Fragment-based discovery of bromodomain inhibitors part 2: Optimization of phenylisoxazole sulfonamides

Article abstract of DOI:10.1021/jm201283q

Bromodomains are epigenetic reader modules that regulate gene transcription through their recognition of acetyl-lysine modified histone tails. Inhibitors of this protein-protein interaction have the potential to modulate multiple diseases as demonstrated by the profound anti-inflammatory and antiproliferative effects of a recently disclosed class of BET compounds. While these compounds were discovered using phenotypic assays, here we present a highly efficient alternative approach to find new chemical templates, exploiting the abundant structural knowledge that exists for this target class. A phenyl dimethyl isoxazole chemotype resulting from a focused fragment screen has been rapidly optimized through structure-based design, leading to a sulfonamide series showing anti-inflammatory activity in cellular assays. This proof-of-principle experiment demonstrates the tractability of the BET family and bromodomain target class to fragment-based hit discovery and structure-based lead optimization.

Full text of DOI:10.1021/jm201283q

Article
pubs.acs.org/jmc
Fragment-Based Discovery of Bromodomain Inhibitors Part
2: Optimization of Phenylisoxazole Sulfonamides
Paul Bamborough,*,‡,∥ Hawa Diallo,† Jonathan D. Goodacre,† Laurie Gordon,§ Antonia Lewis,§
Jonathan T. Seal,† David M. Wilson,† Michael D. Woodrow,† and Chun-wa Chung‡,∥
†Epinova DPU, Immuno-Inflammation Centre of Excellence for Drug Discovery, GlaxoSmithKline R&D, Medicines Research Centre,
Gunnels Wood Road, Stevenage, Hertfordshire SG1 2NY, U.K.
‡Computational & Structural Chemistry, Molecular Discovery Research, GlaxoSmithKline R&D, Medicines Research Centre,
Gunnels Wood Road, Stevenage, Hertfordshire SG1 2NY, U.K.
§Screening & Compound Profiling, Molecular Discovery Research, GlaxoSmithKline R&D, Medicines Research Centre,
Gunnels Wood Road, Stevenage, Hertfordshire SG1 2NY, U.K.
S
* Supporting Information
ABSTRACT: Bromodomains are epigenetic reader modules
that regulate gene transcription through their recognition of
acetyl-lysine modified histone tails. Inhibitors of this protein−
protein interaction have the potential to modulate multiple
diseases as demonstrated by the profound anti-inflammatory
and antiproliferative effects of a recently disclosed class of BET
compounds. While these compounds were discovered using
phenotypic assays, here we present a highly efficient alternative
approach to find new chemical templates, exploiting the
abundant structural knowledge that exists for this target class.
A phenyl dimethyl isoxazole chemotype resulting from a focused fragment screen has been rapidly optimized through structure-
based design, leading to a sulfonamide series showing anti-inflammatory activity in cellular assays. This proof-of-principle
experiment demonstrates the tractability of the BET family and bromodomain target class to fragment-based hit discovery and
structure-based lead optimization.
fragments are not lost during optimization. For this, the binding
pocket must be large and deep enough to allow elaboration that
gains additional productive interactions. Iterative high reso-
lution crystallography of protein−ligand complexes and
molecular modeling predictions greatly assist the process of
FBDD optimization. Structures of many bromodomains have
now been solved, from which it appears that some have AcK
pockets which may indeed be suitable for productive fragment
growth. Here we describe how a structure-derived 3D
pharmacophore was key to the rapid optimization of a weak
BET fragment into compounds with cellular anti-inflammatory
activity.
INTRODUCTION
■
Bromodomain (BD) modules regulate the structure of chroma-
tin, and thereby gene expression, through their ability to “read”
the acetylated-lysine state of histone tails.1 Bromodomain-
containing proteins (BCPs) have been associated with
numerous conditions including diabetes, obesity, oncology,
inflammation, viral infection, metabolic, and cardiovascular
disorders.1−4 The discovery of the first potent bromodomain
inhibitors using cellular phenotypic assays has recently been
reported.5,6 These benzodiazepines (BZDs, Table 1) target the
BET family (BRD2, BRD3, BRD4) tandem bromodomains,4
which showed for the first time that the BET/histone protein−
protein interaction can be disrupted by small molecules. In an
accompanying paper,7 we demonstrate that this interaction
can also be blocked by fragment-sized molecules selected
to target the “hotspot” of the acetyl-lysine (AcK) pocket.
Crystallography confirmed that these chemically diverse
fragments exhibit remarkably similar binding features within
this recognition site.
RESULTS
■
Fragment Screening. Initial screening was performed
with a fluorescence anisotropy (FA) assay against the three
tandem BET bromodomain constructs using a rationally
selected screening set targeting the bromodomain AcK pocket.
Numerous potential starting points were found, some of which
are described elsewhere.7 4-Phenyl 3,5-dimethyl isoxazole 3
was one of the hits (Table 1), showing activity of 32% and 26%
Fragment-based drug discovery (FBDD) uses high-concen-
tration screening to identify low-MW hits8 which, although
weak, are efficient in terms of their binding energy per heavy
atom, usually expressed as ligand efficiency or LE.9 Care must
be taken to ensure that the attractive drug-like properties of the
Received: August 23, 2011
Published: December 5, 2011
© 2011 American Chemical Society
587
dx.doi.org/10.1021/jm201283q | J. Med. Chem. 2012, 55, 587−596

Journal of Medicinal Chemistry
Article
Table 1. Literature BET Bromodomain Inhibitors and Their
Tandem Bromodomain IC50 (μM, Mean of ≥7 Values) in
FA, TR-FRET, and Isothermal Titration Calorimetry
a
Assays
a
Ligand efficiency (LE) calculated as 1.37(pIC50/number of heavy
atoms) is also shown. (+)-JQ1 had IC50 of 77 nM and 33 nM against
isolated N- and C-terminal BRD4 bromodomains in a luminescence
proximity homogeneous assay.6
at 200 μM against BRD3 and BRD4, respectively. While this
was far from the most potent fragment detected, it was small,
efficient, novel, and chemically attractive. It was also one of the
fragments which is less directly related to acetyl-lysine. We
therefore investigated its binding mode in order to optimize the
potency of the series.
