Discovery and characterization of small molecule inhibitors of the BET family bromodomains

Article abstract of DOI:10.1021/jm200108t

Epigenetic mechanisms of gene regulation have a profound role in normal development and disease processes. An integral part of this mechanism occurs through lysine acetylation of histone tails which are recognized by bromodomains. While the biological and structural characterization of many bromodomain containing proteins has advanced considerably, the therapeutic tractability of this protein family is only now becoming understood. This paper describes the discovery and molecular characterization of potent (nM) small molecule inhibitors that disrupt the function of the BET family of bromodomains (Brd2, Brd3, and Brd4). By using a combination of phenotypic screening, chemoproteomics, and biophysical studies, we have discovered that the protein-protein interactions between bromodomains and acetylated histones can be antagonized by selective small molecules that bind at the acetylated lysine recognition pocket. X-ray crystal structures of compounds bound into bromodomains of Brd2 and Brd4 elucidate the molecular interactions of binding and explain the precisely defined stereochemistry required for activity.

Full text of DOI:10.1021/jm200108t

ARTICLE
pubs.acs.org/jmc
Discovery and Characterization of Small Molecule Inhibitors of the BET
Family Bromodomains
Chun-wa Chung,*,† Hervꢀe Coste,§ Julia H. White,† Olivier Mirguet,§ Jonathan Wilde,† Romain L. Gosmini,§
Chris Delves,† Sylvie M. Magny,§ Robert Woodward,† Stephen A. Hughes,† Eric V. Boursier,§ Helen Flynn,†
Anne M. Bouillot,§ Paul Bamborough,† Jean-Marie G. Brusq,§ Franc-oise J. Gellibert,§ Emma J. Jones,†
Alizon M. Riou,§ Paul Homes,† Sandrine L. Martin,§ Iain J. Uings,‡ Jꢀer^ome Toum,§ Catherine A. Clꢀement,§
Anne-Bꢀenꢀedicte Boullay,§ Rachel L. Grimley,† Florence M. Blandel,§ Rab K. Prinjha,‡ Kevin Lee,‡
Jorge Kirilovsky,§ and Edwige Nicodeme§
†Molecular Discovery Research, GlaxoSmithKline R&D, Stevenage SG5 4HT, U.K.
‡Epinova, Discovery Performance Unit, GlaxoSmithKline R&D, Stevenage SG5 4HT, U.K.
§Lipid Metabolism Discovery Performance Unit, GlaxoSmithKline R&D, Les Ulis, France
S Supporting Information
b
ABSTRACT: Epigenetic mechanisms of gene regulation have a
profound role in normal development and disease processes. An
integral part of this mechanism occurs through lysine acetyla-
tion of histone tails which are recognized by bromodomains.
While the biological and structural characterization of many
bromodomain containing proteins has advanced considerably,
the therapeutic tractability of this protein family is only now
becoming understood. This paper describes the discovery and
molecular characterization of potent (nM) small molecule
inhibitors that disrupt the function of the BET family of
bromodomains (Brd2, Brd3, and Brd4). By using a combination
of phenotypic screening, chemoproteomics, and biophysical studies, we have discovered that the proteinꢀprotein interactions
between bromodomains and acetylated histones can be antagonized by selective small molecules that bind at the acetylated lysine
recognition pocket. X-ray crystal structures of compounds bound into bromodomains of Brd2 and Brd4 elucidate the molecular
interactions of binding and explain the precisely defined stereochemistry required for activity.
’ INTRODUCTION
assemble into complexes which can drive chromatin remodelling and
result in transcriptional regulation.3
Regulation of chromatin structure is a key mechanism by
which a conserved genome can propagate distinct patterns of
gene transcription and silencing across different cell lineages.
This layer of epigenetic control is an important contributor to
normal development but also plays a major role in the transfor-
mation events leading to tumorogenesis. Moreover, evidence is
gradually emerging that dysregulation of epigenetic processes may
be associated with the development of a range of human diseases,
including cardiovascular disorders and autoimmunity.1 Epigenetic
information is encoded both by direct methylation of DNA at
particular sites in the genome and by a series of post-translational
marks placed on the tails of histone proteins which protrude from
the nucleosome structure. These marks include methylation, acet-
ylation, phosphorylation, and ubiquitination and are generated and
erased by distinct families of enzymes which target a variety of
residues on these histone tails.2 The resulting pattern of modifica-
tions, or epigenetic code, is sensed by proteins that contain discrete
“reader” modules, which are often found in combination to allow a
complexity of permutations to be interpreted. These proteins then
Significant progress has been made in the development of
compounds that inhibit some of the enzymes responsible for
the transfer of these epigenetic marks including the approval of
Vorinostat (SAHA, a histone deacetylase inhibitor) for treat-
ment of T-cell lymphomas.4 However, there are few available
tools to modulate the many additional enzyme classes that
mediate epigenetic changes (e.g., methyltransferases and
demethylases) and fewer still for the wide variety of reader
domains that exist (e.g., PHD, Tudor, Chromodomains,
Bromodomains). Historical attempts to target proteinꢀpro-
tein interactions, such as that described here between a reader
protein and its chromatin partner, with potent and specific
small molecule inhibitors, have proved challenging.5ꢀ7 On this
basis, reader domains have until now been considered less
viable opportunities for drug discovery.
Received: January 31, 2011
Published: May 13, 2011
r
2011 American Chemical Society
3827
dx.doi.org/10.1021/jm200108t J. Med. Chem. 2011, 54, 3827–3838
|

Journal of Medicinal Chemistry
ARTICLE
Figure 1. Luciferase reporter screen identifies small molecule inducers of ApoA1 expression. (a) The human ApoA1 promoter and 30-UTR were cloned
flanking a luciferase reporter gene and stably transfected into HepG2 cells. (b) High throughput screening identified a series of cell active small molecules
and closely related inactive analogues. (c) Active compounds 1 (circles) and 2 (triangles) were able to dose-dependently increase the ApoA1 reporter
gene (440 ( 60 and 700 ( 60 nM), whereas 3 (squares) failed to do so over the same concentration range. (d) All compounds had no effect on an
analogous LDL-R reporter system showing the effects were specific for the ApoA1 gene. (e) The ability of 2 (triangles) to dose dependently increase
Apo-A1 protein production in HepG2 cells in shown. The levels of secreted mature human ApoA1 protein were measured using 35S pulse chase labeling
in cells. The inactive enantiomer 3 has no effect even at the highest concentrations used. (squares) (f) The domain structure of Brd2, Brd3, Brd4, and
BrdT, highlighting the tandem bromodomains BD1 and BD2 and the ET domain.
The use of cell-based reporter assays provides a means to
identify chemical epigenetic modulators, as evidenced by a
number of recent reports.8ꢀ10 An immense challenge for this
approach is that once compounds have been discovered their
molecular target or targets can be difficult to pinpoint, leading to
compounds of unknown mode of action (MOA). This lack of
understanding limits their utility and confers risks associated with
selectivity and safety as well as potential complexities in under-
standing the chemical structure activity relationships (SAR).
Apolipoprotein A1 (ApoA1) upregulation is associated with
protection from atherosclerosis progression and with anti-in-
flammatory effects.11 Despite much effort, nomolecularmechanisms
3828
dx.doi.org/10.1021/jm200108t |J. Med. Chem. 2011, 54, 3827–3838

