Study of pharmacokinetics of prenylflavonoids and dynamics of estrogen action in sera following ingestion of epimedium using validated, ultra sensitive cell based bioassays 4

185
CHAPTER 4
DISCUSSION

4.1
Development of sensitive in-vitro estrogen-responsive bioassays as
clinical tools to understand biological effects of estrogenic
compounds, alone and in combination
187

4.1.1
Ultra-sensitive bioassays expressing ERα and ERβ can be used
to detect ER-isoform selective activity and study of structure–
function relationships
187

4.1.2
MCF-7 breast cancer cell proliferation assay for quantification of
estrogenic activity
191

4.1.3
Estrogenic activities of Epimedium compounds
193

4.1.4
Estrogenic activities of binary mixtures
195

4.1.5

Application of in-vitro ER-responsive bioassays as clinical tools
to understand estrogenic effects of compounds in serum
197

4.2
Pharmacokinetics and pharmacodynamics of Epimedium
compounds
200

4.2.1
Pharmacokinetics of unconjugated prenylflavonoid aglycones
from aqueous Epimedium decoction in humans
200

4.2.2
Pharmacokinetics and pharmacodynamics of unconjugated
prenylflavonoids after ingestion of an enriched Epimedium
preparation by rats
204

4.2.3
Pharmacokinetics and pharmacodynamics of conjugated
prenylflavonoids after ingestion of an enriched Epimedium
preparation by rats
206

4.2.4
Accumulation of conjugated Epimedium compounds in-vivo
208

186
4.3
Gene expression profiling reveals partially overlapping but distinct
genomic actions of different estrogenic compounds in human breast
cancer cells
209

4.3.1
Convergence of global gene expression profiles MCF-7 breast
cancer cells treated with estrogenic Epimedium
prenylflavonoids, genistein and estradiol
209

4.3.2
Discovery of up-regulation of CYP1A1 transcription by
Epimedium compounds via global gene expression profiling
211

4.3.3
Epimedium prenylflavonoids are dual activators of the AhR and
ER
212

187
4.1 Development of sensitive in-vitro estrogen-responsive bioassays as clinical tools to
understand biological effects of estrogenic compounds, alone and in combination
4.1.1 Ultra-sensitive bioassays expressing ERα and ERβ can be used to detect ER-
isoform selective activity and study of structure–function relationships

A panel of stable human cell lines derived from HeLa human cervical
adenocarcinoma cells has been developed that specifically respond to estrogenic
compounds that interact with human ERα and ERβ proteins thereby allowing the efficient
screening of estrogenic activity of chemicals, alone or in combination, such as in a
complex mixture encountered in botanical extracts and sera.
A yeast-based reporter gene assay was not used as such it does not discriminate
estrogenic and anti-estrogenic substances (Andersen et al., 1999; Jørgensen et al., 1998)
and may not accurately represent human systems, as the permeability of compounds
through the yeast cell wall might be different relative to mammalian cell membranes
(Scrimshaw & Lester, 2004).
HeLa cells were chosen, as in the untransfected form, they endogeneously express
undetectable amounts of ERα and ERβ proteins (Escande et al., 2006). The pERE
4
-
Luc
hygro
reporter gene was first stably incorporated into HeLa cells consists of 4 tandem
copies of consensus vitellogenin ERE cDNA cloned into a pGL3-basic plasmid upstream
of the luciferase reporter gene. Subsequently, cells with pERE
4

-Luc
hygro
in their genome
were stably transfected with coding sequences for either ERα or ERβ. Two highly
inducible clones that stably expressed ERα and ERβ proteins respectively were selected
for the development of ERα and ERβ cell-based bioassays. The production of stable cell
lines eliminates the necessity of constantly producing DNA and variability associated
with transient transfection procedures (Paris et al., 2002).

188
For the validated ERα bioassay, the intra- and inter-assay variations near the EC
50

concentration of the standard estrogen, estradiol, were about 6% and 14% respectively.
For ERβ, the intra- and inter-assay variations were determined to be 8.1% and 16%
respectively. Detection limits for estradiol in ERα and ERβ bioassays were 8.4 and 13.1
pM, respectively. This set of assays is simple and rapid to perform and offers high-
throughput, as cells require only 24 h of incubation with test samples.
The estrogen agonist properties of estradiol and its metabolites, which include
estrone and estriol, were first examined. The rank order in terms of decreasing estrogenic
potency for this series of compounds in both ERα and ERβ bioassays is estradiol >
estriol > estrone. This result is in good agreement with those reported by Paris et al.,
(2002) for their corresponding ERα bioassay.

Both assays are highly specific to estrogens
and they displayed minimal cross reactivity with most steroidal hormones such as the
androgens, progestogens and corticosteroids. However, when testing samples that contain
testosterone, such as sera from males, where testosterone levels can be high, an aromatase
inhibitor, such as DL-aminoglutethimide, would have to be added.
These two assays are also able to distinguish full agonists from partial agonists.