Crystal Structure of Phenyl Isoxazole Fragment 3.
Compounds have previously been shown to bind with similar
affinity and identical binding mode to both N- and C-terminal
bromodomains of the highly homologous BET family
members.10 Therefore we used X-ray structures of compounds
soaked into apo crystals of the N-terminal bromodomain of
BRD2 (BRD2-BD1) to exemplify their interactions within
the AcK pocket7 (Supporting Information Figure S1). The C2
crystal form contains 3 molecules in the asymmetric unit
(ASU) whose AcK pockets are differentially accessible (chains
A−C, Figure 1A).
The dimethyl isoxazole 3 binds in accord with the hypothesis
underlying the focused set selection.7 One of the two methyl
groups of the dimethyl isoxazole occupies the small pocket
adjacent to Phe99, mimicking the terminal methyl of the AcK
side chain (Figure 1B,C, Figure 2A). The second methyl group
overlays with the ε-CH2 group of the AcK side chain,
contacting the side chain of Leu110. The isoxazole nitrogen
and oxygen atoms together mimic the AcK carbonyl group
whose position is roughly midway between the two (Figure
1B). Because the isoxazole ring is almost symmetrical, even at
<2.0 Å resolution the electron density cannot differentiate the
isoxazole heteroatoms unambiguously. The true binding mode
of 3 may be one in which the nitrogen and oxygen exchange
positions or inhibition may result from a mixture of both
binding modes if the energy difference is small.
Figure 1. (A) The ASU of BRD2-BD1 showing 3 (cyan) bound into
chains A (blue), B (orange), and C (green), in two orientations: that
of (B) above, and of (C) below. (B) Chain A (blue) and 3 (cyan),
with the side chain of AcK (orange) taken from histone H4 1−20,
K5Ac K8Ac bound to mouse BRDT (PDB 2wp2).24 (C) As (B), in an
alternative orientation. (D) As (B), showing the red electrostatic
isopotential surface of 3. (E) The complex of 4 (yellow) in BRD2
(orange) with 3 (cyan), in the orientation of (C). (F) As (E), in the
orientation of (B).
structures.7 For each heteroatom, the angle of interaction with
the water deviates by over 50° compared to an ideal sp2 lone
pair. One of the heteroatoms is 3.2 Å away from the conserved
AcK-binding side chain nitrogen of Asn156, but this is also 50°
away from ideality. The isoxazole has been modeled so that the
more electron-rich nitrogen atom is nearest to the water
molecule (Figure 1D), in accord with our finding that for most
fragments the ligand lies closer to the water than to Asn156.7
To attempt to confirm this, the 5-ethyl 3-methyl fragment 4
This uncertainty cannot be resolved by the observed inter-
actions: both heteroatoms are roughly equidistant (2.7−2.9 Å)
from a water molecule that is itself hydrogen-bonded to the
side chain oxygen of Tyr113. This bridging water molecule is
one of the most highly conserved in BRD2/fragment crystal
588
dx.doi.org/10.1021/jm201283q | J. Med. Chem. 2012, 55, 587−596

Journal of Medicinal Chemistry
Article
Figure 2. OMIT (Fo − Fc) difference maps of BRD2-BD1 complexes. (A) Complex with 3; (B) complex with 4; (C) complex with 5a; (D) complex
with 6a. Contours are blue = 3σ, gray = −3σ for (A−C) and blue = 2σ, gray = −2σ for (D).
was synthesized and crystallized with BRD2-BD1. As the longer
ethyl group is visible within the electron density, the binding
mode of this compound can be determined with certainty,
supporting the proposed orientation of the isoxazole in 3
(Figure 1E,F).
The phenyl ring of 3 occupies a tight cleft, sandwiched on
either face by the aliphatic side chains of Pro98 and Leu108
(Figure 1B). The phenyl−isoxazole torsion angle varies between
51 and 59° for the three ligands in the asymmetric unit. While
slightly greater than the quantum mechanics gas-phase minimized
torsion of 44° (see Experimental Section), the calculated ΔE is
negligible.
Initial Optimization. To find active analogues, our
knowledge base of BET X-ray complexes was exploited using
a variety of computational approaches. One method used 3D
pharmacophore models to describe the essentials of the binding
site. Steric and electrostatic features of the AcK headgroup were
encapsulated by features placed upon critical atoms of the
dimethyl isoxazole 3 (Figure 3A). Another key interaction
was with the WPF shelf, a lipophilic region named after the
sequence of residues 97−99 in BRD2, which we previously
showed to bind aromatic rings of 1 and 2.10 Further, in the
complex with 3, this site was occupied by ethylene glycol.
Therefore a lipophilic feature was added to the pharmacophore
at the centroid of the pendant phenyl ring of 1 (Figure 3A).
The pharmacophore was used to search a database of com-
mercially available compounds. Among the search hits was a
cluster of phenyl isoxazole analogues of 3 with sulfonamide
substituents on the phenyl ring meta to the isoxazole. When
fitted to the pharmacophore model, the sulfonamide introduces
a bend into the shape of the molecule that causes the substi-
tuent to project in the direction of the WPF shelf (Figure 3B).
Docking also predicted this binding mode, providing some
validation for this hypothesis. Selected compounds were pur-
chased and screened in dose−response mode, the results
of which are shown in Table 2 and Supporting Information
Figure S2. These compounds showed encouraging potency in
both FA and TR-FRET assays, with IC50 in the low micromolar
range. Compared to the fragment hit 3, which has an IC50 of
above 200 μM, 5a is at least 100-fold more potent. This was
particularly interesting because IC50s from the FA and TR-
FRET assays can underestimate equilibrium affinities obtained
from direct binding methods such as isothermal titration calori-
metry due to the low tight binding limit of these competition
assays (Table 1). The ligand efficiency (LE, approximated as
1.37(pIC50/number of heavy atoms)9) of 5a is 0.39, com-
parable to 3 based on IC50 ≈ 200 μM, and superior to 1 and 2
(Table 1).
