De novo design, synthesis, and evaluation of novel nonsteroidal phenanthrene ligands for the estrogen receptor

Article abstract of DOI:10.1021/jm020536q

Although there are many estrogen receptor antagonists with improved tissue selectivity profiles compared with tamoxifen, optimal tissue selectivity has not yet been demonstrated. As such there is still a need for additional diversity and new chemical scaf

Full text of DOI:10.1021/jm020536q

1408
J . Med. Chem. 2003, 46, 1408-1418
De Novo Design , Syn th esis, a n d Eva lu a tion of Novel Non ster oid a l
P h en a n th r en e Liga n d s for th e Estr ogen Recep tor
J onathan M. Schmidt,† J ulie Mercure,†,‡ Gilles B. Tremblay,§ Martine Page´,§ Aida Kalbakji,§ Miklos Feher,†
Robert Dunn-Dufault,† Markus G. Peter,† and Peter R. Redden*,†
SignalGene Inc., 335 Laird Road, Unit 2, Guelph, Ontario, Canada N1G 4T2, and SignalGene Inc.,
8475 Avenue Christophe-Colomb, bureau 1000, Montreal, Quebec, Canada H2M 2N9
Received November 26, 2002
Although there are many estrogen receptor antagonists with improved tissue selectivity profiles
compared with tamoxifen, optimal tissue selectivity has not yet been demonstrated. As such
there is still a need for additional diversity and new chemical scaffolds to allow for exploration
of improved tissue selectivity. Here, we describe the discovery of a novel phenanthrene scaffold
for estrogen receptor ligands utilizing a ligand based de novo design approach. The nanomolar
binding of phenanthrenes, 12b,c, 14b,c, and 15 against human recombinant ERR indicates
that our ligand based de novo design approach was successful. From a gene transfection assay,
12b,c, 14b,c, and 15 displayed only antagonistic activity with no observable agonistic activity.
The alkyl 9,10-dihydrophenanthrene 16 (presumably a racemic mixture) was a substantially
more potent ER binder than the phenanthrenes. It also displayed only antagonistic activity
and was effective at inhibiting estradiol stimulated MCF-7 cell proliferation. These results
demonstrate that this phenanthrene (and 9,10-dihydrophenanthrene) scaffold warrants further
study as potential selective estrogen receptor modulators and/or pure antiestrogens.
In tr od u ction
Estrogen receptors (ER) are members of a superfamily
of ligand-activated transcription factors, which includes
progesterone, androgen, and glucocorticoid and miner-
alocorticoid receptors, as well as receptors for thyroid
hormone, retinoids, and vitamin D.1 Additionally a large
group of receptors known as ‘orphan receptors’, for
which no ligand has been described, belong to this
superfamily.
Stimulation of estrogen receptors by endogenous
estrogens plays an important role in both male and
female physiology. Estrogens are involved in the regula-
tion of cholesterol and lipid levels, the skeletal system,
the central nervous system, and reproductive func-
tions.2,3 However, estrogen stimulation is also impli-
cated in the development of breast cancer.4 Conse-
quently, many estrogen receptor ligands (see Figure 1)
are being developed with the aim of preventing estrogen-
mediated tumor growth. Tamoxifen, which was origi-
nally developed as an estrogen receptor antagonist, is
currently the hormonal treatment of choice for both pre-
and postmenopausal women with breast carcinoma. It
F igu r e 1. Estrogen receptor ligand structures.
is now known that tamoxifen and other selective
suggest that the tissue selectivity of certain estrogens
estrogen receptor modulators (SERMs) display a broad
may be related, at least in part, to their different effects
range of agonist and antagonist activity dependent on
at the ERR and ERâ subtypes.7,10,11 Other factors that
the tissue and species being evaluated and that several
may contribute to selectivity include the tissue-specific
factors are thought to contribute to this phenomenon.5-8
presence of corepressors and coactivators, and the cell/
Recently, a new estrogen receptor subtype was discov-
tissue/species accessibility of different DNA-response
ered and designated ERâ.9 Experiments based on the
elements.8,12
tissue distribution and pharmacology of ERR and ERâ
Since the tissue selective effects of tamoxifen have
been discovered, other SERMs with improved selectivity
* To whom correspondence should be addressed: Tel (519)-823-9088.
profiles have been or are being developed (Figure 1).
These include the benzothiophenes7,13,14 (raloxifene and
related analogues 1, LY335124 and 2, LY357489) and
triphenylethylene analogues7,15,16 (including tamoxifen),
Fax (519)-823-9401. e-mail peter.redden@signalgene.com.
† SignalGene Inc., Guelph, Ontario, Canada.
‡ Present address: SYNX Pharma Inc. 6354 Viscount Road, Mis-
sissauga, ON, Canada L4V 1H3.
§ SignalGene Inc., Montreal, Quebec, Canada.
10.1021/jm020536q CCC: $25.00 © 2003 American Chemical Society
Published on Web 03/18/2003

Nonsteroidal Phenanthrene Ligands for the Estrogen Receptor
J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 8 1409
along with the indole analogues17 (ERA-923 and TSE-
424). These have estrogenic activity on skeletal and
cardiovascular systems and are being developed as
alternatives to estrogen-replacement therapy. On the
other hand, ligands that have pure antiestrogenic
activity on breast tissues are being developed for the
treatment of breast cancer. These include faslodex and
EM-800 (4) (although recently it has been suggested
that EM-800 may possess more SERM-like activity than
pure antiestrogenic activity18,19), which are currently in
Phase II development.
Although generally these compounds display im-
proved selectivity compared with tamoxifen, optimal
tissue selectivity has not yet been demonstrated. As
such there is still a need for additional diversity and
new chemical scaffolds to allow for exploration of
improved tissue selectivity. Here, we describe the
discovery of a novel phenanthrene scaffold for estrogen
receptor ligands utilizing a ligand based de novo design
approach. We examined the activity of several members
of this novel class in ER binding and cellular assays and
the results demonstrate that this phenanthrene scaffold
warrants further study as potential selective estrogen
receptor modulators and/or pure antiestrogens.
F igu r e 2. Steroids used as input for the de novo design
approach, evolutionary molecular design (EMD).
Ch em istr y
DBU, gave ynones 7. Heating 7 with 5-cyanopyrone22
gave the biphenyl ketones 8. Cyclization of 8 to give the
key intermediate cyano phenanthrenes 9 occurred in a
one-pot reaction by first bromination with NBS, followed
by a Wittig reaction. Grignard treatment of 9 in the
presence of a catalytic amount of cuprous bromide gave
the keto phenanthrenes 10, which were demethylated
with either BBr3 or concentrated hydrogen bromide in
acetic acid to 11. Treatment of 11 with 1-(2-chloroethyl)-
piperidine gave a mixture of the mono- and disubsti-
tuted piperidine phenanthrenes with the desired phenan-
threnes 12 isolated relatively easily. A second Grignard
treatment of 12 generated the alcohol phenanthrenes
13, which were converted to the alkenyl phenanthrenes
14. Initial attempts at selective 9,10-double bond reduc-
tion of 14b gave 15 using Pd/C hydrogenation condi-
tions. However, Birch reduction of 12b reduced the 9,10-
double bond but also reduced the keto group to give the
racemic 9,10-dihydrophenathrene 16.
Com p u ta tion a l Mod elin g. The crystal structure of
ERR only became available during the synthesis and
biological testing of the molecules generated from the
EMD process. Therefore, to gain further insight into the
mode and energetics of binding of our de novo designed
compounds, computational docking studies were con-
ducted. The crystal structure used in the docking studies
was that obtained from the cocrystallization of ERR with
raloxifene as found in the PDB database23 (reference
code: 1ERR). After removal of raloxifene from the
pocket, ligands 12b,c, 14b,c, (1R)-15, (9R)-17b, (9S)-
17b, and the four stereoisomers of 17a were docked. For
comparative purposes, raloxifene was also redocked
using the same conditions as for all the other molecules.
Dock in g Resu lts. The calculated binding energies
from the docking experiments are presented in Table
1. The first column, Ebind,rigid, is the binding energy
obtained with the rigid receptor approximation with the
ligand having full torsional flexibility. The docking
calculations utilize simulated annealing which optimizes
De Novo Design . Our ligand based de novo design
approach is based on the proprietary Evolutionary
Molecular Design (EMD) computational technology.20
Briefly, EMD utilizes structural information and bio-
logical activity of known pharmacologically active and
inactive compounds as input structures to generate new
structures that share and/or improve the biological
activity of the original compounds (see Experimental
Section).
For the present ER case, the steroidal compounds,
estratriene, estratrien-17â-ol, estratriene-1,17â-diol, es-
tradiol, estratriene-3,7â,17â-triol, 11-ketoestratriene-3,-
17â-diol, estratriene-3-ol, 17-oxoestratriene-3-ol, 5R-
androstane-3R,17â-diol, and progesterone, given in Figure
2, which are known to bind to ER21 were used as input
for the EMD process.
The first 40 molecules generated in the de novo design
process were grouped into six structural classes or
scaffolds. The first group contained structures, not
surprisingly, based on estradiol. The second class con-
sisted of the triphenylethylene skeleton, and the third
class a series of flavanoid structures. The remaining
three scaffolds were novel and not known in the
estrogen literature. The result for one of these novel
classes based on phenanthrene is described here. We
will report on the other two scaffolds once synthesis and
biological testing are complete which we believe will
further demonstrate the utility and value of the EMD
process.
Syn th esis. The synthetic route for the preparation
of the phenanthrene ligands is given in Scheme 1.
Substituted anisoles are converted to p-methoxybenzal-
dehydes 5 by treatment with hydrogen chloride and zinc
cyanide followed by aluminum trichloride. Aldol Con-
densation of 5 with p-methoxyacetophenone using po-
tassium tert-butoxide, gave enones 6. Treatment of 6
in a one-pot reaction with bromine, followed by addition
of potassium acetate/acetic acid and then heating with

1410 J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 8
Schmidt et al.