Journal of Medicinal Chemistry
ARTICLE
Scheme 1. Synthetic Routea
a (i) (a) CHCl3, 60 °C, (b) Et3N, DCM, reflux, (c) AcOH, 1,2-DCE, 60 °C; (ii) Lawesson’s reagent, toluene, reflux; (iii) (a) H2NꢀNH2, THF, 0 °C, (b)
Et3N, AcCl, 0 °C to RT, (c) AcOH, THF, RT; (iv) NaOH, THF, RT; (v) ethylamine, HBTU, DIEA, THF, RT. 2 and 3 were obtained from 10 by chiral
HPLC separation. For full details see Supporting Information, Figure S5a.
have yet been elucidated to suggest how this may be achieved in a
therapeutic manner.12 We therefore employed a “chemical
genetic” approach using a reporter cell line to identify small
molecules that were able to enhance ApoA1 expression. These
upregulators specifically regulate the human ApoA1 gene but had
an unknown mechanism of action.
In this manuscript, we describe the use of chemoproteomics,
siRNA, biophysical assays and X-ray crystallography to elucidate
the molecular target of these compounds and their MOA. The
molecules inhibit the function of the tandem bromodomain
containing family of transcriptional regulators known as the
BET proteins13 (Brd2, Brd3, Brd4). They bind to the acetyl-
lysine (AcK) recognition pocket and directly antagonize the
interaction between the bromodomains and acetylated histone
peptides. The liganded crystal structures presented here offer
insights into how selectivity for the BET family is achieved and
the basis of their mimicry of the native AcK peptide. This work
has identified a novel class of chemical tools that can be used to
modulate epigenetic processes and suggests that the proteinꢀ
protein interactions mediated by bromodomains can be targeted
therapeutically using small molecule antagonists.
was a single enantiomer but of undefined chirality at the 4-position
of the benzodiazepine (BZD) ring. To determine whether this
was general transcriptional upregulation or a specific effect on the
ApoA1 locus, we tested the compound’s effect on an LDL-R
promoter reporter construct in the same HepG2cells. Compound
1 had very little effect on LDL-R luciferase activity at the con-
centrations at which it induces ApoA1 expression (Figure 1d),
suggesting that the effect is indeed specific.
A program of medicinal chemistry was carried out to optimize
the potency of 1 in the ApoA1 upregulation assay in the absence
of any knowledge of the molecular target.14 While this is outside
the scope of this manuscript, this work yielded much useful SAR,
some of which it is now possible to rationalize using the X-ray
structures presented here. For example, we found that the
benzodiazepine core was essential for activity and the aryl group
extending from the 6-position on the BZD ring was common to
all active analogues.14 A variety of groups projecting from the
4-position of the BZD core were found to be well tolerated and
could be used to modify the physicochemical and pharmacoki-
netic properties of the molecules. This led to the synthesis of
GSK525762A15 (2, Figure 1b; Scheme 1), which shows similar
potency and specificity to 1 in the reporter gene systems
(Figure 1c, d; EC50 700 nM) but whose properties make it more
suitable for in vivo experiments.14,15 Intriguingly, activity was
found to depend critically upon the stereochemistry at the
4-position. Thus, the R-enantiomer of 2, GSK525768A15 (3,
Figure1b), is unable to alter reporter gene activity (Figure 1c).
The activity of the BZDs was not limited to reporter gene assays, as
2 displayed a similar upregulation of endogenous ApoA1 expres-
sion levels in HepG2 cells (as quantified by 35S pulse chase labeling
of secreted protein in HepG2 cells). In these studies, 3 was also
found to be inactive (Figure 1e). This dependence on stereochem-
istry strongly implied that the observed activity resulted from
specific interactions with defined but unknown molecular targets.
To try to identify these targets, a panel of structurally related
compounds that spanned a full range of activities in the luciferase
’ RESULTS
Discovery of Small Molecule Upregulators of ApoA1 with
Potent and Specific Effects. Efforts to target endogenous
upregulation of ApoA1 expression are hampered by the absence
of an obvious molecular mechanism around which to build a drug
discovery effort. In 2001, we therefore generated a stable human
HepG2 hepatocyte cell-line containing an ApoA1 luciferase
reporter (Figure 1a) and used this to screen compounds in order
to identify molecules with the ability to upregulate reporter gene
activity. From this effort and with some minimal medicinal
chemistry optimization,14 we identified the benzodiazepine (BZD)
GW841819X15 (1, Figure 1b), which shows potent induction of
the ApoA1 reporter gene with an EC50 of 440 nM and a
maximum induction of 4.5-fold (Figure 1c). This compound
3829
dx.doi.org/10.1021/jm200108t |J. Med. Chem. 2011, 54, 3827–3838

Journal of Medicinal Chemistry
ARTICLE
assay (from nM to inactive) were profiled against panels of
kinases, ion channels, nuclear receptors, GPCRs, and other drug-
target protein classes.15 This exercise did not provide any insights
into the mode of action of these compounds beyond suggesting
that these compounds may act through a novel class of targets
not represented in our panels.
proteins, biophysical and structural methods were employed.
Initially, the N-terminal portions of Brd2, Brd3 and Brd4 containing
both bromodomains (BD12) were expressed in E. coli and purified
to homogeneity (Supporting Information, Figure 5i (S22)). Iso-
thermal titration calorimetry (ITC) experiments confirmed a direct
and specific high affinity (19ꢀ80 nM) interaction of 1 with the
tandem bromodomains of the BET family (Figure 2a, Supporting
Information, Figure 5ii (S23)). Interestingly, it appears that two
molecules of 1 bind to each molecule of the BET BD12 protein.
This unexpected stoichiometry indicates that compounds may bind
to a shared motif within each bromodomain and can bind
simultaneously to multiple bromodomains when present in a single
protein. To confirm this, constructs expressing each bromodomain
of Brd2 were generated separately and the proteins purified
(Supporting Information, Figure 5i (S22)). Compound 1 bound
to both the individual BD1 and BD2 domains with a 1:1 stoichio-
metry with very similar affinities, 46 and 52.5 nM, respectively
(Figure 2b,c). The thermodynamic parameters of this binding are
given in Figure 2d. This provides compelling evidence that the BZD
compounds recognize a common feature of the two bromodomains.
The behavior of the single bromodomains of Brd2 was further
investigated using surface plasmon resonance (SPR) at 25 °C.
Titrations of 1 over immobilized Brd2-BD1 (Figure 2e) and
Brd2-BD2 (Figure 2f) gave dissociation constants of 220 and
54 nM, respectively. These affinities are broadly consistent with
the ITC findings and therefore support the observation of specific
interaction of 1 with BET bromodomains using an orthogonal
method. The approximately 5-fold weaker affinity to Brd-BD1 at
25 °C measured by SPR (220 nM) compared to that found at
16 °C by ITC (46 nM) is believed to be a real and significant
difference and due to a strong temperature dependence in the KD for
BD1 binding. This was also observed during the ITC experiments
and led to the Brd2-BD1 titration being carried out at 16 °C rather
than 26 °C as a marked reduction in binding enthalpy and affinity
was seen at 26 °C (Supporting Information, Figure 5ii (S23)).
The binding kinetics of 1 to Brd2-BD1 are also noticeably
more rapid than seen for Brd2-BD2 (Figure 2e,f). It is tempting
to speculate that these differences may translate to the possibility
of sequential and cooperative binding to the individual sites in
tandem domain constructs. However, our experience and the
emerging literature on other epigenetic targets suggests that the
presence or absence of adjacent domains can have a profound
influence on the activity, kinetics, specificity, and function of
truncated proteins18 and it may be difficult to extrapolate the
behavior of truncates to full length proteins with high fidelity.
As bromodomains are known to bind to lysine-acetylated
histone peptides and to function as transcriptional regulators,19,20
we tested the ability of BZD compounds to modulate the interac-
tion of a tetra-acetylated lysine histone 4 peptide (H4AcK4) with
recombinant tandem constructs of Brd2, 3, and 4 in a FRET assay
(Figure 2g,i). The active exemplar compounds exhibit dose
dependent inhibition of peptide binding to the Brd2, 3, and 4
proteins with IC50s of between 15 and 30 nM (Figure 2j). These
compounds therefore function as antagonists of the BET proteins’
role as sensors of histone acetylation.
BZD Compounds Bind to the Tandem Bromodomains of
Brd2,3,4 of the BET Protein Family. Recognizing that the target
for our compounds was probably not among the established drug
classes, we embarked upon a chemoproteomics approach.
Briefly, we generated an affinity matrix by linking a derivatized
version of 2 to a ReactiGel matrix (“active BZD matrix”, see
Supporting Information, Section 4). A control matrix using an
inactive BZD analogue was also generated. We then performed
affinity chromatography by passing lysates from HepG2 cells
over the affinity matrix, followed by extensive washing. A con-
sistent and defined pattern of proteins were retained by the active
BZD matrix but were absent when the inactive BZD matrix or matrix
alone was used (Supporting Information, Figure 4i,ii (S7ꢀS10)).
The binding of these proteins to the active BZD matrix could be
competed away by the addition of excess 1 (100 μM) but was not
altered by the inclusion of DMSO or an inactive analogue. Thus it
appeared that the proteins retained on the active matrix represent
specific protein targets of the BZD compounds. Mass spectro-
metric analysis of all the proteins retained by the active matrix by
LC/MS/MS reproducibly identified them as fragment peptides
from just three proteins, namely Brd2, Brd3, and Brd4, with
sufficient coverage to allow definitive identification (see Sup-
porting Information, Section 4 (S7ꢀS19)).
These proteins belong to the BET family of transcriptional
coregulators, which share a characteristic architecture of a
tandem bromodomain motif at the N-terminus and an “extra-
terminal” (ET) domain at the C-terminus (Figure 1f). A fourth
member of this family, BrdT, has also been previously described,
but whereas Brd2, Brd3, and Brd4 are all ubiquitously expressed,
BrdT is only expressed in testis16 and is therefore not present in
HepG2 cells. No other proteins were detected on the gels,
implying that the BET family of proteins may be the molecular
targets for the BZD compounds.
The N and C-terminal regions of BET proteins have distinct
functions.13,17 The N-terminal tandem bromodomains (BD12)
recognize patterns of acetylated lysines on histone tails, allowing
the specific localization of the BET proteins to defined areas of
chromatin. The C-terminal ET domain is thought to be involved
in scaffolding the multiprotein complexes associated with the
BET proteins to bring about transcriptional regulation. To define
more precisely the region(s) of the BET proteins involved in
compound binding, Flag-tagged Brd2 cDNA constructs of full
length Brd2(1ꢀ801) or N(1ꢀ473) and C(473ꢀ801) terminal
truncates (Supporting Information, Figure 4i(b)) were gener-
ated and transiently transfected into HepG2 cells. Although all
three proteins could be recovered by immunoprecipiation using
an anti-Flag antibody, affinity purification of the same lysates
using the active BZD matrix recovered only the full length and
N-terminal region of Brd2 and not the C-terminal domain construct
(Supporting Information, Figure 4i(c) (S7ꢀS10)). Therefore,
the BZD compounds interact with the BET proteins via their
bromodomain-containing N-terminal region.
Crystallography Shows that BZDs are Orthosteric Antago-
nists Acting at the AcK Recognition Site. To determine
whether compounds act by direct steric competition with the
H4AcK4 peptide or by an indirect allosteric mechanism, struc-
tures of BZD compounds complexed with BD1 and BD2 single
bromodomains of the BET family were determined by X-ray
crystallography(seeSupportingInformation, Section7). Figure 3
BZD Compounds Bind Stoichiometrically to Each BET
Bromodomain with nM Potency and Can Bind the Tandem
Bromodomain at Two Sites Simultaneously. To determine
the specificity and nature of the BZD binding to the BET
3830
dx.doi.org/10.1021/jm200108t |J. Med. Chem. 2011, 54, 3827–3838

