A natural-product switch for a dynamic protein interface

Article abstract of DOI:10.1002/anie.201403773

Small ligands are a powerful way to control the function of protein complexes via dynamic binding interfaces. The classic example is found in gene transcription where small ligands regulate nuclear receptor binding to coactivator proteins via the dynamic activation function 2 (AF2) interface. Current ligands target the ligand-binding pocket side of the AF2. Few ligands are known, which selectively target the coactivator side of the AF2, or which can be selectively switched from one side of the interface to the other. We use NMR spectroscopy and modeling to identify a natural product, which targets the retinoid X receptor (RXR) at both sides of the AF2. We then use chemical synthesis, cellular screening and X-ray co-crystallography to split this dual activity, leading to a potent and molecularly efficient RXR agonist, and a first-of-kind inhibitor selective for the RXR/coactivator interaction. Our findings justify future exploration of natural products at dynamic protein interfaces.

Full text of DOI:10.1002/anie.201403773

Angewandte
Chemie
DOI: 10.1002/anie.201403773
Protein–Protein Interactions
Hot Paper
A Natural-Product Switch for a Dynamic Protein Interface**
Marcel Scheepstra, Lidia Nieto, Anna K. H. Hirsch, Sascha Fuchs, Seppe Leysen,
Chan Vinh Lam, Leslie in het Panhuis, Constant A. A. van Boeckel, Hans Wienk, Rolf Boelens,
Christian Ottmann, Lech-Gustav Milroy,* and Luc Brunsveld*
Abstract: Small ligands are a powerful way to control the
function of protein complexes via dynamic binding interfaces.
The classic example is found in gene transcription where small
ligands regulate nuclear receptor binding to coactivator
proteins via the dynamic activation function 2 (AF2) interface.
Current ligands target the ligand-binding pocket side of the
AF2. Few ligands are known, which selectively target the
coactivator side of the AF2, or which can be selectively
switched from one side of the interface to the other. We use
NMR spectroscopy and modeling to identify a natural product,
which targets the retinoid X receptor (RXR) at both sides of the
AF2. We then use chemical synthesis, cellular screening and X-
ray co-crystallography to split this dual activity, leading to
a potent and molecularly efficient RXR agonist, and a first-of-
kind inhibitor selective for the RXR/coactivator interaction.
Our findings justify future exploration of natural products at
dynamic protein interfaces.
Scheme 1. Regulation of the dynamic nuclear receptor interface—the
activation function 2 (AF2)—by using small synthetic ligands derived
from the same natural product (1): stabilization of the AF2 through
binding at the ligand-binding pocket (2); destabilization of the AF2
through binding at the coactivator side of the interface (3).[11]
Small ligands are a powerful way to control the function of
large protein complexes via the selective modulation of
dynamic binding interfaces.[1] A classic example of this is seen
in eukaryotic gene transcription initiation, and the protein
complex formed between nuclear receptors and coactivator
proteins via the dynamic activation function 2 (AF2) binding
interface.[2] Ligand binding to a hydrophobic pocket located
at the solvent-excluded side of the AF2 (Scheme 1), within
the nuclear receptor ligand-binding domain,[3] allosterically
stabilizes or destabilizes coactivator protein binding at the
opposite, solvent-exposed side of the interface, which in turn
determines the transcriptional output. Ligand binding thus
functions as a molecular switch, where stabilization or
destabilization of the AF2 switches gene transcription either
“on” or “off”.[4]
Ligands targeting the nuclear receptor ligand-binding
pocket continue to be an important source of drug mole-
cules.[5] However, issues of toxicity and drug resistance mean
that ligands with atypical modes-of-action are in urgent
demand.[6] For instance, ligands targeting the ligand binding
pocket but with atypical partial agonist/antagonist behavior—
so-called selective nuclear receptor modulators—are less