Sch em e 1
the torsion angles of the molecules in order to achieve
optimum interaction with the receptor. In this process
two important factors are neglected: potential changes
in bond lengths during rotations around the torsion
angles within the binding pocket and changes in the
position of hydrogen atoms, such as sharing a hydrogen
in hydrogen bonding. These effects have been incorpo-
rated and the results are also given in Table 1. Ebind,lig
gives the binding energy after a force field minimization
of the ligand in the field of the receptor, while Ebind,H,lig
is the binding energy after a simultaneous optimization
of the positions of the ligand and the surface hydrogen
atoms. With all these approaches, the structure of the
receptor is kept rigid and hence where an induced fit
would occur in reality, these calculations predict high
binding energies. This can be remedied by keeping
flexible both the ligand and the receptor atoms in the
vicinity of the ligand. To avoid extreme distortion of the
receptor structure, this was carried out in a stepwise
fashion using a limited number of molecular mechanics
steps (see Experimental Section). The results with 10
and 50 molecular mechanic (MM) steps are shown in
columns Ebind,flex,10 and Ebind,flex,50. We limited the num-
ber of MM steps to 50 since a higher number of MM
steps may lead to an unrealistic distortion of the
receptor structure.
There are a number of assumptions that must be
highlighted with respect to the results of these docking

Nonsteroidal Phenanthrene Ligands for the Estrogen Receptor
J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 8 1411
Ta ble 1. Results of the Docking Experiments with the Rigid
and Flexible Receptor Approximations Using the ERR Binding
Site with Continuum Solvation (energies given in kcal/mol)
a
b
c
d
e
compound
Ebind,rigid Ebind,lig Ebind,H,lig Ebind,flex,10 Ebind,flex,50
14c
12c
14b
12b
(9R)-17b
(9S)-17b
(1R)-15
(1R,9R)-17a
(1S,9R)-17a
(1R,9S)-17a
(1S,9S)-17a
raloxifene
175.0
156.3
54.4
74.4
66.3
34.7
62.8
49.8
52.5
26.5
20.8
41.5
42.6
14.5
-4.8
23.0
-11.3
-0.3
20.1
14.0
16.9
22.8
-2.9
-8.2
19.7
-14.8
-5.3
13.7
8.5
14.4
-15.5
-19.4
-31.5
-12.6
-29.1
-29.9
0.8
-46.4
-42.4
-13.2
-11.9
-14.8
-44.0
-52.0
-53.6
-27.9
-36.8
-41.6
-2.9
-43.1
-46.5
-19.9
-17.7
-20.4
-40.8
-60.1
-53.9
-2.9
-14.8
-25.8
a
The binding energy obtained with the rigid receptor ap-
b
proximation with the ligand having full torsional flexibility. The
binding energy after a force field minimization of the ligand in
the field of the receptor. c The binding energy after a simultaneous
optimization of the positions of the ligand and surface hydrogens.
d
To avoid extreme distortion of the receptor a limited number of
molecular mechanic (MM) steps (10) were used. e 50 MM steps.
See Experimental Section for further details.
experiments. Errors in the calculated binding energies
are partly caused by the approximate nature of molec-
ular mechanics energies and the treatment of solvation
contributions. The continuum solvation terms may be
unsuitable to describe the inside of the binding site and
the effect of single water molecules in the pocket. Also,
the position of bound water molecules might change
from ligand to ligand but is fixed in the rigid receptor
models. Although water molecules can move around in
the flexible binding site model, this approximation
greatly overemphasizes the flexibility of the receptor.
The process is a minimization in Cartesian space and
hence changes the position of both backbone and side
chain protein atoms. However, since the receptor is very
bulky, the backbone of the receptor cannot quickly
assume conformations that are very different from the
original conformation. However, despite the approxi-
mate nature of the calculated binding energies, the
described docking studies provide additional evidence
of the validity of our de novo designed structures with
respect to binding at ERR.
An examination of the docking results leads to some
interesting observations. All protocols predict that ral-
oxifene, 17a , and 17b bind better than any of the other
molecules studied, while their relative order depends
on the actual protocol used. The isopropyl ketone 12c
and alkene 14c do not fare very well in the rigid receptor
docking because of the conserved water molecule that
is likely to be squeezed out in reality but was considered
present in the docking studies presented in the table
below. These, however, fit the pocket much better once
the receptor within 10-Å of the ligand is fully relaxed.
Apart from these, the ketone 12b is predicted to be the
worst binding molecule of all.
F igu r e 3. The Gauss-Conolly surface of the estrogen receptor
(1ERR) binding pocket with docked (1S,9R)-17a . The surface
is colored according to the hydrophobibicity of neighboring
residues (green, hydrophobic; blue, hydrophilic). The residues
nearest to the ligand are also shown (ARG392, GLU353,
HIS524, and ASP351).
it cannot form any H-bond in the tail section due to
steric restrictions and its oxo-group is also unable to
H-bond to HIS524, the oxo-group is at least not forced
close to a hydrophobic residue. However, if the ligand
is relaxed, the predicted binding affinity of (9R)-17b
becomes increasingly similar to (9S)-17b as the effect
of the mismatch is reduced.
The comparison of the four different 17a stereoiso-
mers reveals the same trend. In all experiments (1S,9S)-
17a is predicted to be the best stereoisomer, followed
(1R,9S)-17a . The change in the stereochemistry of the
4-(2-(piperidin-1-yl)ethoxy)phenyl side chain appears to
be critical, (1S,9R)-17a and (1R,9R)-17a are expected
to be substantially less active. In these two cases the
headgroup of the molecule is in a different binding mode
to allow sufficient space for the side chain. Due to the
fact that all four stereoisomers bind (albeit in different
modes), experimentally these might compete with each
other for binding to the receptor. The docked molecule,
(1S,9S)-17a , is shown in Figure 3 where it nicely fits
the pocket. It must be noted, however, that even the
binding of (1S,9S)-17a is not perfect, as it is unable to
hydrogen bond to HIS524.
Biology
Recep tor Bin d in g Assa y. The binding data for
compounds 12b,c, 14b,c, and 15 with human recombi-
nant ERR, and ERâ are given in Table 2. At ERR the
alcohol phenanthrene 15 gave a Ki of 120 nM, the alkene
phenanthrenes 14b,c gave 290 nM and 210 nM, respec-
tively, and the ketones 12b,c displayed lower binding
(the Ki for 12c was 490 nM). The 9,10-dihydrophenan-
threne 16 was a more potent binder at ERR (Ki of 8.69
nM) and showed a 20-fold selectivity for ERR over ERâ.
Gen e Tr a n scr ip tion . Phenanthrenes 12b,c, 14b,c,
and 15 and the dihydrophenanthrene 16 were evaluated
It is interesting to compare the predictions for the two
stereoisomers of 17b. With the rigid receptor ap-
proximation, the binding energy of (9R)-17b is higher
than (9S)-17b. Unfortunately, neither of these ligands
appears to bind in an optimal fashion. While (9R)-17b
establishes a hydrogen bond in the tail section (with
ASP351), its oxo-group is wedged into a hydrophobic
region, unable to establish an H-bond with HIS524.
(9S)-17b is slightly better in this rigid docking. Although

1412 J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 8
Schmidt et al.
Ta ble 2. Summary of In Vitro Activity of Phenanthrene
Analogues
binding (Ki, nM)a
transcriptional
compound
ERR
ERâ
activationb
12b
12c
14b
14c
15
16
>1000
630 + 47
antagonist
antagonist
antagonist
antagonist
antagonist
antagonist
490 ( 130
290 ( 60
>1000
180 ( 17
200 ( 10
190 ( 5
210 ( 60
120 ( 54
8.69 ( 0.33
1.04 ( 0.91
0.25 ( 0.02
190 ( 35
ICI 182780c
4-OH-Tamd
1.39 ( 0.14 antagonist
0.16 ( 0.04 partial agonist/
antagonist
DESe
0.49 ( 0.09
0.63 ( 0.13 agonist
Binding affinity by competition with 0.5 nM [3H]-17â-estra-
diolspecific binding to recombinant human estrogen receptors, ERR
and ERâ. Results represent the mean ( SEM of three separate
experiments. b Effect on transcriptional potential (ERE-stimulated
luciferase activity) of human ERR and ERâ on the vitellogenin A2
a
F igu r e 5. Effect of 16 on the inhibition of E2-stimulated
MCF-7 cell proliferation. Starved cells were treated with 10
nM E2 in the absence or presence of increasing concentrations
of 16 as indicated. The growth of unstimulated cells is shown
(0.1% ethanol, control). Results represent the mean ( SEM of
three separate experiments and are expressed as the percent
of the maximun E2 response in the absence of 16.
c
d
e
ERE. Pure antiestrogen. 4-Hydroxytamoxifen. Diethylstil-
besterol.
proprietary ligand based computational technique, EMD,
with the aim of doing exactly that. To demonstrate the
utility of EMD, we chose ER (ERR) as our first target
since there is in fact a need for new chemical scaffolds
and because the ERR crystal structure (at the time) was
unavailable. The steroids of Figure 2 were chosen as
input structures. These steroids bind to ERR with
varying degrees of affinity irrespective of their agonistic/
antagonistic profile. To incorporate only antagonistic
activity, we incorporated the typical 4-(2-(piperidin-1-
yl)ethoxy)phenyl side chain found in antiestrogens and
SERMs.
As outlined in Scheme 1, the EMD-generated phenan-
threne analogues were synthesized and subsequently
tested for in vitro ER binding and antagonistic activity.
The nanomolar binding of 12b,c, 14b,c, and 15 against
human recombinant ERR indicates that our ligand-based
EMD de novo design approach was successful. This was
particularly gratifying since the synthetic routes for
these compounds, although straightforward, did require
a number of steps.
F igu r e 4. Effect of 16 on ERR-mediated transactivation.
HepG2 cells were transfected with the pCMX-hERR expression
plasmid in the presence of the ERE3bLuc reporter plasmid and
treated with 10 nM E2 in the absence or presence of increasing
concentrations of 16 as indicated. Transfected cells treated
with the vehicle (0.1% ethanol, control) indicate basal levels
of transcription. Results represent the mean ( SEM of three
separate experiments and are expressed as the percent of the
maximun E2 response in the absence of 16.
Since 17-hydroxy-containing estradiol compounds
were used as input (see Figure 1) and bind strongly to
ERR, not surprisingly H-bond donors (e.g., OH) on the
C ring of the phenanthrene scaffold were generated (i.e.,
15). However, compounds containing H-bond acceptors
(e.g., ketone) and hydrophobic (e.g., alkene) groups at
this position, 12b and 14b, respectively, were also
generated. Moreover, compounds 12c and 14c were also
generated, indicating that there was a limited degree
of bulk tolerance available in this C ring region. Once
biological data became available, it was envisioned that
subsequent design cycles using these compounds could
further refine the requirements for this region if re-
quired. As it turned out, with the synthetic route
devised, the conversion to all compounds was carried
out very easily starting with the appropriate ketone
intermediate (10, see Scheme 1).
in a gene transfection assay by monitoring the tran-
scriptional activity of human ERR on a luciferase
reporter gene under the control of the vitellogenin A2
estrogen response element. As given in Table 2, all
phenanthrenes were determined to be antagonists due
to their ability to abrogate the E2-responsive ERR
activity. As expected, DES was determined to be ago-
nistic, and 4-hydroxytamoxifen was determined to be a
partial agonist/antagonist. An example of the transcrip-
tion assay result for 16 is shown in Figure 4.
