice were fed diets with different amounts of Se for 7
weeks starting directly after weaning. We controlled the success of our
feeding approach by the assessment of commonly used markers of the
6.9–7.2 min (95–20% water), 7.2–10 min (isocratically 20% water). dC and 5-mdC co- Se status such as GPx (Fig. 1A) and TrxR (Fig. 1B) activities and selenoprotein
eluted from the separation column with their stable-isotope labeled analogues at 1.6 W (Sepw1) transcript levels (Fig. 1C). All three markers were significantly up-
min and 2.9 min, respectively. The total run time for one analysis was 14 min,
regulated in the liver of both adequately fed (+Se) and supplemented
including re-equilibration of the LC system. The following optimized settings of the
ESI source were used: drying gas temperature=80°C, drying gas flow=15 l/min of (++Se) mice in comparison to the suboptimal group (−Se). Activities of
nitrogen, sheath gas temperature=350°C, sheath gas flow=12 l/min of nitrogen, both enzymes were unchanged in ++Se compared to +Se mice.
nebulizer pressure=26 psi, capillary voltage=2250 V, nozzle voltage=500 V. The Accordingly, the selenoprotein expression appeared to be maximized already
optimized ion funnel parameters were: high pressure RF voltage=95 V and low
in the +Se group. Vice versa, the Nrf2-mediated NQO1 activity was increased
pressure RF voltage=60 V. Quantification of 2′-deoxycytidine (dC) and 5-methyl-2′-
deoxycytidine (5-mdC) in relation to their stable-isotope labeled standards was
in the −Se group (Fig. 1D) in comparison to both the +Se and ++Se groups
carried out using the multiple reaction monitoring (MRM) approach. Thereby, the as shown previously [36]. Also here, no significant difference was detected
following mass transitions (loss of 2′-deoxyribose) were used as quantifiers between the +Se and ++Se groups. Moreover, we found that plasma
(optimized collision energies in parentheses): dC: m/z 228.1N112.0 (8 V); triglyceride levels were increased along with increasing the dietary Se content
[15N2,13C1]dC: m/z 231.1N114.9 (8 V); 5-mdC: m/z 242.1N126.0 (8 V); 5-mdC-d3: m/z
(Fig. 1E) while hepatic triglyceride levels were unchanged (Fig. 1F).
245.1N129.0 (8 V). An additional mass transition per analyte was recorded for
unambiguous identification. The dwell time for each of the eight mass transitions
analyzed was 50 ms. 3.2. Se up-regulates global DNA methylation and SAM/SAH levels in the liver

