Structure-guided synthesis of tamoxifen analogs with improved selectivity for the orphan ERRγ

Article abstract of DOI:10.1016/j.bmcl.2005.11.030

The design and synthesis of 4-hydroxytamoxifen (4-OHT) derivatives are described. The binding affinities of these compounds toward the orphan estrogen-related receptor γ and the classical estrogen receptor α demonstrate that analogs bearing hydroxyalkyl g

Full text of DOI:10.1016/j.bmcl.2005.11.030

Bioorganic & Medicinal Chemistry Letters 16 (2006) 821–824
Structure-guided synthesis of tamoxifen analogs with
improved selectivity for the orphan ERRc
a
a
b
a
a
´
Esther Y. H. Chao, Jon L. Collins, Stephanie Gaillard, Aaron B. Miller, Liping Wang,
Lisa A. Orband-Miller,a Robert T. Nolte,a Donald P. McDonnell,b
Timothy M. Willsona and William J. Zuerchera,*
aDiscovery Research, GlaxoSmithKline, Five Moore Drive, Research Triangle Park, NC 27709, USA
bDepartment of Pharmacology and Cancer Biology, Duke University, Levine Science Research Center,
Research Drive, Durham, NC 27710, USA
Received 4 October 2005; revised 5 November 2005; accepted 7 November 2005
Available online 22 November 2005
Abstract—The design and synthesis of 4-hydroxytamoxifen (4-OHT) derivatives are described. The binding affinities of these com-
pounds toward the orphan estrogen-related receptor c and the classical estrogen receptor a demonstrate that analogs bearing
hydroxyalkyl groups display improved binding selectivity profiles compared with that of 4-OHT. An X-ray crystal structure of
one of the designed compounds bound to ERRc LBD confirms the molecular basis of the selectivity.
Ó 2005 Elsevier Ltd. All rights reserved.
The orphan estrogen-related receptor exists as three
subtypes (ERRa, ERRb, and ERRc) in the human
genome.1 ERRa is expressed in metabolically active
tissues such as muscle and adipose. ERRc is expressed
in the spinal cord and CNS. ERRb appears to play a
role in development, as postnatal expression is highly
restricted, being detected only at low levels in the liv-
er, stomach, skeletal muscle, heart, and kidney. These
orphan nuclear receptors are most closely related in
sequence to the classical estrogen receptors ERa and
ERb but do not bind estradiol or other related steroi-
dal estrogens. In contrast, the ERRs show varying
levels of constitutive transcriptional activity, and no
endogenous ligand has been identified to date.2 Nota-
bly, the fasting-induced cofactor PGC-1a has been
reported as an endogenous protein ligand for the
ERRs, raising the possibility that the receptors may
not be regulated by small molecules.3 Despite these
observations, two lines of evidence demonstrate that
the ERRs have the potential to be chemically tracta-
ble. First, inverse agonists (antagonists of constitutive
ERR activity) have been identified: the potent estro-
gen receptor antagonist 4-hydroxytamoxifen (1,
4-OHT) is an inverse agonist of ERRc.4 Second, we
recently reported the acyl hydrazone GSK4716 (2) as
a submicromolar ERRc/b agonist with efficacy compa-
rable to that of PGC-1a and greater than 50-fold
binding selectivity over the classical estrogen
receptors.5 Despite these advances, the development
of an ERRc inverse agonist with selectivity against
the classical estrogen receptors remains unrealized.
NMe2
O
O
N
N
H
HO
HO
2
1
With the goal of developing a selective ERRc inverse
agonist, we focused on the 4-OHT (1) template be-
cause of its affinity for ERRc and its inverse agonist
functional profile. The key challenge was that 1 has
higher affinity for ERa than ERRc. A number of ana-
logs of 1 and their associated SAR with the classical
estrogen receptor have been reported. Combining this
knowledge with analysis of the X-ray crystal struc-
tures of 1 bound to the ligand binding domains of
ERRc6 and ERa7 provided a basis for design of
ERRc-selective ligands.
Keywords: ERRc chemical tool; Orphan nuclear receptor.
*
Corresponding author. Tel.: +1 919 483 5275; fax: +1 919 315
5668; e-mail: william.j.zuercher@gsk.com
0960-894X/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2005.11.030