Journal of Medicinal Chemistry
ARTICLE
Figure 2. Biophysical data for BZD binding to BET proteins. Isothermal calorimetric titration curves and analysis for the titration of solutions of
(a) Brd2-BD12, (b) Brd2-BD1, and (c) Brd2-BD2 into compound 1, demonstrating stoichiometric binding of one compound per bromodomain
present in the protein. (d) The thermodynamic parameters of 1 binding to the Brd2 constructs. SPR sensorgrams of 1 injected over (e) Brd2-BD1
(KD = 220 ( 50 nM (n = 4)). (f) Brd2-BD2 (KD = 54 ( 14 nM (n = 3)) at tripling dilutions start at 6.7 μM, illustrating the difference in the kinetics of
binding to the two domains. The ability of 1 (black) and 2 (red) to dose-dependently displace a tetra-acetylated H4 peptide from the tandem BET
bromodomains of (g) Brd2, (h) Brd3, and (i) Brd4 in the FRET assay with potencies in the 15ꢀ40 nM range (j).
shows structures of 1 with Brd4-BD1/2, as well as 1 and 2 with
Brd2-BD1.
The BET bromodomains adopt a left-handed four-helix
bundle topology consistent with other members of the bromo-
domain superfamily.21,22 Structures of BET family members in
complex with acetylated lysine-containing peptides have been
determined crystallographically by several groups. These include
Brd2-BD1/BD2,23 Brd4-BD1,24 and BrdT-BD1/BD2.25 These
studies have shown that the acetylated lysine occupies a pocket
formed by the ZA and BC loops at one end of the helical bundle
3831
dx.doi.org/10.1021/jm200108t |J. Med. Chem. 2011, 54, 3827–3838

Journal of Medicinal Chemistry
ARTICLE
Figure 3. Molecular interactions of BZD with BET bromodomains. (a) The four-helix bundle topology of Brd2-BD1 complexed with 1 (orange). The
acetyl-lysine K12Ac histone H4 peptide taken from its complex with Brd2 (PDB entry 2DVQ23) is shown in green. (b) Schematic showing the
interactions between 1, the Brd2-BD1 site and nearby water molecules. (c) Brd2-BD1 (blue) complexed with 1 (orange), superimposed on the
diacetylated K5Ac K8Ac histone H4 peptide (magenta) taken from BrdT-BD1/H4K5AcK8ac complex (PDB entry 2WP225). In this figure and in
Figure 4d, the surface of the protein is shown as a blue mesh. In Figure 4c, a view through the protein enables the shape of the AcK pocket to be clearly
delineated and the conserved waters lining the bottom and their extensive hydrogen bonding network to be visualized. (d) Brd2-BD1 (blue) complexed
with 1 (orange) and 2 (green) showing the similar binding modes of the two compounds. This is a view from above the protein surface down into the
AcK pocket. (e) Superimposed complexes of 1 bound into Brd2-BD1 (blue, with 1 in orange) and Brd4-BD1 (magenta). The two amino acid
substitutions in the shell of residues surrounding the binding site (labeled) do not significantly alter the binding mode of the compound to the BD1
domains of the two proteins. This again is a view from above the protein surface down into the AcK pocket, in an identical manner to 4d and 4f.
(f) Superimposed complexes of 1 bound into Brd4-BD1 (magenta) and Brd4-BD2 (green), viewed from above the AcK pocket, showing the similar
binding modes of the compound to the two bromodomains of Brd4. This is unaffected by the conservative substitutions within the binding site, such as
that of the gatekeeper residue (labeled).
(Figure 3a). The acetyl group on the AcK side chain interacts
with a network of water molecules located at the bottom of the
pocket that participate in a hydrogen-bonding network with the
asparagine and tyrosine residues that are conserved in the
majority of the bromodomains (Figures 3c and 4a). These
residues are essential to the bromodomains’ function of “reading”
the presence of the acetyl-lysine modification.
network, which is maintained with little variation in over a hundred
structures we have determined to date. To explain the features of
compound binding it is instructive to compare the BZD complexes
with the crystal structure of a histone peptide, for example, as in
Figure 3c, which shows an overlay of Brd2-BD1 bromodomain
bound to 1(orange) on the diacetylated H4 peptide (magenta) from
a published structure.25 The methyl group of 1 binds into the small
hydrophobic pocket that is responsible for recognition of the methyl
group of the acetyl headgroup of the acetyl-lysine. This is formed in
The BZD compounds bind to the acetyl-lysine recognition
pocket (Figure 3a, 3b) and interact with the same conserved water
3832
dx.doi.org/10.1021/jm200108t |J. Med. Chem. 2011, 54, 3827–3838

Journal of Medicinal Chemistry
ARTICLE
Figure 4. Basis of BZD selectivity among bromodomains. (a) Multiple sequence alignment of human BET Brd2, 3, 4, and T bromodomains with
GCN5. Highlighted in blue is the I/V to Y sequence difference in the gatekeeper residue at the bottom of the WPF shelf observed between BD1, BD2,
and GCN5, respectively. The WPF motif is highlighted in green, and the conserved tyrosine and asparagine that are key to acetylated lysine recognition
are highlighted in red. (b) The X-ray structure of Brd2-BD1 (blue stick and surface) with 1 (green stick) has been overlaid with GCN5 (PDB entry 1E6I
as yellow stick).25 This illustrates that the change of the gatekeeper residue from isoleucine to tyrosine would likely obstruct the entrance of the WPF
shelf. (c) A surface representation of Brd2-BD1 with 1 (orange stick), overlaid with the acetylated H4 peptide from the PDB entry 2DVQ shown in
green. The accessible areas for ligand binding outside of the conserved acetylated lysine (AcK) binding site are mapped out in this diagram. The
N-terminal portion of the peptide binds in the N-side before the AcK and then exits to make interactions along the C-side. 1 also accesses the N-side but
explores additional interactions in the WPF shelf and the ZA channel.
large part by the side chain of Phe99 in Brd2-BD1, which is perfectly
conserved in the other BET bromodomains (Figure 4a).
good shape complementarity which may in part explain the high
binding potency of these compounds.
The two nitrogen atoms of the 1,2,4-triazolyl ring of the BZD
together mimic the carboxyl group of the acetyl headgroup. One
ring nitrogen (N1) accepts a hydrogen-bond from the water molecule
that bridges to the hydroxyl group of the side chain of the conserved
tyrosine residue (position 113 in Brd2-BD1). The second nitrogen
(N2) accepts a hydrogen bond from the side chain NH2 group of the
conserved asparagine (position 156 in Brd2-BD1). The ability of a
triazolyl ring to mimic the acetyl group in this way was unexpected,
but this discovery was important because it suggested for the first time
that other diverse chemotypes can act as antagonists of the bromo-
domain/AcK peptide interaction.
The fused phenyl ring of the benzodiazepine core of 1 extends
beyond the region occupied by the histone peptide, binding into
a lipophilic pocket we termed the ZA channel (Figure 4b),
sandwiched between side chains of the long ZA loop (notably
Pro98 and Leu108 in Brd2-BD1). The curvature of the benzo-
diazepine is well matched to the shape of the channel, providing
The pendent phenyl group at the 6-position of the benzodia-
zepine interacts with another hydrophobic region of the BC loop
that we have named the WPF shelf (because it includes the
conserved WPF motif present in all of the BET family bromo-
domains), especially residues Trp97 and Pro98 of Brd2-BD1.
Other Brd2-BD1 residues that make up the WPF shelf include
Ile162 and Met165. While this manuscript was in preparation, a
report was published that described the X-ray structures of BD1
and BD2 bromodomains from BrdT in complex with histone
peptides from H4 and H3, respectively.25 One key finding of
Moriniere et al.25 was the binding mode of a diacetylated histone
H4 tail to the single bromodomain of BrdT-BD1, in which the
established AcK pocket binds the first H4AcK mark at position 5.
Interestingly, the second AcK at position 8 lies within the WPF
shelf region of the binding site. The CγꢀCε atoms of the H4AcK8
side chain overlay closely with the positions of the pendant
phenyl ring of the BZD compounds (Figure 3c). This reinforces
3833
dx.doi.org/10.1021/jm200108t |J. Med. Chem. 2011, 54, 3827–3838