Estriol and estrone are found to be full agonists like estradiol. Interestingly, some
compounds elicited superagonist effects. Flavonoids like apigenin and genistein exhibited
superagonist effects in both assays. Kaempferol and luteolin are superagonists in the ERα
bioassay but are partial agonists in the ERβ bioassay.
The development of stable cell lines that separately express only one ER isoform
enables the differentiation of ERα and ERβ bioactivities. We validated this aspect of our
bioassays using well-known ER-isoform selective ligands, such as genistein, PPT and
DPN. Similar to the results reported by Escande et al., (2006), PPT, an ERα-selective

189
ligand, strongly activated our ERα bioassay in a dose-dependent manner but did not
induce detectable luciferase activity in the ERβ bioassay in the range of concentrations
tested. Both genistein and DPN, ERβ-selective ligands, displayed ERβ selectivity when
tested in both systems.
The effect of hydroxylation on the estrogenic activities of four common
flavonoids - apigenin, kaempferol, luteolin, and quercetin was also studied. Previous
studies indicated that the two hydroxyls in position 7 of ring A and 4’ of ring B represent
the minimal hydroxylation pattern for estrogenic activity (Vaya & Tamir, 2004) and these
four phytoestrogens have these two hydroxyl groups. Above this number, an inverse
relationship between the number of hydroxyl groups and ERα bioactivity was observed.
Increasing the number of hydroxyl groups decreased the molecule’s hydrophobicity,
which reduces binding to the hydrophobic ligand-binding pocket.
Addition of a hydroxyl group in the 3’ position in benzene ring B, which converts
the flavone apigenin to luteolin, reduced estrogenic activity, reflected by higher EC
50

values and lower relative peak activities. This structure–bioactivity relationship was also
observed among the flavonols—the presence of a hydroxyl in the same position changed
kaempferol to the less estrogenic quercetin, suggesting a role for the 3’ position of ring B
in ERα activity. Another active position was the hydroxyl in position 3 of ring C of the

chromone backbone: its presence also reduced estrogenic activity (apigenin vs.
kaempferol; luteolin vs. quercetin). This observation was consistent with previous reports
that hydroxyls at positions 3 and 5 decrease reporter gene activities (Le Bail et al., 1998).
Interestingly, the hydroxyl substitutions in position 3 of ring C, which define
flavones and flavonols, suggest that the division of phytoestrogens into these categories
also reflect intrinsic ERα activity. A similar observation was also seen for ERβ, with the

190
exception of kaempferol, which suggests that the mechanisms of ERβ activation by
kaempferol may differ from the rest of the phytoestrogens. Positions 3’ of ring B and 3 of
ring C may have important roles for estrogenic activity, and more detailed investigations
on these two positions may be warranted.
The molecular basis of these structure–function differences is unlikely to be
completely due to differential receptor affinity. Cell culture media used in this study had
dextran-coated charcoal stripped serum added to it and ex-vivo samples from clinical trials
and animal studies use sera samples from volunteers and test animals. In the absence of
serum, binding of quercetin to recombinant ERα was comparable to that of genistein and
kaempferol (Maggiolini et al., 2004; Leung et al., 2004). The presence of serum reduced
ERα binding of quercetin in MCF-7 cells, with relative binding affinities for genistein,
kaempferol, and quercetin being 0.1, 0.012, and 0.001, respectively (Zava & Duwe, 1997).
This apparently low ERα affinity of quercetin in serum is not surprising, considering that
quercetin binds extensively to serum proteins (Boulton et al., 1999). Reflecting these
changes in relative affinity, transactivation effects of quercetin were comparatively higher
in serum-free media compared to experiments done with serum (Maggiolini et al., 2004;
Harris et al., 2005).

191
4.1.2 MCF-7 breast cancer cell proliferation assay for quantification of estrogenic
activity

Unlike HeLa cells, MCF-7 cells are breast cancer cells which express ERα and
ERβ to no relevant extent. They require estrogenic stimulation for transformation from
dormant into proliferating cells. MCF-7 cells have been used widely as a highly sensitive
tool for the detection of even small agonistic effects at ERα. An in-vitro assay based on
the proliferation of MCF-7 cells commonly called the ‘E-screen’ is performed by
incubating cells with test samples over a period of six days. The number of cells is
quantified and comparisons can be made between cells treated with test samples with
those incubated with the vehicle control.
Although the assay procedure is more laborious than ERα and ERβ stable cells,
the MCF-7 cell proliferation assay is more sensitive towards estrogens as it has a lower
detection limit (ERα: 8.45 pM; ERβ: 13.1 pM & MCF-7: 0.112 pM). The rank order in
terms of estrogenicity of estradiol, estrone and estriol were similar to that obtained via the
stable ERα and ERβ HeLa cell lines where estradiol is the most potent, followed by
estriol and estrone.
To develop the MCF-7 cell proliferation assay for quantification of estrogenic
activity of ex-vivo samples, we incubated large doses of steroids such as
dihydrotestosterone, progesterone, and cortisol in the presence of estradiol and found that
these do not significantly affect the bioassay. However, there is a need to include the
aromatase inhibitor DL-aminoglutethimide to prevent conversion of testosterone to
estradiol by aromatase in MCF-7 cells.
Like ERα and ERβ cell-based bioassays, the MCF-7 cell proliferation assay also
possesses the ability of being able to distinguish full agonists from partial agonists.