2.5. Murine LINE1 bisulfite sequencing
Methylation of the murine LINE1 element was quantified using a slightly modified
protocol [34]. Briefly, DNA was extracted from mouse liver using the Invisorb Genomic
DNA Kit II (Stratec Molecular) and 500 ng DNA were bisulfite-converted with the EZ
DNA Methylation-Gold Kit (Zymo Research). The following PCR was carried out with
~50 ng bisulfite-converted DNA, 200 nM forward primer LINE1-F (GGTTGAGGTAG-
TATTTTGTGTG), 200 nM reverse primer LINE1-R (TCCAAAAACTATCAAATTCTCTAAC)
and HotStarTaq (Qiagen) components in a 20 μl reaction volume. The PCR program
consisted of an initial heat activation for 15 min at 95°C, 40 cycles of 1 min at 95°C, 0.5
min 59.2°C, 1 min 72°C, and a final extension for 10 min at 72°C. The reaction mix was
electrophoresed on a 2% agarose gel, the amplicon (195 nt) excised, the DNA purified
and sequenced (LGC Genomics). Sequences were analyzed with the Sequence Scanner
Software 2 (Applied Biosystems) and % cytosine methylation at CpG sites were
determined by relative height of cytosine peaks using the formula C/(C + T).
2.6. SAM and S-adenosylhomocysteine (SAH) quantification by UPLC
Snap frozen liver (approximately 50 mg each) were thawed on ice, extracted by
using 1 N acetic acid (10 μl for each 1 mg tissue). After centrifugation (1000×g, 10 min,
4°C), the supernatant was collected and 50 μl were used for SAM and SAH
measurements. The concentrations of SAM and SAH in liver tissue extracts were
measured by using established methods that depend on labeled isotopes as internal
standards (13C5-SAH and 2H3-SAM) using a Waters 2795 alliance HT, which was
coupled to a MicroMass Quattro Micro API tandem mass spectrometer (Waters
Corporation) as described before [35]. The sample volume of liver extracts was adjusted
to obtain final concentrations within the range of the standard curve (up to 300 nmol/L
SAM and 150 nmol/L for SAH). The between-day coefficient of variations of the SAM and
SAH assays was b10% at a mean SAM concentration of 86 nmol/L and SAH of 13.7 nmol/L.
As marker of total DNA methylation, the amount of 5-mdC in
relation to dC was assessed in the livers of mice. Methylation levels in
the −Se group tended to be slightly lower than in the +Se and ++Se
groups without reaching statistical significance (Fig. 2A). A large
proportion of the human genome is rather stable in terms of its 5-mdC
content [37]. We therefore reasoned that global 5-mdC quantification
might be inappropriate to detect methylation changes at more
flexible genomic sites, e.g. repetitive elements such as LINEs.
LINE1 methylation is commonly used as surrogate marker of global
DNA methylation and has been found to respond to dietary and
metabolic changes [38]. LINE1 methylation was significantly increased
in both the +Se and the ++Se group in comparison to the −Se
group (Fig. 2B) indicating that a suboptimal Se status reduced
the steady-state methylation level of a large fraction of the murine
hepatic genome.
One possible explanation for this phenomenon is that Se interferes
with cellular DNA methylation capacity per se. This capacity essentially
depends on the relative amounts of the methyl donor SAM and its
reaction product SAH. The amounts of hepatic SAM and SAH were
considered as markers for the flow of methyl groups. Liver tissue SAM
did not differ between all feeding groups (Fig. 2C), while SAH was
lower in the +Se group in comparison to the −Se group (Fig. 2D).
Thus, the hepatic SAM/SAH ratio was higher in the +Se versus −Se
group (Fig. 2E).

2.7. MSRE-qPCR 3.3. Se increases hepatic SHMT1 expression

Methylation of a CpG site located within the second exon of Srebf1 (genome assembly:
GRCm38/mm10; chr11:60,206,956) was quantified by a methylation-sensitive restriction
enzyme qPCR assay. Aliquots of 500 ng genomic DNA from hepatocyte cultures were treated
with MspI or HpaII restriction enzymes (New England Biolabs) or left untreated (mock-
treated) at 37°C for 16 h, followed by 20 min at 80°C. 50 ng of these DNA samples was used in
qPCR reactions as described above using the primers CCACTCACCATCCTACAGCC
and CGGCCACTCTTCTTCCATCA. Percent methylation of the CpG site located within the
recognition sequence of the two restriction enzymes was calculated with the equation %
methylation = amplifyableDNAHpaII/(amplifyableDNAmock-amplifyableDNAMspI). As a reference, qPCRs
of totally methylated and unmethylated DNA were performed in parallel.
We speculated that Se increases the cellular methylation capacity
via enhanced conversion of Hcy to SAM, the active methyl donor in
the cell. As shown in Fig. 2F, Se increased the mRNA expression of
serine hydroxy-methyltransferase (SHMT) in a dose-dependent
manner. SHMT is known to catalyze the first step of the formation of
5-methyl tetrahydrofolate, and thus indirectly increases the formation
of SAM. The mRNA expression of Dnmt1 was however unaffected by
the Se status (Fig. S1D).
 B. Speckmann et al. / Journal of Nutritional Biochemistry 48 (2017) 112–119 115