Journal of Medicinal Chemistry
ARTICLE
our hypothesis that occupation of the WPF shelf and surround-
ing regions of the site are critical for binding and for governing
selectivity. It seems that the BZD compounds serendipitously
identified an important hotspot for specific bromodomain
recognition.
The crystal structure clearly shows that the R-enantiomer of 1
binds to the bromodomain. This stereocenter is the position
from which the benzyl carbamate moiety projects and which was
found to be key for Apo-A1 activity. This carbamate group is
mobile, as evidenced by its high B-factors and variable positions
in different molecules within the asymmetric unit, but it extends
into the volume that is occupied by the peptide chain of the AcK
(or the “N-side”) (Figure 3c). The NH group of the carbamate
acts as an additional hydrogen-bond donor to the side chain of
Asn156, which is an interaction that is not made by the AcK peptide.
This partially explains the reduced activity of one enantiomer
of the BZD compounds, which would be unable to make this
interaction. However, a more significant reason for their inability
to bind is that they would clash sterically with the side of the
pocket, particularly the side chains of Leu108 and Leu110.
Other members of the BZD class of compounds bind to Brd2-
BD1 with essentially identical binding modes for their BZD core,
as shown for 2 bound to Brd2-BD1. The superimposition of 1
and 2 is shown in a different orientation in Figure 3d. The change
from the carbamate linker in the 4-position of 1 to an amide in 2
results in the loss of a direct H-bond with Asn156, but this is
replaced with a water mediated interaction between the amide
and N156.
To clarify the way in which these compounds bind to other
members of the BET family, crystal structure complexes with
these bromodomains were also determined. For example, the
structure of 1 with Brd4-BD1 (Figure 3e) shows essentially
identical binding interactions to Brd2-BD1. This is not surpris-
ing, given the relatively high conservation of the active sites
between the BD1 domains of the BET family. The different
positions of the benzyl carbamate substituent visible in Figure 3e
reflect its mobility, consistent with the ability to tolerate mod-
ification at this position as discussed above. The second BD2
bromodomains are more distantly related to the BD1s
(Figure 4a). Despite this, BZD compounds bind to the BD2
domains in a similar manner, as is shown by the complex between
1 and Brd4-BD2 (Figure 3f). This observation is consistent with
the similar experimental binding affinities described earlier.
The complex structures explain the SAR observed within the
BZD series and rationalize the potent nM affinity of these
compounds, as well as the tight requirement for the appropriate
enantiomer within the series. They also elucidate the mode of
action of these compounds: they directly antagonize the pivotal
recognition of acetylated lysine peptides by BET family bromo-
domains and thus influence transcriptional regulation and chro-
matin remodelling.
Figure 5. Cellular ApoA1 upregulation correlates with in vitro binding
affinity to BET proteins. Plot of mean Apo-A1 HepG2 upregulation
pEC170 against Brd4 fluorescence anisotropy pIC50, for 150 BZD
compounds tested in both assays, showing excellent correlation. Data
for Brd2 and Brd3 are shown in Supporting Information, Figure 6i
(S31).
cant increase in ApoA1 protein. (Supporting Information,
Figure 4i (S7), S20ꢀS21). These data clearly suggest that
upregulation of ApoA1 can be driven by disrupting the bromo-
domain binding function of Brd4. However, as only mRNA was
monitored the roles of Brd2 and Brd3 in Apo-A1 regulation
cannot be ruled out, either individually or in combination with
each other and Brd4.
Cellular ApoA1 Upregulation Strongly Correlates with in
Vitro Affinity to BET Proteins. The knowledge that 2 is a potent
ligand for the BET bromodomains, and awareness of the SAR
surrounding this ligand allowed an active fluorescent derivative to
be made and a fluorescent anisotropy (FA) displacement assay to
be established (Supporting Information, Section 6 (S31ꢀS33)).
The results of profiling 150 compounds from the benzodiazepine
series with a broad range of ApoA1 activities in this in vitro binding
assay are given in Figure 5 (also see Supporting Information,
Figure S6i (S31)). The correlation between Brd2, 3, and 4 binding
affinity and ApoA1 upregulation is excellent (R = 0.75, 0.76, and
0.78 respectively). This further supports the siRNA findings that
BET-histone inhibition results in ApoA1 upregulation.
’ DISCUSSION AND CONCLUSION
In this paper, we described how we used a reporter assay to
identify a novel class of ApoA1 upregulators (BZDs) and
epigenetic “reader” inhibitors. The mode of action of these
compounds has been deconvoluted by chemoproteomics and
biophysical experiments to demonstrate unequivocally that they
act as direct functional antagonists of AcK histone binding to the
BET family of bromodomains. Our compounds are pan-BET
active, and therefore we cannot specify whether inhibition of one
or all of these proteins are required for the Apo-A1 activation for
which they were originally discovered. However, the siRNA
studies suggests at least a previously unknown association of
Brd4 with ApoA1 regulation.
While significant progress has been made in the area of
epigenetics through the use of gene knockout mice, siRNA
knockdown, or overexpression of domain deletions, these studies
are generally unable to define the contribution of each individual
domain to the overall function of the protein. The BZD
ApoA1 Upregulation Can Be Induced by siRNA Knock-
down of Brd4. Our experiments established the BZD com-
pounds as inhibitors of the BET proteins, but to confirm that this
function was responsible for the upregulation of ApoA1, we
conducted siRNA studies. As previously observed, treatment of
HepG2 cells with 1 μM 1 was able to induce ApoA1 mRNA
expression by approximately 3-fold (Figure 1e and Supporting
Information (S6ꢀS8, S20ꢀS21)). Treatment of HepG2 cells
with siRNA designed to give specific knockdown of Brd2 and
Brd3 had no effect on ApoA1 expression, whereas progressive
knockdown of Brd4 expression resulted in a gradual and signifi-
3834
dx.doi.org/10.1021/jm200108t |J. Med. Chem. 2011, 54, 3827–3838