192
Superagonism in cell proliferation was not observed in the range of compounds and
concentrations tested. The difference in maximal stimulatory levels in ERα and ERβ
bioassays; and MCF-7 cell proliferation assay could be a reflection of differences in
endogenous versus transfected genes. Superagonistic ER activity exhibited by cells that
have been stably transfected with a reporter gene had also been observed for genistein and
reservatrol (Legler et al., 1999; Harris et al., 2005; Gehm et al., 2004). One possible

mechanism of this superagonist effect may be related to the function of the ER activation
function-1 domain, since its removal prevented enhanced reporter gene activity with
resveratrol (Gehm et al., 2004). Because the ERα activation function-1 domain is
subjected to phosphorylation, superagonist effects could be contributed by the actions of
the AP-1 or the mitogen-activated protein kinase-responsive pathway (Frigo et al., 2002).

193
4.1.3 Estrogenic activities of Epimedium compounds

The estrogenicities of prenylated flavonoid compounds found in the Traditional
Chinese Medicine herb, Epimedium, namely, icariin, icariside I, icariside II, icaritin and
desmethylicaritin, were examined using ERα and ERβ; and MCF-7 cell proliferation
assays in this study.
Out of the five Epimedium compounds, icariin was not estrogenic and this is in
agreement with results reported in an earlier study by Liu et al., 2005. Icariside I, icariside
II, icaritin and desmethylicaritin, were found to be estrogenic in all three assays but were
less potent compared to estradiol and the well-known soy isoflavone, genistein. In terms
of the magnitude of maximal inducible responses, those elicited by genistein,
desmethylicaritin and icaritin were similar to that given by estradiol, albeit at varying
concentrations, indicating these three phytoestrogens are full estrogen agonists in the
MCF-7 cell proliferation assay. This contrasts with results obtained from experiments
involving stable ERα and ERβ HeLa cell lines where genistein was found to be a
superagonist where the maximal luciferase induction of genistein exceeded that achieved
by estradiol. In both stable ERα and ERβ HeLa cell lines, desmethylicaritin and icaritin
were partial agonists where their maximal luciferase inductions were lower than estadiol.
Icariside I and icarside II were also both found to be partial estrogen agonists and less
potent than icaritin and desmethylicaritin in all three assays.
In a paper involving protein–ligand docking simulations published by Wang et al.,
(2006), the prototypical estradiol molecule docked in the ERα ligand-binding domain was
found to be characterized by three hydrogen bonds formed between amino acid residues

Arg-394, Glu-353, and His-524 with estradiol, which are known to be essential for the
molecule’s agonistic activity. Using the same approach on the two aglycones, namely,

194
icaritin and desmethylicaritin, the former compound’s higher estrogenicity could be
attributed to the ability of the hydroxyl group on ring B to interact with Glu-353 and Arg-
394 and another similar interaction involving the hydroxyl group on ring A with His-524.
In contrast, for the case of icaritin, the interaction with Glu-353 and Arg-394 in the ERα
ligand-binding domain was blocked due to the presence of the hydrophobic methyl group
in ring C. The lower estrogenicities of icariside I, icariside II and icariin could be
explained by the steric hindrance produced by the presence of bulky sugar groups which
prevents the required docking to the receptor site and increases the molecules’
hydrophilicity.
Prenylation at the 8-position of the flavonoid backbone was observed to lead to an
enhancement of estrogenicity. Desmethylicaritin is the prenylated version of kaempferol
and the estrogenic potency of the desmethylicaritin is ~ 55 times higher than that of
kaempferol. A similar observation has been reported in studies comparing the
estrogenicities of naringenin and its prenylated derivative, 8-prenylnaringenin, which is
found in hops and beer (Milligan et al., 2000; Schaefer et al., 2003; Kretzschmar et al.,
2010).
Icaritin and icariside I were discovered to be ERα-selective ligands in this study.
This is quite unlike common flavonoids, such as apigenin, kaempferol, luteolin, quercetin
and genistein, which are mainly ERβ-selective. 8-Prenylnaringenin, the prenylated
version of naringenin, was also found to be ERα-selective (Schaefer et al., 2003).