3.4. Se increases gene expression of key regulators of glycolysis and lipogenesis
Differential DNA methylation as well as a supranutritional Se intake
have both been implicated in the development of T2D [23,39]. As we
found modulation of the methylation machinery by Se, we reasoned
whether this might correlate with differential expression of T2D-
associated genes. Thus, we analyzed mRNA levels by means of a
Diabetes RT2 profiler PCR Array in livers of the respective Se-fed mice as
a first screening approach. Out of 84 genes, eight genes were
significantly altered by a Se intervention (either +Se or ++Se) in
comparison to the −Se group, while further seven genes showed a
trend (Pb.1) (Table 2). Out of those, mRNA expression of several
important transcription factors such as C/EBPα, HNF1β, PPARα, and
SREBP1 (Srebf1) was increased in the ++Se compared to the −Se
group. However, mRNA levels of target genes of lipogenic SREBP1 such
as acetyl-CoA carboxylase α (Acaca), fatty acid synthase (Fasn),
and stearoyl-CoA desaturase 1 (Scd1) were unaffected by the Se status
(Fig. S1A-C). Only Enpp1, the ectonucleotide pyrophosphatase/
phosphodiesterase 1, was significantly and consistently down-
regulated both in the +Se and ++Se group in comparison to the −Se
group. Enpp1 has been shown to act as a negative inhibitor of
insulin signaling [40] which could explain an enhanced insulin signaling
under +Se/++Se conditions. In line with that, Gck catalyzing the first
step of glycolysis was up-regulated under +Se and ++Se conditions. To
confirm the results obtained by PCR Array, expression levels of Gck and
Srebf1 were further analyzed in primary murine hepatocytes. Treatment
with sodium selenite or SeMet resulted in higher levels of the Se-sensitive
transcript of selenoprotein H (Selh) (Fig. 3A), and in increased Gck
(Fig. 3B) and Srebf1 mRNA expression (Fig. 3C).
3.5. Selenium-dependent DNA methylation of Srebf1 in hepatocytes
Next, we determined whether the methylation-sensitive Srebf1
gene [41] was differentially methylated in response to Se. We assessed
methylation of one CpG site located within the second exon of Srebf1
by MSRE qPCR. In whole liver samples, the average methylation rate of

Fig. 1. Characterization of the selenium status and of triglyceride levels. Total GPx (A) and TrxR (B) activities were analyzed in liver together with the NQO1 activity (D) expressed as mU/mg
protein. Selenoprotein W (Sepw1) mRNA levels were quantified by qPCR and normalized to the reference genes Hprt and Rpl13a (C). In addition, plasma (E) and hepatic (F) triglyceride
levels were quantified. Data are shown as Scatter dot blot with mean (n=5–6 mice per group). Significant differences were calculated in comparison to the −Se group using a one-way
ANOVA with Bonferroni's post-test: *Pb.05; **Pb.01; ***Pb.001.
116 B. Speckmann et al. / Journal of Nutritional Biochemistry 48 (2017) 112–119

this site was 60% independent of the Se status (Fig. S1E). As effects on
DNA methylation are often cell type-dependent we also analyzed
primary murine hepatocytes. In those samples, methylation of the CpG
site in exon 2 of the Srebf1 gene was significantly reduced by both
selenocompounds (Fig. 3D). Thus, Se reciprocally affects mRNA
expression and methylation of the Srebf1 gene in primary hepato-
cytes. The lack of difference in methylation pattern in liver
homogenates probably results from the impact of non-hepatocyte
cells masking an effect, which takes place in hepatocytes only.
4. Discussion
The current study shows that the hepatic selenium status, which is
highly sensitive to changes in the nutritional Se supply [25], has a relevant
impact on both global and local DNA methylation capacities of liver cells.
Global DNA methylation indicated by LINE1 methylation was increased in
mice with adequate or supranutritional Se intake in comparison to a
suboptimal status. In parallel, the hepatic SAM/SAH ratio was up-
regulated upon Se feeding together with an increased SHMT1 mRNA
expression. In contrast to global DNA methylation, local DNA methylation
of the Srebf1 gene was reduced in hepatocytes after 48 h of Se incubation,
which appears to be essential for Se-dependent changes in Srebf1
expression and down-stream effects on plasma triglyceride levels.
Se has been previously shown to modulate global DNA methylation
using different cell culture and animal models [24]. We addressed this
topic by means of a feeding study in mice with a physiologically relevant
range of Se intakes. Results from related rodent studies are inconsistent
in this regard, which may be explained by chow- and strain-specific
effects [24]. It is obvious that numerous factors affect LINE1 or global
DNA methylation. Interestingly, a sex-specific correlation between
plasma Se status and LINE1 methylation was found in human rectal
mucosa samples, which was positive only for females [42]. Cancer cells
typically have hypomethylated global DNA while local sequences can be
hypermethylated in parallel resulting in e.g. silencing of tumor