Journal of Medicinal Chemistry
ARTICLE
compounds provide a unique opportunity to do so for BET
bromodomains.
“reader” domains in control of epigenetic processes in a way not
currently possible using siRNA or knockout strategies. The
knowledge generated using such compounds will undoubtedly
unravel many of the complex processes regulating transcription
in health and disease.
In conclusion we present the discovery of a class of potent
compounds with a novel epigenetic mechanism, extending the
druggable genome to include bromodomains.
There have been some notable previous attempts to find small
molecule inhibitors of bromodomains6,7 using a structure-based
design approach, especially targeting the bromodomains of PCAF
and CREBBP. These efforts have successful identified compounds
of affinities in the μM or greater range. While these affinities are
comparable to those often measured between isolated bromodo-
mains and their histone partners, the ability to use micromolar
compounds in chemical biological studies can be limited. The BZD
compounds reported here represent a novel chemical template,
distinct from the previously reported simple acetyl containing
templates, with clear mode of action, and demonstrate the ability
to produce inhibitors with high (nM) affinity and cell permeability.
ThebiologicalroleandtherapeuticpotentialoftheBETproteins
is beginning to emerge in the literature. Key observations of their
importance in viral segregation26 and B-cell lymphomagenesis27
suggest that our compounds may have potential in these antiviral
and oncology indications, respectively. Two recent publications
also highlight the potential benefits of BET BZD inhibitors for both
anti-inflammatory diseases and a specific midline carcinoma.15,28
The BZDs bind to both BD1 and BD2 bromodomains of the
BET family with comparable affinity. However, the affinity
purification studies identified only the BET family of proteins
as BZD binders despite the likely presence of many other
bromodomain containing proteins in cell lysates. We hypothe-
size that BET family specificity is achieved by making distinct
contacts outside the acetyl-lysine cavity with the BC and ZA
loops, which are known to differ significantly in both composi-
tion and length between bromodomain containing proteins.21
This hypothesis is strengthened by the inability to detect an
interaction between 1 and two bromodomains, which differ in
the length and characteristics of the ZA and BC loops, CREBBP
and ATAD2 (Figure 4a), suggesting affinities in excess of 50 μM
(Supporting Information, Figure 5iv,v (S25ꢀS26)). This is
supported by recent data on related compounds and their
selectivity for the BET proteins.15,28
In contrast, differences in these loop regions between the BET
domains are relatively minor and are tolerated by the BZD
compounds. For example, BD1 of Brd2, 3, and 4 have an
isoleucine at the position analogous to residue 162 in Brd2. This
“gatekeeper” residue is located at the entrance to the WPF shelf.
The BD2 domains of the BET family have a conservative
replacement of this residue by valine (as shown in Figure 3f
and Figure 4a) which is also well tolerated. CREBBP and ATAD2
also have small gatekeeper residues of valine and isoleucine,
respectively, but this is not the case for all bromodomains as the
structure of closely related bromodomain such as GCN529
(Figure 4a) shows. In this bromodomain, the gatekeeper is a much
larger tyrosine residue. This bulky residue would be expected to
hinder access to the WPF shelf, and thus clash with the binding of
the pendent benzyl of a BZD compound (Figure 4b).
’ METHODS
Chemistry. General. All commercial chemicals and solvents are
reagent grade and were used without further purification unless other-
wise specified. All reactions except those in aqueous media were carried
out with the use of standard techniques for the exclusion of moisture.
Reactions were monitored by thin-layer chromatography on 0.2 mm
silica gel plates (ALUGRAM SIL G/UV254, Macherey-Nagel) and
visualized with UV light. Final compounds were typically purified either
by flash chromatography on silica gel (E. Merck, 40ꢀ63 mm) or by
recrystallization. 1H and 13C NMR spectra were recorded on a Bruker
300 MHz Avance DPX. Chemical shifts are reported in parts per million
(ppm, δ units). Splitting patterns are designed as s, singlet; d, doublet; t,
triplet; q, quartet; m, multiplet; bs, broad singlet. High-resolution mass
spectra were recorded on a Micromass LCT (TOF) spectrometer
coupled to an analytical high performance liquid chromatography
(HPLC) was conducted on a XTERRA-MS C18 column (30 mm ꢁ
3 mm id, 2.5 μm) eluting with 0.01 M ammonium acetate in water and
100% acetonitrile (CH3CN), using the following elution gradient:
0ꢀ100% CH3CN over 4 min and 100% CH3CN over 1 min at
1.1 mL/min at 40 °C. Mass spectra were acquired in either positive or
negative ion mode under electrospray ionization (ESI) method. Specific
rotations were measured on a Perkin-Elmer 343 polarimeter. The purity
of all compounds were >95% as judged by LC-MS and NMR.
Compound 2 2-[(4S)-6-(4-Chlorophenyl)-1-methyl-8-(methyloxy)-
4H-[1,2,4]triazolo[4,3-a][1,4]benzodiazepin-4-yl]-N-ethylacetamide
and Compound 3 2-[(4R)-6-(4-Chlorophenyl)-1-methyl-8-(methyl-
oxy)-4H-[1,2,4]triazolo[4,3-a][1,4]benzodiazepin-4-yl]-N-ethylacet-
amide. These were prepared by chiral separation of compound 10.
The racemic mixture of 2-[(4-chlorophenyl)-1-methyl-8-(methylo-
xy)-4H-[1,2,4]triazolo[4,3-a][1,4]benzodiazepin-4-yl]-N-ethylaceta-
mide 10 (200 mg) was separated by HPLC using a ChiralPack AD
(250 mm ꢁ 20 mm id, 10 μm) column eluting with EtOH/hexane
(60/40) as the mobile phase (flow rate: 17 mL/min). 2 eluted at 5.21
min as the first peak. [a]2D0 = þ88.1 (c = 1.0015/MeOH). 3 came off at
24.28 min. [a]2D0 = ꢀ87.7 (c = 1.0015/MeOH). 3 came off at 24.28
min. [a]2D0 = ꢀ87.7 (c = 1.0015/MeOH).
Compound 10 2-[6-(4-Chlorophenyl)-1-methyl-8-(methyloxy)-
4H-[1,2,4]triazolo[4,3-a][1,4]benzodiazepin-4-yl]-N-ethylacetamide.
To a solution of acid 9 (4.5 g, 11.4 mmol) in THF at RT was added
diisopropylethylamine DIEA (4 mL, 22.7 mmol, 2 equiv), followed by
HBTU (8.6 g, 22.7 mmol, 2 equiv). The reaction mixture was stirred for
3 h at this temperature, and ethylamine (11.3 mL, 2 M in THF, 22.7
mmol, 2 equiv) was added dropwise. The mixture was stirred overnight
before being concentrated under reduced pressure. The crude material
was dissolved in DCM and washed successively with water, 1N NaOH
and 1N HCl. The organic layer was dried over Na2SO4, filtered, and
concentrated in vacuo. The crude solid was recrystallized in CH3CN to
give the title compound 10 (4.1 g, 85% yield) as a white solid. Yield
from 9: 85%, white solid. Rf = 0.48 (DCM/MeOH:90/10); mp
The crystal structures of several BZDs with a number of BET
bromodomains described in this paper yield insights into the
molecular mimicry of the acetylated-lysine interactions. They suggest
opportunities to exploit commonalities and differences in the
bromodomain family to design compounds that achieve distinct
profiles of inhibition. This is likely to be an important requirement for
not only therapeutic indications but also for obtaining chemical tools
that will help to dissect the complex role of these epigenetic targets.
Similarly, small molecule inhibitors of other domains will
provide powerful tools to understand the contributions of the
1
140ꢀ145 °C. H NMR (300 MHz, CDCl3) δ 7.53ꢀ7.47 (m, 2H),
7.39 (d, J = 8.9 Hz, 1H), 7.37ꢀ7.31 (m, 2H), 7.20 (dd, J = 2.9 and 8.9
Hz, 1H), 6.86 (d, J = 2.9 Hz, 1H), 6.40 (m, 1H), 4.62 (m, 1H), 3.80
(s, 3H), 3.51 (dd, J = 7.3 and 14.1 Hz, 1H), 3.46ꢀ3.21 (m, 3H), 2.62
(s, 3H), 1.19 (t, J = 7.3 Hz, 3H). 13C NMR (75 MHz, DMSO-d6)
3835
dx.doi.org/10.1021/jm200108t |J. Med. Chem. 2011, 54, 3827–3838