Crystal Structures of Optimized Sulfonamides. Exam-
ples of the sulfonamide series were soaked into crystals of
BRD2-BD1. In the crystal form used, the Z-helix of chain A,
especially its N-terminus around Asn77, is in close proximity to
the AcK pocket of chain C,7 reducing the accessibility of this
site (Figure 1A). Because of this, the AcK site of chain C in the
complex with 5a is occupied by a DMSO molecule. The
orientation of DMSO in chain C is very similar to that in
589
dx.doi.org/10.1021/jm201283q | J. Med. Chem. 2012, 55, 587−596

Journal of Medicinal Chemistry
Article
molecule, and one of the two methyl groups fits into the methyl
pocket. The sulfonamide substituent can be accommodated
without the need for any change in the binding mode of the
fragment or the interactions that it makes with the acetyl lysine
pocket. The torsion angles between the isoxazole and phenyl
rings are similar to those of 3 (51−56° in both chains). As
predicted, the sulfonamide introduces a bend in the shape of
the molecule that directs the cyclopropyl ring onto the WPF
shelf, occupying the space of the lipophilic pharmacophore
feature. It makes contact with the hydrophobic parts of the
shelf, most extensively with the side chains of residues Trp97,
Pro98, Ile162, and Met165, and the methylene group of the
Asp161 side chain. The sulfonamide itself makes no direct
hydrogen-bonding interactions with the protein but does parti-
cipate in some water-mediated interactions. One of the sulfo-
namide oxygen atoms is close to a well resolved water molecule
that caps the N-terminus of the αC-helix by hydrogen-bonding
to the backbone NH of Ile162. An equivalent water molecule is
also present in the complex with 3 and was always present in
40 crystal structures of BRD2-BD1 with various AcK-site frag-
ments.7 The other oxygen and the nitrogen atom of the sulfon-
amide group are also close to water molecules, but the details
vary from one molecule of the asymmetric unit to another so
we do not believe that these interactions contribute strongly to
binding.
Structure−Activity Relationships. Expanding the cyclo-
propyl ring to cyclopentyl (6a, 6b) and cyclohexyl (7a) gave
compounds with increased activity but with a trend for
decreasing ligand efficiency when normalized for the additional
atoms introduced (Table 2). This suggests that the optimal size
of the lipophilic substituent that can be accommodated by the
WPF shelf is 3−5 heavy atoms. The cyclopentyl analogue 6a
binds to all three molecules of the ASU of the crystal (Figure 2).
Compared to its position in chains A and B, the ligand moves
slightly in chain C, probably as a result of the obstructed nature
of this AcK site. Overall, the binding mode of the cyclopentyl 6a
in chains A and B is very similar to that of the cyclopropyl 5a
(Figure 3C,D). As with 5a, the sulfonamide substituent of 6a
occupies the WPF shelf. The phenyl/isoxazole bond in the
cyclopentyl structure is slightly more in-plane (37−43°) than in
fragment 3 and the cyclopropyl 5a, resulting in the C3 atom of
the cyclopentyl ring overlaying almost perfectly with the C2
atom of the cyclopropyl sulfonamide (Figure 3D). Presumably
this is because of steric contacts which prevent the cyclopropyl
ring from moving any deeper into the pocket (Figure 3E,F).
This rotation also results in a shift of the positions of the two
sulfonamide oxygen atoms by about 1 Å, leading to subtle
changes in the way that they coordinate water molecules. For
example, one SO group moves away from the αC-helix Ile162
capping water, so that the H-bonding interaction takes place
through an extra intervening water molecule (compare parts B
and C of Figure 3).
Even larger sulfonamide substituents, such as those of the
phenyl-substituted 8a and 8b, are accommodated but are not
optimal, as indicated by their decreased potency and LE (Table 2).
Substitution on the phenyl ring at any position lowered the
affinity, regardless of electronic accepting or donating ability
(9a−14a, 9b−14b), which can also be rationalized sterically.
Our expectation was that the introduction of a chain linking the
sulfonamide to the phenyl ring would be detrimental to binding
because this would make it difficult for the phenyl ring to lie
upon the WPF shelf and would increase the entropic penalty to
binding. This was found to be the case for the benzyl-substituted
Figure 3. (A) The complex of BRD2 with 3 (cyan), with 2 (orange)
superimposed in the orientation of Figure 1B. Transparent spheres
represent pharmacophore features: blue = H-bond acceptor (AcK
carbonyl); green = methyl; red = aromatic; yellow = lipophilic (WPF
shelf). (B) BRD2 (olive) in complex with 5a (magenta), superimposed
on 3 (cyan). (C) BRD2 (yellow) in complex with 6a (green), overlaid
on 3 (cyan) and 5a (magenta). (D) As (C), in alternative orientation
looking down the phenyl/isoxazole bond. (E) BRD2 electrostatic
potential surface with 6a (spheres) in orientation of (C). (F) As (E) in
orientation of (D).
CREBBP (PDB entry 3p1e)11 and so will not be discussed
further. In the more accessible A and B chains of the ASU, 5a
binds in an identical manner and as intended by the modeling
hypothesis.
Figures 3B and 2 show the binding mode of 5a in chain B,
highlighting the similarity between its binding mode and that of
fragment 3. The isoxazole oxygen of 5a lies within hydrogen-
bonding distance of both Asn156 and the bridging water
590
dx.doi.org/10.1021/jm201283q | J. Med. Chem. 2012, 55, 587−596

Journal of Medicinal Chemistry
Article
Table 2. (A) Activity of Sulfonamide Analogues against BRD2, BRD3, and BRD4 (TR-FRET Assay, μM) (See Supporting
Information Figure S2 for Complete Data Including n and Standard Deviation; LE = Ligand Efficiency, ΔG ∼ 1.37(pIC50/N
heavy atoms)); (B) Binding Curves of 5a (Triangles), 6a (Pentagons), and 29a (Crosses) in BRD4 FA and TR-FRET Assays
compounds (15a, 15b) and compounds with longer linking
chains (16a, 17a). Similar arguments rationalize the reduced
affinity of flexible straight-chain sulfonamides (18a, 19a), although
the bulky t-butyl group of 20a appears sufficiently rigid and
lipophilic to regain some of the loss.
