Review
For reprint orders, please contact [email protected]
Phenotypic screening and fragment-based
approaches to the discovery of small-molecule
bromodomain ligands
Bromodomains are protein modules that bind to acetylated lysine residues and hence facilitate protein–protein
interactions. These bromodomain-mediated interactions often play key roles in transcriptional regulation and their
dysfunction is implicated in a large number of diseases. The discovery of potent and selective small-molecule
bromodomain and extra C-terminal domain bromodomain ligands, which show promising results for the treatment
of cancers and atherosclerosis, has promoted intense interest in this area. Here we describe the progress that has
been made to date in the discovery of small-molecule bromodomain ligands, with particular emphasis on the roles
played by phenotypic screening and fragment-based approaches. In considering the future of the field we discuss
the prospects for development of molecular probes and drugs for the non-bromodomain and extra C-terminal
domain bromodomains.
Lysine acetylation is a prevalent protein posttranslational modification that regulates many
cellular processes. Consequently, there has been
significant interest in developing small molecules
that manipulate the cellular machinery responsible for modulating lysine acetylation. The clinical use of the HDAC inhibitors vorinostat and
romidepsin demonstrates the possibility of treating cancers by modifying cellular lysine acetylation. Bromodomains (Box 1; Figure 1) are protein
modules that bind to acetylated lysine residues
(KAc) and, in the terminology of the ‘histone
code’ (Box 2; Figure 2), are viewed as ‘readers’
of the acetylated lysine state. Bromodomaincontaining proteins (BCPs) are found exclusively as part of larger protein scaffolds, many
of which are components of transcriptional
regulation. Until recently, there had been little
interest in developing ligands for bromo­domains
or investigating them as therapeutic targets.
However, rapid progress in the development of
small-molecule ligands of the bromodomain
and extra C-terminal domain (BET) family of
bromodomains has stimulated intense interest
in this area and there are now at least four BET
bromodomain ligands in clinical trials.
Zhou and co-workers were the first to report
small-molecule bromodomain ligands, targeting
the bromodomains of PCAF [1] and CREBBP
[2]. These bromodomain ligands were discovered using NMR-based screening of fragments.
Conversely, the first BET bromodomain probes
were discovered using phenotypic screening
(Box 3; Figures 3 & 4)
by the Mitsubishi Tanabe
Pharma Corporation (Osaka, Japan) [101,102],
GlaxoSmithKline (GSK; Brentford, UK) [3,4]
and Resverlogix Corporation (Calgary, Canada) [5,6]. Few details about the discovery of
the Mitsubishi compounds have been revealed,
however, the structures disclosed in their patents
[101,102] are the basis of (+)-JQ1 (3; Figure 5), a
potent and selective BET bromodomain probe
reported by the Structural Genomics Consortium (SGC; Oxford, UK) and the Bradner laboratory at Harvard (MA, USA) [7]. Oncoethix
have in-licensed a compound, OTX015, from
Mitsubishi and have this compound in a Phase I
clinical trial [8–10]. GSK have described how a
phenotypic screen designed to identify upregulators of apolipoprotein A1 (ApoA1), which
raises plasma levels of high-density lipoprotein
cholesterol (HDL-C) and is linked with protection from atherosclerosis progression and antiinflammatory effects, led to the development
of the BET bromodomain probe I-BET762 (2)
[3,4]. The Resverlogix compound, RVX-208 (6),
was identified using a HepG2 cell-based assay
to screen for compounds with the ability to raise
ApoA1 production [5] and has subsequently been
shown to inhibit the BET bromodomains.
GSK [11–13] and our laboratory [14–16] reported
fragment-based approaches (B ox 4; Figure 6)
to the development of 3,5-dimethylisoxazole-based BET bromodomain ligands. Fish
and co-workers also employed a fragmentbased approach to develop PFI-1, which is a
10.4155/FMC.13.197 © 2014 Future Science Ltd
Future Med. Chem. (2014) 6(2), 179–204
Laura E Jennings‡,
Angelina R Measures‡,
Brian G Wilson‡
& Stuart J Conway*
Department of Chemistry, Chemistry
Research Laboratory, University of
Oxford, Mansfield Road, Oxford,
OX1 3TA, UK
*Author for correspondence:
Tel.: +44 1865 285 109
Fax: +44 1865 285 002
E-mail: [email protected]
‡
Authors contributed equally
ISSN 1756-8919
179
Review | Jennings, Measures, Wilson & Conway
Key Terms
Box 1. Bromodomain structure.
Bromodomain and extra
C-terminal domain
bromodomain:
Named after the Drosophila gene brahma where they were first identified [49,50] , bromodomains have
emerged as protein modules that bind to e-N-acetylated lysine residues (KAc) [34,51,66,67] . There are
61 human bromodomains found within 46 proteins in the human proteome, with some proteins
containing more than one distinct bromodomain [24,30,67] . These protein modules comprise
approximately 110 amino acids that form a characteristic, antiparallel, four-helix bundle composed of
helices aA, aB, aC and aZ [34,66,67] . The acetylated lysine residue binds in a pocket that is
predominantly hydrophobic, but which contains four structurally conserved water molecules that act
as the base of the pocket. There are often two additional conserved water molecules that reside in
the loop region between the aA and aZ helices, known as the ZA loop. KAc donates a hydrogen
bond from the e-NH to one of the ZA channel water molecules and accepts a hydrogen bond from a
conserved asparagine residue, which is replaced by a tyrosine or threonine in some bromodomains
[30] . The KAc also forms a water-mediated hydrogen bond to a conserved tyrosine residue within the
pocket. Although these hydrogen-bonding interactions are key for KAc recognition, the inherent
affinity of KAc for the bromodomain is low. In the case of the histones, affinity comes from binding
of the peptide tail at the entrance to the KAc-binding pocket. In the BET bromodomains there is a
clear groove in the bromodomain structure that accommodates the histone peptide. This groove
includes a hydrophobic pocket, defined by residues W97, P98 and F99 [BRD2(2) numbering], known
as the WPF shelf, which has proved important in the binding of synthetic ligands. Other classes of
bromodomains do not have a defined WPF shelf region. Despite high overall structural homology
between different bromodomains, sequence variation in ZA and BC loop regions contributes to
bromodomains displaying selectivity for certain KAc residues located in defined positions in a given
histone peptide [30] .
Bromodomain-containing
proteins have been grouped into
eight sub-families based on
phylogenetic relationships from
structure-based sequence
alignments – bromodomain and
extra C-terminal domain
bromodomains comprise one of
these families.
Atherosclerosis: Form of
arteriosclerosis, or thickening of
the artery walls, which results
from accumulation of calcium
and fatty materials including
cholesterol.
A
B
V103
BC loop
ZA loop
Y113
Y155
P102
Q101
αA
αB
W97
αZ
BRD2(2) bromodomain
N156
P98
F99
αC
KAc recognition site
Figure 1. Bromodomain structure. (A) x-ray crystal structure of the second bromodomain
of BRD2 in complex with KAc (carbon = purple, PDB code 2DVQ) [26] . The bromodomain
structural motif is formed of four helices aZ (blue), aA (yellow), aB (orange) and aC (red), and two
loop regions known as the ZA loop (green) and the BC loop (light green). (B) The KAc residue
binds in a well-defined hydrophobic pocket, forming interactions with a highly conserved
asparagine residue and several tightly bound water molecules (red spheres), which form the base of
the pocket. In BRD2, hydrogen bonding occurs from the carbonyl oxygen to N156 and via water to
Y113. The hydrophobic entrance to the pocket, W97-P98-F99, is common to BET bromodomains
and termed the ‘WPF shelf’.
dihydroquinazolinone-based BET bromodomain ligand [17]. Most recently, Zhao et al.
reported a fragment-based approach leading to a
range of chemically distinct BET bromodomain
ligands [18].
The nature in which bromodomain ligands
have been discovered provides a platform to
180
Future Med. Chem. (2014) 6(2)
compare the approaches of phenotypic screening and fragment-based ligand discovery. Herein
we discuss these two different approaches to
ligand discovery and compare the advantages,
disadvantages and types of molecules that
result in each case. Despite rapid progress in the
bromo­domain field, the literature in this area
future science group
Phenotypic screening & fragment-based discovery of small-molecule bromodomain ligands
is currently still tractable and hence provides a
unique opportunity for discussing the methods
in which small-molecule ligands can be discovered, their subsequent value as chemical probes
and their role in therapeutic target validation.
| Review
Phenotypic screening approaches to
the discovery of BET bromodomain
ligands
The first potent and selective BET bromodomain ligands, which have subsequently
Box 2. Histone structure, the histone code and epigenetics.
In eukaryotic cells, DNA is packaged into units, known as nucleosomes, which comprise approximately 150 DNA base pairs wound
around a core of histone proteins [68] . Chromatin consists of multiple nucleosome units linked by flexible strands of DNA [69] . The histone
core is a dimer of tetramers, comprising two copies of the histones H2A, H2B, H3 and H4 (histones H1 and H5 are also known), with
disordered histone tails extending from the core nucleosome unit [70] . It has long been known that histone proteins are susceptible to
multiple post-translational modifications (PTMs) [71] ; the histone tails contain many sites that can undergo PTM but modifications are also
known in the histone core (Figure 2) [71–74] . These modifications include acetylation, or methylation up to three times, on the e-nitrogen
atom of lysine, methylation of the guanidinyl group of arginine, ubiquitination of lysine, and phosphorylation of serine, threonine and
tyrosine residues [71,73,75] . The effects of these PTMs were initially thought to be purely electrostatic – lysine acetylation neutralizes the
charge of the e-nitrogen atom, reducing the interactions with negatively charged DNA phosphate groups. This reduction in charge gives
rise to the less tightly packaged form of chromatin, euchromatin, which is associated with active gene transcription. Conversely,
trimethylated lysine is positively charged and promotes the tightly packed heterochromatin structure, which is less accessible to
transcription factors and hence is associated with gene silencing [76] . Subsequent research has shown that these PTMs have a second
function, beyond a passive structural role, that also affects gene transcription. Protein domains have been identified that recognize and
bind to specific PTMs on histones (and other proteins), for example, chromodomains bind to e-methylated lysine residues and
bromodomains bind to ε-acetylated lysine residues [71,73,77] . Proteins that contain these domains bind to the modified amino acids in the
histone components of chromatin, leading to the assembly of complex protein scaffolds, many of which are involved in gene
transcription. The frequency and complexity of these histone PTMs, coupled with their ability to modulate gene transcription, has given
rise to the proposal of a signaling role for chromatin [78,79] . This idea has been extended to the concept of a ‘histone code’, in which a
defined pattern of histone modifications or ‘marks’ gives rise to a specific downstream phenotype [71,73,77] . Consequently, enzymes that
add the marks, such as histone acetyl transferases, are viewed as ‘writers’ of the code. Similarly, histone deacetylases are viewed as
‘erasers’ of the code and domains that bind to and recognize histone PTMs, such as bromodomains, are thought of as ‘reading’ the code.
This concept is attractive from a medicinal chemistry perspective as it should, in principle, be possible to interfere with these proteins as
an alternative to targeting the proteins whose expression they control [23] . As has been discussed in the literature, although the concept
of a code has proved inspirational, this metaphor should not be extended ad infinitum [79] .
The fact that histone PTMs provide a mechanism for environmental factors to affect gene transcription, and the fact that at least some
PTMs survive through cell division, has linked histone PTMs with a molecular mechanism for epigenetic memory. Epigenetics, literally
meaning above genetics, has a number of definitions that stem from two different roots [80] . The word epigenetics has, historically at
least, implied memory of a cellular signal, and the use of this word can be contentious in some contexts [81] , in particular for histone
marks that are not known to be copied along with the DNA or histones during cell division. Arrowsmith et al. define epigenetics as
“heritable changes in gene expression or phenotype that are stable between cell divisions, and sometimes between generations, but do
not involve changes in the underlying DNA sequence of the organism” [23] . This is the definition that is usually referred to in the context
of the work described in this review and encompasses lysine acetylation and its effect on gene transcription.
‘Writers’ HATs
H2A N–
H2B N–
P E PA K
SG
RG K Q GG K A RA
KGSKKA
APK
VTK
SAP
S
K A KT R
AQKKD
SRA
R
GKK
G L Q F P V G RV H R L
LR K
RKESYSIYV
YK
KRS
GN
‘Erasers’ HDACs
YA
RD
VLKQVHP
Bromodomains
‘Readers’
L A H Y N K R ST
IT S
RN
GG
KGL
G KGG
AK
RH
RK
VLR
DNIQGIT K
PA I
R
RL
ARR
GGVKR I SGL
IYE
ET
D E E L NK
REIQTA
D L R F Q SS
FKT
AV M
QD
A
H3 N–
H4 N– SGRG K
NKKTRI IPRHLQ LAI
V
LQEAC
RGVLK
Acetylation
Methylation
LL
RL
LL
GK
VTI
PGELA
LV
EAY
AQG GV
KH
AVS
LP
E GT
NIQ
AV L L P K K T E S
KAVTKYT SAK
EDTNLCA
IH
GLF
AK
RV
TIM
HHKA
KGK
–C H2A
–C H2B
PKDIQL
AR
RI
RG
ERA
–C H3
–C H4
Phosphorylation
Ubiquitination
XXX
XXX
Amino acids buried in nucleosome core
Solvent-accessible amino acids
Figure 2. Post-translational modifications to histone proteins, displayed as amino acid sequences. Amino acids in the histone
tails protrude through the gyres of DNA wrapped around the histone core (central cartoon representation). Modifications occur to both
solvent-exposed regions and residues buried within the nucleosome core. Acetyl groups are a common modification to the e-nitrogen
atom of lysine residues, and are added, recognized or removed by histone acetyltransferases, bromodomains and histone deacetylases,
respectively (shown in green).
future science group
www.future-science.com
181
Review | Jennings, Measures, Wilson & Conway
Box 3. Phenotypic screening.
The classical approach to medicinal chemistry and drug discovery, prevalent until the 1970s and 1980s, depended on the assessment of
compound efficacy at producing a certain desired effect in animal models of diseases. Work to understand the mechanism of these drugs
led to a comprehensive understanding of receptor pharmacology in many systems, which in turn prompted a more reductionist approach
to drug discovery. A move to more biochemical-based assays, focusing on compound selectivity and affinity for a defined protein target,
facilitated the use of high-throughput screens to discover lead compounds [82,83] . The compounds that were developed in this way
showed high affinity and exquisite selectivity for a defined protein target, or narrow group of targets. However, this chapter of drug
discovery has been characterized by high attrition of drug candidates in Phase II clinical trials, the point at which compounds are tested
for their efficacy in treating patients with the disease being targeted [65,84] . The reasons for this attrition are complex, but highlight the
ability of biological systems to effectively compensate for the loss of a certain component. This attrition is partly a function of ineffective
target validation and consequently the importance of this component of the drug discovery process has been highlighted.