195
4.1.4 Estrogenic activities of binary mixtures

The ability of ERα and ERβ stable cell lines and MCF-7 breast cancer cells to
detect estrogenic agonistic and antagonistic activities of mixtures was studied. ERα and

ERβ stable cells; and MCF-7 breast cancer cells were first exposed to binary mixtures
containing sub-maximal dose of estradiol and well known ER-antagonists such as 4-
hydroxytamoxifen, raloxifene and and ICI 182,780. All three compounds exhibited
antagonistic activity towards estrogen-induced luciferase activity in a dose-dependent
manner as reported in previous studies (Sonneveld et al., 2005; Escande et al., 2006).
MCF-7 breast cancer cells also showed reduced cell proliferation when incubated with
estradiol in the presence of increasing doses of 4-hydroxytamoxifen in a dose-dependent
manner.
Phytoestrogens have been reported in some studies to possess both additive and
antagonistic properties. First, the effects of the well known soy phytoestrogen, genistein,
on estrogen-induced luciferase transactivation were examined using ERα and ERβ cell
lines and were found to exhibit superagonistic activity in the presence of estradiol in a
dose-dependent manner. In contrast, there was insignificant enhancement of cell
proliferative effects when MCF-7 breast cancer cells were exposed to low doses of
genistein and estradiol. When doses of genistein were increased to 100 nM and greater,
inhibitory effects on estrogen-induced cell proliferation were observed. In all three assays,
low doses of compounds derived from Epimedium did not show any significant additive
estrogenic activity in the presence of estradiol. Above 1 µM in the MCF-7 cell
proliferation assay and 10 µM in ERα and ERβ cell lines, Epimedium compounds were
observed to suppress estrogen-induced bioactivity.

196
Any antagonistic effects against estrogen stimulation exerted by test compounds
need to be examined more closely as these may be brought about by non-specific
inhibition of reporter gene activity or overall cytotoxicity, or more specific effects such as
inhibition of protein synthesis or mRNA transcription (Sonneveld et al., 2005). Genistein
and other flavonoids have also been demonstrated to influence effects on other signaling
pathways, such as tyrosine kinases, mitogen-activated protein kinases, and protein kinase
C inhibition (Akiyama et al., 1987; Kuo & Yang, 1995; Knight & Eden, 1996; Kurzer &
Xu, 1997).

An experiment in the form of an ‘estrogen rescue’ can be used to verify non-
receptor-specific inhibitive effects of the test compound on the reporter gene activation or
MCF-7 cell proliferation. The ‘estrogen rescue’ experiment is easily performed by
increasing the concentration of estradiol so that receptor-specific inhibitive effects can be
reversed. This approach worked well with a range of androgen receptor antagonists tested
in a study involving of an androgen receptor luciferase reporter assay by Sonneveld et al.,
(2005) where all inhibitory responses were reversed by co-incubation with excess
dihydrotestosterone, demonstrating the specificity of the response. In contrast, the
inhibitory effects of high levels of a number of individual brominated flame retardants
congeners could not be reversed by excess dihydrotestosterone, which coincided with
cytotoxicity of these ligands, as assessed through inhibition of expression of a
constitutively expressed reporter gene and a positive response in the MTT assay (Hamers
et al., 2006).


197
4.1.5 Application of in-vitro ER-responsive bioassays as clinical tools to understand
estrogenic effects of compounds in serum

In this study, eight healthy male subjects were enrolled and administered
separately of estradiol valerate and Epimedium pubescens decoction. Serum samples were
obtained and assayed ex-vivo for levels of estrone and estradiol by tandem mass
spectrometry, for ERα and ERβ bioactivity and MCF-7 breast cancer cell proliferative
effects. This study provides new validated assays, which measure global activity in serum,
to evaluate the effects of ligands on ERα, compared to ERβ activity and breast cancer cell
growth.
Regression modelling that was performed indicated that ERα and ERβ bioassays
correlated better with summated estrone and estradiol compared to either estrogen alone,
demonstrating their ability to reflect the global activity of estrogens. In contrast, MCF-7
cell proliferation appeared to be driven mainly by estradiol. This is not surprising

considering that MCF-7 cells express predominantly ERα.
Estradiol is a more potent estrogen than estrone and estrone is quantitatively the
predominant estrogen both endogenously and after ingestion of estradiol valerate. Estrone
is produced by the rapid metabolism of estradiol in the liver (Vree & Timmer, 1998).
However, estrone levels are not measured routinely in clinical practice, and its
contribution to estrogenicity is regarded as negligible (Kuiper et al., 1997). In our study,
coefficient ratios for estrone/estradiol were two-fold higher for ERβ compared to ERα,
suggesting that estrone contributed twice as much to ERβ activity in subjects’ sera.
Estrone can induce steroid receptor coactivator-1 recruitment to ERβ with much higher
efficiency than for ERα (Margeat et al., 2003), and estrone–ERβ complexes can bind
coactivator LXXLL fragments with higher affinity than estrone–ERα complex (Ozers et