Table 3. Activity (TR-FRET and PBMC-IL6 Assays) and
Solubility of Sulfonamide Analogues 31−34
For comparison, we synthesized the unsubstituted and ethyl
sulfonamides 21a and 22a. Although the ethyl is slightly less
potent than the cyclopropyl 5a and the unsubstituted sulfon-
amide is almost 10-fold less potent, they are of comparable
efficiency. The water-mediated interactions of the sulfonamide
described earlier, and potential additional water-mediated inter-
actions of the primary sulfonamide might account for the
improved activity of 21a over 3. Alkylation of the sulfonamide
nitrogen was tolerated with marginally increased potency (23b,
24b) but with a slight decrease in efficiency. This is consistent
with our belief that the water molecule seen near the sulfon-
amide nitrogen in the crystal structures of 5a and 6a is weakly
bound. It also proved possible to homologate the nitrogen into
the ring, producing the pyrrolidine 25a and piperidine 26a.
However, these compounds offered no clear advantage over
the secondary sulfonamides and were slightly lower in ligand
efficiency.
substituent would be deterimental to binding (see electrostatic
potential map, Figure 3E, Figure 3F). This was found to be the
case because the introduction of polar groups led to reduced
potency (for example 27a−30a, 27b−30b). Consequently, we
turned to modifications of the phenyl ring para to the isoxazole.
Diversity at this position in the commercially available com-
pounds was restricted to methyl and methoxy. This position
Improvement in Solubility. Low solubility was a limita-
tion of the series (Table 3). To improve this, charge and
polarity were incorporated at positions guided by the X-ray
structures. As the WPF shelf is mainly lipophilic, we were con-
cerned that introducing these properties into the sulfonamide
591
dx.doi.org/10.1021/jm201283q | J. Med. Chem. 2012, 55, 587−596

Journal of Medicinal Chemistry
Article
points toward the solvent (Figure 2C), and there was little
difference in potency between the methyl and methoxy com-
pounds (Table 2). In fact, the substituent could be removed
entirely with only a small effect on activity (31, Table 3). A
wide range of groups were therefore introduced at this position
to modulate the physicochemical properties. Compounds in-
corporating solubilizing groups were accessed via the phenol
32. As hoped, these analogues (e.g., 33, 34) had much improved
solubility (Table 3) and retained most of the affinity of the simpler
analogues, albeit with lower ligand efficiency.
SPR Kd Determination. To obtain label-free confirmation
of binding to the BET bromodomains and an equilibrium
binding constant, exemplars were tested in a surface plasmon
resonance (SPR) assay. Compound 34 was found to have a Kd
of 2−3 μM for BRD2 and BRD4 (Table 4), consistent with the
micromolar TR-FRET IC50 values in Table 3.
DISCUSSION
■
Inhibiting protein−protein interactions (PPIs) with small
molecules is often perceived as challenging.12−14 Even when
it is possible to find ligand-efficient starting points, the
elaboration of these to yield biologically active compounds
without compromising traditional drug-like properties can
prove problematic. It has even been suggested that to increase
the probability of success, Lipinski’s rules for drug-likeness15
need to be relaxed and modified for molecules that target PPIs.16
However, while some targets are undeniably difficult, not all PPI
interfaces are flat and featureless. In an accompanying manuscript,
we reported the use of focused fragment screening to find diverse
low-MW BET bromodomain inhibitors that bind at one “hotspot”,
the acetyl-lysine pocket.
One common difficulty in interpreting high concentration
biochemical FBDD screening is in verifying that weak
fragments interact directly with the target rather than interfere
with the assay. This may be one reason why apparently promising
hits can prove impossible to optimize. In this example, the
hits were validated by orthogonal assays and X-ray crystallo-
graphy, confirming both their binding mode and the under-
lying knowledge-based strategy used to drive the compound
screening set selection.7
Table 4. Surface Plasmon Resonance Binding Results
Showing Kd for Compound 34 Binding to BRD2 and BRD4
a
Starting from the isoxazole fragment hit, our knowledge of
the BET family active site predicted that the WPF shelf would
be an important region to enhance potency and selectivity.
Analogues were selected to exploit this, leading to an immediate
increase in potency of 2 orders of magnitude. Importantly, this
was accompanied by structure−activity relationships which
could be rationalized crystallographically. After available com-
pounds had been screened, additional molecules were selected
through modeling and synthesized to complete our evalua-
tion. Members of the series achieve higher ligand efficiency
than the BZD and related BET inhibitors discovered through
phenotypic screening. SAR within the series is consistent with
the crystallographic binding mode, and levels of cellular anti-
inflammatory activity correlate with BET binding affinity.
Furthermore, our assessment of selectivity suggests that, despite
their relatively small size, members of this sulfonamide series
bind to the BET family bromodomains preferentially over other
bromodomains.
The resulting series binds using an isoxazole AcK-mimetic
template unknown in the literature until very recent disclosures.
This chemotype appears to have been found in several different
ways, demonstrating the tractability of the BET family to a
variety of lead discovery approaches. The first 3,5-dimethyl
isoxazole BET inhibitors to be published belong to a related
quinoline series.17,18 These were found through phenotypic
screening and were optimized to high levels of potency and
cellular efficancy without any knowledge of the molecular
target.19,20
a
Average BRD2 Kd = 3.14 μM (SD 1.65); BRD4 Kd = 2.75 μM (SD
1.76).
Selectivity. Selectivity profiles of compounds were assessed
by thermal melting (Tm) assays. These were performed against
BET family single and tandem bromodomains as well as eight
other proteins selected to sample other branches of the phylo-
genetic tree. ΔTm results are shown in Figure 4 for compounds
6a and 34. Both compounds stabilized the melting temperature
of the BET bromodomains consistently, most notably in the
tandem BD12 constructs, where a shift of over 5 °C was
observed in the case of BRD4. For the other bromodomains,
the effects were much smaller (≤1.5 °C), suggesting that there
is a window of selectivity between the BET family and other
bromodomains, although this method does not permit a precise
quantitation.