Phenotypic screening represents, to some extent, a return to the classical method of assessing compounds for their potential to act as
drugs. The phenotypic screen will assess the ability of a compound to produce a desired phenotype in cellular, and less frequently animal,
models of disease (Figure 3) . A simple example of phenotypic screening is measuring the ability of compounds to cause cell death. A
more specific example of a phenotypic screen was used in the identification of I-BET762. A cell-based assay was established that
monitored the upregulation of Apolipoprotein A1 (ApoA1), which is known to be associated with protection from atherosclerosis and
anti-inflammatory effects [4] . The reason that phenotypic screening was a particularly good approach in this case was the lack of a
known molecular mechanism for ApoA1 upregulation that could be targeted therapeutically. Therefore a luciferase-based reporter gene
assay system was developed (Figure 4) . In this system, molecules that led to the upregulation of ApoA1 would also cause expression of
firefly luciferase, which generates light as a convenient readout of gene expression. Consequently, this system can be used to conduct an
unbiased screen for small molecules that affect the expression of the desired gene, in this case ApoA1.
A significant challenge with the phenotypic screening approach to drug development is identifying the mechanism or mechanisms of
action used by the hit compound to produce the desired phenotype. Although it is not a regulatory requirement to know the mechanism
of drug action for US FDA approval as a drug, elucidating the compound’s mechanism of action might well be required to understand the
safety profile of the drug [83] . Typically, some form of pull-down experiment or activity-based profiling is employed to determine the
protein target or targets of the active compound [65] . In the case of I-BET762, a series of affinity matrices were developed that ultimately
identified the BET bromodomains as the target of the active compounds. siRNA experiments were used to demonstrate that knockdown
of BRD4 resulted in the same phenotype (upregulation of ApoA1) as evoked by I-BET762 [4] .
Null or undesired
phenotypic
response
R
R
R
Desired
phenotypic
response
R
R
Lead
compound
Optimization
Clinical
candidate
Figure 3. Phenotypic screens can be used to identify particular compounds that enable a
desired phenotypic response, represented by transition from blue square to yellow.
Subsequent optimization of the lead compound will ultimately lead to the clinical candidate.
progressed to clinical evaluation, were identified
using phenotypic screening.
„„The
discovery of I-BET762 using a
cell-based ApoA1 luciferase reporter assay
ApoA1 upregulation is protective against atherosclerosis progression and results in anti-inflammatory effects. However, lack of an obvious molecular mechanism for modulating ApoA1 levels had
previously hindered therapeutic efforts in this
area. A phenotypic approach to identify modulators of ApoA1 expression was therefore ideal
in this setting. Researchers at GSK employed a
‘chemical genetic’ approach to identify small molecules that enhanced ApoA1 expression [3,4]. To
182
Future Med. Chem. (2014) 6(2)
observe ApoA1 upregulation, Nicodeme, Chung
and co-workers produced a human HepG2 hepatocyte cell line with an ApoA1 luciferase reporter
to screen a library of compounds. The phenotypic screen identified a single enantiomer of the
benzodiazepine GW841819X (1; Figure 5) that
caused potent induction of the ApoA1 reporter
gene with an EC50 value of 440 nM. This induction was shown to result from a specific effect
on the ApoA1 locus, but the molecular target of
the compound was unknown. Structure–activity
studies were undertaken to optimize the ApoA1
upregulation activity of GW841819X (1). These
studies revealed that the benzodiazepine core was
essential for activity and a 6-aryl substituent was
future science group
Phenotypic screening & fragment-based discovery of small-molecule bromodomain ligands
3´-UTR
ApoA1
5´-UTR
ApoA1
Human ApoA1 promoter
| Review
Figure 4. The luciferase-based reported gene assay used by GlaxoSmithKline to identify
upregulators of ApoA1 by phenotypic screening.
Figure adapted from [4] .
present in all active benzodiazepine compounds.
A range of groups was tolerated in the 4-position
and they could be used to modulate the physico­
chemical and pharmacokinetic (PK) properties of
the compounds. The resulting optimized compound, GSK525762A (2; I-BET762), displayed
similar potency in the ApoA1 reporter gene assay
(EC50 700 nM), but possessed superior physicochemical and PK properties than 1, making it
more suitable for in vivo experiments. Importantly, it was discovered that having the correct
stereochemistry at the 4-postion was critical to
ApoA1-modulating activity, implying that the
benzodiazepines were interacting with defined
molecular targets.
In an attempt to elucidate the target of these
compounds, a selection of the benzodiazepines,
with EC50 values obtained from the luciferase
reporter assay ranging from nM to inactive, were
evaluated for their activity on panels of kinases,
ion channels, nuclear receptors, GPCRs and
other protein classes known to be drug targets.
This screen did not identify the target for these
compounds, but instead indicated that they
2
1
Me
O
3
N N
O
6
1
Me
4
5
10
NH
2
NH
11 N
3
N N O
8
9
Me
11N
4
5
10
N
6
O
O
7
N N
N
N
6
1S
9
9
4
5N
2
7
8
7
8
Me
H3CO
1
Cl
Cl
3
I-BET762 (GSK525762A)
IC50 = 32.5 nM vs BRD2
LE = 0.35
IC50 = 42.4 nM vs BRD3
LE = 0.34
IC50 = 36.1 nM vs BRD4
LE = 0.35
N N
3
Me
2
(R)-GW841819X
IC50 = 29.9 nM vs BRD2
LE = 0.33
IC50 = 28.4 nM vs BRD3
LE = 0.33
IC50 = 15.5 nM vs BRD4
LE = 0.34
Me
Me
Me
Me
Me
Ph
(+)-JQ1
IC50 = 77 nM vs BRD4(1)
LE = 0.32
IC50 = 33 nM vs BRD4(2)
LE = 0.34
Me
N N
O
Me
N
N
N
MeO
N
N
Cl
Cl
Me
NH
OMe O
Cl
4
Alprazolam
OH
5
Triazolam
6
RVX-208
Figure 5. Bromodomain and extra-C terminal domain ligands. (R)-GW841819X (1) discovered
by phenotypic screening. Subsequently developed molecule GSK525762A ( 2, I-BET762) [3,4] , the
related benzodiazepines alprazolam (4) and triazolam (5) , (+)-JQ1 (3) reported by Filippakopoulos
et al. [7] and the Resverlogix compound RVX-208 (6) [5,6] . IC50 values for 1 and 2 were determined
by a fluorescence resonance energy transfer assay, and for 3 by an amplified luminescent proximity
homogeneous assay (AlphaScreen).
future science group
www.future-science.com
183
Review | Jennings, Measures, Wilson & Conway
Box 4. Fragment-based drug discovery.
Since its inception in the late 1990s, fragment-based drug discovery (FBDD) has evolved into a credible alternative approach to
high-throughput screening (HTS) for the generation of drug leads [85,86] . FBDD differs from standard HTS in a number of important
ways. Most significantly, the molecules (‘fragments’), that are screened are typically smaller than for a standard HTS library and
adhere to the ‘rule of three’ with a relative molecular mass of <300 Da, a calculated log P of ≤3, up to three hydrogen bond acceptors
and up to three hydrogen bond donors [87,88] . The overall FBDD process can be simplified to the following steps: fragment library
design of hundreds or thousands of low-molecular-weight compounds; fragment screening using a range of highly sensitive in vitro
biophysical techniques, including NMR spectroscopy, x-ray crystallography and fluorescence-based thermal shift assays to detect low
affinity hits (0.1–10 mM); fragment elaboration, involving iterative cycles of synthesis, guided by computational docking and
bioaffinity measurement [88] . As a result of characteristics of the fragment library, the hit molecules identified by FBDD are small and
hence their absolute affinity for the target tends to be low. This fact presents a challenge in the detection of fragment binding, which
is why sensitive techniques such as saturation transfer difference NMR- and mass spectrometry-based approaches can be particularly
useful in this context. The concept of ligand efficiency is also important as it identifies molecules that make high-quality interactions
with the target protein, despite their low affinity [43] . Improved potency and selectivity of final compounds can be achieved by
combining interactions observed in multiple fragments that bind to the target in close proximity to each other, building out from an
initial fragment to develop further interactions with the target protein or potentially displacing water molecules that are present in the
binding site. FBDD has two important advantages over techniques that identify lead compounds of a higher molecular mass. First, a
much greater proportion of chemical space can be sampled using a fragment library than with a HTS technique. Second, as the
detected fragment hits must make high-quality interactions with the target for binding to be detected at all, the optimized
compounds tend to have high ligand efficiency. These advantages have attracted interest in this methodology from both the
pharmaceutical industry and academic community alike, with more than ten clinical candidates developed using this strategy [89] .
Notably, in 2011 the first fragment-based drug, vemurafenib, was approved for the treatment of metastatic or unresectable
melanoma [90] .
Fragment library design
Small library
Fragment screening
Biophysical techniques
Low affinity hits
High ligand efficiency
MW <300 Da
Fragment elaboration
Iterative synthesis
High affinity compounds
Maintain ligand efficiency
Figure 6. Summary of stages in fragment-based drug discovery. Initial
small-ligand hits exhibit high-quality interactions, for example, hydrogen bonds to
asparagine residues shown left. Expansion of these hits can increase affinity by
interacting with further residues or displacing water molecules (yellow hexagon
displacing water shown as red sphere), or increase selectivity by interacting with
surrounding variable regions (blue triangle) shown right.
184
Future Med. Chem. (2014) 6(2)
might exert their cellular activity through a
novel class of targets.
In order to define the novel target class a chemoproteomics approach was employed. To facilitate this approach, an affinity matrix that incorporated a derivative of I-BET762 (7; Figure 7) was
produced. A control matrix, which incorporated
an inactive analog of I-BET762, was also made.
Using these matrices, affinity chromatography
was conducted using HepG2 cell lysates. A number of proteins were retained on the active affinity matrix but did not bind to the matrix produced using the inactive analog, indicating that
they bound selectively to the active compound.
In addition, these proteins were competed off
the matrix by addition of 1, suggesting that they
were the specific protein targets of the active
compound. Using LC–MS/MS, these proteins
were identified as BRD2, BRD3 and BRD4,
three members of the bromodomain-containing BET family of proteins. As these proteins
were the main interacting partners detected, the
BET family of BCPs seemed to be the molecular
targets for the identified compounds.
To ascertain which part of the BET BCPs
bound to the benzodiazepine, f lag-tagged
cDNA constructs of BRD2, containing the
entire sequence, the N-terminal residues
(1–473) and the C-terminal residues (473–
801) were transfected into HepG2 cells. Affinity chromatography of the lysates afforded
only the full length and N-terminal region of
BRD2, showing that benzodiazepines bind
future science group
Phenotypic screening & fragment-based discovery of small-molecule bromodomain ligands
to the N-terminal bromodomain-containing
region of the BET proteins. To assess the
selectivity and nature of compound binding,
N-terminal portions containing the tandem
bromodomains, both together and individually,
of BRD2, BRD3 and BRD4 were expressed in
Escherichia coli. Isothermal titration calorimetry and surface plasmon resonance confirmed
direct and selective high affinity simultaneous
2:1 binding of 1 with the tandem bromodomains. Furthermore, a fluorescence resonance
energy transfer assay, used to test the ability of
1 and 2 to compete with tetra-acetylated histone
H4 peptide H4(KAc)4 in recombinant BRD2,
BRD3 and BRD4 constructs, showed dosedependent inhibition of peptide binding to the
bromodomain-containing proteins (IC50 values
of 16–42 nM). This combination of results support the conclusion that the benzodiazepines
act as inhibitors of the bromodomain–histone
acetyl–lysine interaction.
x-ray crystal structures of compounds 1 and
2 bound to both the first and second bromodomains of BRD2 [BRD2(1) and BRD2(2)] and
BRD4 [BRD4(1) and BRD4(2)] showed that
the triazolobenzodiazepines occupy the KAcbinding pocket and peptide-binding groove, and
hence can displace the histone peptide H4(KAc).
This work helped to define the key interactions required for a potent BET bromodomain
N N
Me
NH2
H
N
O
| Review
Key Terms
Affinity matrix: Pull-down
assays are a commonly used
technique to identify binding
partners through attachment of
an affinity label.
N
N
x-ray crystal structures:
H3CO
7
Cl
Figure 7. The structure of the I-BET762
derivative 7 that was attached to agarose
beads for chemoproteomics experiments.
ligand (Figure 8A & B), and this subject has been
extensively reviewed elsewhere [4,19–26].
To confirm that the BET inhibition activity
exhibited by I-BET762 was responsible for the
modulation of ApoA1 levels, siRNA knockdown
of BRD4 was carried out. Progressive knockdown
of BRD4 induced a gradual and significant upregulation of ApoA1, whereas knockdown of BRD2
and BRD3 had no effect on ApoA1 expression.
This suggested that inhibiting the histone–BRD4
interaction in particular is responsible for upregulation of ApoA1. With the structure–activity relationships (SAR) and high affinity of I-BET762
established, an in vitro fluorescence anisotropy
displacement assay was developed to allow profiles of 150 compounds from the benzodiazepine
Protein structures can be
obtained from refining x-ray
diffraction patterns from single
crystals. This requires high
purity protein and often lengthy
trials in crystallization
conditions.
siRNA: siRNA is used to
prevent translation of specific
proteins, and can reinforce an
understanding of mode of action
through replication of resulting
phenotype.