Fig. 2. Selenium-modulated effects on global DNA methylation and levels of the methyl donor SAM. The amount of 5-mdC in relation to dC was quantified by LC–MS/MS (A) and the
percentage methylation rate of CpGs within the murine LINE1 element was assessed by bisulfide sequencing (B) in whole liver. Furthermore, hepatic SAM (C) and SAH (D) levels were
analyzed by UPLC and expressed as hepatic SAM/SAH ratio (E). Shmt1 mRNA levels were quantified by qPCR and normalized to the reference genes Hprt and Rpl13a. Data are shown as
Scatter dot blot with mean (n=4–6 mice per group). Significant differences were calculated in comparison to the −Se group using a one-way ANOVA with Bonferroni's post-test: *Pb.05.
 B. Speckmann et al. / Journal of Nutritional Biochemistry 48 (2017) 112–119 117

suppressor genes. Our finding of increased hepatic global DNA
methylation when comparing an adequate with a suboptimal Se supply
is therefore in agreement with a putative preventive function of Se
against the development of hepatocellular carcinoma, which has been
suggested by results from a recent cohort study [43].
Besides that, global DNA methylation essentially depends on the
availability of methyl donors because DNMTs are SAM-dependent. We
have shown previously that hepatic Hcy levels are drastically reduced
upon adequate Se feeding in comparison to a suboptimal Se supply [25].
In addition, the protein expression of cystathionine beta synthase (CBS)
was up-regulated during an adequate Se status, which indicates an
enhanced flux of Hcy through the TS pathway [25]. Thus, our present
results on low SAH and high expression of SHMT in the +/++Se groups
appear to be the natural and very much expected consequence of the
up-regulation of the TS pathway. Up-regulation of SHMT expression
upon Se feeding would reflect an important metabolic flexibility of one-
carbon metabolism: if more SAH will flow via Hcy and CBS to cysteine
and glutathione, more methyl groups will be derived from folate to
compensate and balance SAM and cell needs for SAM. However, effects
on SHMT1 expression have only been analyzed on mRNA level, which
should be confirmed on protein or activity level in future experiments.
In contrast to our results, Uthus and Ross reported that both SAM and
SAH levels did not change in Se-deficient rodents [44,45], which might
be not directly comparable to the suboptimal Se group studied herein.
In parallel with the observed changes in DNA methylation and one-
carbon metabolism, we showed that dietary Se significantly increased
plasma triglyceride levels, while hepatic triglycerides remained
unaffected. Our gene expression data provides an explanation for
this outcome by showing Se-dependent induction of genes whose
respective proteins catalyze key steps in glycolysis and lipogenesis,
such as Gck and Srebf1. Animal studies wherein Se was supplemented
at supranutritional doses (between 0.4 and 3 ppm) for longer time
periods (≥8 weeks) have partially reported the development of insulin
resistance and traits of the metabolic syndrome (summarized in [39]).
Several rodent studies are in support of our finding of a hyperlipidemic
effect of dietary Se. Female Wistar rats fed 3 versus 0.3 ppm Se starting
from 5 weeks before breeding had increased plasma triglyceride levels
on day 19 of gestation. Their offspring fed the same diet subsequently
developed insulin resistance and glucose intolerance [17]. Similarly,
overexpression of GPx1 caused hyperlipidemia in mice [20], which
was attenuated in the liver by feeding a Se-depleted chow [46].
Increased hepatic triglyceride content was also reported for rats fed
0.15 ppm Se as selenate for 8 weeks compared to rats fed a Se deficient
diet [16]. This study also found higher expression of Srebf1 in livers of
Se-fed rats. Our study showed that this effect became apparent after a
relatively short feeding period and, more importantly, within a range
of Se doses that is of physiological relevance. In addition, we provide
for the first time evidence that Se effects on Srebf1 expression in
hepatocytes can be explained by Se-induced changes in DNA
methylation of the Srebf1 gene. Thus, DNA methylation appears to
link Se (i.e., selenite and SeMet) to hepatic lipogenesis. A recent
feeding study (20 weeks) directly comparing effects of Se (as selenite,
selenate, and SeMet) at concentrations of 0.15 ppm and 0.75 ppm
revealed that plasma triglycerides were increased in comparison to
the −Se group independent of the Se species but the effects were more
pronounced with the two inorganic Se compounds [47]. Regarding
humans, a recent study found significant interactions between plasma
lipid and Se concentrations with variants of genes involved in