822
E. Y. H. Chao et al. / Bioorg. Med. Chem. Lett. 16 (2006) 821–824
The unliganded ERRc LBD crystal structure contains a
NMe2
OH
3
O
˚
small (220 A ) ligand binding pocket (LBP), but reorga-
nizes to accommodate the binding of 1 (Fig. 1).
Although the sequence identity between ERa and ERRc
is only 36%, the residues that line the first shell of the li-
gand binding pocket have a high degree of identity be-
tween ERRc and ERa, and many common elements
critical for ligand recognition are conserved in the two
structures. Notably, the glutamate and arginine residues
in ERa (and ERb) that form the hydrogen bond interac-
tion with the phenolic A-ring of estradiol are found in
all three of the ERRs (ERRc E275 and R316). The basic
side chain of 1 makes a Coulombic interaction with an
aspartate (ERa D351; ERRc D273).
a, b, c
Ph
O
R
HO
PvO
R = (CH2)nCl
or (CH2)nCO2Me
NMe2
NMe2
O
O
d, e
or f
In contrast to the similarity between ERRc and ERa in
the first shell of the ligand binding pocket, a key differ-
ence is located proximal to the ethyl group of 4-OHT. A
pair of phenylalanine residues (F404 and F425) in ERa
corresponds to a tyrosine and asparagine (Y326 and
N346) in ERRc.8 The distances from the ethyl terminal
carbon to the oxygens in Y326 and N346 are 3.2 and
Ph
Ph
R
(CH2)nX
PvO
HO
Scheme 1. Synthesis of 4-OHT analogs designed to interact with
ERRc Y326 and N346. Reagents and conditions: (a) PvCl, NaH,
DMF, 0 °C to rt; (b) PhCOR, TiCl4, Zn, THF, D; (c) Me2NCH2-
CH2OH, PS-PPh2, DEAD, THF; (d) Nuꢀ, D; (e) LiOH, THF/H2O, D;
(f) LiAlH4, THF.
˚
4.3 A, respectively, suggesting that these residues may
be accessible by extension of the ethyl side chain. We
set out to exploit these residues in the design of analogs
with targeted selectivity for ERRc over ERa. Accord-
ingly, we designed 4-OHT analogs with extended ethyl
side chains bearing polar functionality capable of inter-
actions with the Y-N pair in ERRc and unfavorable
interactions with the hydrophobic phenylalanines in
ERa.
>4:1 as determined by NOE measurements. Mitsunobu
reaction with N,N-dimethylaminoethanol was employed
for installation of the basic side chain. With
R = (CH2)nCl, functionalization to the 4-OHT analogs
was accomplished by nucleophilic ꢀdisplacement10 of
the halide with RR0NH, MeSꢀ, CN , or N3 followed
ꢀ
by hydrolysis of the pivaloate ester. Alternatively, the
intermediates with R = (CH2)nCO2Me could be directly
reduced to the desired alcohol compounds with concom-
itant pivaloate deprotection. Compounds were purified
to configurational homogeneity prior to in vitro
profiling.
Synthesis of the 4-OHT analogs was accomplished using
an established methodology (Scheme 1). The triaryleth-
ylene intermediates were synthesized in three steps fol-
lowing the procedure of Gauthier and coworkers.9
Monoprotection of commercially available 4,40-dihydr-
oxybenzophenone with pivaloyl chloride followed by
McMurry coupling with substituted acetophenones
yielded triarylethylenes; stereoselectivity of the coupling
favored the desired E isomer, with typical E:Z ratios
Binding affinity was measured by a scintillation proxim-
ity binding assay using [3H]4-OHT (ERRc) or [3H]estra-
diol (ERa/b) as radioligand. In all cases, the ERb
affinity was not significantly different from ERa. Com-
pounds were compared to 1, which showed high affinity
for both receptors and moderate (sixfold) selectivity for
ERa. 4-Hydroxytoremifene11 (3) had an affinity profile
nearly identical to that of 1. Surprisingly, 4-hydroxy-
homo-toremifene 4 was nearly equipotent on the two
receptors. An effect of chain length upon ERa/ERRc
affinity was also observed with CN and N3. At n = 2,
the nitrile substituent in 5 decreased ERRc affinity by
100-fold and further extension as in 6 increased the affin-
ity by almost 10-fold. In contrast, 7, an azide compound
with n = 2, showed increased ERRc affinity, which was
reduced by chain extension in 8. The SMe group in 9
decreased the ERRc affinity more than the affinity for
ERa. Incorporation of acidic (CO2H), basic (NRR0),
or highly polar (NHAc) groups led to a significant loss
of affinity on both receptors, as no radioligand displace-
ment was observed for 10–13 at the maximum concen-
tration of the assay (3 lM). The potency and
selectivity of alcohol-containing tethers depended on
chain length. Equivalent potency was observed with 14
(n = 2), and the optimal compound in the series was 15
Figure 1. Overlap of ligand binding pocket residues from X-ray crystal
structures of 1 (green) bound to the LBDs of ERa (pink) and ERRc
(yellow) reveals key differences which might be exploited for the
improvement of ERRc selectivity. The ligand 1 and residues of interest
are depicted as sticks: ERRc Y326 and N346, and ERa F404 and
F425.