Journal of Medicinal Chemistry
ARTICLE
δ 169.2, 165.5, 157.4, 155.8, 150.7, 137.3, 135.4, 131.0, 129.3, 128.2,
126.4, 125.6, 117.8, 115.1, 55.8, 53.3, 37.6, 33.4, 14.8, 11.5. HRMS
(M þ H)þ calcd for C22H2335ClN5O2 424.1540; found 424.1525.
prepared as detailed in the Supporting Information (5 mL) were thawed
on ice and cleared by centrifugation (10000g, 10 min.). The lysate was
first precleared with 50 μL of 50% slurry of blocked Reacti-gel 6ꢁ beads
(with rotation, 1 h, 4 °C). The beads were pelleted by centrifugation
(1000g, 1 min, 4 °C) and 1.5 mL of lysate added to a fresh tube
containing 100 μL of 50% slurry of Reacti-gel 6ꢁ beads coupled to the
relevant compound and incubated with rotation for 4 h at 4 °C. For
competition experiments, excess free compound or 5% DMSO control
was added prior to incubation. The beads were washed 4ꢁ with 1%
NP40 lysis buffer with 250 mM salt and eluted with 25 μL of 2ꢁ SDS
sample buffer (10 min, 37 °C). Samples were separated by SDS-PAGE
using NuPAGE 4ꢀ12% gradient gels and then stained using GelCode
Blue. For subsequent analysis by mass spectrometry, bands were excised
from destained gels.
Hek293 lysates were prepared as above. From 1 mL of lysate, one-half
was mixed with 25 μL of Reacti-gel 6ꢁ beads coupled to the relevant
compound and the remaining half mixed with 25 μL Anti FLAG-M2
antibody coupled to sepharose beads. Lysates and beads were incubated
with rotation for 4 h at 4 °C. The beads were washed 4ꢁ with 1% Triton
X100 lysis buffer and eluted with 15 μL of 2ꢁ SDS sample buffer
(10 min, 37 °C). Samples were separated by SDS-PAGE, transferred to
PVDF membrane, and immunoblotted with anti-FLAG M2 antibody.
Mass Spectrometry (Full Methods Can Be Found in Sup-
porting Information, Section 4iv (S16ꢀS19)). The entire lanes
of the chemoproteomic gels were subjected to MS analysis and not just
the visible bands as described in the Supporting Information. The
manually excised gel bands were subject to automated in situ enzymatic
digestion and analyzed by LC/MS/MS. Database searches were carried
out using an in-house MASCOT (Matrix Science, UK)30 server against
the PSR protein database.
Compound
9 [6-(4-Chlorophenyl)-1-methyl-8-(methyloxy)-4H-
[1,2,4]triazolo[4,3-a][1,4] benzodiazepin-4-yl] Acetic Acid. To a solu-
tion of ester 8 (7.4 g, 18.1 mmol) in THF (130 mL) at RT was added 1N
NaOH (36.2 mL, 36.2 mmol, 2 equiv). The reaction mixture was stirred
at this temperature for 5 h before being quenched with 1N HCl
(36.2 mL) and concentrated in vacuo. Water is then added and the
aqueous layer was extracted with DCM (3ꢁ), and the combined organic
layers were dried over Na2SO4, filtered, and concentrated under reduced
pressure to give the title compound 9 (7 g, 98% yield) as a pale-yellow
solid. Yield from 8: 98%, pale-yellow solid. 1H NMR (300 MHz, CDCl3)
δ 7.55ꢀ7.48 (m, 2H), 7.41 (d, J = 8.9 Hz, 1H), 7.38ꢀ7.31 (m, 2H), 7.22
(dd, J = 2.9 and 8.9 Hz, 1H), 6.90 (d, J = 2.9 Hz, 1H), 4.59 (dd, J = 6.9
and 6.9 Hz, 1H), 3.81 (s, 3H), 3.70 (dd, J = 6.9 and 25.7 Hz, 1H), 3.61
(dd, J = 6.9 and 25.7 Hz, 1H), 2.63 (s, 3H). 13C NMR (75 MHz, DMSO-
d6) δ 172.1, 165.6, 157.4, 155.5, 150.8, 137.2, 135.4, 131.1, 129.2, 128.3,
126.3, 125.6, 117.9, 115.2, 55.8, 52.9, 36.5, 11.5. HRMS (M þ H)þ calcd
for C20H1835ClN4O3 397.1067; found 397.1063.
Compound 8 Methyl [6-(4-Chlorophenyl)-1-methyl-8-(methyloxy)-
4H-[1,2,4]triazolo[4,3-a][1,4]benzodiazepin-4-yl]acetate. Yield from
7: 91%. 1H NMR (300 MHz, CDCl3) δ 7.54ꢀ7.47 (m, 2H), 7.40 (d,
J = 8.8 Hz, 1H), 7.37ꢀ7.31 (m, 2H), 7.22 (dd, J = 2.8 and 8.8 Hz, 1H),
6.89 (d, J = 2.8 Hz, 1H), 4.61 (dd, J = 6.4 and 7.8 Hz, 1H), 3.82 (s, 3H),
3.78 (s, 3H), 3.66 (dd, J = 7.8 and 16.9 Hz, 1H), 3.60 (dd, J = 6.4 and 16.9
Hz, 1H), 2.62 (s, 3H). 13C NMR (75 MHz, DMSO-d6) δ 171.1, 166.9,
157.4, 155.3, 150.9, 137.1, 135.5, 131.1, 129.2, 128.3, 126.3, 125.6, 117.8,
115.3, 55.8, 52.8, 51.6, 36.2, 11.5. HRMS (M þ H)þ calcd for
C21H2035ClN4O3 411.1229; found 411.1245.
Compound 7 Methyl [5-(4-Chlorophenyl)-7-(methyloxy)-2-thioxo-
2,3-dihydro-1H-1,4-benzodiazepin-3-yl]acetate. Yield from 6: 81%,
pale-yellow solid. H NMR (300 MHz, CDCl3) δ 10.11 (br s, 1H),
7.48 (m, 2H), 7.34 (m, 2H), 7.18 (d, J = 8.9 Hz, 2H), 7.11 (dd, J = 8.9
and 2.8 Hz, 1H), 6.78 (d, J = 2.8 Hz, 1H), 4.39 (dd, J = 7.0 and 6.8 Hz,
1H), 3.76 (s, 3H), 3.75 (s, 3H), 3.66 (dd, J = 16.8 and 6.8 Hz, 1H), 3.38
(dd, J = 16.8 and 7.0 Hz, 1H). 13C NMR (75 MHz, DMSO-d6) δ 198.6,
Luciferase Reporter Assay (Full Details in Supporting
Information, Figure 3 (S6)). Human hepatocyte HepG2 cells
(stably transfected with an ApoA1 luciferase reporter) were incubated
for 18 h in the presence or absence of freshly diluted compound (0.5%
DMSO) prior to addition of Bright Glo reagent (Promega). Lumines-
cence was read using a Perkin-Elmer Envision plate reader after 10 min
incubation with the substrate. The selectivity of the effect was tested
using an LDL-receptor promoter luciferase reporter (HepG2LDLr
Luciferase clone 2) using the same conditions in parallel.
ApoA1 Secretion Assay. Compounds were tested for their ability
to increase ApoA1 production by HepG2 cells. HepG2 cells were grown
in BME medium containing 10% heat inactivated FCS, glutamine,
sodium pyruvate, nonessential amino acids and penicillin-streptomycin
at 37 °C in a 95% airꢀ5% CO2 humidified atmosphere. The compounds
were first incubated onto confluent HepG2 cells in BME medium
containing 1% FCS and all other additives. After 48 h incubation, the
supernatants were removed and compounds were readded onto cells in
RPMI-1640 medium depleted in methionine and cysteine and supple-
mented with 1% cystine, 1% glutamine, 1% penicillinꢀstreptomycin,
and 35S-methionine. The incubation was maintained for 6 additional
hours at 37 °C. Neosynthesized radiolabeled secreted proteins were then
analyzed on 5ꢀ12% SDS Page. After exposure of gels onto Phosphor-
Imager screens, radiolabeled secreted ApoA1 was quantified using a
Storm 860 (Molecular Dynamics) apparatus.
Isothermal Titration Calorimetry. ITC titrations were carried
out using a Microcal VP-ITC in 20 mM Hepes, pH 7.5, 100 mM NaCl at
the temperatures indicated. Compound concentrations were determined by
NMR and protein concentrations by A280 measurements. Typically,
10ꢀ20 μM compound was placed in the cell and 120ꢀ240 μM protein
was titrated into this at 16ꢀ26 °C to achieve at least a final excess of 2:1
protein:ligand concentrations for single bromodomain constructs and
1.2:1 for tandem bromodomain constructs. Injections were performed
using protein in the syringe rather than compound due to the limited
solubility of 1. The data was fitted within Origin (Microcal version).
1
171.3, 166.2, 155.4, 136.7, 135.4, 133.2, 131.1, 128.5, 128.4, 123.4, 119.0,
113.6, 63.8, 55.6, 51.3, 39.0. HRMS (M þ H)þ calcd for C19H18
-
35
ClN2O3S 389.0727; found 389.0714.
Compound 6 Methyl [5-(4-Chlorophenyl)-7-(methyloxy)-2-oxo-
2,3-dihydro-1H-1,4-benzodiazepin-3-yl]acetate. H NMR (300 MHz,
1
CDCl3) δ 8.73 (br s, 1H), 7.50 (m, 2H), 7.35 (m, 2H), 7.15ꢀ7.07 (m,
2H), 6.76 (m, 1H), 4.18 (dd, J = 7.2 and 6.8 Hz, 1H), 3.75 (s, 3H), 3.74
(s, 3H), 3.42 (dd, J = 16.9 and 7.2 Hz, 1H), 3.42 (dd, J = 16.9 and 6.8 Hz,
1H). 13C NMR (75 MHz, DMSO-d6) δ 171.6, 169.5, 166.6, 154.3,
137.1, 135.2, 132.7, 131.1, 128.4, 126.8, 122.9, 119.2, 113.4, 60.3, 55.5,
51.3, 35.9. HRMS (M þ H)þ calcd for C19H1835ClN2O4 373.0955;
found 373.0957.
Compound 1 was prepared as described in ref 34.
For full details of synthesis and QC, see Supporting Information,
Section 1 (S2ꢀS5).
Reagents and Plasmids (Full Details in Supporting Infor-
mation, Figure 2 (S6)). Human Brd2 was cloned from human heart
Marathon ready cDNA (Clontech), subcloned into pcDNA3.1D
(invitrogen) and verified by full-length DNA sequencing. Flag-Brd2 full
length (1ꢀ2406 bp), Flag-Brd2 N term (1ꢀ1422 bp) and Flag-Brd2 C
term (1423ꢀ2406 bp) were generated using the Gateway system
(Invitrogen, San Diego, CA). Purified PCR products were subcloned
into pENTR-D-TOPO and subsequently transferred into a modified
pDEST16 vector containing a 50 prime FLAG tag sequence. Constructs
were verified by DNA sequencing
Affinity Chromatography and Gel Electrophoresis (See Also
Supporting Information, Section 4 (S16ꢀS19)). HepG2 lysates
3836
dx.doi.org/10.1021/jm200108t |J. Med. Chem. 2011, 54, 3827–3838