MCF -7 Cell P r olifer a tion In h ibition . The 9,10-
dihydrophenanthrene 16 was tested for its ability to
inhibit E2-stimulated human breast cancer (MCF-7) cell
proliferation. The result, shown in Figure 5, demon-
strates that 16 effectively abrogates the E2-stimulated
cell growth.
As shown in Table 2, for the phenanthrene alcohol
15, the binding was better at ERR than for either the
ketones 12b,c or alkenes 14b,c, although the differences
are small. A related observation was seen for analogues
of raloxifene where the ER binding was comparable
when various groups including ethene replaced the 4′-
Discu ssion
The identification of new chemical scaffolds that share
or improve the biological activity of known drug target
ligands is unquestionably of tremendous value in the
drug discovery process. We have developed a novel

Nonsteroidal Phenanthrene Ligands for the Estrogen Receptor
J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 8 1413
hydroxyl group.24 Since 15 contains a secondary alcohol,
as in estradiol, it might be expected to have higher
binding than compounds with ketone or alkene func-
tionality in this region. Moreover, since 15 is presum-
ably a racemate, increased binding (2-fold maximum one
enantiomer is not competing) might be seen with one
of the individual enantiomers. These binding results
tend to corroborate the computational predictions.
Generally for both the rigid and flexible results (Table
2), the calculated binding of 15 (actually (1R)-15) is
similar to the alkenes 14b,c with the differences again
being small. Of all the compounds 12b was predicted
to be the worst ligand at ERR and did turn out to be the
least active experimentally. Additionally, although we
did not use ERâ in docking studies, 12b, experimentally,
was not a good binder at ERâ, being only slightly better
than 12c. 12c was predicted to have binding similar to
the alkenes 14b,c but was somewhat less active experi-
mentally.
and two stereoisomers, respectively, each required
separate docking. Our computational docking studies
using the ERR crystal structure (cocrystallized with
raloxifene) suggests that it is the 9S as opposed to the
9R configuration of the 4-(2-(piperidin-1-yl)ethoxy)-
phenyl side chain of these 9,10-dihydrophenanthrenes
that docks with higher binding to ERR. For instance for
both the rigid and flexible docking experiments, both
17a and 17b with the side chain in the S configuration
are predicted to dock with greater affinity than with the
side chain in the R configuration. Also 17a and 17b with
the side chain in the S configuration were predicted to
bind better to ERR than any of the aromatic phenan-
threnes. However, our initial attempt to generate either
17a or 17b from 12b resulted in both the reduction of
the ketone as well as the 9,10 double bond to give the
alkyl 9,10-dihydrophenanthrene 16. Although 16 is
presumably a racemic mixture, it was a substantially
more potent ER binder than the phenanthrenes (see
Table 2). This was gratifying since, although we did not
study 16 computationally, this experimental result
indirectly corroborates our computational studies re-
garding the 9,10-dihydrophenanthrenes. 16 also dis-
played only antagonistic activity and was effective at
inhibiting E2 stimulated MCF-7 cell proliferation. Clearly,
additional agonist/antagonist profiling of 16 and other
9,10-dihydrophenanthrene analogues as well as the
separation of the enantiomers is required to fully
explore this scaffold for an anti-estrogen/SERM indica-
tion.
In summary, a series of analogues based on the
phenanthrene scaffold were identified using our pro-
prietary EMD de novo design method. After synthesis
and testing, all compounds with the exception of 12b
demonstrated affinity for ERR as well as ERâ. 12b,c,
14b,c, and 15 did not show any agonistic activity in the
gene transcription assay, which is very encouraging in
the quest for alternate antiestrogenic or SERM-like
compounds. The 9,10-dihydrophenanthrene 16 demon-
strated even higher affinity for ER than the phenan-
threnes, was antagonistic, and inhibited E2-stimulated
MCF-7 cell growth. Overall this experimental in vitro
evidence suggests that the de novo designed phenan-
threne (as well as the 9,10-dihydrophenanthrene) scaf-
fold analogues are worthy of further study as anties-
trogens or SERMs and is the first example that
demonstrates the utility of our EMD approach.
Generally SERM or antiestrogenic compounds consist
of a flat aromatic scaffold with either two phenolic
groups13 as is the case for raloxifene, its fixed ring
analogues (1, LY335124 and 2, LY357489), TSE-424,
and EM652, as shown in Figure 1, or a single phenolic
group and a phenyl group in place of the second phenol
such as lasofoxifene (CP336,156) and levormeloxifene25
plus a side-chain imparting antiestrogenicity. It has
been suggested that it is the orientation of the basic
amino-containing side chain of tamoxifen and raloxifene
that is responsible for their different biological activi-
ties.13
With tamoxifen, the side chain is coplanar with the
stilbene plane, giving rise to partial agonist/antagonist
activity, whereas in raloxifene it is orthogonal with
respect to the benzothiophene moiety conferring an-
tagonism. Structurally related fixed ring analogues of
raloxifene13 (1, LY335124 and 2, LY357489) and the
benzopyran, EM-800, both which have orthogonal side
chains, seem to corroborate this hypothesis. In particu-
lar it is the S configuration of EM-800 that appears to
hold the biological activity.18 Moreover, TSE-424, which
is in clinical studies as a SERM for postmenopausal
osteoporosis,17 has the side chain orthogonal with the
indole moiety.
Therefore, on the basis of the hypothesis regarding
the amino-containing side chain geometry, the phenan-
threne structures, having a coplanar 4-(2-(piperidin-1-
yl)ethoxy)phenyl side chain, should behave more like
tamoxifen than raloxifene with regards to antagonism.
However, 12b,c, 14b,c, and 15 along with the pure
antagonist ICI 182780 demonstrated only antagonist
activity in the gene transfection assay. DES and 4-hy-
droxytamoxifen behaved as expected as an agonist and
partial agonist/antagonist, respectively, in the assay. It
is possible that any partial agonist/antagonist activity
will only show up in uterine tissue, as was the case for
N-arylbenzophenanthridines.26
Exp er im en ta l Meth od s
Com p u ta tion a l Mod elin g. Ligand docking was performed
with full torsional flexibility of the ligands into both a rigid
receptor and a flexible receptor using algorithms within the
MOE suite along with algorithms developed during the course
of this work. In all calculations the Born continuum solvation
model was applied27-29 in assessing solvation-desolvation with
a constant (distance independent) dipole term, as implemented
in the Molecular Operating Environment (MOE) suite of
programs.30
F lexible Dock in g of Molecu les in to a Rigid Recep tor .
The docking algorithm utilized was a multiple start Monte
Carlo method.31,32 First, a box is defined around the ligand,
such that it is sufficiently large to encompass the active site
fully but small enough to prevent docking on the receptor
surface. In the docking process different orientations and
conformations of the ligand are tested, in which all of its atoms
are inside the docking box. Each docking run begins with the
ligand in a random conformation and orientation. Some or all
Since an orthogonal side chain suggested antago-
nist activity we also considered the possibility that the
9,10-dihydrophenanthrenes as exemplified by 17a ,b
would give antagonist activity. We chose the hypotheti-
cal compounds 17a ,b to dock since it was anticipated
that these would be the compounds that would be
generated from 12b. Since these compounds contain four

1414 J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 8
Schmidt et al.
of the rotatable bonds of the ligand are then randomly
perturbed, while the bond lengths are held fixed. The energy
of a given configuration is evaluated using the Merck force
field (MMFF94), using the terms described below. Although
calculating force field energies for docking purposes is not
widely used due to lengthy computer time, the accuracy of the
method has been demonstrated.33,34 Each step is evaluated and
is accepted or rejected on the basis of the Metropolis criterion.35
Favorable binding modes are generated in repeated simulated
annealing runs. A simulated annealing run consists of a
sequence of Monte Carlo cycles: the ligand temperature within
each cycle is kept constant but is systematically reduced from
one cycle to the next (in our case, 1000 to 100 K in six cycles).
This simulated annealing process is repeated a number of
times, leading to different docked orientations and conforma-
tions. In all cases, at least 100 such configurations were
produced.
The potential changes in bond lengths during rotations
around the torsion angles were taken into account by the
optimization of ligand geometry, shown as column Ebind,lig in
Table 1. To account for the changes in the position of hydrogen
atoms as a result of hydrogen bonding the positions of surface
hydrogens were also optimized (shown as Ebind,lig,H in Table 1).
Only those hydrogens were considered that were within 3 Å
of any of the ligand atoms. The 3-Å radius ensures that only
four residues (i.e., ARG394, GLU353, ASP351, and HIS524)
of ERR directly interacting with the ligand and the preserved
water molecule are affected. In both cases, the optimizations
were terminated after 50 molecular mechanics steps, and it
was made sure in all cases that the calculations were suf-
ficiently close to convergence.