E. Y. H. Chao et al. / Bioorg. Med. Chem. Lett. 16 (2006) 821–824
823
where n = 3 in which a 25-fold improvement in selectiv-
120
100
80
60
40
20
0
ity for ERRc over ERa was realized relative to 1, albeit
at a lower overall affinity. Further extension of the chain
bearing the hydroxyl (as in 16) reversed the selectivity
profile in favor of ERa (Table 1).
In addition to the radioligand displacement assay, an
ERRc-RIP140 HTRF (FRET) assay was utilized to
measure changes in coactivator interaction caused by
compound treatment. A basal level of interaction be-
tween ERRc and a peptide fragment from NR cofactor
RIP140 was observed in the absence of added ligand and
was inhibited by 1 (IC50 = 20 nM). The interaction was
decreased by a similar amount with 15, albeit at lower
potency (IC50 = 250 nM), demonstrating that 15 would
function as an ERRc inverse agonist.
0
0.1
1
5
µM 1
Figure 2. Normalized luciferase activity of ERa + 100 nM estradiol
(black) and ERRc (white) as a function of treatment with 1. Data are
expressed as percent activity relative to the absence of compound.
Given the improved selectivity in binding affinity of 15,
we sought validation of its utility in a cellular context.
HeLa cells were transfected with an expression plasmid
encoding full-length ERRc or ERa and a reporter plas-
mid 3xERE-TATA-LUC containing three copies of the
vitellogenin ERE fused upstream of a TATA box se-
quence linked to luciferase (LUC). When normalized
to transcriptional activity in the absence of compound
treatment, 1 shows dose-dependent repression of the
activity of ERRc and competitive antagonism of
100 nM estradiol on ERa (Fig. 2). In contrast, there is
no effect on ERa transcriptional activity by 1 lM 15,
but the ERRc activity is repressed by 69% (Fig. 3). At
120
100
80
60
40
20
0
Table 1. ERRc and ERa binding affinities of 4-OHT analogs
NMe2
O
0
0.1
1
5
µM 15
Figure 3. Normalized luciferase activity of ERa + 100 nM estradiol
(black) and ERRc (white) as a function of treatment with 15. Data are
expressed as percent activity relative to the absence of compound.
(CH2)nX
HO
Compound
x
n
ERRc
ERa
R.S.b
IC50 (lM)a IC50 (lM)a
1 (4-OHT)
H
Cl
2
2
3
2
3
2
3
3
3
3
3
3
2
3
4
0.013
0.013
0.13
1.3
0.002
0.004
0.10
0.20
0.025
0.001
0.013
0.010
i.a.
1.0
2.0
5.0
1.0
0.8
1.6
0.5
0.1
—
3
4
Cl
CN
5
6
CN
N3
0.20
0.005
0.16
1.0
7
8
N3
9
SMe
CO2H
NHMe
NMe2
NHAc
OH
10
11
12
13
14
i.a.
i.a.
i.a.
i.a.
—
—
i.a.
i.a.
i.a.
—
0.25
0.079
0.79
0.25
0.32
0.079
6.3
25
15 (GSK5182) OH
16 OH
0.6
a IC50 of binding inhibition of [3H]4-OHT (ERRc) or [3H]estradiol
(ERa); i.a. = inactive at 3 lM; all data n = 2 in triplicate, variance
<15%.
Figure 4. Compound 15 (GSK5182) in ligand binding pocket of ERRc
shows hydrogen bonding interactions with Y326 and N346. The
coordinates have been deposited in the Brookhaven Protein Data
Bank, Accession No. 2EWP.
b Relative selectivity (R.S.) ꢁ [(ERRc compound IC50)/(ERa com-
pound IC50)]/[(ERRc 4-OHT IC50)/(ERa 4-OHT IC50)].