toxic due to tissue-selective behavior.[7] Alternatively, modi-
fied peptides derived from the binding epitopes of coactivator
proteins or phage peptides selectively target the coactivator
side of the AF2.[8] Small non-peptidic ligands[9] are arguably
better suited than peptides as coactivator inhibitors due to
their high ligand efficiency, metabolic stability and cell
permeability, and some have shown promising NR-selective
behavior.[9e,f,j] Natural products, though well investigated at
[*] M. Scheepstra, Dr. L. Nieto, Dr. S. Fuchs, Dr. S. Leysen, C. V. Lam,
L. in het Panhuis, Prof. Dr. C. A. A. van Boeckel, Dr. C. Ottmann,
Dr. L.-G. Milroy, Prof. Dr. L. Brunsveld
Laboratory of Chemical Biology and Institute of Complex Molecular
Systems (ICMS), Department of Biomedical Engineering
Technische Universiteit Eindhoven
Den Dolech 2, 5612 AZ Eindhoven (The Netherlands)
E-mail: l.milroy@tue.nl
l.brunsveld@tue.nl
Homepage: http://www.tue.nl/cb
Dr. A. K. H. Hirsch
Stratingh Institue for Chemistry, University of Groningen
Nijenborgh 7, 9747AG Groningen (The Netherlands)
Dr. H. Wienk, Prof. Dr. R. Boelens
Bijvoet Center for Biomolecular Research, NMR Spectroscopy
Utrecht University, Padualaan 8, 3584CH Utrecht (The Netherlands)
[**] We thank Nicky Hoek, Dr. Eric Kalkhoven and Dr. Arjen Koppen
(UMC, Utrecht, The Netherlands) for their scientific support.
Funding was granted by the Netherlands Organisation for Scientific
Research via Gravity program 024.001.035, ECHO grant 711011017,
VENI grant to A.K.H.H., and a Marie Curie Action (PIEF-GA-2011-
298489 to LN).
Supporting information for this article (synthesis protocols, ana-
lytical data for all novel compounds, biochemical and cellular
evaluation of compounds, including cofactor recruitment, fluores-
cence polarization, molecular modeling, STD-NMR, CORCEMA-ST,
and X-ray co-crystallography) is available on the WWW under http://
dx.doi.org/10.1002/anie.201403773.
Angew. Chem. Int. Ed. 2014, 53, 6443 –6448
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
6443

.
Angewandte
Communications
the
ligand-binding
pocket, have been
underexplored at the
coactivator-side of the
AF2, and are well
suited for this purpose
due to their biological
relevance and the
unique and diverse
chemical space they
populate.[10]
Herein, we report
the development of
two different ligand
types, originating from
the same natural prod-
uct, targeting different
sides of
interface, and with
opposite stabilizing/
a dynamic
destabilizing proper-
ties. We used co-factor
recruitment screening
and a combination of
STD-NMR spectros-
copy, molecular dock-
ing, and CORCEMA-
ST calculations to
show that honokiol (1,
Scheme 1) targets both
sides of the AF2 of the
retinoid
(RXR).[12] We then
applied rational
X
receptor
Figure 1. Honokiol (1) binds to both sides of the dynamic AF2 interface of RXR. a) Four selected examples from
a fluorescence-based co-factor-recruitment assay showing inhibition of coactivator protein binding at 100 mm of 1 in
the presence of a potent full agonist, 4. b) Fluorescence polarization data showing that full agonist 4 induces
binding of a fluorescently labeled coactivator peptide in a concentration-dependent manner. Repeating the assay in
the presence of increasing concentrations of 1 (10, 100 mm) resulted in a progressive decrease in the maximum
polarization signal, but without changing the EC50 of 4, thus showing an alternative binding mode. c) A summary of
a
chemical-biology
approach, with an effi-
cient synthesis proto-
col at its core, to split
1
1D-1H NMR data revealing the atypical dual binding of 1: line broadening of the H resonances of 1 in the presence
the
dual-binding
and
of protein and agonist ligand; recovery of sharp intense signals on addition of a competitor LXXLL peptide; d)
1H NMR spectrum of honokiol (1; bottom) and STD (saturation transfer difference) spectrum of the system 1/RXR
(top) show rapid ligand exchange. e) The sum of separate theoretical CORCEMA-ST[13] values at the ligand binding
pocket and coactivator side of the AF2 interface compare well with the experimental STD data for the system 1/
RXR, thus indicating the dual ligand binding mode.