Journal of Medicinal Chemistry
ARTICLE
Crystal Structure of BET Bromodomain Proteins with 1
and 2 (Full Details Are Available in Supporting Information,
Section 7 (S34ꢀS39)). Crystals were generated by cocrystallization
and/or compound soaks into apo crystals. In some instances, structures
could be generated using both methods, and as there were no significant
differences in binding mode observed, structures obtained by both
methods were treated equally. Crystals were generating by vapor
diffusion using sitting or hanging drops. Data was collected from
cryoprotected single crystals, and typically the data was processed and
refined using the CCP431 suite of software, with rounds of model
building using Coot.32 Parts a and bꢀf of Figure 3 and Figure 4 were
prepared using PyMOL (Schrodinger Inc.) and Figure 3b using MOE
(Chemical Computing Group).
material and methods, siRNA material and methods; bio-
physical characterization: constructs and production of single
and tandem bromodomain proteins, example ITC data, KD
determination of FRET H4 peptide for the tandem BET
proteins, thermal shift and ITC selectivity data, matrix of
sequence identify between BET bromodomains and CREBBP,
ATAD2, biacore methods and analysis of cmpd1 binding to Brd-
BD1/2; fluorescence anisotropy assay: correlation of FA BET
binding activity with Apo-A1 upregulation, FA material and
methods; X-ray crystallography: X-ray refinement statistics,
Omit maps of the ligands, crystallization and crystallography
material and methods, electrostatic surface of Brd2. This material
is available free of charge via the Internet at http://pubs.acs.org.
SPR Experiments (See Supporting Information, Figure 5vi
(S27ꢀS31)). Real-time monitoring of the interactions between 1 and
BET bromodomains were performed on a BIAcore S51 (GE
Healthcare) at 25 °C. Flow cells of a CM5 chip were first activated
with 0.2 M N-ethyl-N0-(diethylaminopropyl)-carbodiimide (EDC) and
0.05 M N-hydroxysuccimide (NHS). Typically 0.15 mg/mL inhibited
Brd2-BD1 and Brd2-BD2 at pH 6.5 in NaAc was then injected at 10 μL/
min for 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 10ꢀ20 μM (For Figure 3e,f
the concentrations were: pink = 6.7 μM, red = 2.3 μM, cyan = 0.74 μM,
green = 0.25 μM, blue = 0.08 μM, gray = 0.03 μM, respectively). Binding
curves were analyzed with BIAevaluation (GE Healthcare).
Accession Codes
Atomic coordinates of Brd2-BD1:compound 1 (2ydw), Brd2-
BD1:compound 2 (2yek), Brd4-BD1:compound 1 (2yel) and
Brd4-BD2:compound 1 (2yem) have been deposited withthe
Protein Data Bank.
’ AUTHOR INFORMATION
Corresponding Author
*Phone: þ44 1438 763342. Fax: þ44 1438 763352. E-mail:
Chun-Wa.H.Chung@gsk.com.
FRET Titrations. Compounds were titrated against BD12 of Brd2
(200nM), Brd3 (100nM), and Brd4 (50nM) in 50 mM HEPES pH7.5,
50 mM NaCl, 0.5 mM CHAPS in the presence of tetra-acetylated
histone H4 peptide (200 nM) (Millipore, cat. 12-379). After equilibrat-
ing for 1 h, the bromodomain protein:peptide interaction was detected
using FRET following the addition of 2 nM europium cryptate labeled
streptavidin (Cisbio610SAKLA) and 10 nM XL-665-labeled anti-6His
antibody (Cisbio 61HisXLA) in assay buffer containing 0.05% (v/v)
BSA and 400 mM KF. Plates were read using an Envision Plate reader
(excitation 320 nm, emission 615 and 665 nm).
RNAi (Full Details in Supporting Information, Section 4
(S7ꢀS8,S20). HepG2 cells were grown in Dulbeco’s Modified Eagles
Medium (DMEM) supplemented with 10% FCS, 1% (W/V) glutamine,
and 1% pen/strep. Cells were plated at a density of 2 ꢁ 104cells per 96-
well plate, 24 h pretransfection, and siRNAs were transfected with the
Gemini lipid GSK34723233 at 7.5ug/mL final concentration. ApoA1
induction was measured relative to the control siRNA treated cells.
Specific custom designed siRNAs targeting Brd2, Brd3, and Brd4 were
used that were found to have no mRNA knockdown effect on the other
family members.
’ ACKNOWLEDGMENT
We thank staff at the ESRF at Genoble for beamline assistance.
The Structure Genomics Consortium kindly provided cDNA for
the Brd4-BD1 and Brd4-BD2 constructs. Single bromodomain
construct design was performed by Pam Thomas. Mike
Woodrow, Jon Seal, and Dave Wilson helped us gather together
the chemical QC data. Champa Patel conducted the initial FRET
assay development and members, of the Screening and
Compound Profiling Department ran the BET FA assay. Finally,
we acknowledge the many colleagues that have contributed to
the ApoA1 program and those that have taken time to pass on
useful comments in the preparation of this manuscript.
’ DEDICATION
This paper is dedicated to the memory of Dr. Franc-ois Hyafil,
who died on 26 May 2006.
’ ABBREVIATIONS USED
Fluorescence Anisotropy (FA) Binding Assays (Full Details
in Supporting Information, Section 6 (S31ꢀS33)). Com-
pounds were tested in dose response mode using an Alexa Fluor 488
derivative of 2 (Figure 5). Then 75 nM of Brd2, 3, or 4, 5 nM fluorescent
ligand and compound were incubated for at least 1 h in 50 mM Hepes
pH 7.4, 150 mM NaCl, 0.5 mM CHAPS prior to the FA being read on an
Envision ((λex = 485 nm, λEM = 530 nm; dichroic ꢀ505 nM).
AcK, acetylated lysine; ApoA1, apolipoprotein A1; BET, bromo-
domain and ET domain; BD1,BD2,BD12, bromodomain
1,2,1and2; BZD, benzodiazepine; chromodomain, chromatin or-
ganization modifier domain; FA, fluorescence anistropy; FRET,
fluorescence resonance energy transfer; H4AcK4, tetra-acety-
lated lysine histone 4 peptide; ITC, isothermal titration calorim-
etry; MOA, mode of action; PHD, plant homeodomain; SAHA,
suberoylanilide hydroxamic acid; SAR, structureꢀactivity rela-
tionships; siRNA, short interfering ribonucleic acid; SPR, surface
plasmon resonance
’ ASSOCIATED CONTENT
S
Supporting Information. Additional figures, experimen-
b
tal data, material and methods. Chemical synthesis and char-
acterization of all intermediate compounds to 1, 2, 3: biological
reagents and plasmids details; luciferase reporter assay methods;
chemoproteomic and siRNA: chemoproteomic and siRNA re-
sults, example gels from chemoproteomics experiments, table of
proteins identified by MS on analysis of gels, chemoproteomic
’ REFERENCES
(1) Goldberg, A. D.; Allis, C. D.; Bernstein, E. Epigenetics: a land-
scape takes shape. Cell 2007, 128, 635–638.
(2) Kouzarides, T. Chromatin modifications and their function. Cell
2007, 128, 693–705.
3837
dx.doi.org/10.1021/jm200108t |J. Med. Chem. 2011, 54, 3827–3838