198
al., 2005). Our data indicated that consumption of estradiol valerate induced 600% higher
incremental mean estrone AUC compared to the 72% increase observed with estradiol.
The higher concentrations of estrone, plus the distinct affinity of the estrone–ERβ
complex for coactivator peptides, may mean that estrone can exert significant estrogenic
effects following ingestion of estradiol valerate. The associations between high serum
estrone levels with breast cancer risk in postmenopausal women (Miyoshi et al., 2003;
Missmer et al., 2004) and with bone health in elderly men (van den Beld et al., 2000)
underline the need for further evaluation of the relative significance of this estrogen to
health in aging populations.
There was a consistent pattern of diurnal variation of estrogenicity in the subjects,
consistent with previous studies that reported night time fall in testosterone and estradiol
levels in pubertal males (Albertsson-Wikland et al., 1997). This intrinsic rhythm with a
clear nocturnal and early morning rise is driven by changes in GnRH pulse generator and
LH secretion from the pituitary gland. After adjustment for diurnal variation, the effects
of estradiol valerate on ERα, ERβ, and MCF-7 AUC profiles paralleled that of estrone
and estradiol, confirming that these were indeed the main bioactive products causing
estrogenicity in sera. Ingestion of estradiol valerate resulted in a 67% increase in adjusted

AUC over baseline for both ERα and ERβ, comparable to the 72% observed for estradiol.
The AUC increase for MCF-7 was 23%, reflecting a saturation effect limiting the
proliferation of breast cancer cells. The data indicated that ERα, ERβ, and MCF-7 cell
proliferation bioassays are useful instruments to measure changes in estrogenicity in sera
following ingestion of a standard dose of a common estrogenic drug.


199
The use of poorly defined botanical entities has resulted in controversial results
despite many studies of estrogenic flavonoids, including soy products (Williamson &
Manach, 2005). Cell-based bioassays developed in the present study were used to
examine the bioavailability and biological effectiveness in sera of the decoction when
administered orally. Ingestion of a defined Epimedium decoction resulted in a 6%
increase in ERα AUC effect over baseline in the subjects. The increase in Epimedium
ERα estrogenicity was contributed to a large extent by the 24 h time period and was more
than 10-fold lower than that of observed following ingestion of estradiol valerate. A
single dose of this traditionally prepared Epimedium pubescens decoction may not to
exert physiological effects. The development of viable and robust bioassays facilitates the
rapid screening of similar herbal extracts with purported estrogenic properties and the
measurement of estrogenic activity ex-vivo at close time points for pharmacokinetics and
pharmacodynamics studies.


200
4.2 Pharmacokinetics and pharmacodynamics of Epimedium compounds
4.2.1 Pharmacokinetics of unconjugated prenylflavonoid aglycones from aqueous
Epimedium decoction in humans

In this pilot study, only the concentrations of icaritin and desmethylicaritin were
measured as they are prenylflavonoids unique to Epimedium. In addition, icaritin and

desmethylicaritin are both aglycones that exert stronger estrogenic activities than their
glycosides and flavonoids such as apigenin, kaempferol, luteolin and quercetin (Table 4).
Human sera, obtained from healthy male volunteers over a 48 h period following
ingestion of a traditional Epimedium decoction, were analyzed for their icaritin and
desmethylicaritin content. Icaritin was first detected at 1 h where the mean value was 0.24
nM, which suggests that very little of the aglycone was directly absorbed. In contrast,
genistein or diosmin, which were shown to be directly absorbed exhibited high
concentrations in blood within 1 h (Kano et al., 2006). This was also consistent with the
low levels of the aglycone in Epimedium decoction (119.9 μg/g). The level of icaritin was
also observed to rise rapidly after 4 h, reaching a peak at 8 h (1.51±1.6 nM) and returning
to almost baseline (0.25±0.17 nM) after 48 h.
The estrogenicity of these trace amounts of unconjugated icaritin was well below
the detection limits of the ERα and ERβ, as well as, MCF-7 cell-based bioassays. The low
amounts of the compound were likely to be largely bound to serum proteins, particularly,
human serum albumin but not sex hormone-binding globulin, which would influence its
bioavailability. A large proportion of proteins found in serum are made up of albumin,
which binds and transports acidic and neutral bioactive molecules and food nutrients
(Xiao et al., 2011b). Studies that examined the bioavailability of flavonoids in human
serum albumin reported that flavonoid bioavailability is often poor due to interaction with