Table 2
Fold changes (FC) of mRNA levels of type 2 diabetes-associated genes in the liver of mice fed +Se and ++Se compared to −Se diets with p-values b0.1 (n=4). P-values b0.05 are
printed in bold. FCN1.2 are highlighted in green, FCb−1.2 in red

Group 1 (+Se) Group 2 (++Se)

Symbol Gene name
FC p–value FC p–value

Cebpa CCAAT/enhancer binding protein alpha 1.089 0.524 1.404 0.075

ectonucleotide pyrophosphatase/
Enpp1 –1.175 0.047 –1.340 0.010
phosphodiesterase 1

Fbp1 fructose bisphosphatase 1 1.232 0.018 1.277 0.018

Gck glucokinase 1.411 0.069 1.552 0.036

Hnf1b hepatocyte nuclear factor 1 beta –1.305 0.159 1.364 0.014

Mapk8
mitogen–activated protein kinase 8 –1.331 0.275 –1.322 0.080
(JNK)

Nos3 nitric oxide synthase 3, endothelial cell 1.625 0.014 1.636 0.095

Pfkfb3 fructose–2,6–biphosphatase 3 1.334 0.317 2.497 0.017

phosphatidylinositol 3–kinase, regulatory

Pik3r1 1.131 0.292 1.439 0.035
subunit, polypeptide 1
peroxisome proliferator activated receptor
Ppara 1.032 0.799 1.251 0.064
alpha
Ppargc1a peroxisome proliferative activated
–1.765 0.074 –1.322 0.213
(Pgc1a) receptor, gamma, coactivator 1 alpha

Sell selectin L 2.363 0.037 2.374 0.042

sterol regulatory element binding

Srebf1 1.306 0.411 1.747 0.071
transcription factor 1

Trib3 tribbles pseudokinase 3 –1.235 0.349 –1.556 0.078

Vapa vesicle–associated membrane protein A –1.123 0.270 –1.181 0.095

118 B. Speckmann et al. / Journal of Nutritional Biochemistry 48 (2017) 112–119

Fig. 3. Selenium treated hepatocytes have higher Srebf1 expression and lower methylation of the Srebf1 gene. Primary murine hepatocytes were incubated for 48 h with 200 nM selenite
or 1 μM selenomethionine (SeMet). Both RNA and DNA were isolated from harvested cells. Expression of Selh (A), Gck (B), and Srebf1 (C) was analyzed by qPCR and normalized to the
reference genes Hprt and Rpl13a. Methylation of a CpG site located within the second exon of Srebf1 was quantified by a methylation-sensitive restriction enzyme qPCR assay (D). Data
are shown as mean + SD (n=3). Significant differences were calculated in comparison to the −Se group using a one-way ANOVA with Bonferroni's post-test: *Pb.05; **Pb.01.