ZA channel
D104
D104
V103
P102
Y113
Peptide
KAc
I-BET762
Q101
Y113
Q101
N156
I-BET762
P98
F99
W97
W97
P98
N156
F99
WPF shell
Figure 8. Binding mode of I-BET762 in BRD2. (A) x-ray crystal structure of I-BET762 (2) bound to
human BRD2(1) (PDB ID: 2YEK; carbon = yellow [4] ) showing the occupancy of the e-N-acetylated
lysine residues (KAc)-binding pocket by 2. The structure is overlaid with an x-ray crystal structure of
human BRD2(1) in complex with the diacetylated histone peptide H41–15KAc12 (PDB ID: 2DVQ;
carbon = purple [26] ), demonstrating that 2 occupies part of the peptide-binding groove. (B) The
3-methyl-1,2,4-triazole of 2 acts as an effective KAc mimic and forms similar interactions to KAc
(carbon = purple). N(2) is positioned to form a water-mediated hydrogen bond to Y113 and N(3)
accepts a hydrogen bond from N156. Structures were aligned using the ‘cealign’ command in PyMOL.
future science group
www.future-science.com
185
Review | Jennings, Measures, Wilson & Conway
series to be determined. This displayed excellent
correlation between BET bromodomain binding
affinity and ApoA1 upregulation.
„„
The
discovery of I-BET151
A second chemotype of ApoA1 upregulators was
discovered by GSK and developed in parallel to
I-BET762. This series of compounds (including 8–10) was based on the isoxazoloquinoline
scaffold, and it was rapidly established that the
3,5-dimethylisoxazole motif was essential for
activity (Figure 9) [27,28].
x-ray crystal structures of 10 bound to BRD4
demonstrated that the 3,5-dimethylisoxazole
motif resided in the KAc-binding pocket of
BRD4 [11,29]. The nitrogen atom of the isoxazole ring forms a water-mediated hydrogen bond
with Y97 [BRD4(1) numbering] and the oxygen atom forms a direct hydrogen bond with
the highly conserved N140 residue (Figure 10A)
[30]. The effective occupancy of the WPF shelf
region (Box 1) by the 4-aniline group is believed
to contribute significantly to binding affinity
(Figure 10B). Additional packing of the methoxy
group against side chain I162 was also evident.
Differential scanning fluorimetry (DSF) ana­
lysis demonstrated that 10 was selective for the
BET bromodomains, with only a small change
in thermal shift (ΔTm) observed for CREBBP,
and no change observed for a number of other
phylogenetically diverse bromodomains that
were tested [11]. Compound 10 displayed broad
anti-inflammatory properties, and therapeutic
efficacy against clonogenic potential of mixedlineage leukemia-fusion-driven leukemic cell
lines was established in vivo in both murine
and human models, with abrogation of BRD3/4
binding to chromatin leading to apoptosis. Mirguet et al. have also reported optimization of
I-BET151 with the synthesis of a series of napthyridine derivatives, some of which displayed
favorable rat PK properties and efficacy in a
mouse model of inflammation [29].
„„Mitsubishi
compounds, (+)-JQ1
& OTX015
Research by Mitsubishi disclosed in two patents
published in 2006 [101] and 2009 [102] identified
compounds with a very similar chemotype to
I-BET762 (2) using a phenotypic screen. It is
possible that the similarity in chemotype results
from both companies using similar screening files
in the phenotypic assays [31]. The inclusion of
benzodiazepine-type scaffolds in screening files is
common, as these compounds are recognized as
‘privileged structures’, which have a high propensity to yield biologically active compounds [32].
Molecular modeling using the thienotriazolodiazepine scaffold disclosed by Mitsubishi and the
apo-structure of BRD4(1), led to the design of
the derivative (+)-JQ1 by Filippakopoulos et al.
[7]. The triazolodiazepine moiety is a key feature in the US FDA-approved drugs alprazolam
(4; Figure 5) and triazolam (5), which bind to
N
N
O
O
HN
O
H2N
Me
OMe
Me
H 2N
N
O
Me
N
HN
OMe
Me
N
N
O
O
8
GW694481
EC170 = 0.5 µM
Me
N
HN
Me
9
EC170 = 0.22 µM
Me
N
N
10
I-BET151
EC170 = 0.09 µM
IC50 = 0.79 µM vs BRD4(1)
LE = 0.28
Figure 9. The optimization of GW694481 (8) to give the potent and selective
3,5-dimethylisoxazole-based BET bromodomain probe I-BET151 (10) [11,27,28] . Compound 8
was identified in a high throughput screen to evaluate upregulated ApoA1 expression in HepG2
cells. The 3,5-dimethylisoxazole motif, indicated in green, occupies the KAc-binding pocket of BRD4
and acts as a KAc mimic. Optimization of the compounds included modification of the aniline
nitrogen atom and cyclisation of the 3-position carboxamide giving the ring, resulting in 10. The red
substituents occupy the WPF shelf. EC170 values (effective concentration required to increase
luciferase activity by 70% after 18 h) were determined using an ApoA1 luciferase reporter gene
assay and IC50 values were determined using a fluorescence anisotropy assay.
186
Future Med. Chem. (2014) 6(2)
future science group
Phenotypic screening & fragment-based discovery of small-molecule bromodomain ligands
Peptide
ZA channel
KAc
D88
| Review
Y97
D88
I-BET151
Y97
V87
W81
N140
Q85
N140
P86
I-BET151
Q85
P82
F83
F83
P82
W81
WPF shell
Figure 10. Binding mode of I-BET151 in BRD4. (A) The isoxazole of I-BET151 (10) forms similar
interactions with the KAc-binding pocket to KAc and acts as an effective KAc mimic [11] . (B) x-ray
crystal structure of 10 bound to human BRD4(1) (PDB ID: 3ZYU; carbon = yellow [11] ). The structure
is overlaid with the x-ray crystal structure of human BRD4(1) in complex with the diacetylated
histone peptide H41–12KAc5KAc8 (PDB ID: 3UVW; carbon = purple [30] ), demonstrating that the
isoxazole moiety resides in the KAc-binding pocket, the pyridyl group binds to the WPF shelf and
the quinoline nitrogen atom accepts a hydrogen bond from one of the ZA channel water molecules.
Structures were aligned using the ‘cealign’ command in PyMOL.
g-aminobutyric acid receptors and are used in
the treatment of anxiety neurosis and hypnotic
therapy. To reduce the likelihood of the compounds binding to g-aminobutyric acid receptors,
a tBu ester was incorporated at the C6 position,
which is known to reduce this activity [33]. DSF
and isothermal titration calorimetry showed that
(+)-JQ1 was highly selective for the BET bromodomains, displaying K D values of approximately
50 nM and 90 nM for BRD4(1) and BRD4(2),
respectively. No binding to the bromodomains
of CREBBP or WDR9 was observed. In addition, the enantiomer of the active compound,
(−)-JQ1, showed no affinity for a wide range of
bromodomains. A peptide displacement-based
amplified luminescent proximity homogeneous
assay (AlphaScreen), used to measure the ability of (+)-JQ1 to compete with a H4(KAc)4
peptide for BRD4 binding, gave IC50 values for
(+)-JQ1 of 77 nM and 33 nM for BRD4(1) and
BRD4(2), respectively. Co-crystallization of (+)JQ1 with BRD2(2) and BRD4(1) revealed that
the triazoles bound in a very similar fashion to
that of I-BET762 and KAc, forming hydrogen
bonds to the conserved asparagine residue and
the KAc-binding pocket water molecule. (+)-JQ1
also forms similar interactions to I-BET762 with
the WPF shelf region of the BET bromodomains.
A fluorescence recovery after photo bleaching assay was used to show that (+)-JQ1 inhibits
future science group
the interaction of BRD4 with chromatin, and
could directly target the BRD4-NUT onco­
protein, responsible for NUT midline carcinoma (NMC), a rare but highly lethal cancer.
(+)-JQ1 induced dose- and time-dependent differentiation of NMC cell lines. Furthermore,
there was significant reduction in 18F-fluorodeoxyglucoseuptake when xenograft NMC mice
were treated with (+)-JQ1, without toxicity
and weight-loss. The mice also showed marked
tumor regression and prolonged overall survival
after treatment with (+)-JQ1.
OTX015 is thought to be structurally similar to (+)-JQ1 and is based on structures in the
Mitsubishi patents. This compound has been
in-licensed by Oncoethix from Mitsubishi.
„„RVX-208
In a similar manner to Nicodeme et al., Bailey
and co-workers from Resverlogix used phenotypic screening with a HepG2 cell-based assay
to identify compounds that upregulated ApoA1
expression [5]. The quinolinone-based compound RVX-208 (6; Figure 5) was identified,
which significantly increased ApoA1 mRNA and
high-density lipoprotein (HDL) mass in a dosedependent manner. PK and bioavailability assays
in cynomolgus monkeys showed that RVX-208
had favorable pharmacological properties including low systemic clearance, a moderate volume
www.future-science.com
187
Review | Jennings, Measures, Wilson & Conway
of distribution, a short half-life and good oral
absorption. Furthermore, mouse studies showed
localization of RVX-208 in ApoA1-expressing
tissues. Pulse-chase ana­lysis was carried out to
monitor movement of ApoA1 upon treatment
with RVX-208, as observed by autoradiography.
A 10-min intracellular pulse of [35S]-methionine
and [35S]-cysteine was delivered. This pulse
was chased by unlabelled methionine and cysteine. Production of ApoA1 enriched with the
[35S]-label was observed at the time of the pulse
in RVX-208 treated cells, indicating that RVX208 is indeed increasing production of ApoA1.
The unlabelled chase enabled the extracellular
passage of labeled ApoA1 to be monitored over
time; after 150 min intracellular labeled ApoA1
was reduced but extracellular labeled ApoA1 was
increased, consistent with the belief that cholesterol flux is responsible for the protective role of
HDL against atherosclerosis. Furthermore, dosedependent increases in ApoA1 and HDL cholesterol levels in a preclinical model of atherosclerosis
(RVX-208-treated male African green monkeys)
agreed with the phenotypic screening results in
cells. In addition, a combination of 2D-PAGGE,
high-performance liquid chromatography, nondenaturing PAGGE and densitometric scanning
demonstrated that RVX-208 increased the size
of HDL particles, indicating good therapeutic
potential for RVX-208. Following the promising
results in African green monkeys, and the observation of larger HDL particles, RVX-208 was
tested in healthy humans, showing a 10% rise in
ApoA1 levels in comparison with placebo. It was
only subsequent to the development of this molecule that its mode of action was discovered to be
mediated by BET bromodomain inhibition[6,31].
Since then, there has been some suggestion that
NO2
H
N
+
NH3
Me
NO2
H
N
+
NH3
H2N
11
IC50 = 1.6 µM vs PCAF
LE = 0.54
(ELISA peptide competition assay)
EC50~10 µM
(HIV-1 LTR-luciferase reporter gene assay)
EC50~2.8 µM
(syncytial assay in C8166 cells)
CC50~17.7 µM
(MTT cytotoxicity assay in C8166 cells)
12
EC50 = 0.63 µM
(syncytial assay in C8166 cells)
CC50~8.0 µM
(MTT cytotoxicity assay in C8166 cells)
Figure 11. The PCAF bromodomain ligand 11 reported by Zhou et al. [1,36]
and the cytotoxic compound 12 reported by Zheng, Wang et al. [37] .
ELISA: Enzyme-linked immunosorbent assay; LTR: Long terminal repeat.
188
Future Med. Chem. (2014) 6(2)
RVX-208 is a weak BET bromodomain ligand,
but interestingly shows some selectivity for the
second bromodomain over the first [31].
Fragment-based approaches to the
discovery of bromodomain ligands
„„Fragment-based approaches to the
development of ligands for the PCAF
bromodomain
The first structural information on both bromo­
domains and their ligands was published by
Zhou and colleagues during their investigation
aimed at the development of new HIV therapeutics [34]. Initial work showed that Tat transcriptional coactivator recruitment requires binding
of KAc50 (Tat-K50Ac) to the bromodomain of
the coactivator PCAF [35]. Having established
the importance of this interaction, efforts then
focused on the development of a small-molecule
inhibitor of the Tat-KAc50–PCAF interaction
leading to disruption of the viral replication pathway [1]. To identify lead compounds, an NMRbased screen of small-molecules selected from
the ChemBridge Corporation collection was
conducted. Analysis of ligand-induced protein
signal changes in 2D 15N HSQC spectra, with an
emphasis placed on compounds that bound the
bromodomain near to the KAc pocket, identified an amino nitrophenyl derivative as a starting
point for SAR studies. An ELISA assay, in which
the ability of ligands to inhibit the interaction
of PCAF with a biotinylated Tat-KAc50 peptide, was used to determine the IC50 values for a
set of 24 compounds, of which 11 was the most
effective (Figure 11) [1,36,37].
An NMR structure of compound 11 bound
to the bromodomain of PCAF suggests that the
2-nitro group forms a hydrogen bond with the
phenol of Y802 and possibly Y809. The terminal ammonium group interacts electrostatically
with the side chain carboxylate of E750. The
methyl group of 11 sits in the hydrophobic KAc
binding pocket and this interaction is thought
to contribute to the affinity of 11 for the PCAF
bromodomain.
This work was extended by Wang et al. who
synthesized a range of compounds that included
those compounds previously reported by Zhou,
and some novel additions [37]. These compounds
were investigated in two assays. In the first assay
C8166 cells were infected with HIV-1IIIB. The
ability of compounds to inhibit the cytopathic
effect of the virus was measured by counting
the number of syncytium (multinucleated giant
cells) and a percentage inhibition of syncytial cell
future science group
Phenotypic screening & fragment-based discovery of small-molecule bromodomain ligands
Me
Me
N
O
Me
13
MS2126
Me
O
N
O
N
O
14
MS7972
| Review
O
Me
Me
Me
NH
O
15
MS9802
Me
16
MS0433
Figure 12. The CREBBP bromodomain ligands 13–16 reported by Sachchidanand et al. [2] .
formation was quoted. The cytotoxicity of the
compounds in C8166 cells was also measured
using a tetrazolium dye (MTT) colorimetric assay.
Compound 12 showed improved activity over
compound 11 in these assays. However, it should
be noted that there is no direct evidence that this
increase in activity can be attributed to inhibition of the PCAF bromodomain. Indeed, members of a second series of compounds, which had
previously been shown by Zhou to be essentially
inactive in binding the PCAF bromo­domain,
were observed to be effective in the syncytial cell
formation and MTT assays [1,37]. Hence caution
should be exercised when interpreting these data.