From this, the nonbonded ligand-receptor interaction en-
ergy, Ebind, is calculated as follows:
Ebind ) Ecpx - (Elig + Eprot
)
(1)
where the E denotes the total energies with the superscripts
corresponding to the complex, the ligand and the protein,
respectively. The energy terms can be partitioned into bonding
(subscript bd) and nonbonding (subscript nb) terms:
cpx
nb
lig
nb
prot
nb
cpx
bd
lig
bd
prot
bd
Ebind ) [E - (E + E )] + [E - (E + E )] (2)
To perform the docking calculations fast, the energy is
referenced to the energy of the ligand in the geometry it is
found in the complex (Elig,cpxgeom). This is different from the
actual lowest energy conformer by a constant term (∆Elig). The
bonding and nonbonding terms can then be expressed as
follows:
lig
nb
lig,cpxgeom
nb
lig
nb
lig,cpxgeom
nb
lig,cpxgeom
nb
lig
nb
E
) E
) E
+ (E - E
) ) E
) ) E
+ ∆E
+ ∆E
(3)
lig
bd
lig,cpxgeom
bd
lig
lig,cpxgeom
bd
lig,cpxgeom
bd
lig
bd
E
+ (E - E
bd
(4)
From equations 2-4, the binding energy is:
cpx
nb
lig,cpxgeom
nb
prot
nb
lig,cpxgeom
bd
Ebind ) [E - (E
+ E )] -
lig
cpx
prot
lig
∆E + [E - (E
+ E )] - ∆E (5)
bd bd
nb
bd
Ap p r oxim a tin g th e F lexibility of th e Recep tor . The
starting points for the ‘flexible receptor’ experiments were
those orientations and conformations that were already docked
into the rigid ERR binding site from above. Previously a partial
receptor flexibility protocol has been described.36 In this
procedure, those residues that contained atoms within 10 Å
of the ligand were optimized first, while the ligand itself and
the rest of the protein were held fixed. Next, the geometry of
the ligand was optimized, followed by the optimization of both
the ligand and the 10-Å vicinity of the binding site. In our case,
the docking calculations in MOE apply simulated annealing,
and this process optimizes the torsion angles of the molecules
in order to achieve optimum interaction with the receptor. As
described above, the changes in bond lengths of the ligand and
the position of hydrogen atoms are also important to consider.
In view of this fact, it was thought to be unwise to consider
the receptor atoms first, since the ligand is usually much more
easily deformed than the receptor. Hence, the protocol was
changed in such a way that first the ligand position was
optimized, followed by the neighboring site and then finally
the ligand and the neighboring site together. This modified
protocol leads to substantially lower energies with the same
number of optimization steps than the reference.36 All mini-
mizations were performed using the MMFF94 force field37 with
the Born continuum solvation model. Technically this is not
the same as docking ligands into a flexible binding site, since
the receptor structure was not being optimized while the best
conformation and orientation of the ligand were sought.
However, as this procedure was undertaken for all low energy
conformations/orientations of the ligand, it is likely that the
found solutions would correspond to those obtained from a
docking process to a flexible receptor. Both 10 and 50 were
tested as the number of MM steps in the optimization of
receptor atom positions. The interaction radius of receptor
atoms considered around the ligand was 10 Å.
The first square bracket is calculated as the total interaction
energy, Eint. The second square bracket is zero as a result of
the rigid receptor approximation. From this, the following
formula can be derived:
Ebind ) E{int} + Elig,cpxgeom - (∆Elig - Elig,cpxgeom) )
E{int} + Elig,cpxgeom - Elig (6)
Here the first two terms are obtained directly from the force
field calculations. The third term, the bonded internal energy
of the unbound ligand (Elig), was calculated from a stochastic
conformational search38 using continuum solvation (100 fail-
ures in a row, using a 7-kcal/mol energy window).
When comparing binding energies in the table, it must also
be born in mind that the experimentally measurable quantity
is the free energy of binding that would also include entropic
terms. No attempts have been made here to allow for entropic
contributions, although it is likely that the studied ligands
would have similar entropic contributions. Obviously, all of
the calculated binding energies must be viewed as approxima-
tions at different levels of accuracy to the true binding energy.
The rigid receptor approximation entirely neglects induced fit
effects, while the flexible receptor approximation at the MM
) 50 steps, r ) 10 Å grossly overestimates them. It is likely
that at the time-scale of the binding event the receptor shape
cannot change extensively, and therefore it is expected that
the true binding energy should lie somewhere between the
rigid and flexible receptor values.
Com p etitive Bin d in g. The human ERR and ERâ proteins
were in vitro transcribed-translated using the rabbit reticu-
locyte lysate (Promega, Madison, WI) with pCMX-hERR and
pCMX-hERâ templates, respectively, as previously described.39
Kis were calculated using Prism (Graphpad Software, Inc.).
The cDNAs encoding the full length ERs were generous gifts
from Dr. V. Gigue`re, McGill University Health Center.
Cell Cu ltu r e, DNA Tr a n sfection , a n d Lu cifer a se As-
sa y. For transient transfections, Cos-1 and HepG2 cells
(ATCC, Manassas, VA) were seeded in 12-well plates in phenol
red-free DMEM supplemented with 10% charcoal-treated FBS,
100 µg/mL penicillin, and 100 µg/mL streptomycin. At 50-
75% confluence, cells were transfected with 1.0 µg of luciferase
reporter plasmid, 0.1 µg of receptor expression plasmid, and
Th e Ca lcu la tion of Bin d in g En er gy. As described above,
the calculated binding energy is obtained from the full force
field energy of the docked molecule and the receptor, including
solvation contribution. Only residues that have atoms within
10-Å of the ligand were taken into account in the energy
calculations. The energy balance for the binding process is
calculated assuming the following single-step process:
ligand + substrate f complex

Nonsteroidal Phenanthrene Ligands for the Estrogen Receptor
J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 8 1415
0.5 µg of pCMX-â-galactosidase expression plasmid using the
Polyfect reagent as described by the manufacturer (Qiagen
Inc., Mississauga, ON). Cell treatment, lysis, and assays for
luciferase and â-galactosidase were as described previously.39
Values were expressed as arbitrary light units normalized to
the â-galactosidase activity of each sample.
and maintained at 17 °C. Zn(CN)2 (37.1 g, 0.316 mol) was
added, and HCl gas was bubbled through the mixture with
stirring. The rate of HCl gas addition was adjusted to allow
for HCl absorption. After 1 h of HCl gas addition, the rate of
absorption significantly decreased and AlCl3 (37.2 g, 0.279 mol)
was added. A slow rate of HCl gas flow was maintained. The
temperature was increased to 55 °C and the reaction main-
tained at this temperature for 3 h. The reaction mixture was
then poured onto a mixture of ice (800 mL) and concentrated
HCl (800 mL). The content of the reaction vessel was then
rinsed twice with CHCl3 (200 mL) and added to the aqueous
layer. The resulting biphasic layer was stirred at 60-65 °C
overnight. The organic layer was separated and the aqueous
layer washed with 200 mL and 150 mL of the organic layer.
The combined extract was washed with deionized water (3 ×
200 mL), and the organic solvents were removed by evapora-
tion. The concentrate was transferred to a distillation flask
equipped with a Vigreux column and distilled at 110 °C (0.7
mmHg) to give a distillate (29.2 g) containing residual (CHCl2)2
and a 2:1 mixture of the desired 4-methoxy-2,6-dimethylbenz-
aldehyde and 6-methoxy-2,4-dimethylbenzaldehyde isomer.
The distillate was added to methyl tert-butyl ether and
crystallized overnight at 4 °C. The crystals were filtered and
washed with 5 mL of a mixture of ethyl acetate:hexane (1:12)
to give 7.84 g of 5b. The mother liquor was concentrated and
chromatographed on silica gel with ethyl acetate/hexane
(1:12) to give an additional 10.74 g (18.58 g total, 61%). 1H
NMR (400 MHz, CDCl3) δH 2.62 (6H, s, CH3), 6.59 (2H, s, ArH),
10.48 (1H, s, CHO).
Br ea st Ca n cer Cell Gr ow th In h ibition . MCF-7 cells
(ATCC, Manassas, VA) were seeded in 48-well plates in
Iscove’s phenol red-free medium containing 10% FBS and 10
µg/mL insulin. The next day, the cells were starved for 72 h
in medium containing 10% dextran-coated charcoal treated
serum in the absence of insulin. The cells were treated with
10 nM E2 in the absence or presence of increasing concentra-
tions of 16. Cell number was determined after 96 h using the
MTS assay as described by the manufacturer (Promega,
Madison, WI).
Evolu tion a r y Molecu la r Design (EMD). Briefly, EMD
utilizes structural information and biological activity of known
pharmacologically active compounds as input structures to
generate new structures that share and/or improve the biologi-
cal activity of the original compounds. This process takes place
in two stages.
The first stage of the design process begins with the
construction of virtual receptors (VRs) that identify the
structural features of ligands, which are required for binding
to specific biological receptors. These VRs act as computational
mimics of their biological counterparts and are mathematical
objects and not molecular models of the biological binding site.
As VRs are constructed, features that are essential for high
binding affinity are identified including location and type of
functional groups, requirements for appropriate solvation
effects, and space filling properties. VRs are created through
an iterative process using structural and biological data from
known active and inactive compounds. These compounds
comprise a training set that is used to optimize the ability of
the VR to mimic the binding properties of a corresponding
biological receptor. Each VR is represented by a unique string
of characters that encodes a complex three-dimensional sur-
face. Accurate estimates of binding affinities between the VR
surface and the members of the training set are rapidly
calculated using pairwise contributions to molecular interac-
tions including solvation effects, electrostatic interactions, and
induced fitting of flexible ligands, as well as entropic effects.
The calculated binding affinity is compared to the experimen-
tal values for the members in the training set to obtain an
initial threshold fitness score. The VRs that meet this initial
criterion are then further mutated and recombined to give
optimized VRs that must be able to reproduce the experimental
binding free energies with a predetermined quality.
3-(4-Met h oxy-2-m et h ylp h en yl)-1-(4-m et h oxyp h en yl)-
p r op en on e (6a ). 5a (41.8 g, 279 mmol) was reacted as
1
described for 6b to give 51 g (65%) of 6a . H NMR (400 MHz,
CDCl3) δH 2.48 (3H, s, CH3), 3.85 (3H, s, OCH3), 3.90 (3H, s,
OCH3), 6.76-6.82 (2H, s, ArH), 6.99 (2H, d, ArH), 7.40, (1H,
d, COCH), 7.70 (1H, d, ArH), 8.08 (1H, d, ArCH),8.50 (2H, d,
ArH).
3-(4-Meth oxy-2,6-dim eth ylph en yl)-1-(4-m eth oxyph en yl)-
p r op en on e (6b). Compound 5b (18.58 g, 0.113 mol) and
4-methoxyacetophenone (17.42 g, 0.116 mol) were added to
anhydrous ethanol (110 mL) and NaOH (2.5 g) and stirred
overnight at room temperature. The precipitate that formed
was filtered, washed with water (3 × 50 mL), and dried
overnight under high vacuum to give 29.75 g of 6b. The filtrate
was stirred overnight to give an additional 1.1 g (30.85 g, 92%).
1H NMR (400 MHz, CDCl3) δH 2.44 (6H, s, CH3), 3.83 (3H, s,
OCH3), 3.90 (3H, s, OCH3), 6.66 (2H, s, ArH), 6.99 (2H, d, ArH),
7.16, (1H, d, ArCH), 7.96 (1H, d, COCH), 8.01 (2H, d, ArH).