behavior of
1
create a potent and
molecularly efficient
RXR agonist and an
atypical RXR-selec-
tive antagonist (2 and
3, Scheme 1).
The RXR–coactivator interaction is important for the
development of cancer,[14] metabolic disorder,[15] and Alz-
heimerꢀs disease.[16] Current RXR ligands target the ligand-
binding pocket, for which a rigid and bulky hydrocarbon-rich
moiety is typically needed for potent binding (e.g., LG100268,
4, Figure 1b).[17] Non-peptidic ligands targeting the coactiva-
tor side of the RXR AF2 are at present non-existent. For
these reasons we became interested in the atypical RXR
activity of honokiol (1, Scheme 1).[18] Alongside related
natural products isolated from the Magnolia tree bark,[19]
1 displays an array of biological properties, including neu-
rite-growth induction and anti-angiogenic effects. It is likely
that 1 targets multiple proteins,[20] in light of its fragment-like
profile (MW = 266 Da),[10,21] and because of the privileged
nature of the biaryl structural motif.[22] Importantly, 1 has also
shown evidence of partial activity in a luciferase-based screen
using U2OS cells overexpressing RXR.[18]
We profiled the RXR-activity of 1 alongside analogous
natural and synthetic biaryl ligands using a fluorescence-
based co-factor-recruitment assay (Figure 1a and Supporting
Information), where an increase in fluorescence signal would
correspond with ligand binding at the RXR ligand binding
pocket.[23] In contrast to the other ligands tested, and contrary
to our initial expectations, 1 inhibited coactivator-binding at
100 mm ligand concentration in the presence and absence of
100 mm of a potent full agonist, 4 (Figure 1a). At 100 mm, 4
fully saturates the ligand binding pocket, and thus excludes
6444 www.angewandte.org
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2014, 53, 6443 –6448

Angewandte
Chemie
the binding of 1 at the ligand binding pocket. In support of
this, a separate fluorescence polarization binding assay was
performed, in which increasing concentrations of 4 predict-
ably induced recruitment of a fluorescently labelled coacti-
vator peptide (Figure 1b). Repeating the assay in the
presence of increasing amounts of 1 resulted in a concentra-
tion-dependent decrease in the maximum polarization signal,
but without detectable changes in the EC50 for 4 (Figure 1b).
The Ki for 1 was determined to be 94 mm ꢀ 9 mm (Supporting
Information). This atypical behavior of 1 indicated to us
a mode-of-binding distinct from the ligand binding pocket.
Despite a growing number of non-peptidic coactivator
inhibitor ligands,[6a,c,9] only a few are reported to be selective
for one nuclear receptor over others,[9e,f,j] of which none are
natural products and none selective for RXR. Importantly,
therefore, we found 1 to be selective for RXR over the
estrogen and androgen receptors (ER and AR) in a fluores-
cence-based co-factor-recruitment assay (Supporting Infor-
mation). Repeating the competitive fluorescence polarization
assay in the presence of detergent did not alter the binding
profile (Supporting Information),[25] which, coupled with the
inactivity of 1 towards ER and AR confirms the physiological
significance of the interaction between 1 and RXR. In
conclusion, 1 selectively inhibits the RXR–coactivator inter-
action in a physiologically significant manner and via an
atypical mechanism, which is independent of the ligand-
binding pocket.