Cellular Activity. BET inhibitors show potent effects in
cellular systems, including inhibition of pro-inflammatory
cytokine production.5 In an LPS-stimulated human PBMC
assay, examples of the sulfonamide series inhibited IL-6 produc-
tion at about the micromolar level (Tables 2, 3). In line with
results from the BZD series,10 the IC50s observed in the in vitro
binding assays translated to very similar potencies in cellular
readouts of cytokine reduction. The correlation between the
BRD4 FA binding assay and the PBMC IL-6 assay is consistent
with BET inhibition being the mode of action of the cellular
efficacy (Figure 5, r2 = 0.57). For comparison, the mean IC50 of
2 in the IL-6 assay is 400 nM, consistent with its greater
binding assay potency.
After the first draft of this manuscript was submitted, a paper
by Hewings et al. appeared online in this journal describing
BET inhibitors also containing 3,5-dimethyl isoxazole as the
acetyl-lysine mimic.21 The discovery was serendipitous, arising
from screening analogues of a dihydroquinazolinone hit. In the
crystal structure of the complex between BRD4-BD1 and the
dihydroquinazoline, the positioning of the dimethyl isoxazole
group is very similar to that of 3 in BRD2-BD1 reported above.
As with our work, optimization led to phenyl 3-substituted
compounds, the most potent of which was compound 4d of
Hewings et al. (35, Figure 6). The X-ray structure of 35 in
Example sulfonamides were also tested for their ability to
inhibit TNFα release from LPS-stimulated PBMCs. Compound
6a had an IC50 of ∼200 nM in this assay (pIC50 = 6.9, mean of
8 values, standard deviation 0.56). The potency was reduced in
an assay measuring the suppression of TNFα production in
LPS-stimulated human whole blood (5 μM for 6a), probably
reflecting the suboptimal physicochemical characteristics of this
compound.
592
dx.doi.org/10.1021/jm201283q | J. Med. Chem. 2012, 55, 587−596

Journal of Medicinal Chemistry
Article
Figure 4. Bromodomain selectivity of representative compounds 6a and 34 using a thermal shift assay. The average temperature shifts (Tm) in °C
upon addition of 100 μM compound are represented by color-coded spheres as indicated in the key. The thermal shift was virtually identical for
these two molecules and so can be adequately represented by a single summary. The one exception is for the CREBBP domain, where compound 34
showed a 1.5 °C shift and compound 6a gave a < 1 °C shift. Bromodomains not assessed as part of the selectivity panel are shown in gray. Proteins
were expressed as N-terminally His6 or His6-FLAG tagged constructs. Constructs comprising more than one bromodomain are designated as, e.g.,
BD12, denoting that the construct contains two bromodomains BD1 and BD2.
Figure 6. Comparison between the BRD2-BD1 crystal structure of 6a
(green, with BRD2 in yellow) superimposed on that of 3 (cyan). The
BRD4-BD1 complex with 35 is also shown (deep red).21
BRD4 showed that the terminal methyl group of its (R)-
configuration secondary alcohol occupies the WPF shelf.
Unexpectedly, a comparison of this structure with our BRD2
Figure 5. Correlation between 36 sulfonamides in the PBMC IL-6
assay and the BRD4 FA binding assay: pIC50 = log10 IC50 (r2 = 0.57).
593
dx.doi.org/10.1021/jm201283q | J. Med. Chem. 2012, 55, 587−596

Journal of Medicinal Chemistry
Article
exclusively substituted meta to the dimethyl isoxazole. The sulfonamide
products were obtained by addition of the amine under basic con-
ditions with subsequent Boc deprotection with TFA where required.
For route B, 3-bromosulfonamides were prepared from commercially
available bromo-aryl- sulfonyl chlorides and the appropriate amine.
Suzuki coupling of these products with 3,5-dimethyl-4-(4,4,5,5-
tetramethyl-1,3,2-dioxaborolan-2-yl) isoxazole afforded the isoxazole-
aryl sulfonamides 5b, 6b, and 31. Further derivatization was achieved
through alkylation of the phenol 32, formed via boron tribromide
mediated demethylation of 6a, to yield the ethers 33 and 34. Alkyl-
ation of the sulfonamide nitrogen could be achieved through NaOH
deprotonation and subsequent reaction with the appropriate alkyl
halide to yield the tertiary sulfonamides 23b and 24b. See Supporting
Information for full details.
For example, 5b was prepared as follows. Cyclopropylamine (1.21 mL,
17.5 mmol), 5-bromo-2-methoxy benzene sulfonyl chloride (2.00 g,
7.0 mmol), and DIPEA (1.84 mL, 10.5 mmol) were combined as a
solution in DCM (50 mL) and stirred at room temperature under N2
overnight. The reaction mixture was washed with water (50 mL),
dried, and concentrated in vacuo to yield crude desired product.
Purification by flash column chromatography (silica gel, 7−100% ethyl
acetate/cyclohexane) yielded 5-bromo-N-cyclopropyl-2-(methyloxy)-
benzenesulfonamide S3a. S3a (1.94 g, 6.3 mmol), Pd(dppf)Cl2 (463
mg, 0.63 mmol), and potassium carbonate (1.75 g, 12.7 mmol) were
suspended in 1,4-dioxane (14 mL) and water (3.5 mL). The reaction
mixture was heated at 100 °C for 45 min in a microwave reactor.
The mixture was concentrated and ethyl acetate added (100 mL).
The organics were washed with aqueous sodium carbonate solution
(100 mL), dried, and concentrated to yield a brown residue.
Purification by flash column chromatography (silica gel, 12−60%
ethyl acetate/cyclohexane) yielded N-cyclopropyl-5-(3,5-dimethyl-4-
isoxazolyl)-2-(methyloxy) benzenesulfonamide 5b (1.38 g, 4.07 mmol,
64%) as a yellow solid. 1H NMR (400 MHz, Chloroform-d) δ 7.88 (d,
J = 2.5 Hz, 1H), 7.45 (dd, J = 2.5, 8.5 Hz, 1H), 7.14 (d, J = 8.5 Hz,
1H), 5.31 (s, 1H), 4.05 (s, 3H), 2.41 (s, 3H), 2.27 (s, 3H), 2.16 (m,
1H), 0.73 (m, 2H), 0.58 (m, 2H). LCMS: (ES+) 323, retention
time 0.90 min, 97% pure.
complexes reveals that relative to the position of the fragment
3, the substituted phenyl isoxazole of 35 moves some distance
from the acetyl lysine pocket toward the WPF shelf.21 When
the protein backbones are superimposed, the rmsd between the
phenyl 3,5-dimethyl isoxazole core of 35 and that of 3 is 1.5 Å
(Figure 6). This movement toward the WPF shelf suggests that
the secondary alcohol substituent of 35 may not be quite large
enough and that a longer group might allow it to fulfill
the WPF binding interaction while maintaining the optimal
placement of the fragment 3 within the AcK site (as found
for 6a).