Journal of Medicinal Chemistry
ARTICLE
(3) Taverna, S. D.; Li, H.; Ruthenburg, A. J. How chromatin-binding
modules interpret histone modifications: lessons from professional
pocket pickers. Nature Struct. Mol. Biol. 2007, 14, 1025–1040.
(4) Grant, S.; Easley, C.; Kirkpatrick, P. Vorinostat. Nature Rev. Drug
Discovery 2007, 6, 21–22.
(5) Sperandio, O.; Reynes, C. H.; Camproux, A. C.; Villoutreix, B. O.
Rationalizing the chemical space of proteinꢀprotein interaction inhibi-
tors. Drug Discovery Today 2009, 15, 220–229.
(6) Zeng, L.; Li, J.; Muller, M.; Yan, S.; Mujtaba, S.; Pan, C.; Wang,
Z.; Zhou, M.-M. Selective Small Molecules Blocking HIV-1 Tat and
Coactivator PCAF Association. J. Am. Chem. Soc. 2005, 127, 2376–2377.
(7) Sachchidanand; Resnick-Silverman, L.; Yan, S.; Mutjaba, S.; Liu,
W.-j.; Zeng, L.; Manfredi, J. J.; Zhou, M.-M. Target Structure-Based
Discovery of Small Molecules that Block Human p53 and CREB Binding
Protein Association. Chem. Biol. 2006, 13, 81–90.
(8) Koeller, K. M.; Haggarty, S. J.; Perkins, B. D.; Leykin, I.; Wong,
J. C.; Kao, M. C.; Schreiber, S. L. Chemical genetic modifier screens:
small molecule trichostatin suppressors as probes of intracellular histone
and tubulin acetylation. Chem. Biol. 2003, 10, 397–410.
(9) Johnson, R. L.; Huang, R.; Jadhav, A.; Southall, N.; Wichterman,
J.; MacArthur, R.; Xia, M.; Bi, K.; Printen, J.; Austin, C. P.; Inglese, J. A
quantitative high-throughput screen for modulators of IL-6 signaling: a
model for interrogating biological networks using chemical libraries.
Mol. BioSyst. 2009, 5, 1039–1050.
(10) Johnson, R. L.; Huang, W.; Jadhav, A.; Austin, C. P.; Inglese, J.;
Martinez, E. D. A quantitative high-throughput screen identifies poten-
tial epigenetic modulators of gene expression. Anal. Biochem. 2008,
375, 237–248.
(11) Barter, P. J.; Nicholls, S.; Rye, K.; Anantharamaiah, G. M.; Navab,
M.; Fogelman, A. M. Antiinflammatory properties of HDL. Circ. Res. 2004,
95, 764–772.
(12) Wong, N. C. Novel therapies to increase apolipoprotein AI and
HDL for the treatment of atherosclerosis. Curr. Opin. Invest. Drugs 2007,
8, 718–728.
(13) Florence, B.; Faller, D. V. You bet-cha: a novel family of
transcriptional regulators. Front. Biosci. 2001, 6, D1008–D1018.
(14) Mirguet, O.; Ajakane, M.; Boullay, A. M.; Boursier, E.; Brusq,
J.-M.; Clꢀement, C.; Costaz, A.; Daugan, A.; Dudit, Y.; Gosmini, R.; Huet,
P.; Martin, S.; Pineau, O.; Riou, A.; Toum, J.; Trottet, L.; Nicodꢁeme, E.
Discovery of benzodiazepine derivatives as Apo A-1 upregulators.
Abstracts of Papers, 241st ACS National Meeting & Exposition, Anaheim,
CA, United States, March 27ꢀ31, 2011; American Chemical Society:
Washington DC, 2011; MEDI-76.
(15) Nicodeme, E.; Jeffrey, K. L.; Schaefer, U.; Beinke, S.; Dewell, S.;
Chung, C.; 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.
(22) Nakamura, Y.; Umehara, T.; Nakano, K.; Jang, M. K.; Shirouzu,
M.; Morita, S.; Uda-Tochio, H.; Hamana, H.; Terada, T.; Adachi, N.;
Matsumoto, T.; Tanaka, A.; Horikoshi, M.; Ozato, K.; Padmanabhan, B.;
Yokoyama, S. Crystal structure of the human BRD2 bromodomain:
insights into dimerization and recognition of acetylated histone H4.
J. Biol. Chem. 2007, 282, 4193–4201.
(23) Umehara, T.; Nakamura, Y.; Jang, M. K.; Nakano, K.; Tanaka,
A.; Ozato, K.; Padmanabhan, B.; Yokoyama, S. J. Biol. Chem. 2010,
285, 7610–7618.
(24) Vollmuth, F.; Blankenfeldt, W.; Geyer, M. Structures of the dual
bromodomains of the P-TEFb-activating protein Brd4 at atomic resolu-
tion. J. Biol. Chem. 2009, 284, 36547–36556.
(25) 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.; Kochbin, S.; Muller, C. W.; Petosa, P. Cooperative binding
of two acetylation marks on a histone tail by a single bromodomain.
Nature 2009, 461, 664–668.
(26) You, J.; Croyle, J. L.; Nishimura, A.; Ozato, K.; Howley, P. M.
Interaction of the bovine papillomavirus E2 protein with Brd4 tethers
the viral DNA to host mitotic chromosomes. Cell 2004, 117, 349–360.
(27) Greenwald, R. J.; Tumang, J. R.; Sinha, A.; Currier, N.; Cardiff,
R. D.; Rothstein, T. L.; Faller, D. V.; Denis, G. V. E mu-BRD2 transgenic
mice develop B-cell lymphoma and leukemia. Blood 2004, 103, 1475–1484.
(28) 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 Thangue,
N.; French, C. A.; Wiest, O.; Kung, A. L.; Knapp, S.; Bradner, J. E. Selective
inhibition of BET bromodomains. Nature 2010, 468, 1067–1073.
(29) Owen, D. J.; Ornaghi, P.; Yang, J.-C.; Lowe, N.; Evans, P. R.;
Ballario, P.; Neuhaus, D.; Filetici, P.; Travers, A. A. The structural basis
for the recognition of acetylated histone H4 by the bromodomain of
histone acetyltransferase gcn5p. EMBO J. 2000, 19, 6141–6149.
(30) Perkins, D. N.; Pappin, D. J.; Creasy, D. M.; Cottrell, J. S.
Probability-based protein identification by searching sequence databases
using mass spectrometry data. Electrophoresis 1999, 20, 3551–3567.
(31) Collaborative Computational Project Number 4. The CCP4
suite: programs for protein crystallography. Acta Crystallogr., Sect. D:
Biol. Crystallogr. 1993, 50, 760–763.
(32) Emsley, P.; Cowtan, K. Coot: model-building tools for molecular
graphics. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2004, 60, 2126–2132.
(33) Castro, M.; Griffiths, D.; Patel, A.; Pattrick, N.; Kitson, C.;
Ladlow, M. Effect of chain length on transfection properties of spermine-
based gemini surfactants. Org. Biomol. Chem. 2004, 2, 2814–2280.
(34) Freidinger, R. M.; Evans, B. E.; Bock, M. G. Triazolobenzodia-
zepines. U.S. Patent US5185331 A1, 1993.
(16) Pivot-Pajot, C.; Caron, C.; Govin, J.; Vion, A.; Rousseaux, S.;
Khochbin, S. Acetylation-dependent chromatin reorganization by BRDT, a
testis-specific bromodomain-containing protein. Mol. Cell. Biol. 2003,
23, 5354–5365.
(17) Huang, H.; Zhang, J.; Shen, W.; Wang, X.; Wu, J.; Wu, J.; Shi,
Y. Solution structure of the second bromodomain of Brd2 and its
specific interaction with acetylated histone tails. BMC Struct. Biol.
2007, 7, 57.
(18) Horton, J. R.; Upadhyay, A. K.; Qi, H. K.; Zhang, X.; Shi, Y.;
Cheng, X. Enzymatic and structural insights for substrate specificity of a
family of jumonji histone lysine demethylases. Nature Struct. Mol. Biol.
2010, 17, 38–43.
(19) LeRoy, G.; Rickards, B.; Flint, S. J. The double bromodomain
proteins Brd2 and Brd3 couple histone acetylation to transcription. Mol.
Cell 2008, 30, 51–60.
(20) Wu, S. Y.; Chiang, C. M. The double bromodomain-containing
chromatin adaptor Brd4 and transcriptional regulation. J. Biol. Chem.
2007, 282, 13141–13145.
(21) Mujtaba, S.; Zeng, L.; Zhou, M. M. Structure and acetyl-lysine
recognition of the bromodomain. Oncogene 2007, 26, 5521–5527.
3838
dx.doi.org/10.1021/jm200108t |J. Med. Chem. 2011, 54, 3827–3838