In-depth ligand-detected NMR studies were performed to
further elucidate the RXR-binding mode of 1 (Figure 1c,
1
Supporting Information). Severe line broadening of the H
resonances of 1 was observed in the presence of the RXR
protein, which could be explained by the moderate binding
affinity of the compound, as evidenced by our fluorescence
polarization data, and the rapid ligand exchange. Line
broadening of 1 was also observed in the presence of protein
and an excess of potent ligands 2 and 4 (Figure 1c, Supporting
Information). However, signal intensity and sharpening of
1 recovered upon addition of a coactivator-derived peptide
(Figure 1c), indicating competitive inhibition of 1 by the
peptide. STD experiments on the system 1/RXR revealed the
rapid exchange between free and bound states (Figure 1d).
These data, combined with detailed competition STD and tr-
NOESY NMR experiments (Supporting Information) sug-
gest a novel binding mode for 1 at the coactivator side of the
AF2. Moreover, theoretical CORCEMA-ST[13] data show
that the experimental STD data collected for the 1/RXR
system correspond with a dual ligand binding mode (Fig-
ure 1d and e, Supporting Information) at both the ligand-
binding pocket and the coactivator side.
To capitalize on the dual-binding properties of 1, we
developed an orthogonal pair of RXR ligands capable of
selectively targeting opposite sides of the dynamic AF2
surface. We had reason to believe that 1 inhibits coactivator-
binding by mimicking the LXXLL binding motif, which is
highly conserved throughout coactivator proteins. Indeed, an
overlay of the energy-minimized state of 1 and the co-crystal
structure of an a-helical coactivator peptide bound to RXR
(PDB ID: 2P1T)[26] identified a strong overlap of the
interacting Leu residues at positions i and i + 4 of the a-
Figure 2. a) Overlay of honokiol (1; orange) and LXXLL mimic 3 (cyan)
with the LXXLL-coactivator peptide (red, PDB ID: 2P1T) bound to the
RXR AF2.[26] b) Cellular activities of 1 and 3 measured in a mammalian
two-hybrid luciferase assay in which increasing concentrations of the
ligand are co-incubated with a fixed, 100 nm, concentration of full RXR
agonist, 4.
helix and the allylic side-chains of 1 (Figure 2a). Furthermore,
the 4’-hydroxy functional group (para to the biaryl bond)
makes
a stabilizing hydrogen-bonding interaction with
Glu453—a charge-clamp residue important for selective
binding of the helical LXXLL motif—on molecular docking
of the LXXLL-aligned model of 1 to the AF2.
To favor selective AF2 binding, we synthesized isobutyl
analog 3 (Scheme 1; 25%, 3 steps, Supporting Information),
which we initially hypothesized would serve as a better mimic
than 1 of the LXXLL motif. Experimentally, analog 3 was
inactive as agonist in a mammalian two-hybrid assay (M2H)
up to 50 mm (Table 1). In the same agonistic assay, however,
1 alone elicited a complex response, which we explain by the
dual binding properties delineated in Figure 1c–e. Although
an EC50 value could not be determined for 1 in this case,
nevertheless at concentrations between 1–25 mm, 1 induced
partial activation of luciferase expression followed by inhib-
ition at the highest 50 mm test concentration.[18a] The catalytic
activity of the luciferase protein was unchanged, in the
Angew. Chem. Int. Ed. 2014, 53, 6443 –6448
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 6445

.