It is difficult to say whether this structural difference is
accompanied by a difference in potency because the IC50 value
(and thus ligand efficiencies) of 35 were measured with an
AlphaScreen assay and are not directly comparable with ours.
Hewings et al. reported IC50s for 3 of 34 μM (BRD2) and
84 μM (BRD4), lower than the value of >200 μM measured in
our FA assay. Relative to 3, the optimized secondary alcohol 35
is ∼20-fold more potent, while the cyclopentyl sulfonamide 6a
is over 100-fold more potent. Interestingly, Hewings et al. also
reported that some dimethyl isoxazoles can bind to CREBBP,
which could be consistent with the marginally significant Tm
shift seen for 34 (Figure 4).
CONCLUSION
■
These are not the first bromodomain inhibitors reported, but
for the first time we have demonstrated the process of X-ray
structure-guided optimization of a bromodomain fragment into
a series of drug-like molecules with anti-inflammatory activity.
Crystallography has subsequently validated the predictions
made during each step of the design process. We conclude that
the BET bromodomains are highly tractable both to the FBDD
approach to lead discovery and to structure-based rational opti-
mization of the resulting fragments. Because of the amenability
of the wider bromodomain target class to crystallographic
studies, it is likely that similar approaches will be applicable
to other members. The discovery that drug-like inhibitors of
this histone reader module can be found in this way presents
exciting new opportunities for targeted intervention spanning
multiple disease areas.
Binding Assays and Fragment Screening. The tandem
bromodomain constructs of BRD2, BRD3, and BRD4 were screened
in two formats.10 The FA (fluorescence anisotropy) assay mea-
sures displacement of a labeled analogue of 1 and 2, while the TR-
FRET (time-resolved fluorescence resonance energy transfer) assay
measures displacement of a tetra-acetylated biotinylated peptide.
The set of fragments was constructed and screened as previously
described.7
IL-6 and TNF-α PBMC Assays. Peripheral blood mononuclear
cells (PBMCs) purified from whole blood were incubated with LPS
and varying concentrations of test compounds for 18−24 h. Inhibi-
tion of IL-6 or TNF-α was assessed as described in Supporting
Information.
EXPERIMENTAL SECTION
■
General Experimental. All solvents were purchased from Fisher
Scientific Ltd. and commercially available reagents were used as
received. All reactions were followed by TLC analysis (TLC plates
GF254, Merck) or LCMS (liquid chromatography mass spectrometry)
using a Waters ZQ instrument. All microwave reactions were carried
out in a Biotage Initiator 60 focused beam microwave (400 W).
Surface Plasmon Resonance. Binding experiments with BRD4-
BD12 and BRD2-BD12 were conducted at 25 °C on at least three
occasions on a T200 BIAcore instrument. In all cases, a CM5 chip
with amine coupled protein was used. Flow cells of a CM5 chip were
first activated with 0.2 M N-ethyl-N′-(diethylaminopropyl)-carbodii-
mide (EDC) and 0.05 M N-hydroxysuccimide (NHS). Typically
0.1−2 mg/mL BRD4-BD12/BRD2-BD12 at pH 4.5−5.5 in NaAc was
then injected at 10 μL/min for 1−2 min, resulting in >5kRU of
protein immobilized on the surface. The surface was neutralized
with ethanolamine and extensively washed in the running buffer
50 mM HEPES pH7.5, 150 mM NaCl. Compounds were titrated as a
tripling dilution starting at 10−20 μM. Sensorgrams and equilibrium
binding curves were analyzed with BIAevaluation (GE Healthcare)
using a 1:1 binding model. Parameters were derived from an equili-
brium binding analysis. The equilibrium Kd was calculating using a 1:1
binding model, response = concentration × Rmax/(Concentration +
Kd) + offset.
1
Purity of all compounds was determined by LCMS and H NMR.
NMR spectra were recorded on a Bruker AVANCE 400 spectrometer
or a Bruker Ultrashield 400 MHz instrument. Abbreviations for
multiplicities observed in NMR spectra: s; singlet; br s, broad singlet;
d, doublet; t, triplet; q, quadruplet; p, pentuplet; spt, septuplet; m,
multiplet. Column chromatography was performed on prepacked silica
gel columns (30−90 mesh, IST) using a Biotage SP4. Mass spectra
were recorded on a Waters ZQ (ESI-MS) spectrometer. The UPLC
analysis was conducted on an Acquity UPLC BEH C18 column
(50 mm × 2.1 mm, i.d. 1.7 μm packing diameter) at 40 °C. See Supporting
Information for full details.
Synthetic Procedures. Compounds 5a−17a, 6b−15b, 20a, 22a,
25a, 26a, 28a, and 28b were purchased. Other compounds were
prepared by one of two routes (Scheme 1). For route A, Suzuki
coupling of 4-iodo-3,5-dimethyl isoxazole with the appropriate aryl
boronic acid yielded 4-aryl 3,5-dimethyl isoxazoles. Treatment of these
with chlorosulfonic acid and PCl5 gave the desired sulfonyl chlorides
Thermal Shift Assay (Tm). Thermal shifts of bromodomains
marked in Figure 4 were analyzed on a Bioneer RT-PCR instrument.