„„Fragment-based
approaches to the
development of ligands for the CREBBP
bromodomain
CREBBP and its paralog, p300, interact with
at least 400 protein partners and are key nodes
in the mammalian protein−protein interactome
[38]. CREBBP possesses both a bromodomain
and histone acetyltransferase (HAT) catalytic
activity, and hence can be viewed as both a
‘reader’ and ‘writer’ of the histone code. Work by
Mujtaba et al. demonstrated that the CREBBP
bromodomain binds to KAc382 of p53, linking it to the DNA damage repair mechanisms
[39]. To develop small molecules that inhibit
the p53–CREBBP bromodomain interaction,
Sachchidan et al. used a fragment-based screening approach [2]. A knowledge-based library of
approximately 200 compounds was constructed
from a approximately 14,000-member ChemBridge Corporation collection. Electron-rich
small molecules were favored in order to promote
interactions with the positively charged rim of
the CREBBP KAc-binding pocket. By monitoring chemical shift changes of protein backbone
amide resonances in 2D 1H-15N-HSQC spectra
acquired in the presence and absence of a mixture of eight different fragment sets, 14 CREBBP
bromo­domain ligands were identified and classified into four groups (one compound from each
future science group
group is shown in Figure 12). Small-molecule
inhibition of the CREBBP bromodomain was
then measured using an assay in which a chemical ligand competes against binding of a biotinylated p53-AcK382 peptide immobilized to
the GST-fusion CREBBP bromodomain, as
assessed by western blot. All but one of these
14 compounds showed selectivity for CREBBP
over PCAF as determined by weighted chemical shift perturbations. Two of the compounds,
MS2126 (13; Figure 12) and MS7972 (14), were
found to almost completely block the CREBBP
bromodomain–p53 interaction at 100 µM and
50 µM, respectively. Results from a cell-based
study showed that compounds 13 and 14 dramatically decreased p53 expression levels in
response to doxorubicin stimulation, validating the CBP–p53 interaction as an interesting
therapeutic target.
The same group screened a structurally
diverse library of 3000 compounds using an
NMR-based assay and identified the azobenzene moiety as a new scaffold for the inhibition
of the CREBBP bromodomain [40]. This screen
generated ten hits, some of which contained the
azobenzene motif, suggesting that compounds
containing this moiety were good candidates
for optimization. With the aim of improving
the potency and selectivity for the CREBBP
bromodomain, 26 derivatives of compound 17
(Figure 13) were synthesized and evaluated in
SO3H
SO3H
Me
N
N
N
Me
OH
17
MS456
N
Me
H2N
OH
18
Ischemin
Figure 13. The CREBBP bromodomain ligands 17 and 18 reported by Borah
et al. [40] .
www.future-science.com
189
Review | Jennings, Measures, Wilson & Conway
a p53-dependent, p21 luciferase reporter gene
assay, which measured the ability of each compound to modulate DNA-damage-induced p53
activation in U2OS cells. Each compound had
a stronger affinity than the initial hit 17, which
gave 4.6% inhibition of p53 activity at 50 µM
in U2OS cells. A tryptophan fluorescence-based
binding assay was then used to determine the
K D for the six compounds that showed >80%
inhibition of the interaction between a biotinylated p53-AcK382 peptide and the GST-tagged
CREBBP bromodomain. Of these, ischemin (18)
exhibited a K D = 19 µM and cellular activity with
IC50 = 5 µM in the luciferase assay. Selectivity of
18 against CREBBP was assessed using the tryptophan fluorescence-based binding assay, which
revealed that this compound was up to fivefold
selective over the bromodomains of BRD4(1),
PCAF, BAZ1B and BAZ2B.
Ischemin (18) was found to completely prevent U2OS cells from undergoing doxorubicininduced cell cycle arrest. Further studies on the
effects of 18 on p53 stability and function as a
transcription factor revealed its mode of action
involves alteration of post-transcriptional modification states on p53 and histones. These results
confirmed that upon doxorubicin exposure, ischemin inhibits p53-induced p21 activation by preventing recruitment of CREBBP [40]. RT-PCR
array experiments indicated that 18 can reduce
doxorubicin-induced expression of p53 target
genes. Finally, studies involving primary neonatal rat cardiomyocytes and TUNEL assay measurements indicated that 18 inhibits doxorubicininduced apoptosis, demonstrating that 18 is cellpermeable and functions as a cellular protective
agent against myocardial ischemic stress.
Fragment-based discovery of the
3,5-dimethylisoxazole moiety as a
KAc mimic
The bromodomain ligands described above
have been shown to reside predominantly in
O
Me
H
N
O
N
Me
N
H
N
O
Me
Me
N
O
Me
19
IC50 = 1.9 mM vs CREBBP
LE = 0.54
20
21
IC50 = 4.8 µM vs BRD4(1)
LE = 0.39
Figure 14. The development of the lead BET bromodomain ligand 21
reported by Hewings et al. [14] .
190
N
Future Med. Chem. (2014) 6(2)
the peptide-binding groove rather than the
bromodomain KAc-binding pocket. However,
a number of fragments have now been identified that reside in the KAc-binding pocket
and closely mimic the interactions formed by
KAc and the bromodomain [12–14,17]. Of these
fragments, the 3,5-dimethylisoxazole motif
has been independently discovered by a number of groups and has emerged as an excellent
KAc mimic that can form the basis of ligands
for the BET [11,13,14,27,29,41] and CREBBP [14,15]
bromodomains.
Development of an AlphaScreen-based
peptide displacement assay by Philpott et al.
revealed that the solvent N-methylpyrrolidone
(19; Figure 14) binds to the CREBBP bromodomain in a weak, but ligand-efficient, manner
[14,16,42,43]. Screening of a range of N-methylpyrrolidone analogs identified the dihydroquinazolinone (DHQ; 20) core as a promising
KAc mimic and the 3,5-dimethylisoxazolecontaining DHQ derivative (21) displayed an
unexpectedly low IC50 value of approximately
7 µM against BRD4(1) in the initial assay [14].
Although 21 was later shown to inhibit
BRD4(1) with an IC50 value of 4.8 µM, this
was not the compound bound in an x-ray crystal
structure with BRD4(1). The crystal structure
indicated that the DHQ unit was susceptible to
oxidation and had conjugated with ethylene glycol from the crystallization buffer (22; Figure 15),
via presumed formation of an iminium ion
[14]. Consequently, the dihydroquinazolinone
unit was too large to access the KAc-binding
pocket and hence the 3,5-dimethylisoxazole was
observed to bind in the pocket and act as a KAc
mimic. Additionally, the ethylene glycol unit
occupied the WPF shelf in BRD4(1) suggesting that this might be a favorable interaction for
BET affinity. Initial docking studies indicated
that a 3,5-substituted phenyl ring would project substituents into the WPF shelf and the ZA
channel. Compound 23, which possesses secondary alcohol and ethoxy substituents showed an
IC50 value of 4.8 µM for BRD4(1) and increased
selectivity for the BET bromodomains over
CREBBP compared with 22 . An x-ray crystal
structure of 23 bound to BRD4(1) indicated
that the methyl group of the secondary alcohol
occupies the WPF shelf while the ethoxy group
is oriented into the ZA channel.
Optimization of the methyl group to a phenyl
ring improved the interaction with the WPF
shelf, giving 24. Removing the ethyl group to
give the phenol (25) allowed hydrogen bonding
future science group
| Review
Phenotypic screening & fragment-based discovery of small-molecule bromodomain ligands
with one of the ZA channel water molecules
and provided a BET bromodomain ligand with
an IC50 value of 384 nM (average value of the
(R)- and (S )-enantiomers). Addition of an
acetate group on the phenol (26) gave a slightly
enhanced IC50 value of 371 nM for BRD4(1)
and improved the selectivity over CREBBP
[16]. Compounds 25 and 26 were evaluated for
their cytotoxic effects on MV4;11 cells and
were shown to have IC50 values of 794 nM and
616 nM, respectively.
The 3,5-dimethylisoxazole moiety was independently identified by Bamborough et al.
[13], ultimately leading to the development of a
range of potent and selective BET bromodomain
ligands. The 3,5-dimethyl-4-phenylisoxazole
fragment (27) showed 32% inhibition of BRD3
and 26% inhibition of BRD4 at 200 µM. A
comparison of the x-ray crystal structures of the
fragment 27 (Figure 16), and I-BET762 led to
the development of a 3D-pharmacophore model
for BET bromodomain binding. A search of
commercially available compounds using the
pharmacophore gave a number of hits based
on the 3,5-dimethylisoxazole with a sulfonamide substituent on the phenyl ring meta to
the 3,5-dimethylisoxazole [13]. Both the pharmacophore and docking models indicated that
the sulfonamide was effective at directing the
attached lipophilic substituent into the WPF
shelf. x-ray crystal structures of these ligands
bound to BRD2(1) confirmed that the sulfonamide and its substituent were oriented as
expected. Initial hits possessed a cyclopropyl
ring binding to the WPF shelf (28) but this was
optimized to a cyclopentyl ring (29). A substituent para to the 3,5-dimethylisoxazole was
found to be beneficial, with methyl, methoxy
and hydroxyl all tolerated. A limitation of the
sulfonamide series was their low solubility. This
problem was addressed by the incorporation of
a morpholine solubilizing group attached to an
oxygen atom para to the 3,5-dimethylisoxazole, giving compound 30. This compound has
low µM IC50 values for the BET bromodomains
(determined using a time-resolved fluorescence
resonance energy transfer assay). DSF ana­lysis
indicated reasonable selectivity for the BET bromodomains over other BCPs evaluated. Compound 30 showed an IC50 value of 3.0 µM in
a cellular assay based on inhibiting release of
IL-6, a cytokine shown to be regulated by BET
ligands [3].
Hay et al. developed a series of isoxazoles
linked to a 5,6-fused ring system targeted at the
future science group
H
N
O
Me
OEt
Me
Me
N
O
Me
HO
HO
O
N
Me
O
N
Me
23
IC50 = 4.8 µM vs BRD4(1)
LE = 0.39
22
OR
OEt
Me
HO
Ph
Me
Me
O
N
HO
O
Ph
25 IC50 = 384 nM vs BRD4(1)
LE = 0.41
R = Ac 26 IC50 = 371 nM vs BRD4(1)
LE = 0.36
R=H
Me
N
24
IC50 = 640 nM vs BRD4(1)
LE = 0.36
Figure 15. The optimization of the lead compound 22 to give the BET
bromodomain ligands 25 and 26 reported by Hewings et al. [14,16] . The
3,5-dimethylisoxazole motif, indicated in green, occupies the KAc-binding pocket
of BRD4 and acts as a KAc mimic. The red substituents occupy the WPF shelf and
the blue substituents are oriented into the ZA channel. IC50 values were
determined using an AlphaScreen assay. Ligand efficiency was calculated using
the equation (pIC50 × 1.4 kcal/mol)/heavy atom count [43] .
Me
O S
O
Me
Me
O
NH
N
O
Me
N
28
IC50 = 3.2 µM vs BRD4
LE = 0.27
27
O
N
O
Me
O
O S
NH
O
Me
N
30
IC50 = 2.6 µM vs BRD4
improved solubility
LE = 0.24
HO
Me
O
O S
NH
Me
O
N
29
IC50 = 2.5 µM vs BRD4
LE = 0.34
Figure 16. The optimization of the lead fragment 27 to give the BET
bromodomain ligand 30 reported by Bamborough et al. [13] . The green
substituent occupies the e-N-acetylated lysine residues-binding pocket and acts as
aKAc mimic, the red substituents occupy the WPF shelf. IC50values were
determined using a time-resolved fluorescence resonance energy transfer assay.
www.future-science.com
191
Review | Jennings, Measures, Wilson & Conway
BET bromodomains [15]. Introduction of a fused
5-membered ring to the 4-phenyl-3,5-dimethylisoxazole scaffold 27, afforded molecules possessing a new substitution pattern in this class
of bromodomain ligands.
An AlphaScreen assay identified compound
31 (Figure 17) as a modest BRD4(1) bromodomain with low affinity for the CREBBP bromodomain, whereas compound 33 was more
potent but had comparable affinity for both
the BRD4(1) and CREBBP bromodomains.
Optimization of the indanone core led to the
identification of the indanol 32 as a more potent
BRD4(1) ligand with some selectivity over the
CREBBP bromodomain. The x-ray crystal
structure of 32 bound to BRD4(1) shows that
the aryl ring occupies the WPF shelf. However,
attempts to further optimize this scaffold led
to dehydration, forming the corresponding,
inactive, indenes.
Optimization of the benzimidazole series
first required development of a regioselective synthesis to allow both regioisomers to
be obtained separately. These regioisomers
result from substitution of one of the two
imidazole nitrogen atoms, labeled either g or
e for clarity. The g-substituted benzimidazole
derivatives were generally more potent against
BRD4(1) than the e-substituted compounds.
N
Me
O
Me
N
Me
O
HO
Me
N
32
IC50 = 1.3 µM vs BRD4(1)
LE = 0.36
IC50 = 20 µM vs CREBBP
LE = 0.29
Me
Nγ
O
N
O
N
Me
Cl
33
IC50 = 6.3 µM vs BRD4(1)
LE = 0.46
31
IC50 = 63 µM vs BRD4(1)
LE = 0.33
Me
N
Me
N
H
O
The 4-cyano-derivative 36 showed an IC50
value of 200 nM for BRD4(1) and low affinity for the CREBBP bromodomain. Analysis by
DSF showed a ΔTm = 3.2°C against BRD4(1)
and ΔTm = 1.1°C against the CREBBP bromodomain, with thermal shifts of ≤0.1°C
against a panel of phylogenetically diverse
bromodomains.
Researchers at Constellation Pharmaceuticals
(MA, USA) have recently reported an isoxazole-based BET bromodomain ligand that was
developed using hits from a fragment screen [41].
This screen identified several compounds with
potencies in the micromolar range. Analysis of
co-crystals of these compounds with BRD4(1)
led to the selection of 37 (Figure 18), an analog
of 27, as the lead to be progressed, as it mimicked the key interactions formed between
BRD4(1) and KAc. Based on the structure of
(+)-JQ1 [7], compound 38 was synthesized, in
which the triazole is replaced with the isoxazole ring. This compound (38) possessed a lower
IC50 value for BRD4(1) than (+)-JQ1 when
evaluated in the same assay. Changing the t Bu
ester to a carboxamide (39) resulted in an IC50
value of 26 nM, which was similar to the values
displayed by (+)-JQ1 and I-BET151 in the same
assay. A subsequent ana­lysis of the SAR around
the aryl ring that occupies the WPF shelf did
35
IC50 = 790 nM vs BRD4(1)
LE = 0.36
Cl
N
ε
N
Nγ
Me
N
Me
O
N
34
IC50 = 2.5 µM vs BRD4(1)
LE = 0.33
Me
NC
Me
O
N
36
IC50 = 200 nM vs BRD4(1)
LE = 0.38
Figure 17. The 3,5-dimethylisoxazole-based BET bromodomain ligands reported by Hay et al. [15] . The green substituent
occupies the KAc-binding pocket and acts as a KAc mimic, the red substituent occupies the WPF shelf. IC50 values were determined
using an AlphaScreen assay.