Angewandte
Communications
Table 1: Summary of fluorescence polarization (FP) and mammalian
two-hybrid (M2H) data for synthetic ligands vs. honokiol (1) and LG
100268 (4).[a]
and Supporting Information) in which the key biaryl bond
was formed under Buchwald-modified Suzuki conditions.[27]
Whereas 15 and 17 were only weakly active in both
fluorescence polarization and M2H assays, analog 18
Compound[b]
FP/EC50 [mm][c]
M2H (Luciferase)/EC50 [mm][d]
showed
significant
activity
(EC50(FP) = 8.3 ꢀ 2.2 mm;
LG100268 (4)
1
2
3
15
17
18
19
0.15ꢀ0.04
inactive
0.26ꢀ0.11
inactive
>250
0.0051ꢀ0.002
[e]
EC50(M2H) = 0.31 ꢀ 0.04 mm, Table 1). Our binding model
–
hinted at further activity gains by removing the 4’-hydroxy
0.0063ꢀ0.004
group (Figure 3, R2 = OH!H). Therefore, analog
2
inactive
>50
>50
(Scheme 1 and Figure 3) was prepared via a similar synthetic
>250
route, and, was gratifyingly 40-fold more active than 18
8.3ꢀ2.2
0.31ꢀ0.04
(EC50(FP) = 0.26 ꢀ 0.06 mm;
EC50(M2H) = 0.063 ꢀ 0.004 mm,
1.2ꢀ0.48
6.2ꢀ1.6
Table 1). The 20- to 30-fold difference between the FP and
M2H data is a common phenomemon,[9k] which can be
explained by intrinsic differences between the two different
assay formats, in particular, the different protein and peptide
concentrations used.
[a] Please refer to Supporting Information for activity curves. [b] See
Figure 3 for further synthesis details. [c] Fluorescence polarization (FP)
assay to determine direct binding (EC50). [d] Agonistic mammalian two-
hybrid (M2H) luciferase assay (EC50). See the Supporting Information for
details about the different assay formats. [e] Partial activity measured,
see Supporting Information and reference [18a].
To gain further molecular insight at the ligand-binding
pocket, the X-ray co-crystal of 2 bound to the RXR ligand-
binding domain was solved at 2.6 ꢁ resolution (Figure 4). The
carboxylate group of 2 is seen making a canonical interaction
with Arg116, while the flexible allylic side-chain occupies the
lipophilic region of the binding pocket. A molecular overlay
with known RXR co-crystal structures (PDB IDs: 2P1T and
4K6I) did not reveal any significant differences in global
protein conformation. We could not find electron density in
the X-ray structure, nor evidence from MS data (Supporting
Information) to suggest covalent attachment of 2 to the RXR
protein, thus ruling out irreversible inhibition as a possible
mode-of-action. Combined with the biochemical and cellular
results, this data suggests that, in contrast to current RXR
absence or presence of 1 and 3 (Supporting Information), thus
ruling out direct inhibition of luciferase as a possible mode-of-
action. Similar to 1, analog 3 also inhibited coactivator
binding to the AF2 in the fluorescence polarization assay,
albeit with slightly reduced affinity (Ki = 199 ꢀ 2 mm). Impor-
tantly though, analog 3 was more effective than 1 at
suppressing the full agonistic activity of 4 in the mammalian
two-hybrid assay (Figure 2b). The improved cellular activity
of 3 can be explained by an improved selectivity for the
solvent-exposed coactivator side of the AF2. Indeed molec-
ular modeling suggested that the isobutyl substituents on 3
disfavor binding at the ligand bind-
ing pocket due to additional repul-
sive interactions (Supporting Infor-
mation). We conclude therefore
that compared to 1, LXXLL
mimic 3 selectively inhibits the
RXR–coactivator interaction via
a more preferential binding at the
coactivator side of the dynamic
AF2 interface.
Our next aim was to switch
selectivity from the coactivator
side of the AF2 to the ligand-bind-
ing pocket. Potent RXR ligands
targeting
the
ligand-binding
pocket (e.g., 4, Figure 1b) typically
require a carboxylate group, which
forms a salt-bridge with residue
Arg116 in the hydrophilic region
of the binding pocket. Our model-
ing data (Supporting Information)
indicated that modifying one of the
allylic side-chains of 1 to a carbox-
ylate group would favor binding at
the ligand-binding pocket. Unsure
of the binding preference, we syn-
thesized analogs 15, 17, and 18
using an efficient palladium-cata-
Figure 3. Synthesis of ligands targeting the ligand-binding pocket. Reagents and conditions: a) [Pd2-
(dba)3], SPhos, 1,4-dioxane/H2O, 1108C, 18 h; b) aq. NaOH; c) BBr3. Details see Supporting
lyzed cross-coupling route (Figure 3 Information.