824
E. Y. H. Chao et al. / Bioorg. Med. Chem. Lett. 16 (2006) 821–824
5 lM 15, however, a partial repression of ERa activity is
observed. These results suggest that there may be a
workable concentration range in which the ERRc and
ERa effects of 15 are separable.
References and notes
`
1. Giguere, V. Trends Endocr. Metab. 2002, 13, 220.
2. Horard, B.; Vanacker, J.-M. J. Mol. Endocrinol. 2003, 31,
349.
3. (a) Hentschke, M.; Susens, U.; Borgmeyer, U. Biochem.
¨
To confirm the molecular basis of the improved selectivity
of 15, its complex with ERRc LBD was crystallized. The
Biophys. Res. Commun. 2002, 299, 872; (b) Huss, J. M.;
Kopp, R. P.; Kelly, D. P. J. Biol. Chem. 2002, 277, 40265;
(c) Kamei, Y.; Ohizumi, H.; Fujitani, Y.; Nemoto, T.;
Tanaka, T.; Takahashi, N.; Kawada, T.; Miyoshi, M.;
Ezaki, O.; Kakizuka, A. Proc. Natl. Acad. Sci. U.S.A.
2003, 100, 12378.
˚
X-ray structure was refined to a resolution of 2.30 A with
a final R factor of 18.7% and a free R factor of 24.7%.
(Fig. 4).12 The ligand binds in a mode similar to that ob-
served in the 4-OHT/ERRc structure where the phenol
interacts with E275 and R316, and the basic side chain
of the inhibitor interacts with D273. Distances from the
alcohol functionality to Y326 and N346 indicate hydro-
gen bonding interactions with each residue: Y326 (O–O:
4. Coward, P.; Lee, D.; Hull, M. V.; Lehmann, J. M. Proc.
Natl. Acad. Sci. U.S.A. 2001, 98, 8880.
5. Zuercher, W. J.; Gaillard, S.; Orband-Miller, L. A.; Chao,
E. Y. H.; Shearer, B. G.; Jones, D. G.; Miller, A. B.;
Collins, J. L.; McDonnell, D. P.; Willson, T. M. J. Med.
Chem. 2005, 48, 3107.
˚
˚
2.6 A) and N346 (O-O: 2.4 A). The X-ray crystal structure
is consistent with our original design hypothesis.
6. Greschik, H.; Flaig, R.; Renaud, J.-P.; Moras, D. J. Biol.
Chem. 2004, 32, 33639.
We have described the structure-based design of ERRc
selective inverse agonists. The 4-OHT scaffold has been
modified to improve selectivity for ERRc. A set of com-
pounds was designed and prepared to utilize a polar
interaction possible in ERRc but not in ERa. When
compared with the starting compound 1, a 25-fold
improvement in binding selectivity for ERRc over
ERa has been achieved with 15, and the presence of
the predicted hydrogen bonding interactions has been
confirmed by the X-ray cocrystal structure of 15 and
ERRc. The observed improvement in binding selectivity
is reflected in the cellular activity of 15, which has an im-
proved ERRc selectivity profile relative to 1. As such, 15
has potential for use as an inverse agonist chemical tool
in the elucidation of ERRc function.
7. Shiau, A. K.; Barstad, D.; Loria, P. M.; Cheng, L.;
Kushner, P. J.; Agard, D. A.; Greene, G. L. Cell 1998, 95,
927.
8. Moreover, the nature of this pair of residues is different in
ERRa (F344 and G364) and ERRb (Y301 and Y321).
Moras and coworkers (Ref. 6) have previously proposed
exploiting these differences for subtype selective ERR
ligands.
9. Gauthier, S.; Mailhot, J.; Labrie, F. J. Org. Chem. 1996,
61, 3890.
10. (a) Harrold, M. W.; Chang, Y. A.; Wallace, R. A.;
Farooqui, T.; Wallace, L. J.; Uretsky, N.; Miller, D.
D. J. Med. Chem. 1987, 30, 1631; (b) Shrestha-
Jawadi, P. B.; Lugtenburg, J. Eur. J. Org. Chem.
2003, 4654.
11. Simberg, N. H.; Murai, J. T.; Siiteri, P. K. J. Steroid
Biochem. 1990, 36, 197.
12. X-ray diffraction data were collected on the IMCA
beamline 17ID at the Advanced Photon Source,
Argonne, IL. The structure was solved by standard
molecular replacement methods and refined with
programs from the CCP4 project [Collaborative
Computational Project Number 4, Acta Crystallogr.,
Acknowledgments
The authors thank Randy Bledsoe and Thomas Consler
for protein expression and Anthony Handlon and Ryan
Trump for helpful discussions and critical review. S.G.
and D.P.M. acknowledge the support of the NIH
(NURSA U19 DK062434).
Sect.
D
1994, 50, 760]. Coordinates have been
deposited in the Brookhaven Protein Data Bank with
code 2EWP.