192
Future Med. Chem. (2014) 6(2)
future science group
Phenotypic screening & fragment-based discovery of small-molecule bromodomain ligands
Cl
Cl
NH2
O
Me
O
Me
N
Me
Me Me
Me
S
O
O
N
Me
S
N
Me
38
IC50 = 290 nM vs BRD4(1)
LE = 0.30
NH2
N
Me
O
Me
37
IC50 = 33 µM vs BRD4(1)
LE = 0.48
| Review
O
N
39
IC50 = 26 nM vs BRD4(1)
LE = 0.39
Figure 18. The optimization of the aminoisoxazole (37) fragment to give the
isoxazoleazepine BET bromodomain ligand 39, reported by Gehling et al. [41] . The green
substituent occupies the KAc-binding pocket and acts as a KAc mimic, the red substituent occupies
the WPF shelf. IC50 values were determined using an AlphaScreen assay.
not improve the compound potency, indicating that the original 4-chlorophenyl group
was optimal. An x-ray crystal structure of 39
bound to BRD4(1) confirmed that the compound bound as expected. The isoxazole resides
in the KAc-binding pocket and acts as a KAc
mimic, and the 4-chlorophenyl group binds to
the WPF shelf. The carboxamide NH 2 group
forms a water-mediated hydrogen bond with the
carbonyl oxygen atom of N140. Compound 39
was shown to be a potent BET bromodomain
probe in cellular assays and PK profiling in rat
and dog demonstrated suitable characteristics
for further in vivo experiments.
H
N
H
N
S
O
N
The DHQ fragment (20; Figure 19), which was
identified as a potentially useful bromodomain
ligand by a number of groups [12,14,42], formed
the basis of work by Fish et al. to develop a BET
bromodomain ligand, PFI-1 (43) with a chemotype distinct from other reported compounds.
Crystallization of the 6-bromo-DHQ (41)
derivative confirmed that the DHQ moiety can
act as a KAc mimic and that substitution at the
6-position is tolerated by BRD4(1). Although
the bromine atom was a potentially useful synthetic handle via a variety of coupling reactions,
this approach would predominantly produce
H
N
H
N
Me
O
N
O O
H
N
6
Br
Me
N
Me
41
IC50 = 23 µM vs BRD4(1)
LE = 0.50
O
N
O
Me
20
IC50 = 39 µM vs BRD4(1)
LE = 0.51
40
IC50 = 4.4 µM vs BRD4(1)
LE = 0.34
O O
S
N
H
„„
DHQs
O O
S
N
H
H
N
O
N
Me
OMe
43
IC50 = 220 nM vs BRD4(1)
LE = 0.39
42
IC50 = 880 nM vs BRD4(1)
LE = 0.39
Figure 19. The optimization of the lead fragment 20 to give the BET bromodomain ligand
reported by Fish et al. [17] . The green substituent occupies the KAc-binding pocket and acts as
aKAc mimic, the red substituents occupy the WPF shelf. IC50 values were determined using an
AlphaScreen assay.
future science group
www.future-science.com
193
Review | Jennings, Measures, Wilson & Conway
molecules with a linear relationship between the
DHQ core and the substituent. Fish et al. predicted that addition of a tetrahedral component
in the substituent would most effectively orient
groups into the WPF shelf. The sulfonamide
group was selected as the tetrahedral component
and two series of compounds were made, one
with the sulfur atom attached to the DHQ core,
similar to the compounds reported by Bamborough et al. [13], and the other with the nitrogen
atom attached to the core.
Evaluation of these compounds revealed that
the series in which the nitrogen atom is attached
to the DHQ core (42) showed greater affinity for
BRD4(1) compared with the alternative series
(40). Addition of an ortho-methoxy substituent to the WPF-shelf-occupying aryl group was
observed to be favorable from the perspective
of BRD4(1) affinity and compound solubility.
An x-ray crystal of 43 bound to BRD4(1)
reveals that this compound binds as predicted,
with the DHQ core acting as an effective KAc
mimic and the sulfonamide group directing
the methoxyphenyl ring into the WPF shelf.
The sulfonamide NH forms a water-mediated
hydrogen bond with the carbonyl group of L92.
It is also possible that the oxygen atom of the
methoxy group interacts both with this water
molecule and the NH of the sulfonamide.
However, in both cases, the distance between
the oxygen atom and the other atom is observed
as over 3 Å in the x-ray crystal structure, and
in the case of the NH, the orientation of the
atoms looks to be sub-optimal. Analysis of the
x-ray crystal structure does indicate that the
methoxy group plays an important role in aiding the molecule to adopt a conformation that is
favorable for BRD4(1) binding. The x-ray crystal structure shows that, at least in the binding
conformation of the molecule, rotation around
the aryl-sulfur bond of the sulfonamide will be
restricted. Compound 43 shows good selectivity
for the BET bromodomains over the bromodomain of CREBBP (surface plasmon resonance
K D = 49 µM). DSF ana­lysis over a broader panel
of phylogenetically diverse bromodomains indicated that 43 is a selective BET bromodomain
ligand. Furthermore, 43 showed <50% inhibition against a panel of kinases, GPCRs and ion
channels. The cellular activity of 43 was assessed
by demonstrating inhibition of production, and
was confirmed using a lipopolysaccharide challenge assay in peripheral blood mononuclear
cells (EC50 1.89 µM). Rat studies indicated
attractive oral bioavailability and PK properties.
194
Future Med. Chem. (2014) 6(2)
A separate publication has highlighted the value
of this probe in cancer research [44]. Exposing
sensitive cell lines to PFI-1 resulted in downregulation of aurora B kinase, providing an
alternative method to inhibit the action of this
oncology target.
„„
2-thiazolidinone
A fragment-based approach employed by Zhao
et al. identified the 2-thiazolidinone core as
another KAc-mimicking chemotype [18]. Using
the ZINC database, 487 fragments were selected
using human expertise and parameters including molecular weight, number of rotatable bonds
and log P. Docking using the x-ray crystal structure of (+)-JQ1 bound to BRD4(1) was used
to select 41 fragments for crystallization trials.
These trials provided structures of nine fragments bound to BRD4(1) with several chemotypes identified as KAc mimics, including the
2-thiazolidinone moiety, which was selected for
further optimization. Comparison of the binding modes of 44 (Figure 20) with (+)-JQ1 and
fragment 27 (3,5-dimethyl-4-phenylisoxazole;
Figure 16) suggested that affinity gains could
be made by the addition of meta- and/or parasubstituents on the phenyl ring of 44. These substituents allowed the addition of hydrophobic
moieties to effectively occupy the WPF shelf.
Thus a range of sulfonamides based on scaffold 44 were synthesized and screened against
BRD4(1) using a fluorescence anisotropy assay.
The sulfonamides 45 and 46 had identical
affinities (BRD4(1) IC50 4.1 µM) but comparison of their crystal structures with BRD4(1)
revealed distinct binding modes of the sulfonamide substituents. As expected, the thiophenyl
ring of 46 occupies the WPF shelf, but the larger
benzyl group of 45 sits in the ZA channel. This
led to speculation that the sulfonamide group
of 45 was an unnecessary component for binding and prompted investigations into whether a
1,3,5-trisubstituted phenyl ring would lead to
increased affinities for BRD4(1). These investigations led to the development of 47, which
has an IC50 value of 230 nM for BRD4(1) and
is 17-fold more potent than 45 and 46. An x-ray
crystal structure of the trisubstituted analog 47
indicated that the sulfonamide directs the thiophene group into the WPF shelf, with the amide
group extending into the ZA channel, making
extensive hydrogen bonding interactions with
residues in this region.
Compound 47 was found to have good metabolic stability profiles, determined by in vitro
future science group
Phenotypic screening & fragment-based discovery of small-molecule bromodomain ligands
| Review
O O
S
NH
H
N
H
N
F
O
S
S
45
IC50 = 4.1 µM vs BRD4(1)
LE = 0.33
O
S
44
44% inhibition
@ 100 µM vs BRD4(1)
O
O O
S
N
H
S
S
O O
S
N
H
H
N
NH
O
H
N
O
S
47
IC50 = 230 nM vs BRD4(1)
LE = 0.32
S
46
IC50 = 4.1 µM vs BRD4(1)
LE = 0.36
Figure 20. The optimization of the 2-thiazolidinone-based fragment 44 to give the potent
BET bromodomain ligand 47 reported by Zhao et al. [18] . The green substituent occupies the
KAc-binding pocket and acts as a KAc mimic, red substituents occupy the WPF shelf and blue
substituents occupy the ZA channel. IC50 values were determined using a fluorescence
anisotropy assay.
liver microsome stability and cytochrome P450
enzyme inhibition assays. A range of the 2-thiazolidinones caused growth inhibition of the
human colon cancer HT-29 cell line.
„„
Other
KAc-mimicking fragments
Chung et al. [12] have reported a number of BET
bromodomain-binding fragments in addition to
those described above. Using knowledge gained
from ana­lysis of x-ray crystal structures of existing ligands bound to bromodomains, a library
of 1376 compounds, including hydrogen bonding functionality and a small alkyl substituent as
a KAc mimic, was assembled. This library was
screened using a fluorescence anisotropy assay
for activity against BRD2, BRD3 and BRD4.
Of these compounds, 132 showed >30% displacement of the fluorogenic ligand from at
least one of the tandem BET proteins. From the
132 fragments identified, 40 produced x-ray
crystal structures in complex with BRD2(1)
[12]. Chung et al. reported six of these structures,
two of which include fragment chemotypes
previously not known to bind to bromodomains.
Three representative fragments (Figure 21)
were reported which had never been observed
bound in crystallized BET-bromodomains,
each containing an acetylsubstituent located
in the KAc pocket. The tetrahydroquinoline
derivative 13 was an attractive starting point
for optimization due to its rigidity and a greater
future science group
number of interactions with the bromodomain
than the native KAc side chain. Other than the
amide substituent, fragment 13 only engages in
lipophilic interactions with the protein. This
small molecule showed reproducible inhibition
of IL-6 release from lipopolysaccharide-stimulated peripheral blood mononucleated cells, with
IC50 = 14 µM. The similar structural features of
48 and the commonly used drug acetaminophen
(49, paracetamol) directed screening of this compound and the results identified it as another
small, highly tractable starting point. Although
further development of these fragments has not
been reported, they represent excellent starting points for further development. The fact
that some fragments are also shown to bind the
CREBBP bromodomain indicates that these
fragments are potentially useful for development
of non-BET bromodomain ligands.
OH
N
N
O
Me
Me
13
HN
N
Me
Me
O
O
48
49
Figure 21. The three fragments reported by Chung et al. that had not
previously been crystallized bound to a bromodomain [12] .
www.future-science.com
195
Review | Jennings, Measures, Wilson & Conway
Recently, Vidler et al. demonstrated the value
of structure-based virtual screening, which led
to the discovery of four novel KAc mimics [45].
Clinical studies with BET
bromodomain ligands
The work described above has shown that the
BET BCPs, at least, are ‘ligandable’ targets, that
is, potent and selective ligands have been developed for these proteins. The true challenge of
determining BCPs as ‘druggable’, namely, therapeutically relevant targets with sufficiently potent
ligands for in vivo efficacy, remains an open question [46]. Several clinical trials that are currently
underway will help to address this question and
aspects of these trials are outlined below.
„„
RVX-208
Given that bromodomain ligands function
through a clinically novel mode of action,
defining clinical trials as successful is contingent on establishing measurable end points that
relate inhibition of targets to clinical benefits.
Although not developed as a BET bromodomain ligand, RVX-208 (6) became the first BET
probe in clinical trials. Following promising preclinical results, a small Phase I study in healthy
humans was conducted [5]. Translation of the
preclinically observed phenotype was seen in a
statistically significant increase in ApoA1 and
a trend towards increased HDL-C. Following
this successful Phase I trial, two Phase II trials
(ASSERT and SUSTAIN) were undertaken,
with the principle aim of investigating the
appropriate dose range, safety and efficacy of
RVX-208. ASSERT measured the percentage
change in ApoA1 production with 12 weeks of
treatment of patients with stable coronary artery
disease who were on statin therapy. Three different doses plus placebo were administered over a
12-week period, and although there were modest dose-dependent increases in levels of ApoA1
expression during this time, these increases
were not statistically significant. Consequently,
a study of 24 weeks was carried out in the
SUSTAIN clinical trial [47,48].
One problem observed for a large number of
patients treated with RVX-208 in the ASSERT
trial was reversible elevated levels of transaminases [46]. Although this effect is often experienced with other forms of treatment, the short
length and small scale of the study means that
adverse hepatic effects cannot be ruled out during long-term treatment. Levels of transaminase
were elevated up to three-times the normal upper
196
Future Med. Chem. (2014) 6(2)
limit, most frequently in patients treated with the
cardiovascular disease therapeutic simvastatin,
on high-dose treatment of alternative statins, or
who had elevated enzyme levels at baseline [48].
Therefore, the SUSTAIN trial was limited to
patients being treated with only atorvastatin or
rosuvastatin, two alternatives to simvastatin, for
the treatment of cardiovascular disease. Furthermore, the study excluded high doses of these therapeutics and patients with pre-existing elevations
in enzyme levels or high baseline HDL-C levels.
As well as HDL-C levels, percentage change in
ApoA1 and other biomarkers such as triglycerides, abnormal levels of which contribute to cardiovascular disease, were also monitored in this
trial. The SUSTAIN trial reached completion in
August 2012 but results have yet to be published.
An additional Phase II trial of RVX-208,
ASSURE, was carried out to measure the effect
of the compound on atherosclerotic plaque burden in patients with coronary artery disease
and low HDL-C levels. The aim of this study
was primarily to measure percentage change in
atheroma volume, but also look at correlations
between atheroma composition and other biomarkers. The trial reached completion in June
2013 and Resverlogix announced -0.4% plaque
regression, which failed to meet the trial’s primary endpoint of a -0.6% change in atheroma
volume. However, patients receiving combination therapy of RVX-208 and rosuvastatin displayed a statistically significant reduction in atheroma volume of -1.43%. By contrast, patients
receiving combination therapy of RVX-208
and atorvastatin displayed atheroma progression. Two patent applications have been made
as a result of RVX-208’s synergistic effect in
combination with rosuvastatin [201]. Increases
in ApoA1 and HDL-C levels implicate BET bromodomain inhibition in the observed clinical
efficacy.