6446 www.angewandte.org
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2014, 53, 6443 –6448

Angewandte
Chemie
[1] a) C. Y. Majmudar, J. W. Højfeldt, C. J. Arevang, W. C. Pomer-
antz, J. K. Gagnon, P. J. Schultz, L. C. Cesa, C. H. Doss, S. P.
Rowe, V. Vꢂsquez, et al., Angew. Chem. 2012, 124, 11420 –
11424; Angew. Chem. Int. Ed. 2012, 51, 11258 – 11262; b) N.
Wang, C. Y. Majmudar, W. C. Pomerantz, J. K. Gagnon, J. D.
Sadowsky, J. L. Meagher, T. K. Johnson, J. A. Stuckey, C. L.
Brooks, J. A. Wells, A. K. Mapp, J. Am. Chem. Soc. 2013, 135,
3363 – 3366.
[2] P. Huang, V. Chandra, F. Rastinejad, Annu. Rev. Physiol. 2010,
72, 247 – 272.
[3] J.-M. Wurtz, W. Bourguet, J.-P. Renaud, V. Vivat, P. Chambon, D.
Moras, H. Gronemeyer, Nat. Struct. Mol. Biol. 1996, 3, 87 – 94.
[4] L. Nagy, J. W. R. Schwabe, Trends Biochem. Sci. 2004, 29, 317 –
324.
[5] a) J. T. Moore, J. L. Collins, K. H. Pearce, ChemMedChem 2006,
1, 504 – 523.
[6] a) T. W. Moore, C. G. Mayne, J. A. Katzenellenbogen, Mol.
Endocrinol. 2010, 24, 683 – 695; b) J. H. Choi, A. S. Banks, T. M.
Kamenecka, S. A. Busby, M. J. Chalmers, N. Kumar, D. S.
Kuruvilla, Y. Shin, Y. He, J. B. Bruning, et al., Nature 2011,
477, 477 – 481; c) P. Sadana, Future Med. Chem. 2012, 4, 1307 –
1333.
Figure 4. Ribbon representation of the X-ray co-crystal structure (PDB
ID: 4OC7) of 2 bound to the RXR ligand-binding domain: a) protein
(green), TIF2-derived coactiavator peptide (red), 2 (orange).
b) Zoomed-in view of the RXR ligand-binding pocket with amino acid
side-chains represented as sticks, and the electron density map of 2.
[7] C. L. Smith, B. W. OꢀMalley, Endocr. Rev. 2004, 25, 45 – 71.
[8] a) T. R. Geistlinger, R. K. Guy, J. Am. Chem. Soc. 2003, 125,
6852 – 6853; b) C. Phillips, L. R. Roberts, M. Schade, R. Bazin,
A. Bent, N. L. Davies, R. Moore, A. D. Pannifer, A. R. Pickford,
S. H. Prior, et al., J. Am. Chem. Soc. 2011, 133, 9696 – 9699.
[9] a) V. Azzarito, K. Long, N. S. Murphy, A. J. Wilson, Nat. Chem.
2013, 5, 161 – 173; b) L. A. Arnold, E. Estꢃbanez-PerpiÇꢂ, M.
Togashi, N. Jouravel, A. Shelat, A. C. McReynolds, E. Mar, P.