The temperature of the protein sample was ramped from 4 to 94 °C in
594
dx.doi.org/10.1021/jm201283q | J. Med. Chem. 2012, 55, 587−596

Journal of Medicinal Chemistry
Article
a
Scheme 1. Synthetic Routes
a
(A) R1 = H or OMe. (a) R1B(OH)2, Pd(PPh3)4, Na2CO3 (aq), DME; (b) ClSO3H, PCl5, DCM; (c) (i) NHR2R3, NEtiPr2, DCM; (ii) TFA,
DCM. (B) (d) NHR2R3, NEtiPr2, DCM; (e) Pd(dppf)Cl2, K2CO3, 1,4-dioxane/water (4/1). (C) (f) BBr3, DCM; (g) alcohol, PPh3, DEAD, THF;
(h) NaH, R5-I, DMF. For full details, see Supporting Information.
the presence or absence of 100 μM of each compound in at least dupli-
cate. Denaturation was visualized using 1:1000 dilution of a fluorescent
Sypro orange dye in a buffer of PBS, 10% glycerol. The temperature
was ramped at a rate of 1 °C/min and fluorescence readings taken
every 0.5 °C. Bromodomains tested included ATAD2B, BAZ2B, BPTF,
BRD1, BRD2 (bromodomain 1 and tandem), BRD3 (tandem), BRD4
(bromodomain 1 and tandem), CREBBP, PB1 (bromodomain 4),
PHIP (bromodomain 2), and isoform B of TRIM33.
X-ray Crystallography. His-tagged BRD2-BD1 (67−200) was
expressed in E. coli, purified, and crystallized. Compounds were soaked
into apo crystals prior to crystal freezing and data collected as
described in Supporting Information. Data collection and refinement
statistics are given in Supporting Information Figure S1 and OMIT
(Fo − Fc) maps in Figure 2.
Pharmacophore Design. The 3D pharmacophore model was
constructed based on X-ray complexes of BRD2-BD1 with 1 and 3.
Building on findings that structurally distinct inhibitors share common
steric and electrostatic features at the bottom of the AcK-pocket,7
these were modeled using the structure of 3 bound into BRD2. To
model the interaction with either the conserved asparagine Asn156 or
the conserved water molecule near Tyr113, a hydrogen-bond acceptor
pharmacophore feature was placed on the isoxazole nitrogen. A methyl
feature was placed upon the isoxazole 3-methyl to model the presence
of a methyl group near Phe99. Another interaction with the lipo-
philic WPF shelf bordered by the WPF motif (BRD2 97−99), the
“gatekeeper” Ile162, Met165, and Asp161 was modeled with a lipo-
philic feature near the centroid of the pendant phenyl ring of 2.
Pharmacophore building and searching was carried out using MOE.22
Quantum Mechanical Calculations. Gas phase quantum
mechanical calculations were performed at the B3LYP/6-31G** level
using Jaguar.23
ASSOCIATED CONTENT
■
S
* Supporting Information
Supplementary methods and chemistry; X-ray statistics;
complete screening data. This material is available free of charge
via the Internet at http://pubs.acs.org.
Accession Codes
Coordinates have been deposited with the Protein Data Bank
(4a9m, 4a9n, 4a9o, 4a9p) and will be released immediately on
publication.
595
dx.doi.org/10.1021/jm201283q | J. Med. Chem. 2012, 55, 587−596

Journal of Medicinal Chemistry
Article
Uings, I. J.; Toum, J.; Clement, C. A.; Boullay, A. B.; Grimley, R. L.;
Blandel, F. M.; Prinjha, R. K.; Lee, K.; Kirilovsky, J.; Nicodeme, E.
Discovery and Characterization of Small Molecule Inhibitors of the
BET Family Bromodomains. J. Med. Chem. 2011, 54, 3827−3838.
(11) The Structural Genomics Consortium. http://www.thesgc.org,
2011.
AUTHOR INFORMATION
Corresponding Author
*Phone: +44 (0)20 8047 5000. Fax: +44 1438 763352. E-mail:
Paul.a.Bamborough@gsk.com.
■
Author Contributions
∥These authors contributed equally to this work.
(12) Wells, J. A.; McClendon, C. L. Reaching for high-hanging fruit
in drug discovery at protein-protein interfaces. Nature 2007, 450,
1001−1009.
ACKNOWLEDGMENTS
■
(13) Domling, A. Small molecular weight protein−protein interaction
̈
We thank Stuart Baddeley, Matt Lindon, Peter Soden, Iain Uings,
all other members of the bromodomain teams, the Apo-A1 team
in GlaxoSmithKline, Les Ulis, Jason Witherington, Kevin Lee, and
Epinova DPU, and the departments of Biological Reagents &
Assay Development and Screening & Compound Profiling.
antagonists: an insurmountable challenge? Curr. Opin. Chem. Biol. 2008,
12, 281−291.
(14) Meireles, L. M.; Mustata, G. Discovery of modulators of protein−
protein interactions: current approaches and limitations. Curr. Top. Med.
Chem. 2011, 11, 248−257.
(15) Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J.
Experimental and computational approaches to estimate solubiltiy and
permeability in drug discovery and development settings. Adv. Drug
Delivery Rev. 1997, 23, 3−25.
ABBREVIATIONS USED
■
AcK, acetyl-lysine (AcK); ApoA1, apolipoprotein A-1; ATAD2B,
ATPase family AAA domain containing 2B; BAZ2B, bromodo-
main adjacent to zinc finger domain 2B; BET, bromodomain
and extra-terminal; BD, bromodomain; BPTF, bromodomain
PHD finger transcription factor; BRD1/2/3/4, bromodomain-
containing protein 1/2/3/4; COX2, cyclooxygenase-2;
CREBBP, cAMP response element-binding binding protein;
HSQC, heteronuclear single-quantum correlation spectroscopy;
IL-6, interleukin-6; LPS, lipopolysaccharide; NUT, nuclear
protein in testis; PB1, polybromo 1; PCAF, P300/CBP-
associated factor; PHIP, pleckstrin homology domain interact-
ing protein; TRIM33, tripartite motif-containing 33; TR-FRET,
time-resolved fluorescence resonance energy transfer
(16) Morelli, X.; Bourgeas, R.; Roche, P. Chemical and structural
lessons from recent successes in protein−protein interaction inhibition
(2P2I). Curr. Opin. Chem. Biol. 2011, 15, 475−481.
(17) Bouillot, A.-M.; Donche, F.; Gellibert, F.; Lamotte, Y.; Mirguet,
O. Imidazo[4,5-c]quinoline derivatives as bromodomain inhibitors
and their preparation and use in the treatment of cancer, chronic
autoimmune and inflammatory disease. WO2011054846, 2011.