Based on data that suggest HDL directly
controls glucose metabolism, an additional
Phase II trial, announced in November 2012,
will investigate the effect of RVX-208 on males
with impaired fasting glucose or impaired glucose tolerance [202]. The study will primarily
measure postprandial blood glucose levels, but
will also investigate changes in insulin secretion
levels and insulin sensitivity.
„„OTX015
Following completion of a Phase I trial, OTX015,
was in-licensed to Oncoethix from Mitsubishi in
March 2012. OTX015 is currently in a Phase I
future science group
Phenotypic screening & fragment-based discovery of small-molecule bromodomain ligands
clinical trial to determine the dose for treatment of acute leukemia and other hematological
malignancies [203].
„„CPI-0610
Constellation Pharmaceuticals have recently
announced the entry of their novel BET bromo­
domain ligand CPI-0610 into Phase I clinical
trials for patients with aggressive lymphoma [204].
„„I-BET762
I-BET762 (2)
is currently in Phase I clinical trials for NMC, a rare but lethal form of lung cancer arising from a genetic translocation, with an
estimated completion date of December 2014 [205].
Discussion
„„Phenotypic screening identified the BET
bromodomains as important therapeutic
targets
Bromodomains were first discovered in 1992
when the Drosophila gene brahma was found
to encode for a protein, brm, which contained a
motif common to the SNF2, fsh, SPT7, RING3
and CCG1 proteins. This motif, the bromodomain, was thought to mediate protein–protein
interactions involved in transcriptional regulation [49,50]. In 1999, the NMR structure of
the PCAF bromodomain was obtained and
KAc was identified as the endogenous ligand
for bromodomains [34]. These findings led to
the idea that the bromodomain was a module
involved in targeting proteins to chromatin
by binding to K Ac on histone proteins [51].
This work provided the foundations for the
discovery of the first unnatural bromodomain
ligands. Based on the structural studies, NMR
screening was used to identify ligands for the
PCAF bromodomain that block binding of
PCAF to an acetylated lysine residue on the Tat
co-activator, a process that is important in HIV
transcription and replication [1]. Despite showing a clear effect in vitro, structural ana­lysis of
these molecules bound to the PCAF bromodomain suggest that they did not directly mimic
the KAc interactions with the bromodomain
and, perhaps consequently, only showed low
micromolar IC50 values. In addition, although
some selectivity data for these compounds are
disclosed, it seems unlikely that these lowmolecular-weight molecules only show affinity for their intended target. Despite further
work on these molecules being published, they
have not been developed for therapeutic application [36] and, although they are pioneering
future science group
| Review
tool compounds, they are not highly attractive
drug leads.
It was the development of the BET bromodomain probes I-BET762 and (+)-JQ1 that demonstrated potent small-molecule bromodomain
ligands could be identified [3,7]. Both of these
molecules were discovered as bromodomain
ligands by phenotypic screening, in the case of
I-BET762, to identify upregulators of ApoA1. By
using affinity chromatography, GSK were able to
identify members of the BET bromodomain family as the main target of I-BET762. Subsequent
siRNA studies indicated that inhibition of BRD4
mainly responsible for the phenotype caused by
I-BET762, with the binding to BRD2 and BRD3
less important for the biological effect of the
compound. Since I-BET762 had proven cellular
activity, as a result of being identified in a phenotypic screen, it was quickly entered into SAR
optimization studies, leading to a first-in-class
molecule. The fact that both I-BET762 and RVX208 were developed to upregulate ApoA1 levels,
and were only subsequently shown to exert their
action via BET bromodomain inhibition, indicates that ligands of the BET bromo­domains are
an effective route to regulate ApoA1 expression.
In some respects, this work is a classic example of how phenotypic screening is a powerful
method for identifying compounds with a particular biological activity, without prior knowledge of a molecular target. Phenotypic screening typically identifies compounds that show
a strong cellular effect. This effect often arises
from favorable polypharmacology, in which
multiple synergistic events combine to evoke the
observed activity. Although I-BET762 caused
a strong phenotype, this arose, not from polypharmacology, but from a single protein target, BRD4, which was revealed to modulate the
effects of the I-BET762. Therefore, unusually,
this work is an example of a phenotypic screen
revealing a single protein that can be targeted to
give the desired phenotypic effect. Given their
fundamental role in transcription, it is unlikely
that the BET bromodomains would have been
viewed as viable therapeutic targets prior to
the identification of I-BET762 and (+)-JQ1 as
potent and selective ligands for these proteins.
The elucidation of the BET bromodomains as
ligandable entities has catalyzed intense interest in bromodomains more generally as potential therapeutic targets. Indeed, there is much
interest in developing ligands for a wide variety
of bromodomains, which will in turn, help to
reveal the importance of their biology.
www.future-science.com
197
Review | Jennings, Measures, Wilson & Conway
„„The
target information obtained from
phenotypic studies enabled & enhanced
fragment-based bromodomain ligand
discovery
Given that one main protein target, BRD4, was
revealed as being pivotal for the activity of the
phenotypically discoveredApoA1 regulators,
it was tractable to investigate the molecular
basis of bromodomain inhibition. This information subsequently allowed structure-based
optimization of compound affinity and physical properties. This strategy would have been
less appealing had activity at several important
targets been required to achieve the observed
phenotype. Although the bromodomains exist
exclusively as part of more complex proteins,
the BET bromodomains, at least, are stable and
can be expressed separately from the rest of their
parent protein, which is convenient for conducting biophysical assays and obtaining x-ray
crystal structures. Efforts to elucidate the mode
of ligand binding to the BET bromodomains
were assisted by high-resolution x-ray crystal
structures of these proteins in both ligand free
and ligand-bound forms. Comparison of the
protein–ligand interactions of I-BET762 and
(+)-JQ1 allowed the key structural elements
responsible for their high affinity to be identified. The availability of this information has
been fundamental in underpinning the rapid
structure-guided progress made in the development of potent and selective BET bromodomain
ligands based on known chemotypes. In addition, fragment-based approaches to bromodomain ligands from novel chemotypes can only
be carried out with access to robust biophysical
assays, including NMR techniques to identify
weak-binding ligands and, ideally, structural
information on the mode of compound binding. Such fragment screening approaches have
resulted in the discovery of a second generation
of BET bromodomain ligands based on a number of novel chemotypes. The 3,5-dimethylisoxazole motif, in particular, was identified as an
effective KAc mimic. Importantly, in addition
to forming the basis for potent BET bromodomain ligands, 3,5-dimethylisoxazole-based
compounds were also shown to bind the bromodomain of CREBBP [14], indicating that some of
the KAc mimics that are effective for the BET
bromodomains are potentially transferable to
other classes of bromodomains.
Partly as a result of the success with obtaining
x-ray crystal structures for the BET bromodomains, structural information is now available
198
Future Med. Chem. (2014) 6(2)
for over 30 bromodomains from all of the eight
phylogenetically different classes [30]. This work
provides vital information to guide structurebased development of selective ligands for all of
these bromodomains, and it is likely that selective probes for a wide variety of bromo­domains
will emerge in the near future. One interesting
challenge is the development of probes for bromodomains that do not possess the standard
features present in most bromodomains. Notably, 13 bromodomains, including ASH1L, PB1
and the SP proteins, do not contain the KAcbinding asparagine residue, which is conserved
in the remaining BCPs [30]. It seems likely that
screening a small library based on existing
KAc mimics that are known have affinity for
the ‘standard’ bromodomain structure will not
yield ligand hits for these proteins. Screening of
more diverse libraries will be important for generating new lead compounds for these unusual
bromodomains.
„„
Rapid
progress in the development of BET
bromodomain ligands & applicability of the
knowledge gained to development of
non-BET bromodomain ligands
The phenotypic-based programs to identify
efficient upregulators of ApoA1 have been
long-term commitments for a number of pharmaceutical companies. However, once the BET
bromo­domains were identified as potential
therapeutic targets, progress in the identification and application of potent and selective
ligands for these proteins has been unusually
quick. There are a number of factors that have
contributed to the rapid development of potent
and selective BET bromodomain ligands
and consideration of these factors might aid
the development of similar ligands for other
bromodomains. Development of ligands for
a given a class of proteins is often driven by
their involvement in a therapeutically significant biological pathway. One aspect that was
crucial to the rapid development of ligands for
BET bromodomains has been the combination of the interest generated by demonstration that phenotypically discovered BET bromodomain ligands can affect inflammation,
leukemias and modulate myc levels [24,52]. The
BET bromodomain ligands provide an alternative to developing drugs that target myc.
Such compounds have been a long-standing
medicinal chemistry target, which has proved
difficult to achieve. The availability of the phenotypically discovered tool compounds led to
future science group
Phenotypic screening & fragment-based discovery of small-molecule bromodomain ligands
intensive application of these molecules and
consequently a large number of high profile
publications on the biology of the BET bromodomains. Importantly, the availability of
tool compounds allowed investigations by labs
in both academia [53] and industry [54] to rapidly conduct and publish investigations on the
merits of targeting the BET bromodomains.
Once the BET bromodomains had been
identified as therapeutically promising targets,
a number of fragment-based discovery programs
began. These programs benefited from much of
the information discovered in the phenotypic
screening projects (vide supra), especially the
availability of purified proteins for biophysical
and structural work. However, the fragmentbased discovery programs also benefited from the
availability of a range of cellular assays that were
used to characterize the bromodomain ligands
discovered by phenotypic screening. Both the
fluorescence recovery after photobleaching assay
employed by the SGC [7], and investigating the
viability of MV4;11 acute myeloid leukemia
cells, have become standard methods for assessing the cellular activity of BET bromodomain
ligands [11,16,24,53,54].
The fast pace of BET bromodomain ligand
development is perhaps more remarkable given
that these compounds are targeting a protein–
protein interaction (PPI). PPIs are often challenging to inhibit selectively, because their interfaces
are usually heavily solvent-exposed, cover large
areas, and are structurally featureless or poorly
defined [55]. Bromodomains are different from
many PPIs in some respects, as although they
have large, solvent-exposed grooves, in which
the histone tail, or other partner protein, binds,
the acetylated lysine residue invariably binds in a
defined pocket. The defined and well-structured
nature of the KAc-binding pocket is similar to
an enzyme active site or receptor ligand-binding
pocket, and has consequently allowed the design
of drug-like small molecules that are able to
inhibit the bromodomain–histone interaction.
As far as we know, the KAc-binding pocket
is well defined for all of the bromodomains,
although structural information is only available
for approximately 50% of the bromodomains in
the human proteome.
To investigate the tractability of developing
small-molecule ligands of non-BET bromodomains, Vidler et al. have conducted an ana­lysis
of the druggability of 33 bromodomain KAcbinding sites, for which structural information is available [56], using SiteMap and 105
future science group
| Review
PDB entries. An interesting component of this
ana­lysis was on the change in predicted druggability of bromodomains when in the apo
form compared with when a ligand (either a
small molecule or a peptide) is bound. With
the exception of CREBBP, the pocket was not
found to be more druggable when a ligand was
bound. The bromodomains were grouped by
common binding site features, providing useful
information on developing compounds that are
selective for one bromodomain over another.
As would be expected, all bromodomains were
classed as more druggable than PPIs with less
defined binding pockets available for ligand
binding. The ana­lysis indicated that 13 bromodomains are druggable; this includes the BET
family of BCPs. Five additional bromodomains
were viewed as being intermediate in their druggability and the remaining bromodomains were
classified as being hard to drug [56]. It should
be noted that this categorization should really
be viewed as a relative comparison between the
druggability of bromodomains rather than an
absolute measure of how easy it is to develop
small-molecule ligands for each bromodomain.
Zhang et al. point out that our view of the
druggability of a given bromodomain depends
to some extent on the bromodomain ligands
that are already available [57]. Consequently,
our perspective on which bromodomains are
druggable might change as new bromodomain
ligands are developed. Therefore, the ana­lysis
by Vidler et al. ranks the bromodomains in
order of their likely druggability based on our
current knowledge of ligands, but we should
also consider the therapeutic potential of all
bromodomains when seeking to develop novel
bromodomain ligands.
Given the relatively sparse knowledge of biological pathways involving other sub-families of
bromodomains – in some cases little is known
beyond a genetic prediction of the incorporation
of a bromodomain – and associated difficulties
in prediction of phenotypic consequence [58],
it seems likely that determining the clinical
benefits of targeting a given bromodomain will
require deliberate attempts to generate probes
for prospective targets to elucidate a phenotypic
response. However, the current work in the field
suggests that the increased risk in focusing on
targets without identified therapeutic potential
is mitigated by the rapid success in developing BET bromodomain ligands, coupled with
the emerging importance of bromodomains
generally in a range of biological processes.
www.future-science.com
199
Review | Jennings, Measures, Wilson & Conway
„„ADME
& PK properties of BET
bromodomain ligands
With only a few BET bromodomain ligands in
clinical trials it is difficult to draw broad conclusions on the characteristics of these compounds.
(+)-JQ1, which was one of the first BET bromodomain probes reported, has relatively high clearance in mouse, and a relatively short half-life, but
a moderate volume of distribution [7]. I-BET762
showed favorable PK in mouse, dog and monkey
with moderate oral bioavailability, a large volume
of distribution in mouse and a long half-life in dog,
however, sub-optimal PK data was observed in rat
[59]. The related compound reported by Gehling
et al. displays a long half-life in dog, low clearance
and a moderate volume. However, this compound
(39, Figure 18) also has a comparatively low oral
bioavailability in rats [41]. I-BET151 displays low
clearance in mouse and rat but high clearance in
dog. It also has a good half-life in mouse, rat and
dog, and good bioavailability in mouse and rat.
RVX-208 displays low clearance but a relatively
short half-life and volume of distribution. PFI-1
is the only fragment-based BET bromodomain
ligand for which PK and ADME data are available. This compound has moderate clearance, a
short-half life and small volume of distribution,
and a relatively low oral bioavailability, which is
comparable with that of compound 39.