Nguyen, J. D. Baxter, R. J. Fletterick, et al., J. Biol. Chem. 2005,
280, 43048 – 43055; c) L. A. Arnold, A. Kosinski, E. Estꢃbanez-
PerpiÇꢂ, R. J. Fletterick, R. K. Guy, J. Med. Chem. 2007, 50,
5269 – 5280; d) J. Becerril, A. D. Hamilton, Angew. Chem. 2007,
119, 4555 – 4557; Angew. Chem. Int. Ed. 2007, 46, 4471 – 4473;
e) J. R. Gunther, A. A. Parent, J. A. Katzenellenbogen, ACS
Chem. Biol. 2009, 4, 435 – 440; f) P. Nandhikonda, W. Z. Lynt,
M. M. McCallum, T. Ara, A. M. Baranowski, N. Y. Yuan, D.
Pearson, D. D. Bikle, R. K. Guy, L. A. Arnold, J. Med. Chem.
2012, 55, 4640 – 4651; g) A. L. Rodriguez, A. Tamrazi, M. L.
Collins, J. A. Katzenellenbogen, J. Med. Chem. 2004, 47, 600 –
611; h) E. Estꢃbanez-PerpiÇꢂ, L. A. Arnold, A. A. Arnold, P.
Nguyen, E. D. Rodrigues, E. Mar, R. Bateman, P. Pallai, K. M.
Shokat, J. D. Baxter, et al., Proc. Natl. Acad. Sci. USA 2007, 104,
16074 – 16079; i) A. Sun, T. W. Moore, J. R. Gunther, M.-S. Kim,
E. Rhoden, Y. Du, H. Fu, J. P. Snyder, J. A. Katzenellenbogen,
ChemMedChem 2011, 6, 654 – 666; j) L. Caboni, G. K. Kinsella,
F. Blanco, D. Fayne, W. N. Jagoe, M. Carr, D. C. Williams, M. J.
Meegan, D. G. Lloyd, J. Med. Chem. 2012, 55, 1635 – 1644;
k) A. L. LaFrate, J. R. Gunther, K. E. Carlson, J. A. Katzenel-
lenbogen, Bioorg. Med. Chem. 2008, 16, 10075 – 10084.
[10] B. Over, S. Wetzel, C. Grꢄtter, Y. Nakai, S. Renner, D. Rauh, H.
Waldmann, Nat. Chem. 2013, 5, 21 – 28.
ligands, a rigid and bulky hydrophobic moiety is not necessary
for potent binding at the ligand-binding pocket. However,
analog 19 lacking the hydroxy groups at the 2- and 4’-positions
(Figure 3) was significantly less active than 2 (> 200-fold),
highlighting the importance of the 2-hydroxy group for
activity,[26] by simultaneously restricting the rotational free-
dom about the biaryl bond and through formation of a hydro-
gen-bonding interaction with residue Asn306. Thus, by
rational design and using a short and focused synthetic
route, we managed a selective switch of the targeting proper-
ties of 1 from one side of the dynamic AF2 interface of RXR
to the other—from the solvent-exposed side to the ligand-
binding pocket—and most notably with improved ligand
efficiency (BEI)[24] compared to known RXR ligands (2,
BEI(FP) = 23.5 vs. 4, BEI(FP) = 18.8).
In summary, we demonstrate the rational splitting of the
dual-binding properties of a natural product at a dynamic
protein interface. This outcome has resulted in two distinct
and molecularly efficient ligand types targeting opposite sides
of the activation function 2 (AF2) of the retinoid X receptor
(RXR). The first ligand type, represented by 3, exhibits an
atypical behavior, inhibiting coactivator binding at the
solvent-exposed side of the AF2 interface. Notably, ligand 3
is the first of its kind selective for RXR. The second type,
represented by 2, potently binds to the ligand-binding pocket,
thereby inducing coactivator binding via an established
mechanism. Our findings justify the future exploration of
natural products at dynamic protein interfaces.
[11] The PyMOL Molecular Graphics System, Version 1.5.0.4
Schrçdinger, LLC.