201
this protein (Bolli et al., 2010). Walle et al. (2001) demonstrated that binding of the
flavonoid chrysin can be as high as > 99%. Via fluorescence spectroscopy, Dufour &
Dangles (2004) reported that flavonoids display moderate affinities for albumins and
flavones and flavonols being most tightly bound. Methylation of hydroxyl groups
improved the affinities for human serum albumin by 2 to 16-fold (Xiao et al., 2011a)
whereas glycosidation and sulfation could lower the affinity to albumin by one order of
magnitude depending on the conjugation site (Dufour & Dangles, 2004).
Globulins can also be found in serum and are used in the transport of ions,
hormones and lipids responsible for immune function (Xiao et al., 2011b). The sex

hormone-binding globulin transports and regulates the access of sex steroids such as
testosterone and estradiol to their target tissues. The number of steroid-binding sites in the
sex hormone-binding globulin far exceeds the molar concentrations of sex steroids and
can accommodate ligands such as phytoestrogens and fatty acids (Hodgert Jury et al.,
2000). In a study by Hodgert Jury et al. (2000), the relative binding affinities (RBAs) of
various compounds in comparison with estradiol (RBA= 100) to sex hormone-binding
globulin determined using competitive displacement assays. Some phytoestrogens bound
with RBAs of 0.12 (coumestrol) to 0.04 (naringenin). Many compounds did not bind to
sex hormone-binding globulin with sufficient affinity to allow RBA measurements and
these include several phytoestrogens, such as genistein and kaempferol. An earlier study
by Déchaud et al. (1999) reported binding affinity constants of a range of xenoestrogens
measured in equilibrium conditions via solid phase binding assay, also had relatively
similar findings. Flavonoids such as genistein and naringenin were showed weak binding
affinities that were at least three orders of magnitude lower than testosterone and/or
estradiol. Flavonoid glucoside derivatives, genistin and naringin do not bind to the sex

202
hormone-binding globulin. Hence, it is not likely that icaritin will bind to the sex
hormone-binding globulin protein found in serum.
The late appearance of icaritin in sera of human volunteers was hypothesized to be
due to the metabolism of icariin to icaritin which occurs mainly in the intestine. This view
is supported by animal data showing that icariin is stable in gastric juice, and that
hydrolysis of icariin to icaritin occurs in the intestine (Qiu et al., 1999). The aglycone is
formed through successive hydrolysis of the diglycoside icariin and triglycosides by
intestinal enzymes. Significant differences in peak concentrations between subjects were
observed ranging from 5.11 nM in subject E to 0.61 nM in subject G. These may reflect
differences in enzymatic activity contributed by inter-individual variability in bacteria and
enzyme activities (Walle et al., 2005).
Polymorphisms in uridine diphosphate glucuronosyltransferases which attach a
glucuronide residue to flavonoids may contribute to variability in flavonoid

bioavailability but the effects of these polymorphisms have not been widely studied. The
selectivity of conjugation of the flavonoids by UDP-glucuronosyltransferases depends on
structure of the flavonoid and the enzyme. Studies on UDP-glucuronosyltransferase
polymorphisms, to date, mainly focus on the glucuronidation of drugs and various
xenobiotics and similar effects may be seen with ingested flavonoids (Lampe & Chang,
2007). A minute fraction of ingested flavonoids are conjugated with a sulfate residue by
sulfatases found in the liver and gastrointestinal tract. Single-nucleotide polymorphisms
in sulfatases 1A1 and 2A1 have been reported and were found to be associated with
altered response to therapeutic agents and sex steroid concentrations, respectively and this
could influence the disposition of phytochemicals metabolized by sulfatases (Nowell &
Falany, 2006).

203
Flavonoids are also extensively metabolized by intestinal bacteria that reside in
the gut. Notable examples in literature include the conversions of isoflavone daidzein to
equol, lignans, such as, secoisolariciresinol diglucoside and matairesinol to enterolignans
which include enterodiol and enterolactone, and hop-derived prenylchalcone
xanthohumol to 8-prenylnaringenin by intestinal bacteria. The reasons for such inter-
individual differences still remain unknown and diet and host genetics have been thought
to be the contributory factors (Lampe & Chang, 2007).
Desmethylicaritin was not detected in any one of the human serum samples. One
contributory reason may be because the amount of desmethyicaritin (0.19 mg) in the
water decoction was only a quarter that of icaritin (0.75 mg). The low concentrations of
unconjugated icaritin and desmethylicaritin could also be brought about by extensive
conjugation to form sulfates, glucuronides and/or methylated conjugates via the
respective action of sulfatases, uridine diphosphate glucuronosyltransferases and
catechol-O-methyltransferases found in the small intestines and liver.