3-(4-Met h oxy-2-m et h ylp h en yl)-1-(4-m et h oxyp h en yl)-
p r op yn on e (7a ). (i) 2,3-Dibr om o-3-(4-m eth oxy-2-m eth -
ylp h en yl)-1-(4-m eth oxyp h en yl)p r op a n -1-on e. 6a (51.0 g,
182 mmol) was reacted as described for 7b, step i, to give 2,3-
dibromo-3-(4-methoxy-2-methylphenyl)-1-(4-methoxyphenyl)-
1
In the second stage of the EMD process, information
obtained from the VRs is used in a de novo design process
(called a molecular assembler) to generate new structures.
These new structures must fit into the VR and hence satisfy
the identified requirements for biological activity as deter-
mined from the training set. Using an iterative procedure, the
strings corresponding to the proposed ligands are subjected
to mutations and crossovers, till the designs satisfy the
predetermined fitness score. These designed molecules are
subjected to retrosynthetic analysis to quantify their synthetic
feasibility and their physicochemical properties (e.g., logP, pKa)
are evaluated using computational approaches prior to selec-
tion for synthesis. Further details on the method have been
given elsewhere.20
propan-1-one that was used without further purification. H
NMR (400 MHz, CDCl3) δH 2.48 (3H, s, CH3), 3.84 (3H, s,
OCH3), 3.92 (3H, s, OCH3), 5.90 (1H, d, ArCHBr), 6.00 (1H, d,
COCHBr), 6.74 (1H, d, ArH), 6.88 (1H, dd, ArH), 7.03 (2H, d,
ArH), 7.54 (1H, d, ArH), 8.07 (2H, d, ArH).
(ii) Acetic Acid , 2-Br om o-1-(4-m eth oxy-2-m eth ylp h e-
n yl)-3-(4-m eth oxyp h en yl)-3-oxop r op yl Ester . Crude 2,3-
dibromo-3-(4-methoxy-2-methylphenyl)-1-(4-methoxyphenyl)-
propan-1-one was reacted as described for 7b, step ii, to give
79 g of acetic acid, 2-bromo-1-(4-methoxy-2-methylphenyl-)-3-
(4-methoxyphenyl)-3-oxopropyl ester that was used without
1
further purification. H NMR (400 MHz, CDCl3) δH 1.88 (3H,
s, COCH3), 2.60 (3H, s, CH3), 3.81 (3H, s, OCH3), 3.91 (3H, s,
OCH3), 5.36 (1H, d, COCHBr), 6.60 (1H, d, COCHAr), 6.36
(1H, d, ArH), 6.81 (1H, dd, ArH), 7.00 (2H, d, ArH), 7.35 (1H,
d, ArH), 8.05 (2H, d, ArH).
Syn th esis. 4-Meth oxy-2-m eth ylben zaldeh yde (5a). 3-Me-
thylanisole (36.6 g, 300 mmol) was reacted as described for
(iii) 3-(4-Meth oxy-2-m eth ylp h en yl)-1-(4-m eth oxyp h e-
n yl)p r op yn on e (7a ). Crude acetic acid, 2-bromo-1-(4-meth-
oxy-2-methylphenyl)-3-(4-methoxyphenyl)-3-oxopropyl ester was
reacted as described for 7b, step iii, to give 13 g (26%) of 7a .
1H NMR (400 MHz, CDCl3) δH 2.59 (3H, s, CH3), 3.85 (3H, s,
OCH3), 3.91 (3H, s, OCH3), 6.82-8.21 (7H, ArH).
1
5b to give 41.8 g (93%) of 5a . H NMR (400 MHz, CDCl3) δH
2.65 (3H, s, CH3), 3.87 (3H, s, OCH3), 6.74 (1H, d, ArH), 6.85
(1H, dd, ArH), 7.76 (1H, d, ArH), 10.11 (1H, s, CHO).
4-Meth oxy-2,6-d im eth ylben za ld eh yd e (5b). 3,5-Dime-
thylanisole (25.3 g, 0.186 mol) was added to (CHCl2)2 (180 mL)

1416 J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 8
Schmidt et al.
3-(4-Meth oxy-2,6-dim eth ylph en yl)-1-(4-m eth oxyph en yl)-
p r op yn on e (7b). (i) 2,3-Dibr om o-3-(4-m eth oxy-2,6-d im e-
th ylp h en yl)-1-(4-m eth oxyp h en yl)p r op a n -1-on e. To a solu-
tion of 6b (30.9 g, 104 mmol), dissolved in CH2Cl2 (200 mL)
and cooled in an ice bath, was added a solution of Br2 (16.7 g,
5.37 mL, 104 mmol) in CH2Cl2 (100 mL) over 105 min with
additional stirring for 2 h. A second portion of Br2 (1.2 mL) in
CH2Cl2 (50 mL) was added over 30 min and the reaction left
at room-temperature overnight, followed by the evaporation
of the solvent to give 2,3-dibromo-3-(4-methoxy-2,6-dimeth-
ylphenyl)-1-(4-methoxyphenyl)propan-1-one that was used
without further purification. 1H NMR (400 MHz, CDCl3) δH
2.51 (3H, s, CH3), 2.71 (3H, s, CH3), 3.81 (3H, s, OCH3), 3.92
(3H, s, OCH3), 6.16 (1H, d, COCHBr), 6.24 (1H, d, ArH), 6.34
(1H, d, ArCHBr), 6.66 (1H, d, ArH), 7.03 (2H, d, ArH), 8.07
(2H, d, ArH).
(ii) Acet ic Acid , 2-Br om o-1-(4-m et h oxy-2,6-d im et h -
ylph en yl-)-3-(4-m eth oxyph en yl)-3-oxopr opyl Ester . Crude
2,3-dibromo-3-(4-methoxy-2,6-dimethylphenyl)-1-(4-methoxy-
phenyl)propan-1-one from step i was added to acetic acid (550
mL) and KOAc (12.5 g, 130 mmol) and stirred for 6 h.
Additional KOAc (3.0 g) was added and the mixture stirred
overnight. The HOAc was evaporated and the residue dissolved
in water (300 mL) and extracted with CHCl3 (300 mL), and
the CHCl3 extract was washed with water (3 × 150 mL) and
concentrated to give acetic acid, 2-bromo-1-(4-methoxy-2,6-
dimethylphenyl)-3-(4-methoxyphenyl)-3-oxopropyl ester that
was used without further purification. 1H NMR (400 MHz,
CDCl3) δH 1.88 (3H, s, COCH3), 2.58 (3H, s, CH3), 2.65 (3H, s,
CH3), 3.80 (3H, s, OCH3), 3.91 (3H, s, OCH3), 5.76 (1H, d,
COCHBr), 6.60 (2H, s, ArH), 6.87 (1H, d, ArCHO), 6.66 (1H,
d, ArH), 7.02 (2H, d, ArH), 8.04 (2H, d, ArH).
(iii) 7-Meth oxy-10-(4-m eth oxyp h en yl)p h en a n th r en e-2-
ca r bon itr ile. Crude 5-methoxy-2-[2′-(4-methoxybenzoyl)-4′-
cyanophenyl]benzyltriphenylphosphonium bromide was re-
acted as described for 9b, step iii, to give 2.8 g 64% of 9a . H
NMR (400 MHz, CDCl3) δH 3.94 (3H, s, OCH3), 3.99 (3H, s,
OCH3), 7.08-8.73 (11H, ArH).
1
7-Me t h oxy-10-(4-m e t h oxyp h e n yl)-5-m e t h ylp h e n a n -
th r en e-2-ca r bon itr ile (9b). (i) 6′-Meth yl-2′-br om om eth yl-
2-(4-m eth oxyben zoyl)bip h en yl-4-ca r bon itr ile. To a re-
fluxing solution of 8b (45.5 g, 122 mmol) and AIBN (1.0 g) in
CCl4 (2.5 L) was added a mixture of NBS (23.0 g, 129 mmol)
and AIBN (1.0 g) in 10 equal portions. The reaction was
refluxed for 2 h after the addition of the last portion of NBS/
AIBN. The mixture was concentrated to approximately 1 L
and washed three times with water (1 L). The organic layer
was dried with anhydrous Na2SO4 and filtered and the solvent
removed. Toluene (200 mL) was added to the residue and then
evaporated. Repeating this procedure gave 48 g of 6′-methyl-
2′-bromomethyl-2-(4-methoxybenzoyl)biphenyl-4-carbonitrile
that was used without further purification.
(ii) 5-Meth oxy-2-[2′-(4-m eth oxyben zoyl)-4′-cya n op h e-
n yl]-(3-m eth ylben zyl)tr ip h en ylp h osp h on iu m Br om id e.
The crude 6′-methyl-2′-bromomethyl-2-(4-methoxybenzoyl)-
biphenyl-4-carbonitrile from step i was dissolved in DMF (200
mL), triphenyphosphine (48 g) added, and the reaction stirred
at 100 °C for 5 h. The hot reaction was slowly poured into
vigorously stirring methyl tert-butyl ether (4 L) and then
stirred for an additional 30 min and filtered. The crude
5-methoxy-2-[2′-(4-methoxybenzoyl)-4′-cyanophenyl]-(3-meth-
ylbenzyl)triphenylphosphonium bromide was washed twice
with methyl tert-butyl ether (200 mL) and immediately dis-
solved in CH2Cl2 (1 L).
(iii) 3-(4-Meth oxy-2,6-d im eth ylp h en yl)-1-(4-m eth oxy-
p h en yl)p r op yn on e (7b). DBU 36.2 g, 238 mmol) and crude
acetic acid, 2-bromo-1-(4-methoxy-2,6-dimethylphenyl)-3-(4-
methoxyphenyl)-3-oxopropyl ester from step ii were added to
THF (350 mL) and heated to 55 °C overnight. The reaction
mixture was filtered and the precipitate washed with THF (2
× 100 mL). The filtrate was evaporated and the residue
dissolved in CHCl3 (300 mL) and washed with water (2 × 150
mL), 8% HCl (pH 2, 150 mL), and water (2 × 150 mL). It was
noted that the final separation was made easier if an aqueous
solution of NaHCO3 (15 mL) was added to the last extraction.