transport and transfer of lipids [10], suggesting a robust cross-talk
between Se and lipid metabolism. Whether this interaction involves
epigenetic mechanisms also in humans needs to be studied in more
detail in the future.
In conclusion, previous studies have provided evidence that
selenium modulates epigenetic processes. Herein, we connect a
physiologically relevant increase of the Se status to both hepatic global
hypermethylation but also local hypomethylation of the Srebf1 gene.
Acknowledgement
The authors thank Stefanie Deubel, Swetlana Kohse, and Lisa
Richter for excellent technical assistance. This work was supported by
the German Research Foundation (KI 1590/2-1) and NutriAct –
Competence Cluster Nutrition Research Berlin-Potsdam funded by
the Federal Ministry of Education and Research (FKZ: 01EA1408A-B).
Appendix A. Supplementary data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.jnutbio.2017.07.002.
References
[1] Hatfield DL, Tsuji PA, Carlson BA, Gladyshev VN. Selenium and selenocysteine:
roles in cancer, health, and development. Trends Biochem Sci 2014;39:112–20.
[2] Kipp AP, Strohm D, Brigelius-Flohé R, Schomburg L, Bechthold A, Leschik-Bonnet
E, et al. Revised reference values for selenium intake. J Trace Elem Med Biol 2015;
32:195–9.
[3] Stranges S, Marshall JR, Natarajan R, Donahue RP, Trevisan M, Combs GF, et al.
Effects of long-term selenium supplementation on the incidence of type 2
diabetes: a randomized trial. Ann Intern Med 2007;147:217–23.
[4] Thompson PA, Ashbeck EL, Roe DJ, Fales L, Buckmeier J, Wang F, et al. Selenium
supplementation for prevention of colorectal adenomas and risk of associated
type 2 diabetes. J Natl Cancer Inst 2016;108.
[5] Bleys J, Navas-Acien A, Guallar E. Serum selenium and diabetes in U.S. adults.
Diabetes Care 2007;30:829–34.
[6] Laclaustra M, Navas-Acien A, Stranges S, Ordovas JM, Guallar E. Serum selenium
concentrations and diabetes in U.S. adults: National Health and nutrition examination
survey (NHANES) 2003-2004. Environ Health Perspect 2009;117:1409–13.
[7] Laclaustra M, Navas-Acien A, Stranges S, Ordovas JM, Guallar E. Serum selenium
concentrations and hypertension in the US population. Circ Cardiovasc Qual
Outcomes 2009;2:369–76.
[8] Yin Y, Han W, Wang Y, Zhang Y, Wu S, Zhang H, et al. Identification of risk factors
affecting impaired fasting glucose and diabetes in adult patients from Northeast
China. Int J Environ Res Public Health 2015;12:12662–78.
[9] Rotter I, Kosik-Bogacka D, Dolegowska B, Safranow K, Lubkowska A, Laszczynska M.
Relationship between the concentrations of heavy metals and bioelements in aging
men with metabolic syndrome. Int J Environ Res Public Health 2015;12:3944–61.
[10] Galan-Chilet I, Guallar E, Martin-Escudero JC, De Marco G, Dominguez-Lucas A,
Gonzalez-Manzano I, et al. Do genes modify the association of selenium and lipid
levels? Antioxid Redox Signal 2015;22:1352–62.
[11] Laclaustra M, Stranges S, Navas-Acien A, Ordovas JM, Guallar E. Serum selenium
and serum lipids in US adults: National Health and nutrition examination survey
(NHANES) 2003-2004. Atherosclerosis 2010;210:643–8.
[12] Bleys J, Navas-Acien A, Stranges S, Menke A, Miller III ER, Guallar E. Serum
selenium and serum lipids in US adults. Am J Clin Nutr 2008;88:416–23.
[13] Christensen K, Werner M, Malecki K. Serum selenium and lipid levels:
associations observed in the National Health and nutrition examination survey
(NHANES) 2011-2012. Environ Res 2015;140:76–84.
[14] Su L, Gao S, Unverzagt FW, Cheng Y, Hake AM, Xin P, et al. Selenium level and