[12] A. R. de Lera, W. Bourguet, L. Altucci, H. Gronemeyer, Nat.
Rev. Drug Discovery 2007, 6, 811 – 820.
Received: March 27, 2014
[13] a) V. Jayalakshmi, N. R. Krishna, J. Magn. Reson. 2002, 155,
106 – 118; b) N. R. Krishna, V. Jayalakshmi, Top. Curr. Chem.
2008, 273, 15 – 54.
Published online: May 12, 2014
[14] K. T. Liby, M. M. Yore, M. B. Sporn, Nat. Rev. Cancer 2007, 7,
357 – 369.
[15] L. Altucci, M. D. Leibowitz, K. M. Ogilvie, A. R. de Lera, H.
Gronemeyer, Nat. Rev. Drug Discovery 2007, 6, 793 – 810.
Keywords: drug discovery · natural products ·
nuclear receptors · protein–protein interactions ·
retinoid X receptor
.
Angew. Chem. Int. Ed. 2014, 53, 6443 –6448
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 6447

.
Angewandte
Communications
[16] P. E. Cramer, J. R. Cirrito, D. W. Wesson, C. Y. D. Lee, J. C.
Karlo, A. E. Zinn, B. T. Casali, J. L. Restivo, W. D. Goebel, M. J.
James, et al., Science 2012, 335, 1503 – 1506.
[21] C. W. Murray, D. C. Rees, Nat. Chem. 2009, 1, 187 – 192.
[22] D. A. Horton, G. T. Bourne, M. L. Smythe, Chem. Rev. 2003, 103,
893 – 930.
[17] J. H. Barnard, J. C. Collings, A. Whiting, S. A. Przyborski, T. B.
Marder, Chem. Eur. J. 2009, 15, 11430 – 11442.
[23] A. Koppen, R. Houtman, D. Pijnenburg, E. H. Jeninga, R.
Ruijtenbeek, E. Kalkhoven, Mol. Cell. Proteomics 2009, 8,
2212 – 2226.
[24] C. Abad-Zapatero, Expert Opin. Drug Discovery 2007, 2, 469 –
488.
[25] B. Y. Feng, B. K. Shoichet, Nat. Protoc. 2006, 1, 550 – 553.
[26] V. Nahoum, E. Pꢃrez, P. Germain, F. Rodrꢅguez-Barrios, F.
Manzo, S. Kammerer, G. Lemaire, O. Hirsch, C. A. Royer, H.
Gronemeyer, et al., Proc. Natl. Acad. Sci. USA 2007, 104,
17323 – 17328.
[27] K. L. Billingsley, T. E. Barder, S. L. Buchwald, Angew. Chem.
2007, 119, 5455 – 5459; Angew. Chem. Int. Ed. 2007, 46, 5359 –
5363.
[18] a) H. Kotani, H. Tanabe, H. Mizukami, M. Makishima, M. Inoue,
J. Nat. Prod. 2010, 73, 1332 – 1336; b) C.-G. Jung, H. Horike, B.-
Y. Cha, K.-O. Uhm, R. Yamauchi, T. Yamaguchi, T. Hosono, K.
Iida, J.-T. Woo, M. Michikawa, Biol. Pharm. Bull. 2010, 33,
1105 – 1111.
[19] Y.-J. Lee, Y. M. Lee, C.-K. Lee, J. K. Jung, S. B. Han, J. T. Hong,
Pharmacol. Ther. 2011, 130, 157 – 176.
[20] a) P. Banerjee, A. Basu, J. L. Arbiser, S. Pal, Cancer Lett. 2013,
338, 292 – 299; b) T. Singh, S. K. Katiyar, PLoS ONE 2013, 8,
e60749; c) A. Kumar, U. Kumar Singh, A. Chaudhary, Future
Med. Chem. 2013, 5, 809 – 829.
6448 www.angewandte.org
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2014, 53, 6443 –6448