After evaporation of the solvent and drying overnight under
high vacuum, the crude material was crystallized from anhy-
drous toluene (30 mL) by cooling to room temperature and then
to 4 °C to give 22.4 g, (73%) of 7b. 1H NMR (400 MHz, CDCl3)
δH 2.55 (6H, s, CH3), 3.82 (3H, s, OCH3), 3.89 (3H, s, OCH3),
6.65-8.22 (6H, ArH).
(iii) 7-Meth oxy-10-(4-m eth oxyph en yl)-5-m eth ylph en an -
th r en e-2-ca r bon itr ile (9b). The crude 5-methoxy-2-[2′-(4-
methoxybenzoyl)-4′-cyanophenyl]-(3-methylbenzyl)triphe-
nylphosphonium bromide from step ii was added over 5 h to a
vigorously stirred mixture of CH2Cl2 (2 L) and 50% aqueous
NaOH (1 L) at 35 °C. Stirring was continued for an additional
3 h at 35-40 °C. The organic layer was separated and washed
consecutively with 1 N HCl (200 mL), water (200 mL), and
saturated NaHCO3 (200 mL). After drying the organic layer
over anhydrous Na2SO4 and removing the solvent, the residue
was crystallized from CH2Cl2/hexanes/ethanol (80/60/10, 150
mL) to give 24 g of 9b. The filtrated was concentrated and
chromatographed on silica using CH2Cl2/hexanes (50/50) to
give an additional 5.5 g (total 29.5 g, 68.5%). 1H NMR (400
MHz, CDCl3) δH 3.12 (3H, s, CH3), 3.94 (3H, s, OCH3), 3.97
(3H, s, OCH3), 7.07-8.94 (10H, ArH).
1-[7-Meth oxy-10-(4-m eth oxyp h en yl)p h en a n th r en -2-yl]-
eth a n on e (10a ). 9a (2.8 g, 8.3 mmol) and methylmagnesium
bromide (3 M) in diethyl ether (6.1 mL, 18 mmol) were reacted
4′-Meth oxy-2′-m eth yl-2-(4-m eth oxyben zoyl)bip h en yl-
4-ca r bon itr ile (8a ). 7a (5.61 g, 20.0 mmol) was reacted as
described for 8b to give 8a in quantitative yield. 1H NMR (400
MHz, CDCl3) δH 2.14 (3H, s, CH3), 3.74 (3H, s, OCH3), 3.84
(3H, s, OCH3), 6.56-7.79 (10H, ArH).
1
as described for 10b to give 10a . H NMR (400 MHz, CDCl3)
δH 2.61 (3H, s, COCH3) 3.94 (3H, s, OCH3), 3.99 (3H, s, OCH3),
7.07-8.73 (11H, ArH).
4′-Meth oxy-2′,6′-d im eth yl-2-(4-m eth oxyben zoyl)bip h e-
n yl-4-ca r bon itr ile (8b). 7b (4.20 g, 15.7 mmol) and 4-cyan-
opyrone22 (1.90 g, 15.7 mmol) were heated at 190 °C for 17 h.
The crude reaction was chromatographed on silica gel with
ethyl acetate/hexane (1:6) then ethyl acetate/hexane (1:2) to
give 4.4 g (75%) of 8b. 1H NMR (400 MHz, CDCl3) δH 1.96 (6H,
s, CH3), 3.75 (3H, s, OCH3), 3.88 (3H, s, OCH3), 6.54-7.83 (9H,
ArH).
7-Meth oxy-10-(4-m eth oxyp h en yl)p h en a n th r en e-2-ca r -
bon itr ile (9a ). (i) 2′-Br om om eth yl-2-(4-m eth oxyben zoyl)-
bip h en yl-4-ca r bon itr ile. 8a (4.6 g, 13 mmol) was reacted
as described for 9b, step i, to give 2′-bromomethyl-2-(4-
methoxybenzoyl)biphenyl-4-carbonitrile that was used without
further purification.
(ii) 5-Meth oxy-2-[2′-(4-m eth oxyben zoyl)-4′-cya n op h e-
n yl]ben zyltr ip h en ylp h osp h on iu m Br om id e. Crude 2′-
bromomethyl-2-(4-methoxybenzoyl)biphenyl-4-carbonitrile was
reacted as described for 9b, step ii, to give 5-methoxy-2-[2′-
(4-methoxybenzoyl)-4′-cyanophenyl]benzyltriphenylphospho-
nium bromide that was used without further purification.
1-[7-Meth oxy-10-(4-m eth oxyp h en yl)-5-m eth ylp h en a n -
th r en -2-yl]eth a n on e (10b). Following the procedure outlined
for 10c, compound 9b (1.22 g, 3.46 mmol) and methylmagne-
sium bromide (3 M) in diethyl ether (2.5 mL) were reacted to
give 1.42 g of 10b in quantitative yield. H NMR (400 MHz,
CDCl3) δH 2.61 (3H, s, CH3), 3.15 (3H, s, COCH3), 3.94 (3H, s,
OCH3), 3.97 (3H, s, OCH3), 7.08-8.94 (10H, ArH).
1-[7-Meth oxy-10-(4-m eth oxyp h en yl)-5-m eth ylp h en a n -
th r en -2-yl]-2-m eth ylp r op a n -1-on e (10c). 9b (10.0 g, 28.3
mmol) was added to anhydrous THF (200 mL), and isopropy-
lmagnesium chloride (2.0 M) in THF (15.6 mL) was added with
stirring under Ar. After 10 min, crystalline CuBr (72 mg) was
added, and the reaction stirred for 9 h. TLC analysis indicated
that there was still starting material present, and an ad-
ditional 8 mL of the Grignard reagent was added and the
reaction stirred overnight. To the reaction was added 10% H2-
SO4 (150 mL), and the mixture was placed on a rotary
evaporator to remove the THF. The remaining aqueous layer
was stirred overnight at room temperature then extracted with
CHCl3 (3 × 250 mL), washed with water (3 × 150 mL). After
1

Nonsteroidal Phenanthrene Ligands for the Estrogen Receptor
J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 8 1417
removing the CHCl3, the residue was dried under high vacuum
to give 11.2 g (quantitative yield) of 10c. H NMR (400 MHz,
CDCl3) δH 1.22 (6H, d, CH3), 3.15 (3H, s, CH3), 3.54 (1H, septet,
COCH), 3.94 (3H, s, OCH3), 3.96 (3H, s, OCH3), 7.09-8.93
(10H, ArH).
(+)-7-(1-H yd r oxy-1-m et h ylet h yl)-4-m et h yl-9-[4-(2-(p i-
p er id in -1-yl)eth oxy)p h en yl]p h en a n th r en -2-ol (13b). 12b
(0.65 g, 1.4 mmol) in anhydrous THF (150 mL) was cooled to
0 °C. Methylmagnesium bromide (3 M) in diethyl ether (4.8
mL, 14 mmol) was added. The reaction was stirred overnight
at 0 °C then cold 1 N HCl (35 mL) was added. Ethyl acetate
(100 mL) was added and the aqueous phase saturated with
Na2CO3. The organic layer was separated and dried and the
solvent removed to yield a yellow, glassy residue (0.83 g) of
13b that was used without further purification.
1
1-[7-Hyd r oxy-10-(4-h yd r oxyp h en yl)p h en a n th r en -2-yl]-
eth a n on e (11a ). 10a was reacted as described for 11b to give
1
11a . H NMR (400 MHz, DMSO-d6) δH 2.59 (3H, s, COCH3),
6.95-8.87 (11H, ArH), 9.7 (1H, br s, OH), 10.1 (1H, br s, OH).
1-[7-Hyd r oxy-10-(4-h yd r oxyp h en yl)-5-m eth ylp h en a n -
th r en -2-yl]eth a n on e (11b). BBr3 (100 mL, 1 M in CH2Cl2)
was added to 10b (6.76 g, 18.3 mmol) in CH2Cl2 (200 mL) at
-70 to - 60 °C. The reaction was stirred overnight at -60 to
3 °C and then poured into water/ice (500 mL) and extracted
with CHCl3/acetone (2.5 L, 5/1). After the solvent was evapo-
rated, the residue was chromatographed on silica gel using a
gradient of ethyl acetate:methanol (95:5) to ethyl acetate:
(+)-7-(1-Hyd r oxy-1,2-d im eth ylp r op yl)-4-m eth yl-9-[4-(2-
(p ip er id in -1-yl)eth oxy)p h en yl]p h en a n th r en -2-ol (13c).
12c (0.570 g, 1.2 mmol) and methylmagnesium iodide (3 N)
in THF (3.0 mL) was added and the mixture stirred overnight
at room temperature. After the THF was removed, CHCl3 (50
mL) was added to the residue along with H2SO4 (0.675 g)
dissolved in water (20 mL, pH 1). The residue completely
dissolved. Excess NaHCO3 was added and the organic layer
separated, evaporated, and dried under high vacuum to give
1
methanol (80:20) to give 11b. H NMR (400 MHz, DMSO-d6)
δH 2.16 (3H, s, CH3), 2.67 (3H, s, COCH3), 6.52-8.48 (10H,
ArH), 8.81 (1H, br s, OH), 9.91 (1H, br s, OH).
1
13c in quantitative yield. H NMR (400 MHz, CDCl3) δH 0.77
(3H, d, CH3), 0.83 (3H, d, CH3), 1.47 (3H, s, CH3), 1.53 (2H,
m, NCH2CH2CH2), 1.72 (4H, m, NCH2CH2CH2), 2.01 (1H,
septet, CH(CH3)2), 2.67 (4H, br s, NCH2CH2CH2), 2.90 (2H, t,
OCH2CH2N), 3.09 (3H, s, CH3), 4.21 (2H, t, OCH2CH2N), 6.88-
8.80 (10H, ArH).