dyslipidemia in rural elderly Chinese. PLoS One 2015;10:e0136706.
[15] Labunskyy VM, Lee BC, Handy DE, Loscalzo J, Hatfield DL, Gladyshev VN. Both
maximal expression of selenoproteins and selenoprotein deficiency can promote
development of type 2 diabetes-like phenotype in mice. Antioxid Redox Signal
2011;14:2327–36.
[16] Mueller AS, Klomann SD, Wolf NM, Schneider S, Schmidt R, Spielmann J, et al. Redox
regulation of protein tyrosine phosphatase 1B by manipulation of dietary selenium
affects the triglyceride concentration in rat liver. J Nutr 2008;138:2328–36.
 B. Speckmann et al. / Journal of Nutritional Biochemistry 48 (2017) 112–119
[17] Zeng MS, Li X, Liu Y, Zhao H, Zhou JC, Li K, et al. A high-selenium diet induces
insulin resistance in gestating rats and their offspring. Free Radic Biol Med 2012;
52:1335–42.
[18] Pinto A, Juniper DT, Sanil M, Morgan L, Clark L, Sies H, et al. Supranutritional
selenium induces alterations in molecular targets related to energy metabolism in
skeletal muscle and visceral adipose tissue of pigs. J Inorg Biochem 2012;114:
47–54.
[19] Steinbrenner H. Interference of selenium and selenoproteins with the insulin-
regulated carbohydrate and lipid metabolism. Free Radic Biol Med 2013;65:
1538–47.
[20] McClung JP, Roneker CA, Mu W, Lisk DJ, Langlais P, Liu F, et al. Development of
insulin resistance and obesity in mice overexpressing cellular glutathione
peroxidase. Proc Natl Acad Sci U S A 2004;101:8852–7.
[21] Morris AP, Voight BF, Teslovich TM, Ferreira T, Segre AV, Steinthorsdottir V, et al.
Large-scale association analysis provides insights into the genetic architecture
and pathophysiology of type 2 diabetes. Nat Genet 2012;44:981–90.
[22] Yuan W, Xia Y, Bell CG, Yet I, Ferreira T, Ward KJ, et al. An integrated epigenomic
analysis for type 2 diabetes susceptibility loci in monozygotic twins. Nat Commun
2014;5:5719.
[23] de Mello VD, Pulkkinen L, Lalli M, Kolehmainen M, Pihlajamaki J, Uusitupa M. DNA
methylation in obesity and type 2 diabetes. Ann Med 2014;46:103–13.
[24] Speckmann B, Grune T. Epigenetic effects of selenium and their implications for
health. Epigenetics 2015;10:179–90.
[25] Geillinger KE, Rathmann D, Köhrle J, Fiamoncini J, Daniel H, Kipp AP. Hepatic
metabolite profiles in mice with a suboptimal selenium status. J Nutr Biochem
2014;25:914–22.
[26] Krehl S, Loewinger M, Florian S, Kipp A, Banning A, Wessjohann L, et al.
Glutathione peroxidase-2 and selenium decreased inflammation and tumors in a
mouse model of inflammation-associated carcinogenesis whereas sulforaphane
effects differed with selenium supply. Carcinogenesis 2012;33:620–8.
[27] Hesse D, Jaschke A, Kanzleiter T, Witte N, Augustin R, Hommel A, et al. GTPase
ARFRP1 is essential for normal hepatic glycogen storage and insulin-like growth
factor 1 secretion. Mol Cell Biol 2012;32:4363–74.
[28] Kipp A, Banning A, van Schothorst EM, Méplan C, Schomburg L, Evelo C, et al. Four
selenoproteins, protein biosynthesis, and Wnt signalling are particularly sensitive
to limited selenium intake in mouse colon. Mol Nutr Food Res 2009;53:1561–72.
[29] Vandesompele J, De Preter K, Pattyn F, Poppe B, Van Roy N, De Paepe A, et al.
Accurate normalization of real-time quantitative RT-PCR data by geometric
averaging of multiple internal control genes. Genome Biol 2002;3 [epub].
[30] Müller M, Banning A, Brigelius-Flohé R, Kipp A. Nrf2 target genes are induced
under marginal selenium-deficiency. Genes Nutr 2010;5:297–307.
[31] Florian S, Krehl S, Loewinger M, Kipp A, Banning A, Esworthy S, et al. Loss of GPx2