Given the lack of data it is difficult to compare the ADME and PK characteristics of phenotypically discovered BET bromodomain
ligands with those of fragment-based BET bromodomain ligands, although the available data
are summarized in Supplementary Tables 1 & 2 .
The phenotypically discovered ligands have better bioavailability than PFI-1, but it is not clear
whether this observation will form the basis of a
trend. Interestingly the mean ligand efficiency
for the two sets of compounds is very similar,
although there is a wider variation of values for
the fragment-based compounds. Overall these
data suggest that the PK and ADME properties
of the fragment-based compounds are likely to be
at least comparable to those of the phenotypically
discovered compounds.
Future perspective
Despite rapid progress in the development of BET
bromodomain ligands, the therapeutic potential
of these compounds is currently unproven and,
like the field of epigenetic medicine generally, this
area is in its infancy. Some have likened the intense
interest in bromodomains to the initial period of
research on kinase ligands, but it remains to be
200
Future Med. Chem. (2014) 6(2)
seen whether bromo­domain ligands progress to
become important therapeutic compounds in the
same way as kinase inhibitors have. The clinical trials that are underway will provide some indication
of the answers to these questions for the BET bromodomains in the next 12–24 months. As a result
of the interest in the BET bromodomains, there
are now a number of high-quality ligands for these
proteins and hence there is little incentive to start a
program of ligand discovery aimed at these bromodomain at this time. One challenge associated with
the BET bromodomain ligands that has yet to be
addressed is the development of ligands that selectively inhibit one of the BET BCPs over the other
family members, or one of the two tandem bromodomains in a given BET BCP (e.g., BRD4(1) over
BRD4(2)). Without selective ligands it is not clear
whether this level of discrimination is required for
therapeutically important compounds, but one can
speculate that such ligands might possess enhanced
safety profiles over less selective compounds. The
longer-term future of bromodomain research will
be concerned with ligand discovery for non-BET
bromodomains. Although much work in this area
is underway, there are currently no publications
detailing potent and selective ligands for the nonBET bromodomains. The ana­lysis published by
Vidler et al. clearly indicates that bromodomain
families other than BETs are potentially ligandable and druggable (vide supra) and it is likely
that we will see potent and selective non-BET
bromodomain ligands disclosed soon.
Given that bromodomains affect transcriptional regulation, they have the potential to be
important in a wide variety of indications. This
point is illustrated to some extent by the application of (+)-JQ1 as a prototype male contraceptive
[61] and the effect of PCAF ligands on HIV infection [1]. Supplementary Table 3 lists non-oncology
indications and therapeutically important pathways that are affected by removal or malfunction of a BCP. This list indicates that there is
significant potential for bromodomain ligands
to be therapeutically relevant for a wide range of
clinical indications. However, it should be noted
that although the BCPs are implicated in the diseases or pathways shown, little is known about
the specific role played by the bromodomain in
these cases. Research to understand this role will
be greatly aided by the development of potent and
selective small-molecule bromodomain ligands,
similar to those compounds developed for the
BET bromodomains.
The success in developing small-molecule
ligands for the BET bromodomains inspires
future science group
Phenotypic screening & fragment-based discovery of small-molecule bromodomain ligands
confidence that it will be possible to repeat this
achievement for other bromodomain families. As
we have discussed, the availability of structural
information proved key to the rapid progress made
with the BET bromodomain ligands. x-ray crystal
or NMR structures are available for 37 bromodomains with all eight families represented [30], which
will underpin the development of ligands for these
bromodomains. Additionally, the availability of
stable constructs of these proteins is essential for
use in the biophysical assays that are necessary to
generate SAR for rational ligand design. Given the
knowledge developed in obtaining structures of
these proteins, the elucidation of structures for the
remaining bromodomains seems a tractable target.
Hence, should a therapeutic indication implicate a
BCP for which structural information is not available, existing work should facilitate researchers to
obtain these data in an expeditious manner.
A key component of the BET bromodomain
programs was a clearly defined cellular phenotype,
which resulted from the phenotypic discovery of
the initial ligands. It will be important to develop
such assays to assess the cellular effects of nonBET bromodomain ligands. Poor translation
from preclinical to clinical settings has renewed
the demand for more complex phenotypic cellular assays, which better represent diseased states.
In the case of BET bromodomains, correlation
between MYCN amplification and sensitivity to
inhibition was identified through integration of
genetic ana­lysis with chemosensitivity data from
a cell-based screen [62]. The apparent plurality of
epigenetic recognition, alongside relatively weak
binding, ideally calls for in vitro assays involving
larger substrates. For instance, reconstituted modified histones can be refolded into octamers for use
in nucleosomal arrays to detect multisite binding
of epigenetic readers [63]. This would provide an
intermediary for fragment-based development to
be more closely connected to phenotypic screens.
A significant challenge for those developing
bromodomain ligands and epigenetic medicine
generally is the transgenerational heritability of
epigenetic marks, especially when developing compounds to treat diseases with a better prognosis
than terminal cancer. During mammalian development, accurate replication of both the DNA and
epigenetic state are required for maintenance of cell
fate [64]. Although germline cells are able to erase
the epigenetic marks and ‘reset’ their epigenome
at key points, certain marks evade this reprogramming and are instead transmitted to offspring. The
marks that are transmitted and the mechanisms by
which this occurs is largely unknown [64]. Clearly
future science group
| Review
these fundamental biological questions will have
to be answered before we have a full understanding
of the safety considerations for medicines that alter
the pattern of epigenetic marks.
Conclusion
Bromodomains have emerged as exciting and
ligandable protein modules that might become
therapeutically important targets. Although early
approaches to developing bromodomain ligands
were structure-based, it was the identification of
potent and selective BET bromodomain ligands in
a phenotypic screen for ApoA1 modulation that
has ignited intense research in this field. Both phenotypic and rational design approaches have led
to the development of potent and selective BET
bromodomain ligands. With three BET bromodomain ligands in clinical trials, the potential of these
compounds as drugs will soon be revealed. The
future of this field will focus on non-BET bromodomain targets and the success in the development
of BET bromodomain ligands means that confidence is high that similar ligands will be developed
for a diverse range of other bromodomains. The
rapid progress in the understanding of bromodomain biology and their emergence as potential
therapeutic targets has been underpinned by the
development of potent and selective probes or tool
compounds. This approach is vital for effective
target validation [65] and it seems certain that this
approach will play a key role in understanding
the biological role of other bromodomains and
validating them as therapeutic targets.
Supplementary data
To view the supplementary data that accompany this paper
please visit the journal website at: www.future-science.
com /doi /full /10.4155/FMC.13.197
Acknowledgements
The authors are grateful to D Hewings and T Rooney for
critical reading of the manuscript.
Financial & competing interests disclosure
LE Jennings holds a studentship that is supported by GlaxoSmithKline. AR Measures holds a studentship that is supported by Pfizer-Neusentis. BG Wilson holds a studentship
that is supported by UCB. SJ Conway has given invited
lectures at Genentech, GlaxoSmithKline and Novartis, for
which he has received honoraria. The authors have no other
relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial
conflict with the subject matter or materials discussed in the
manuscript apart from those disclosed. No writing assistance
was utilized in the production of this manuscript.
www.future-science.com
201
Review | Jennings, Measures, Wilson & Conway
Executive summary
Bromodomains are KAc-binding protein modules
Recognition of KAc mediates protein–protein interactions, notably, but not exclusively, between histones and transcriptional machinery.
Phenotypic screening identified BET bromodomains as important therapeutic targets
„„
„„
In combination with chemoproteomics, phenotypic screening identified small molecules that modulate ApoA1 levels. These compounds
are effective at reversing the effects of leukemias, atherosclerosis and inflammation in preclinical models.
The compounds identified as upregulators of ApoA1 worked by inhibiting the BET bromodomains, revealing BET
bromodomain-containing proteins as novel therapeutic targets.
Target information enhanced fragment-based approaches to ligand discovery
„„
Obtaining x-ray crystal structures of these ligands bound to the BET bromodomains has underpinned subsequent fragment-based ligand
discovery programs for bromodomain ligands.
Therapeutic potential of targeting bromodomains
„„
„„
Clinical trials are underway on four BET bromodomain ligands. These studies will give the first indications of whether the powerful tool
ligands that have been developed for the BET bromodomains will translate into useful drugs.
„„
This work has stimulated interest in non-BET bromodomains as therapeutic targets with the potential for addressing a wide range of
diseases.
„„
The development of selective bromodomain ligands as molecular probes is an important strategy for understanding the biology and
therapeutic potential of bromodomain-containing proteins.
References
Papers of special note have been highlighted as:
n of interest
nn of considerable interest
1
n
2
3
n
4
nn
Zeng L, Li J, Muller M et al. Selective small
molecules blocking HIV-1 Tat and coactivator
PCAF association. J. Am. Chem. Soc. 127(8),
2376–2377 (2005).
The development of the first inhibitor of a
bromodomain–KAc interaction, identified by
NMR screening methods against PCAF, as a
potential anti-HIV therapeutic target.
Sachchidanand, Resnick-Silverman L, Yan S
et al. Target structure-based discovery of small
molecules that block human p53 and CREB
binding protein association. Chem. Biol. 13(1),
81–90 (2006).
Nicodeme E, Jeffrey KL, Schaefer U et al.
Suppression of inflammation by a synthetic
histone mimic. Nature 468(7327), 1119–1123
(2010).
One of the first reported BET bromodomain
inhibitors, I-BET762, inhibited
lipopolysaccharide-inducible gene expression,
an effect attributed to a high affinity
interaction with BET bromodomains. This
small molecule was optimized from a hit in a
high-throughput phenotypic assay and
highlighted the anti-inflammatory potential
of targeting this class of protein.
Chung C-W, Coste H, White JH et al.
Discovery and characterization of small
molecule inhibitors of the BET family
bromodomains. J. Med. Chem. 54(11),
3827–3838 (2011).
A comprehensive report of the phenotypic,
chemoproteomic, biophysical and structural
202
studies conducted to elucidate the molecular
interactions of I-BET762 with BET
bromodomains.
5
6
7
n
8
9
Bailey D, Jahagirdar R, Gordon A et al. RVX208: a small molecule that increases
apolipoprotein A-I and high-density lipoprotein
cholesterol in vitro and in vivo. J. Am. Coll.
Cardiol. 55(23), 2580–2589 (2010).
Khmelnitsky YL, Mozhaev VV, Cotterill IC
et al. In vitro biosynthesis, isolation, and
identification of predominant metabolites of
2-(4-(2-hydroxyethoxy)-3,5-dimethylphenyl)5,7-dimethoxyquinazolin-4(3H)-one (RVX208). Eur. J. Med. Chem. 64C, 121–128 (2013).
Filippakopoulos P, Qi J, Picaud S et al.
Selective inhibition of BET bromodomains.
Nature 468(7327), 1067–1073 (2010).
12 Chung C-W, Dean AW, Woolven JM,
Bamborough P. Fragment-based discovery of
bromodomain inhibitors part 1: inhibitor
binding modes and implications for lead
discovery. J. Med. Chem. 55(2), 576–586
(2012).
n
13 Bamborough P, Diallo H, Goodacre JD et al.
Fragment-based discovery of bromodomain
inhibitors part 2: optimization of
phenylisoxazole sulfonamides. J. Med. Chem.
55(2), 587–596 (2012).
14 Hewings DS, Wang M, Philpott M et al.
3,5-dimethylisoxazoles act as acetyl-lysinemimetic bromodomain ligands. J. Med.
Chem. 54(19), 6761–6770 (2011).
JQ1 has been used in many studies of BET
bromodomain activity.
Bonetti P, Ponzoni M, Tibiletti MG et al. The
BRD-inhibitor OTX015 shows pre-clinical
activity in diffuse large B-cell lymphoma
(DLBCL). Eur. J. Cancer. 48(Suppl. 6),
163–163 (2012).
Bonetti P, Boi M, Ponzoni M et al. The BRDinhibitor OTX015 is active in pre-clinical
models of mature B-cell lymphoid tumors.
Blood 120(21), abstract 1657 (2012).
10 Boi M, Bonetti P, Ponzoni M et al. The BRD-
inhibitor OTX015 shows pre-clinical activity
in anaplastic large T-cell lymphoma (ALCL).
Blood 120(21), abstract 4872 (2012).
11 Dawson MA, Prinjha RK, Dittmann A et al.
Inhibition of BET recruitment to chromatin
as an effective treatment for MLL-fusion
leukemia. Nature 478(7370), 529–533 (2011).
Future Med. Chem. (2014) 6(2)
An example of the use of crystal structures
generated during fragment-based screens to
expose key interactions, resulting in several
diverse KAc mimicking fragments.
n
An early example of a fragment-based
approach to BET bromodomain ligand
discovery.
15 Hay D, Fedorov O, Filippakopoulos P et al.
The design and synthesis of 5 and
6-isoxazolylbenzimidazoles as selective
inhibitors of the BET bromodomains. Med.
Chem. Commun. 4, 140–144 (2013).
16 Hewings DS, Fedorov O, Filippakopoulos P
et al. Optimization of 3,5-dimethylisoxazole
derivatives as potent bromodomain ligands.
J. Med. Chem. 56(8), 3217–3227 (2013).
17 Fish PV, Filippakopoulos P, Bish G et al.
Identification of a chemical probe for bromo
and extra c-terminal bromodomain inhibition
through optimization of a fragment-derived
hit. J. Med. Chem. 55(22), 9831–9837 (2012).
future science group
Phenotypic screening & fragment-based discovery of small-molecule bromodomain ligands
n
Development of a novel chemotype of BET
bromodomain probe, from a screen of
fragment-sized potential KAc mimics.
Structure–activity relationships draw out
the importance of subtle variations in
connectivity, through an alternative
arrangement of a sulfonamide linkage
yielding an increase in affinity.
18 Zhao L, Cao D, Chen T et al. Fragment-based
drug discovery of 2-thiazolidinones as
inhibitors of the histone reader BRD4
bromodomain. J. Med. Chem. 56(10),
3833–3851 (2013).
19 Chung C-W, Witherington J. Progress in the
discovery of small-molecule inhibitors of
bromodomain–histone interactions. J. Biomol.
Screen. 16(10), 1170–1185 (2011).
20 Chung C-W, Tough DF. Bromodomains: a
new target class for small molecule drug
discovery. Drug Discov. Today: Ther. Strateg.
9,(2–3), e111–e120 (2012).