1-[7-Hyd r oxy-10-(4-h yd r oxyp h en yl)-5-m eth ylp h en a n -
th r en -2-yl]-2-m eth ylp r op a n -1-on e (11c). 10c (5.53 g, 13.9
mmol), 48% HBr (44 mL), and acetic acid (55 mL) were heated
(oil bath at 126 °C) in a closed thick-walled pressure flask for
3.5 h. The reaction was then poured into water (300 mL),
extracted with ethyl acetate (3 × 250 mL), and washed with
water (3 × 150 mL). After the solvent was removed, the residue
7-Isop r op en yl-4-m eth yl-9-[4-(2-(p ip er id in -1-yl)eth oxy)-
p h en yl]p h en a n th r en -2-ol (14b). Crude 13b was dissolved
in CH2Cl2 (20 mL), 10-camphorsulfonic acid (1 g) was added,
and the mixture was refluxed for 1 h. After the solvent was
removed, the residue was dissolved in ethyl acetate (100 mL),
a saturated solution of Na2CO3 (25 mL) was added, and the
mixture was stirred for 15 min. The organic layer was
separated, dried, and concentrated and the residue chromato-
graphed on silica gel using hexanes:acetone (60:40) to yield
0.412 g (65%) of 14b. 1H NMR (400 MHz, CDCl3) δH 1.52 (2H,
br s, NCH2CH2CH2), 1.75 (4H, m, NCH2CH2CH2), 2.05 (3H, s,
CH3), 2.71 (4H, br s, NCH2CH2CH2), 2.92 (2H, t, OCH2CH2N),
3.10 (3H, s, CH3), 4.24 (2H, t, OCH2CH2N), 5.05 (1H, s, dCH),
5.36 (1H, s, dCH), 6.84-8.78 (10H, ArH). Anal. C31H33NO2
(C, H, N).
1
was dried under high vacuum to yield 11c quantitatively. H
NMR (400 MHz, DMSO-d6) δH 1.10 (6H, d, CH3), 3.03 (3H, s,
CH3), 3.56 (1H, septet, COCH), 6.96-8.87 (10H, ArH), 9.65
(1H, br s, OH), 10.1 (1H, br s, OH).
1-{7-Hyd r oxy-10-[4-(2-(p ip er id in -1-yl)eth oxy)p h en yl]-
p h en a n th r en -7-yl}eth a n on e (12a ). 11a (1.35 g, 4.12 mmol)
was reacted as described for 12b to give 0.175 g (9.7%) of 12a .
1H NMR (400 MHz, DMSO-d6) δH 1.40 (2H, br s, NCH2-
CH2CH2), 1.53 (4H, m, NCH2CH2CH2), 2.49 (4H, br s, NCH2-
CH2CH2), 2.57 (3H, s, CH3), 2.70 (2H, t, OCH2CH2N), 4.16 (2H,
t, OCH2CH2N), 7.11-8.86 (11H, ArH), 10.12 (1H, br s, OH).
1-{7-Hydr oxy-5-m eth yl-10-[4-(2-(piper idin -1-yl)eth oxy)-
p h en yl]p h en a n th r en -2-yl}eth a n on e (12b). 11b (0.135 g,
0.394 mmol) was dissolved in anhydrous DMF (5 mL). NaH
(75 mg, 60% in oil) was added and the reaction stirred for 10
min. at room temperature. A solution of 1-(2-chloroethyl)-
piperidine hydrochloride (75 mg, 0.40 mmol) in DMF (4.5 mL)
was added slowly over 2 h. After the addition was complete,
the reaction was stirred for 2 h at 40-50 °C then overnight at
room temperature. The reaction mixture was added to water
(50 mL), extracted with ethyl acetate (75 mL), dried, concen-
trated, and chromatographed on silica using ethyl acetate:
methanol (9:1) to give 12b. 1H NMR (400 MHz, CDCl3) δH 1.53
(2H, br s, NCH2CH2CH2), 1.76 (4H, m, NCH2CH2CH2), 2.51
(3H, s, COCH3), 2.72 (4H, br s, NCH2CH2CH2), 2.93 (2H, t,
OCH2CH2N), 3.10 (3H, s, CH3), 4.24 (2H, t, OCH2CH2N), 6.83-
8.87 (10H, ArH). Anal. C30H31NO3 (C, H, N).
7-(1,2-Dim eth ylp r op en yl)-4-m eth yl-9-[4-(2-(p ip er id in -
1-yl)eth oxy)p h en yl]p h en a n th r en -2-ol (14c). 13c (0.587 g,
1.2 mmol) was dissolved in glacial acetic acid (7.0 mL), and
concentrated H2SO4 (1 drop) was added. The mixture was
heated for 20 min at 75 °C with stirring. The reaction was
neutralized with NaHCO3 (10.5 g in 25 mL of water) and
extracted with methyl tert-butyl ether (3 × 25 mL), evaporated,
and dried under high vacuum. The residue was chromato-
graphed on silica gel with acetone/hexane (1:2) to give 0.407 g
(85%) of 14c. 1H NMR (400 MHz, CDCl3) δH 0.77 (3H, d, CH3),
0.83 (3H, d, CH3), 1.47 (3H, s, CH3), 1.55 (2H, m, NCH2-
CH2CH2), 1.76 (4H, m, NCH2CH2CH2), 2.72 (4H, br s, NCH2-
CH2CH2), 2.95 (2H, t, OCH2CH2N), 3.11 (3H, s, CH3), 4.24 (2H,
t, OCH2CH2N), 6.87-8.78 (10H, ArH). Anal. C33H37NO2 (C, H,
N).
(+)-7-(1-H yd r oxyet h yl)-4-m et h yl-9-[4-(2-(p ip er id in -1-
yl)eth oxy)p h en yl]p h en a n th r en -2-ol (15). 12b (0.550 g, 1.2
mmol) was dissolved in methanol (45 mL). NaBH4 (0.175 g, 4
equiv) was added in portions over 1 h at -5 to -3 °C. The
reaction was brought to room temperature and stirred for 2
h. The solvent was removed, and water and ethyl acetate were
added. The mixture was acidified with concentrated HCl to
pH 2, and then excess Na2CO3 was added. The organic layer
was separated and the aqueous layer extracted with ethyl
acetate. The combined organic layers were washed with brine,
dried, and concentrated. The residue was chromatographed
on silica with hexanes:acetone (1:1) to yield 0.447 g, 81%) of
1-{7-Hydr oxy-5-m eth yl-10-[4-(2-(piper idin -1-yl)eth oxy)-
p h en yl]p h en a n th r en -7-yl}-2-m eth ylp r op a n -1-on e (12c).
11c (1.1 g, 3.0 mmol), Aliquat (3.03 g, 7.5 mmol), NaOH (1.2
g, 30 mmol), and 1-(2-chloroethyl)piperidinium hydrochloride
(0.552 g, 3.0 mmol) were added to TMU (70 mL) and heated
at 50 °C for 11 h. The reaction was added to water (500 mL),
acidified with concentrated HCl (3 mL), and then neutralized
with an excess of NaHCO3 until pH 8. The reaction was
extracted with ethyl acetate (3 × 250 mL), and the combined
extracts were washed with water (2 × 150 mL) and then
concentrated. The residue was chromatographed on silica gel
with acetone/hexane (1:2) to give 0.383 g (26.5%) of 12c. 1H
NMR (400 MHz, CDCl3) δH 1.15 (6H, d, CH3), 1.53 (2H, br s,
NCH2CH2CH2), 1.75 (4H, m, NCH2CH2CH2), 2.71 (4H, br s,
NCH2CH2CH2), 2.92 (2H, t, OCH2CH2N), 3.07 (3H, s, CH3),
3.44 (1H, septet, COCH), 4.23 (2H, t, OCH2CH2N), 6.83-8.88
(10H, ArH). Anal. C32H35NO3 (C, H, N).
1
15. H NMR (400 MHz, DMSO-d6) δH 1.34 (3H, d, CH3), 1.40
(2H, m, NCH2CH2CH2), 1.52 (4H, m, NCH2CH2CH2), 2.47 (3H,
s, CH3), 2.47 (4H, br s, NCH2CH2CH2), 2.71 (2H, t, OCH2CH2N),
3.01 (3H, s, CH3), 4.16 (2H, t, OCH2CH2N), 4.79 (1H, dq, CH),
5.19 (1H, d, OH), 7.05-8.76 (10H, ArH), 9.75 (1H, s, OH).
(+)-7-E t h yl-4-m et h yl-9-[4-(2-(p ip er id in -1-yl)et h oxy)-
p h en yl]p h en a n th r en -2-ol (16). 12b (0.091 g, 0.2 mmol) was

1418 J ournal of Medicinal Chemistry, 2003, Vol. 46, No. 8
Schmidt et al.
(19) Labrie, F.; Labrie, C.; Belanger, A.; Simard, J .; Gauthier, S.;
Luu-The, V.; Merand, Y.; Giguere, V.; Candas, B.; Luo, S.;
Martel, C.; Singh, S. M.; Fournier, M.; Coquet, A.; Richard, V.;
Charbonneau, R.; Charpenet, G.; Tremblay, A.; Tremblay, G.;
Cusan, L.; Veilleux, R. EM-652 (SCH 57068), a Third Generation
SERM Acting as Pure Antiestrogen in the Mammary Gland and
Endometrium. J . Steroid Biochem. Mol. Biol. 1999, 69, 51-84.
(20) Schmidt, J . M., University of Guelph, US Patent 5699268, 1995.
(21) Wiese, T. E.; Polin, L. A.; Palomino, E. and Brooks, S. K.
Induction of the Estrogen Specific Response of MCF-7 Cells by
Selected Analogues of Estradiol-17â: A 3D QSAR Study. J . Med.
Chem. 1997, 40, 3659-3669.
dissolved in dry THF (5 mL). Liquid NH3 (6 mL) was added
via a cannula at -78 °C. Li (4 mg) was then added and the
reaction stirred for 30 min. Additional Li (10 mg) was added,
and the reaction stirred for an additional 1 h. The reaction
was quenched with solid NH4Cl (3.3 g). The NH3 was allowed
to boil off and the residue partitioned between CH2Cl2 and
water. The water layer was separated and extracted with CH2-
Cl2, and the combined organic layers were dried, concentrated,
and chromatographed on silica using a gradient of methanol:
CH2Cl2 (5:95 to 30:70) to give 0.035 g (40%) of 16. 1H NMR
(400 MHz, CDCl3) δH 1.16 (3H, t, CH3), 1.47 (2H, m, NCH2-
CH2CH2), 1.67 (4H, m, NCH2CH2CH2), 2.54 (2H, q, CH2), 2.55
(3H, s, CH3), 2.56 (4H, br s, NCH2CH2CH2), 2.82 (2H, t,
OCH2CH2N), 2.87 (1H, dd, ArCH), 3.00 (1H, dd, ArCH) 3.92
(1H, dd, ArCHAr), 4.10 (2H, m, OCH2CH2N), 6.50-7.56 (9H,
ArH). Anal. C30H35NO2 (C, H, N).
(22) Kvita, V.; Sauter, H. The Synthesis of Some 2H-Pyran-2-one
Derivatives. Helv. Chim. Acta 1990, 73, 883.
(23) Berman, H. M.; Westbrook, J .; Feng, Z.; Gilliland, G.; Bhat, T.