204
4.2.2 Pharmacokinetics and pharmacodynamics of unconjugated prenylflavonoids

after ingestion of an enriched Epimedium preparation by rats

In view of the low bioactivity measured in a traditionally prepared, aqueous
Epimedium pubescens extract, a prenylflavonoid-enriched extract based on Epimedium
brevicornu was subsequently formulated and fed to female ovariectomized Sprague–
Dawley rats. The concentrations of icariin, as well as two other estrogenic glycosides,
namely, icariside I and icariside II, together with the two aglycones, icaritin and
desmethylicaritin were measured in rat sera. The quantification of prenylflavonoid
glycosides, such as, icariin, icariside I and icariside II, will allow the understanding on
whether the lack of estrogenic activity was due to conjugation of prenylflavonoids.
In this second study, an early phase during which unconjugated icariin and
icariside II reached peak levels in 0.5 to 1 h and a late phase wherein unconjugated
icariside I, icaritin and desmethylicaritin peaked at 8 h. Treatment with Epimedium extract
resulted in non-linear increases in AUC and delayed peaks.
Icariin was the first prenylflavonoid to be detected at t
max
0.5 h which increased in
a non-linear dose-dependent manner after oral administration of the extract. This was
similar to that observed when icariin was administered as a pure compound (Liu & Lou,
2004) and consistent with direct absorption (Xu et al., 2007). Although icariin was the
main flavonoid glycoside in the extract (14% by weight), peak concentrations of this
glycoside in serum were in the low nanomolar range, suggesting poor bioavailability of
the intact molecule.
On the other hand, icariside II was detected at several fold higher concentrations
than icariin despite of its lower concentration in the extract (icariin:icariside II,
7.8:1.0;w:w). Compared to icariin, icariside II lacks a glucose moiety at position 7 on ring

205
A and this may make the latter more bioavailable. There is evidence from the rat intestinal
model that apical to basolateral permeability of monoglycosides can be more than 2-fold

greater than prenylflavonoids with 2 or more sugar moieties due to higher absorptive
permeability and carrier-mediated transport in the intestine (Chen et al., 2008). When
icariin was administered alone, icariside II can be rapidly derived from icariin by first-
pass deglycosylation and displays an early t
m a x
~ 1 h (Xu et al., 2007), similar to that
observed in our study. Thus, the early peak of icariside II with t
m a x
1 h is likely to be the
sum of absorption of icariside II in the extract and metabolism of icariin.
At the highest Epimedium dose, a secondary delayed icariside II peak with t
m a x
of
8 h was observed. This delayed peak of icariside II was not seen when it was derived from
icariin administered alone (Xu et al., 2007) but was reminiscent of the delayed re-entry
peak observed when biochanin A was administered as a mixture with quercetin and (−)-
epigallocatechin-3-gallate (Moon & Morris, 2007).
Remarkably, icariside I, icaritin and desmethylicaritin also exhibited a major peak
at t
m a x
8 h. At t
m a x
8 h, the most abundant flavonoid was icariside I with an AUC of 538
nMh
−1
at the highest drug dose. In comparison, AUC of the aglycones icaritin and
desmethylicaritin were an order of magnitude lower.
Rat sera were examined ex-vivo for ERα bioactivity. Rats administered with low
doses of Epimedium did not display any significant ERα activity. At the highest
Epimedium dose (600 mg/kg), ER bioactivity was observed at 8 h. This peak was

comparable in size to that of the minor peak after estradiol benzoate administration. Thus,
estrogenic activity in sera after ingestion of Epimedium lagged by several hours compared
to estradiol benzoate, and corresponded to the appearance of the bioactive metabolites
desmethylicaritin, icaritin and icariside I at 8 h.

206
4.2.3 Pharmacokinetics and pharmacodynamics of conjugated prenylflavonoids
after ingestion of an enriched Epimedium preparation by rats

To measure the amounts of conjugated flavonoids in test sera, analyses were
repeated after digestion by β-glucuronidase and sulfatase. The concentrations of three
compounds, namely, icariside II, icaritin and desmethylicaritin, could be studied in rat
serum samples after enzyme hydrolysis as under these conditions, icariin and icariside I
were deglycosylated completely to give icariside II and icaritin respectively, and could
not be detected in enzyme-treated sera.
A 0.5 h, a peak for icariside II was detected which was consistent with rapid
absorption and conjugation (Xu et al., 2007). The AUC value of icariside II after
digestion was observed to be several orders of magnitude higher than non-digested
samples and this increase reflected the sum of deconjugated icariside II and
deglycosylated icariin.
On the other hand, both desmethylicaritin and icaritin exhibited two peaks, a
minor one at 0.5 h and a higher one at 8 h where the C
m a x
values of icaritin and
desmethylicaritin reached ~ 2 and ~ 0.25 µM, respectively. The higher concentrations of
both compounds after treatment with β-glucuronidase and sulfatase that were two to three
orders of magnitude higher than before exposure to enzyme treatment, suggested that the
greater proportion of these two compounds and their precursors existed in conjugated
forms. This was consistent with data with the flavanone, naringenin, whereby > 95% of
the compound exists in the conjugated form (Ma et al., 2006).