21 Prinjha RK, Witherington J, Lee K. Place
your BETs: the therapeutic potential of
bromodomains. Trends Pharmacol. Sci. 33(3),
146–153 (2012).
22 Conway SJ. Bromodomains: are readers right
for epigenetic therapy? ACS Med. Chem. Lett.
3, 691–694 (2012).
23 Arrowsmith CH, Bountra C, Fish PV, Lee K,
Schapira M. Epigenetic protein families: a
new frontier for drug discovery. Nat. Rev.
Drug Disc. 11(5), 384–400 (2012).
24 Hewings DS, Rooney TPC, Jennings LE et al.
Progress in the development and application
of small molecule inhibitors of bromodomainacetyl-lysine interactions. J. Med. Chem.
55(22), 9393–9413 (2012).
25 Furdas SD, Carlino L, Sippl W, Jung M.
Inhibition of bromodomain-mediated
protein–protein interactions as a novel
therapeutic strategy. Med. Chem. Commun.
3(2), 123–134 (2012).
bromodomain inhibitors. ChemMedChem
doi:10.1002/cmdc.201300259 (2013) (Epub
ahead of print).
41 Gehling VS, Hewitt MC, Vaswani RG et al.
Histone recognition and large-scale structural
ana­lysis of the human bromodomain family.
Cell 149(1), 214–231 (2012).
Discovery, design, and optimization of
isoxazole azepine BET inhibitors. ACS Med.
Chem. Lett. 4(9), 835–840 (2013).
nn
A useful introduction to bromodomains,
containing a thorough ana­lysis of the
protein family: structure-based sequence
alignments used to generate a phylogenetic
tree, crystal structures across several defined
sub-families, and affinities to a panel of
modified histone peptide sequences.
31 Owen DR, Trzupek JD. Epigenetic drugs that
do not target enzyme activity. Drug Discov.
Today Technol. doi:10.1016–j.ddtec.10.008
(2012) (In press).
32 Wu X, Schultz PG. Synthesis at the interface
of chemistry and biology. J. Am. Chem. Soc.
131(35), 12497–12515 (2009).
33 Filippakopoulos P, Picaud S, Fedorov O et al.
Benzodiazepines and benzotriazepines as
protein interaction inhibitors targeting
bromodomains of the BET family. Bioorg.
Med. Chem. 20(6), 1878–1886 (2012).
34 Dhalluin C, Carlson JE, Zeng L, He C,
Aggarwal AK, Zhou MM. Structure and
ligand of a histone acetyltransferase
bromodomain. Nature 399(6735), 491–496
(1999).
35 Mujtaba S, He Y, Zeng L et al. Structural
basis of lysine-acetylated HIV-1 Tat
recognition by PCAF bromodomain. Mol.
Cell. 9(3), 575–586 (2002).
36 Pan C, Mezei M, Mujtaba S et al. Structure-
guided optimization of small molecules
inhibiting human immunodeficiency virus 1
Tat association with the human coactivator
p300/CREB binding protein-associated
factor. J. Med. Chem. 50(10), 2285–2288
(2007).
37 Wang Q, Wang R, Zhang B, Zhang S, Zheng
Structural basis for acetylated histone H4
recognition by the human BRD2
bromodomain. J. Biol. Chem. 285(10),
7610–7618 (2010).
Y, Wang Z. Small organic molecules targeting
PCAF bromodomain as potent inhibitors of
HIV-1 replication. Med. Chem. Commun.
4(4), 737–740 (2013).
ApoA1 upregulation to BET family
bromodomain inhibition: discovery of
I-BET151. Bioorg. Med. Chem. Lett. 22(8),
2963–2967 (2012).
28 Seal J, Lamotte Y, Donche F et al.
Identification of a novel series of BET family
bromodomain inhibitors: binding mode and
profile of I-BET151 (GSK1210151A). Bioorg.
Med. Chem. Lett. 22(8), 2968–2972 (2012).
29 Mirguet O, Lamotte Y, Chung C-W et al.
Naphthyridines as novel BET family
future science group
CREB-binding protein blocks apoptosis in
cardiomyocytes. Chem. Biol. 18(4), 531–541
(2011).
30 Filippakopoulos P, Picaud S, Mangos M et al.
26 Umehara T, Nakamura Y, Jang MK et al.
27 Mirguet O, Lamotte Y, Donche F et al. From
| Review
38 Bedford DC, Kasper LH, Fukuyama T,
Brindle PK. Target gene context influences
the transcriptional requirement for the KAT3
family of CBP and p300 histone
acetyltransferases. Epigenetics 5(1), 9–15
(2010).
39 Mujtaba S, He Y, Zeng L et al. Structural
mechanism of the bromodomain of the
coactivator CBP in p53 transcriptional
activation. Mol. Cell. 13(2), 251–263 (2004).
40 Borah JC, Mujtaba S, Karakikes I et al. A
small molecule binding to the coactivator
www.future-science.com
42 Philpott M, Yang J, Tumber T et al.
Bromodomain-peptide displacement assays
for interactome mapping and inhibitor
discovery. Mol. BioSyst. 7(10), 2899–2908
(2011).
43 Hopkins AL, Groom CR, Alex A. Ligand
efficiency: a useful metric for lead selection.
Drug Discov. Today. 9(10), 430–431 (2004).
44 Picaud S, Da Costa D, Thanasopoulou A et al.
PFI-1, a highly selective protein interaction
inhibitor, targeting BET Bromodomains.
Cancer Res. 73(11), 3336–3346 (2013).
45 Vidler LR, Filippakopoulos P, Fedorov O
et al. Discovery of novel small-molecule
inhibitors of BRD4 using structure-based
virtual screening. J. Med. Chem. 56(20),
8073–8088 (2013).
46 Edfeldt FNB, Folmer RHA, Breeze AL.
Fragment screening to predict druggability
(ligandability) and lead discovery success.
Drug Discov. Today 16(7–8), 284–287 (2011).
47 Nicholls SJ, Gordon A, Johansson J et al.
Efficacy and safety of a novel oral inducer of
apolipoprotein a-I synthesis in statin-treated
patients with stable coronary artery disease a
randomized controlled trial. J. Am. Coll.
Cardiol. 57(9), 1111–1119 (2011).
48 Nicholls SJ, Gordon A, Johannson J et al.
ApoA-I induction as a potential
cardioprotective strategy: rationale for the
SUSTAIN and ASSURE studies. Cardiovasc
Drugs Ther. 26(2), 181–187 (2012).
49 Tamkun JW, Deuring R, Scott MP et al.
brahma: a regulator of Drosophila homeotic
genes structurally related to the yeast
transcriptional activator SNF2/SWI2. Cell
68(3), 561–572 (1992).
50 Haynes SR, Dollard C, Winston F, Beck S,
Trowsdale J, Dawid IB. The bromodomain:
a conserved sequence found in human,
Drosophila and yeast proteins. Nucleic Acids
Res. 20(10), 2603 (1992).
51 Winston F, Allis C. The bromodomain: a
chromatin-targeting module? Nat. Struct.
Biol. 6(7), 601–604 (1999).
52 Chung C-W. Small molecule bromodomain
inhibitors: extending the druggable genome.
Prog. Med. Chem. 51, 1–55 (2012).
53 Zuber J, Shi J, Wang E et al. RNAi screen
identifies BRD4 as a therapeutic target in
acute myeloid leukemia. Nature 478(7370),
524–528 (2011).
203
Review | Jennings, Measures, Wilson & Conway
54 Mertz JA, Conery AR, Bryant BM et al.
Targeting MYC dependence in cancer by
inhibiting BET bromodomains. Proc. Natl
Acad. Sci. USA 108(40), 16669–16674 (2011).
55 Wilson AJ. Inhibition of protein–protein
interactions using designed molecules. Chem.
Soc. Rev. 38(12), 3289–3300 (2009).
56 Vidler LR, Brown N, Knapp S, Hoelder S.
Druggability ana­lysis and structural
classification of bromodomain acetyl-lysine
binding sites. J. Med. Chem. 55(17),
7346–7359 (2012).
57 Zhang G, Sanchez R, Zhou M-M. Scaling the
druggability landscape of human
bromodomains, a new class of drug targets.
J. Med. Chem. 55(17), 7342–7345 (2012).
58 Brennan P, Filippakopoulos P, Knapp S. The
therapeutic potential of acetyl-lysine and
methyl-lysine effector domains. Drug Discov.
Today: Ther. Strategies 9(2–3), e101–e110
(2012).
59 Mirguet O, Gosmini R, Toum J et al.
Discovery of epigenetic regulator I-BET762:
lead optimization to afford a clinical
candidate inhibitor of the BET
bromodomains. J. Med. Chem. 56(19),
7501–7515 (2013).
60 Leeson PD, Springthorpe B. The influence of
drug-like concepts on decision-making in
medicinal chemistry. Nat. Rev. Drug Disc.
6(11), 881–890 (2007).
61 Matzuk MM, McKeown MR,
Filippakopoulos P et al. Small-molecule
inhibition of BRDT for male contraception.
Cell 150(4), 673–684 (2012).
62 Puissant A, Frumm SM, Alexe G et al.
Targeting MYCN in neuroblastoma by BET
bromodomain inhibition. Cancer Discov. 3(3),
308–323 (2013).
63 Fierz B, Kilic S, Hieb AR, Luger K, Muir TW.
Stability of nucleosomes containing
homogenously ubiquitylated H2A and H2B
prepared using semisynthesis. J. Am. Chem.
Soc. 134(48), 19548–19551 (2012).
64 Lange UC, Schneider R. What an epigenome
remembers. Bioessays 32(8), 659–668 (2010).
65 Bunnage ME. Getting pharmaceutical R&D
back on target. Nat. Chem. Biol. 7(6),
335–339 (2011).
66 Zeng L, Zhou M. Bromodomain: an acetyl-
lysine binding domain. FEBS Lett. 513(1),
124–128 (2002).
67 Filippakopoulos P, Knapp S. The
bromodomain interaction module. FEBS Lett.
586(17), 2692–2704 (2012).
204
68 Kornberg RD, Thomas JO. Chromatin
structure; oligomers of the histones. Science
184(4139), 865–868 (1974).
69 Kornberg RD. Chromatin structure: a
repeating unit of histones and DNA. Science
184(4139), 868–871 (1974).
70 Luger K, Mäder AW, Richmond RK, Sargent
2011–2012. Nat. Rev. Drug Disc. 12(8), 569
(2013).
85 Murray CW, Verdonk ML, Rees DC.
Experiences in fragment-based drug
discovery. Trends Pharmacol. Sci. 33(5),
224–232 (2012).
86 Congreve M, Chessari G, Tisi D, Woodhead
DF, Richmond TJ. Crystal structure of the
nucleosome core particle at 2.8 Å resolution.
Nature 389(6648), 251–260 (1997).
AJ. Recent developments in fragment-based
drug discovery. J. Med. Chem. 51(13),
3661–3680 (2008).
71 Allfrey V, Faulkner R, Mirsky A. Acetylation
87 Congreve M, Carr R, Murray C, Jhoti H. A
and Methylation of Histones and their Possible
Role in Regulation of RNA Synthesis. Proc.
Natl Acad. Sci. USA 51(5), 786–794 (1964).
72 Strahl BD, Allis CD. The language of
covalent histone modifications. Nature
403(6765), 41–45 (2000).
73 Turner BM. Histone acetylation and an
epigenetic code. Bioessays 22(9), 836–845
(2000).
74 Schreiber SL, Bernstein BE. Signaling
network model of chromatin. Cell 111(6),
771–778 (2002).
75 Gardner KE, Allis CD, Strahl BD. Operating
on chromatin, a colorful language where
context matters. J. Mol. Biol. 409(1), 36–46
(2011).
76 Grunstein M. Histone acetylation in
chromatin structure and transcription. Nature
389(6649), 349–352 (1997).
77 Jenuwein T, Allis CD. Translating the histone
code. Science 293(5532), 1074–1080 (2001).
78 Bernstein BE, Schreiber SL. Global
approaches to chromatin. Chem. Biol. 9(11),
1167–1173 (2002).
79 Turner BM. The adjustable nucleosome: an
epigenetic signaling module. Trends Genet.
28(9), 436–444 (2012).
80 Bird A. Perceptions of epigenetics. Nature
447(7143), 396–398 (2007).
81 Ptashne M. Epigenetics: core misconcept.
Proc. Natl Acad. Sci. USA 110(18), 7101–7103
(2013).
82 Bakker A, Caricasole A, Gaviraghi G et al.
How to achieve confidence in drug discovery
and development: managing risk (from a
reductionist to a holistic approach).
ChemMedChem 4(6), 923–933 (2009).
83 Lipinski CA. Phenotypic and In Vivo
Screening: Lead Discovery and Drug
Repurposing. Royal Society of Chemistry,
London, UK, 86 (2012).
‘rule of three’ for fragment-based lead
discovery? Drug Discov. Today 8(19),
876–877 (2003).
88 Scott DE, Coyne AG, Hudson SA, Abell C.
Fragment-based approaches in drug
discovery and chemical biology. Biochemistry
51(25), 4990–5003 (2012).
89 Baker M. Fragment-based lead discovery
grows up. Nat. Rev. Drug Disc. 12(1), 5–7
(2013).
90 van Rooij E, Olson EN. MicroRNA
therapeutics for cardiovascular disease:
opportunities and obstacles. Nat. Rev. Drug
Disc. 11(11), 860–872 (2012).
„„Patents
101 Adachi K, Hikawa H, Hamada M et al.:
WO2006129623 (2006).
102 Miyoshi S, Ooike S, Iwata K, Hikawa H,
Sugahara K: WO2009084693 (2009).
„„Websites
201 Resverlogix News Release.
www.resverlogix.com/media/press-release.
html?id=492
202 Clinical trial identifier: NCT01728467.
http://clinicaltrials.gov/ct2/show/
NCT01728467?term=rvx208&rank=2
203 Oncoethix.
www.oncoethix.com/product/otx015
204 Constellation pharma.
www.constellationpharma.com/2013/09/
constellation-pharmaceuticals-initiatesclinical-development-of-cpi-0610-a-novel-betprotein-bromodomain-inhibitor-in-patientswith-lymphoma
205 Clinical trial identifier: NCT01587703.
http://clinicaltrials.gov/ct2/show/NCT01587
703?term=gsk525762A&rank=1
84 Arrowsmith J, Miller P. Trial Watch:
Phase II and Phase III attrition rates
Future Med. Chem. (2014) 6(2)
future science group