increases apoptosis, mitosis, and GPx1 expression in the intestine of mice. Free
Radic Biol Med 2010;49:1694–702.
[32] Schumacher F, Herrmann K, Florian S, Engst W, Glatt H. Optimized enzymatic
hydrolysis of DNA for LC-MS/MS analyses of adducts of 1-methoxy-3-
indolylmethyl glucosinolate and methyleugenol. Anal Biochem 2013;434:4–11.
119
[33] Schiesser S, Pfaffeneder T, Sadeghian K, Hackner B, Steigenberger B, Schroder AS,
et al. Deamination, oxidation, and C-C bond cleavage reactivity of 5-
hydroxymethylcytosine, 5-formylcytosine, and 5-carboxycytosine. J Am Chem
Soc 2013;135:14593–9.
[34] Newman M, Blyth BJ, Hussey DJ, Jardine D, Sykes PJ, Ormsby RJ. Sensitive
quantitative analysis of murine LINE1 DNA methylation using high resolution
melt analysis. Epigenetics 2012;7:92–105.
[35] Kirsch SH, Knapp JP, Geisel J, Herrmann W, Obeid R. Simultaneous quantification
of S-adenosyl methionine and S-adenosyl homocysteine in human plasma by
stable-isotope dilution ultra performance liquid chromatography tandem mass
spectrometry. J Chromatogr B Analyt Technol Biomed Life Sci 2009;877:3865–70.
[36] Muller M, Banning A, Brigelius-Flohe R, Kipp A. Nrf2 target genes are induced
under marginal selenium-deficiency. Genes Nutr 2010;5:297–307.
[37] Ziller MJ, Gu H, Muller F, Donaghey J, Tsai LT, Kohlbacher O, et al. Charting a
dynamic DNA methylation landscape of the human genome. Nature 2013;500:
477–81.
[38] Delgado-Cruzata L, Zhang W, McDonald JA, Tsai WY, Valdovinos C, Falci L, et al.
Dietary modifications, weight loss, and changes in metabolic markers affect global
DNA methylation in Hispanic, African American, and afro-Caribbean breast cancer
survivors. J Nutr 2015;145:783–90.
[39] Zhou J, Huang K, Lei XG. Selenium and diabetes – evidence from animal studies.
Free Radic Biol Med 2013;65:1548–56.
[40] Dimatteo C, Marucci A, Palazzo A, Cisternino C, Marsano RM, Trischitta V, et al.
Role of somatomedin-B-like domains on ENPP1 inhibition of insulin signaling.
Biochim Biophys Acta 1833;2013:552–8.
[41] Adaikalakoteswari A, Finer S, Voyias PD, McCarthy CM, Vatish M, Moore J,
et al. Vitamin B12 insufficiency induces cholesterol biosynthesis by limiting
s-adenosylmethionine and modulating the methylation of SREBF1 and LDLR
genes. Clin Epigenetics 2015;7:14.
[42] Tapp HS, Commane DM, Bradburn DM, Arasaradnam R, Mathers JC, Johnson IT,
et al. Nutritional factors and gender influence age-related DNA methylation in the
human rectal mucosa. Aging Cell 2013;12:148–55.
[43] Hughes DJ, Duarte-Salles T, Hybsier S, Trichopoulou A, Stepien M, Aleksandrova K,
et al. Prediagnostic selenium status and hepatobiliary cancer risk in the European
prospective investigation into cancer and nutrition cohort. Am J Clin Nutr 2016;
104:406–14.
[44] Uthus EO, Ross SA. Dietary selenium affects homocysteine metabolism differently
in fisher-344 rats and CD-1 mice. J Nutr 2007;137:1132–6.
[45] Uthus EO, Yokoi K, Davis CD. Selenium deficiency in fisher-344 rats decreases
plasma and tissue homocysteine concentrations and alters plasma homocysteine
and cysteine redox status. J Nutr 2002;132:1122–8.
[46] Yan X, Pepper MP, Vatamaniuk MZ, Roneker CA, Li L, Lei XG. Dietary selenium
deficiency partially rescues type 2 diabetes-like phenotypes of glutathione
peroxidase-1-overexpressing male mice. J Nutr 2012;142:1975–82.
[47] Lennicke C, Rahn J, Kipp AP, Dojcinovic BP, Müller AS, Wessjohann LA, et al.
Individual effects of different selenocompounds on the hepatic proteome and
energy metabolism of mice. Biochim Biophys Acta 1861;2017:3323–34.