Non-linear dose-dependent AUC increases were observed and saturation effects at
higher Epimedium doses were most evident for icaritin, possibly due to either reduced
absorption or increased metabolism. The isoflavone, genistein, also displayed non-linear

207
pharmacokinetics, attributed to reduced absorption at high doses (Zhou et al., 2008).
Remarkably, conjugated icaritin can be detected at micromolar levels up to 72 h after
administration of the high dose Epimedium extract. This has not been observed when
genistein alone was administered, where conjugated genistein reached baseline levels
after 36 h (Zhou et al., 2008).
Sera from rats fed Epimedium exhibited strong ERα bioactivity throughout most
of the 72 h study period. Dose-dependent increases in ERα activity were observed with
AUC of 374, 543 and 771 pM E2 h
−1
with increasing doses of Epimedium. The peaks of
ERα activity for Epimedium were biphasic with a small first peak a 0.5 to 1 h and larger
second peak at 8 h. This biphasic pattern reflected the appearance of icaritin,
desmethylicaritin, and icariside II at 0.5 and 8 h. Concentrations of desmethylicaritin and
icaritin reached 0.2 and 0.4 µM respectively at 0.5 h, corresponding values at the 8 h
time-point being 0.6 and 1.5 µM. Since the EC
50
of desmethylicaritin and icaritin was
0.07 and 0.83 µM respectively, it is plausible that estrogenic activity of Epimedium is
contributed by desmethylicaritin and icaritin.




208
4.2.4 Accumulation of conjugated Epimedium compounds in-vivo

A unique feature of Epimedium is its prolonged effect, with strong estrogenic
activity (40% of peak activity) being detected up to 72 h. This prolonged estrogenic
activity at higher Epimedium doses correlated with the persistence of micromolar levels of
icaritin at the 48 to 72 h sampling period.
It is relevant to note that pharmacological doses of flavonoids were administered
in this study. At the level of the enterocyte, cycling of flavonoid conjugates back to the
intestinal lumen at high doses may be markedly decreased, reflecting a saturation of the
intestinal conjugation pathways (Silberberg et al., 2006). The long elimination time for
the highest dose of icaritin may also be due to differences in the type of Phase II enzymes
that are activated. The possibility exists that sulfatases, rather than
glucuronosyltransferases, may be preferentially recruited at higher doses, resulting in the
production of very hydrophilic metabolites which may be less efficiently eliminated by
the kidney (Liu & Hu, 2007), thereby contributing to the prolonged presence of icaritin in
sera even 72 h after dosing. Since intestinal and hepatic metabolism of flavonoids at high
doses may differ from more physiological doses, our results may not be directly
applicable to all doses of flavonoids. Nevertheless, similar depot effects have been
observed in ERE-luc transgenic mice, whereby a single physiological dose of the
isoflavone was able to induce significant ER/ERE-driven activity in diverse target organs
such as liver, cerebral cortex and testis for over 24 h (Montani et al., 2008). Such delayed
effects may persist for at least 2 weeks, were unmasked by fasting and are thought to be
due to bioaccumulation in the intestines, liver and reproductive tissues (Penza et al.,
2007).

209
4.3 Gene expression profiling reveals partially overlapping but distinct genomic
actions of different estrogenic compounds in human breast cancer cells
4.3.1 Convergence of global gene expression profiles MCF-7 breast cancer cells
treated with estrogenic Epimedium prenylflavonoids, genistein and estradiol

Illumina's bead-based DNA microarray platform was used in this study to
understand the biological effects and their underlying mechanisms in MCF-7 breast
cancer cells that have been treated with a standardised, prenylflavonoid-enriched
Epimedium extract and a series of Epimedium-derived compounds. Each gene on the
microarray acts as a reporter gene and gene expression profiling was used to identify
specific genes in a biological pathway that are involved in the mode of action.
To date, this study is the first to obtain gene expression profiles of a standardized
Epimedium extract and its compounds. The expression profiles obtained for the range of
treatments were analyzed via two ways - responses of genes by hierarchial clustering and
correlation analysis of gene expression profiles. Hierarchial clustering grouped the range
of treatments into two main categories. One group consisted of well-known estrogenic
compounds such as estradiol and genistein, together with icariside I, icaritin,
desmethylicaritin and the standardized Epimedium extract. The other group comprised
ER-antagonist, 4-hydroxytamoxifen, the vehicle control, together with icariin and
icariside II, which were non-estrogenic at the dose tested.
Correlation analyses of expression profiles was performed to compare the gene
expression profiles of estradiol, an established estrogen, with that of genistein, as well as,
the various estrogenic treatments encountered in this study. The overall gene expression
profiles of estradiol and soy isoflavone, genistein, were found to be largely similar due to
the high correlation coefficient value that was obtained and this result was similar to those