N.; Weissig, H.; Shindyalov, I. N.; Bourne, P. E. The Protein Data
Bank. Nucl. Acid Res. 2000, 28, 235-242.
(24) Grese, T. A.; Cho, S.; Finley, D. R.; Godfrey, A. G.; J ones, C. D.;
Lugar, C. W., III; Martin, M. J .; Matsumoto, K.; Pennington, L.
D.; Winter, M. A.; Adrian, M. Dee; Cole, H. W.; Magee, D. E.;
Phillips, D. Lynn; Rowley, E. R.; Short, L. L.; Glasebrook, A. L.;
Bryant, H. U. Structure-Activity Relationships of Selective
Estrogen Receptor Modulators: Modifications to the 2-Arylben-
zothiophene Core of Raloxifene. J . Med. Chem. 1997, 40, 146-
167.
(25) Sata, M.; Grese, T. A.; Dodge, J . A.; Bryant, H. U.; Turner, C.
H. Emerging Therapies for the Prevention or Treatment of
Postmenopausal Osteoporosis. J . Med. Chem. 1999, 42, 1-24.
(26) Grese, T. A.; Adrian, M. Dee; Phillips, D. Lynn; Shelter, P. K.;
Short, L. L.; Glasebrook, A. L.; Bryant, H. U. Photochemical
Synthesis of N-Arylbenzophenanthridine Selective Estrogen
Modulators (SERMs). J . Med. Chem. 2001, 44, 2857-2860.
(27) Still, W. C.; Tempczyk, A.; Hawley, R. C.; Hendrickson, T.
Semianalytical Treatment of Solvation for Molecular Mechanics
and Dynanmics. J . Am. Chem. Soc. 1990, 112, 6127-6129.
(28) Qiu, D.; Shenkin, S.; Hollinger, F. P.; Still, W. C. The GB/SA
Continuum Model for Solvation. A Fast Analytical Method for
the Calculation of Approximate Born Radii. J . Phys. Chem. A
1997, 101, 3005-3014.
Refer en ces
(1) Carson-J urica, M. A.; Schrader, W. T.; O’Malley, B. W. Steroid
Receptor Family: Structure and Functions. Endocrinol. Rev.
1990, 11, 201-220.
(2) J ordan, V. C. Studies on the Estrogen Receptor in Breast Cancer
- 20 Years as a Target for the Treatment and Prevention of
Breast Cancer. Breast Cancer Res. Treat. 1995, 36, 267-285.
(3) Gradishar, W. J .; J ordan, V. C. Clinical Potential of New
Antiestrogens. J . Clin. Oncol. 1997, 15, 840-852.
(4) J ordan, V. C. Molecular Mechanisms of Antiestrogen Action in
Breast Cancer. Breast Cancer Res. Treat. 1994, 31, 41-52.
(5) Mitlak, B. H.; Cohen, F. J . In Search of Optimal Long-Term
Female Hormone Replacement: The Potential of Selective
Estrogen Receptor Modulators. Horm. Res. 1997, 48, 155-163.
(6) Smith, C. L.; O’Malley, B. W. Evolving Concepts of Selective
Estrogen Receptor Action: From Basic Science to Clinical
Applications. Trends Endocrinol. Metab. 1999, 10, 299-300.
(7) McDonnell, D. P. The Molecular Pharmacology of SERMs.
Trends Endocrinol. Metab. 1999, 10, 301-311.
(8) MacGregor, J .; J ordan, V. G. Basic Guide to the Mechanism of
Antiestrogen Action. Pharmacol. Rev. 1998, 50, 151-196.
(9) Mosselman, S.; Polman, P.; Dijkema, R. ER beta: Identification
and Characterization of a Novel Human Estrogen Receptor.
FEBS Lett. 1996, 392, 49-53.
(10) Hall, J . M.; McDonnell, D. P. The Estrogen Receptor beta-Isoform
(ERbeta) of the Human Estrogen Receptor Modulates ERalpha
Transcriptional Activity and is a Key Regulator of the Cellular
Response to Estrogens and Antiestrogens. Endocrinology 1999,
140, 5566-5578.
(11) Pennisi, E. Differing Roles Found for Estrogen’s Two Receptors.
Mol. Endocrinol. 1997, 277, 1439-1510.
(12) Mitlak, B. H.; Cohen, F. J . Selective Estrogen Receptor Modula-
tors: A Look Ahead. Drugs 1999, 57, 653-663.
(13) Grese, T. A.; Pennington, J . P.; Sluka, J . P.; Adrian, M. D.; Cole,
H. W.; Fuson, T. R.; Magee, D. E.; Phillips, D. L.; Rowley, E. R.;
Shetler, P. K.; Short, L. L.; Venugopalan, M.; Yang, N. N.; Sato,
M.; Blasebrook, A. L.; Bryant, H. U. Synthesis and Pharmacology
of Conformationally Restricted Raloxifene Analogues: Highly
Potent Selective Estrogen Receptor Modulators. J . Med. Chem.
1998, 41, 1272-1283.
(14) Black, L. J .; J ones, C. D.; Falcone, J . F. Antagonism of Estrogen
Action with a New Benzothiophene Derived Antiestrogen. Life
Sci. 1983, 28, 1031-1036.
(15) Foster, A. B.; J arman, M.; Leung, O. T.; McCague, R.; Leclercq,
G.; Devleeschouwer, N. Hydroxy Derivatives of Tamoxifen. J .
Med. Chem. 1985, 28, 1491-1497.
(29) Schaefer, M.; Karplus, M. A. Comprehensive Analytical Treat-
ment of Continuum Electrostatics. J . Phys. Chem. 1996, 100,
1578-1599.
(30) Molecular Operating Environment, Version 2001.01. Chemical
Computing Group Inc., Montreal, Quebec, Canada. (http://
www.chemcomp.com).
(31) Hart, T. N.; Read, R. J . A Multiple-Start Monte Carlo Docking
Method. PROTEINS: Struct. Func. Genet. 1992, 13, 206-222.
(32) Morris, G. M.; Goodsell, D. S.; Huey, R.; Olsen, A. J . Distributed
automated docking of flexible ligands to proteins: Parallel
applications of AutoDock 2.4. J . Comput. Aided Mol. Design.
1996, 10, 293-304.
(33) Holloway, M. K.; Wai, J . M.; Halgren, T. A.; Fitzgerald, P. M.
D.; Vacca, J . P.; Dorsey, B. D.; Levin, R. B.; Thompson, W. J .;
Chen, L. J .; deSolms, S. J .; Gaffin, N.; Ghosh, A. K.; Giuliani,
E. A.; Graham, S. L.; Guare, J . P.; Hungate, R. W.; Lyle, T. A.;
Sanders, W. M.; Tucker, T. J .; Wiggins, M.; Wiscount, C. M.;
Woltersdorf, O. W.; Young, S. D.; Darke, P. L. and Zugay, J . A.
A Priori Prediction of Activity for HIV-1 Protease Inhibitors
Employing Energy Minimization in the Active Site. J . Med.
Chem. 1995, 38, 305-317.
(34) Shen, J .; Pearlstein, R. A.; Vaz, R. J .; Kominos, D.; Effland, R.
C. The Use of Interaction Energy and Solvation Entropy to Rank
Order Docked Ligand-Protein Complexes for Lead Optimiza-
tion: Application to Thermolysin Inhibitors. Med. Chem. Res.
1999, 9, 501-512.
(35) Metropolis, N.; Rosenbluth, A. W.; Rosenbluth, M. N.; Teller, A.
H. Equation of State Calculations by Fast Computing Machines.
J . Chem. Phys. 1953, 21, 1087-1092.
(36) Fink, B. E.; Mortensen, D. S.; Stauffer, S. R.; Aron, Z. D.;
Katzenellenbogen, J . A. Research Paper-Novel Structural Tem-
plates for Estrogen-Receptor Ligands and Prospects for Combi-
natorial Synthesis of Estrogens. Chem. Biol. 1999, 6, 205-219.
(37) Halgren, T. A. The Merck Molecular Force Field. I. Basis, Form,
Scope, Parametrization, and Performance of MMFF94. J . Com-
put. Chem. 1996, 17, 490-519.
(38) Ferguson, D. M.; Raber, D. J . A New Approach to Probing
Conformational Space with Molecular Mechanics: Random
Incremental Pulse Search. J . Am. Chem. Soc. 1989, 111, 4371-
4378.
(16) Chander, S. K.; McCague, R.; Luqmani, Y.; Newton, C.; Dowsett,
M.; J arman, M.; Coombes, R. C. Pyrrolidino-4-Iodotamoxifen and
4-Iodotamoxifen, New Analogues of the Antiestrogen Tamoxifen
for the Treatment of Breast Cancer. Cancer Res. 1991, 51, 5851-
5858.
(17) Miller, C. P.; Collini, M. D.; Tran, B. D.; Harris, H. A.; Kharode,
Y. P.; Marzolf, J . T.; Moran, R. A.; Henderson, R. A.; Bender, R.
H. W.; Unwalla, R. J .; Greenberger, L. M.; Yaardley, J . P.; Abou-
Gharbia, M. A.; Lyttle, C. R.; Komm, B. S. Design, synthesis
and Preclinical Characterization of Novel, Highly Selective
Indole Estrogens. J . Med. Chem. 2001 44, 1654-1657.
(18) Gauthier, S.; Caron, B.; Cloutier, J .; Dory, Y. L.; Favre, A.;
Larouche, D.; Mailhot, J .; Ouellet, C.; Schwerdtfeger, A.; Le-
blanc, G.; Martel, C.; Simard, J .; Merand, Y.; Belanger, A.;
Labrie, C.; Labrie, F. (S)-(+)-4-[7-(2,2-dimethyl-1-oxopropoxy)-
4-methyl-2-[4-[2-(1-piperidinyl)-ethoxy]phenyl]-2H-1-benzopyran-
3-yl]-phenyl-2,2-dimethylpropanoate (EM-800): A Highly Potent,
Specific, and Orally Active Nonsteroidal Antiestrogen. J . Med.
Chem. 1997, 40, 2117-2122.
(39) Tremblay, A.; Tremblay, G. B.; Labrie, C.; Labrie, F.; and
Giguere, V. EM-800, a Novel Antiestrogen, Acts as a Pure
Antagonist of the Transcriptional Functions of Estrogen Recep-
tors R and â. Endocrinology 1998, 139, 111-118.
J M020536Q