(18) Dawson, M. A.; Prinjha, R. K.; Dittmann, A.; Giotopoulos, G.;
Bantscheff, M.; Chan, W. I.; Robson, S. C.; Chung, C. W.; Hopf, C.;
Savitski, M. M.; Huthmacher, C.; Gudgin, E.; Lugo, D.; Beinke, S.;
Chapman, T. D.; Roberts, E. J.; Soden, P. E.; Auger, K. R.; Mirguet,
O.; Doehner, K.; Delwel, R.; Burnett, A. K.; Jeffrey, P.; Drewes, G.;
Lee, K.; Huntly, B. J.; Kouzarides, T. Inhibition of BET recruitment to
chromatin as an effective treatment for MLL-fusion leukaemia. Nature
2011, 478, 529−533.
REFERENCES
■
(19) Mirguet, O.; Lamotte, Y.; Donche, F.; Toum, J.; Gellibert, F.;
Bouillot, A.; Gosmini, R.; Nguyen, V. L.; Delannee, D.; Seal, J.;
́
Blandel, F.; Boullay, A. B.; Boursier, E.; Martin, S.; Brusq, J. M.; Krysa,
G.; Riou, A.; Tellier, R.; Costaz, A.; Huet, P.; Dudit, Y.; Trottet, L.;
Nicodeme, E. From ApoA1 upregulation to BET family Bromodomain
inhibition: Discovery of I-BET151 (GSK1210151A). Submitted for
publication.
(20) Lamotte, Y.; Donche, F.; Bouillot, A.; Mirguet, O.; Gellibert, F.;
Nicodeme, E.; Krysa, G.; Beinke, S.; McCleary, S.; Rioja, I.; Bamborough,
P.; Chung, C.; Gordon, L.; Lewis, A.; Walker, A.; Cutler, L.; Lugo, J.; Seal,
J. Identification of a novel series of BET family Bromodomain inhibitors:
binding mode and profile of IBET-151 (GSK1210151A). Submitted for
publication.
(1) Sanchez, R.; Zhou, M. M. The role of human bromodomains in
chromatin biology and gene transcription. Curr. Opin. Drug Discovery
Dev. 2009, 12, 659−665.
(2) Denis, G. V. Bromodomain coactivators in cancer, obesity, type 2
diabetes, and inflammation. Discoveries. Med. 2010, 10, 489−499.
(3) Denis, G. V.; Nikolajczyk, B. S.; Schnitzler, G. R. An emerging
role for bromodomain-containing proteins in chromatin regulation and
transcriptional control of adipogenesis. FEBS Lett. 2010, 584, 3260−
3268.
(4) Florence, B.; Faller, D. V. You bet-cha: a novel family of
transcriptional regulators. Front. Biosci. 2001, 6, D1008−D1018.
(5) Nicodeme, E.; Jeffrey, K. L.; Schaefer, U.; Beinke, S.; Dewell, S.;
Chung, C. W.; Chandwani, R.; Marazzi, I.; Wilson, P.; Coste, H.;
White, J.; Kirilovsky, J.; Rice, C. M.; Lora, J. M.; Prinjha, R. K.; Lee, K.;
Tarakhovsky, A. Suppression of inflammation by a synthetic histone
mimic. Nature 2010, 468, 1119−1123.
(6) Filippakopoulos, P.; Qi, J.; Picaud, S.; Shen, Y.; Smith, W. B.;
Fedorov, O.; Morse, E. M.; Keates, T.; Hickman, T. T.; Felletar, I.;
Philpott, M.; Munro, S.; McKeown, M. R.; Wang, Y.; Christie, A. L.;
West, N.; Cameron, M. J.; Schwartz, B.; Heightman, T. D.; La, T. N.;
French, C. A.; Wiest, O.; Kung, A. L.; Knapp, S.; Bradner, J. E.
Selective inhibition of BET bromodomains. Nature 2010, 468, 1067−
1073.
(21) Hewings, D. S.; Wang, M.; Philpott, M.; Fedorov, O.; Uttarkar,
S.; Filippakopoulos, P.; Picaud, S.; Vuppusetty, C.; Marsden, B.;
Knapp, S.; Conway, S.; Heightman, T. D. 3,5-Dimethylisoxazoles act
as acetyl-lysine-mimetic bromodomain ligands. J. Med. Chem. 2011, 54,
6761−6770.
(22) Molecular Operating Environment (MOE), version 2008.10;
Chemical Computing Group (CCG), 2011.
(23) Jaguar, version 7.8; Schrodinger, LLC, New York, 2011.
(24) Moriniere, J.; Rousseaux, S.; Steuerwald, U.; Soler-Lopez, M.;
Curtet, S.; Vitte, A. L.; Govin, J.; Gaucher, J.; Sadoul, K.; Hart, D. J.;
Krijgsveld, J.; Khochbin, S.; Muller, C. W.; Petosa, C. Cooperative binding
of two acetylation marks on a histone tail by a single bromodomain.
Nature 2009, 461, 664−668.
(7) Chung, C.; Dean, T.; Woolven, J.; Bamborough, P. Fragment-
Based Discovery of Bromodomain Inhibitors Part 1: Inhibitor Binding
Modes and Implications for Lead Discovery. J. Med. Chem. 10.1021/
jm201320w.
(8) Schulz, M. N.; Hubbard, R. E. Recent progress in fragment-based
lead discovery. Curr. Opin. Pharmacol. 2009, 9, 615−621.
(9) Hopkins, A. L.; Groom, C. R.; Alex, A. Ligand efficiency: a useful
metric for lead selection. Drug Discovery Today 2004, 9, 430−431.
(10) Chung, C.; Coste, H.; White, J. H.; Mirguet, O.; Wilde, J.;
Gosmini, R. L.; Delves, C.; Magny, S. M.; Woodward, R.; Hughes, S.
A.; Boursier, E. V.; Flynn, H.; Bouillot, A. M.; Bamborough, P.; Brusq,
J. M.; Gellibert, F. J.; Jones, E. J.; Riou, A. M.; Homes, P.; Martin, S. L.;
596
dx.doi.org/10.1021/jm201283q | J. Med. Chem. 2012, 55, 587−596