BET bromodomain ligands: Probing the WPF shelf to improve BRD4 bromodomain affinity and metabolic stability

Article abstract of DOI:10.1016/j.bmc.2018.05.003

Ligands for the bromodomain and extra-terminal domain (BET) family of bromodomains have shown promise as useful therapeutic agents for treating a range of cancers and inflammation. Here we report that our previously developed 3,5-dimethylisoxazole-based BET bromodomain ligand (OXFBD02) inhibits interactions of BRD4(1) with the RelA subunit of NF-κB, in addition to histone H4. This ligand shows a promising profile in a screen of the NCI-60 panel but was rapidly metabolised (t_? = 39.8 min). Structure-guided optimisation of compound properties led to the development of the 3-pyridyl-derived OXFBD04. Molecular dynamics simulations assisted our understanding of the role played by an internal hydrogen bond in altering the affinity of this series of molecules for BRD4(1). OXFBD04 shows improved BRD4(1) affinity (IC₅₀ = 166 nM), optimised physicochemical properties (LE = 0.43; LLE = 5.74; SFI = 5.96), and greater metabolic stability (t_? = 388 min).

Full text of DOI:10.1016/j.bmc.2018.05.003

Accepted Manuscript
BET bromodomain ligands: Probing the WPF shelf to improve BRD4 bromo-
domain affinity and metabolic stability
Laura E. Jennings, Matthias Schiedel, David S. Hewings, Sarah Picaud,
Corentine M.C. Laurin, Paul A. Bruno, Joseph P. Bluck, Amy R. Scorah, Larissa
See, Jessica K. Reynolds, Mustafa Moroglu, Ishna N. Mistry, Amy Hicks, Pavel
Guzanov, James Clayton, Charles Evans, Giulia Stazi, Philip C. Biggin, Anna
K. Mapp, Ester M. Hammond, Philip G. Humphreys, Panagis Filippakopoulos,
Stuart J. Conway
PII:
S0968-0896(18)30382-1
https://doi.org/10.1016/j.bmc.2018.05.003
BMC 14345
DOI:
Reference:
To appear in:
Bioorganic & Medicinal Chemistry
Received Date:
Revised Date:
Accepted Date:
23 February 2018
30 April 2018
2 May 2018
Please cite this article as: Jennings, L.E., Schiedel, M., Hewings, D.S., Picaud, S., Laurin, C.M.C., Bruno, P.A.,
Bluck, J.P., Scorah, A.R., See, L., Reynolds, J.K., Moroglu, M., Mistry, I.N., Hicks, A., Guzanov, P., Clayton, J.,
Evans, C., Stazi, G., Biggin, P.C., Mapp, A.K., Hammond, E.M., Humphreys, P.G., Filippakopoulos, P., Conway,
S.J., BET bromodomain ligands: Probing the WPF shelf to improve BRD4 bromodomain affinity and metabolic
stability, Bioorganic & Medicinal Chemistry (2018), doi: https://doi.org/10.1016/j.bmc.2018.05.003
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

BET bromodomain ligands: Probing the WPF shelf to improve BRD4 bromodomain affinity
and metabolic stability
Authors
a†
a†
a
b
Laura E. Jennings , Matthias Schiedel , David S. Hewings , Sarah Picaud , Corentine M. C.
a
c,d,e
a,f
a
a
a
Laurin , Paul A. Bruno , Joseph P. Bluck , Amy R. Scorah , Larissa See , Jessica K. Reynolds ,
a
g
a
a
a
Mustafa Moroglu , Ishna N. Mistry , Amy Hicks , Pavel Guzanov , James Clayton , Charles
a
h
f
c,d,e
g
Evans , Giulia Stazi , Philip C. Biggin , Anna K. Mapp , Ester M. Hammond , Philip G.
i
b
a,
Humphreys, Panagis Filippakopoulos , Stuart J. Conway *
a
Department of Chemistry, Chemistry Research Laboratory, University of Oxford, Mansfield Road, Oxford,
OX1 3TA, United Kingdom.
b
Nuffield Department of Clinical Medicine, Structural Genomics Consortium, University of Oxford, Old Road
Campus Research Building, Roosevelt Drive, Oxford, OX3 3TA, United Kingdom.
c
Department of Chemistry, University of Michigan, Ann Arbor, Michigan 48109-1055, United States
Life Sciences Institute, University of Michigan, Ann Arbor, Michigan 48109-2216, United States
Program in Chemical Biology, University of Michigan, Ann Arbor, Michigan 48109-2216, United States
Department of Biochemistry, University of Oxford, South Parks Road, Oxford, OX1 3QU, United Kingdom.
d
e
f
g
CRUK/MRC Oxford Institute for Radiation Oncology, Department of Oncology, University of Oxford, Old
Road Campus Research Building, Oxford, OX3 7DQ, United Kingdom.
h
Department of Chemistry and Technologies of Drugs, Sapienza University of Rome,
Piazzale Aldo Moro 5, 00185 Rome, Italy.
i
Epigenetics Discovery Performance Unit, GlaxoSmithKline R&D, Stevenage, Hertfordshire, SG1 2NY,
United Kingdom
†
These authors contributed equally to this work.
*To whom correspondence should be addressed: stuart.conway@chem.ox.ac.uk, telephone: +44 (0)1865
285 109, fax: +44 (0)1865 285 002.
1

Abstract
Ligands for the bromodomain and extra-terminal domain (BET) family of bromodomains have
shown promise as useful therapeutic agents for treating a range of cancers and inflammation. Here
we report that our previously developed 3,5-dimethylisoxazole-based BET bromodomain ligand
(
OXFBD02) inhibits interactions of BRD4(1) with the RelA subunit of NF-κB, in addition to histone
H4. This ligand shows a promising profile in a screen of the NCI-60 panel but was rapidly
metabolised (t = 39.8 mins). Structure-guided optimisation of compound properties led to the
½
development of the 3-pyridyl-derived OXFBD04. Molecular dynamics simulations assisted our
understanding of the role played by an internal hydrogen bond in altering the affinity of this series
of molecules for BRD4(1). OXFBD04 shows improved BRD4(1) affinity (IC50 = 166 nM), optimised
physicochemical properties (LE = 0.43; LLE = 5.74; SFI = 5.96), and greater metabolic stability (t
388 mins).
½
=
2

1
. Introduction
Lysine acetylation is a prevalent protein post-translational modification (PTM) that occurs
1
throughout the proteome and is similar to phosphorylation in its ability to regulate protein
2
function. The role of acetyl-lysine (KAc) has been heavily studied in histone proteins, with KAc
3, 4
recognised as one of the key “marks” proposed to comprise the epigenetic code. Lysine
acetylation state is regulated by lysine acetyl transferases (KATs) and lysine deacetylases
(KDACs), while bromodomains are viewed as readers of KAc marks, and mediate chromatin-
5
protein interactions that are frequently involved in transcriptional regulation. There are 61
bromodomains found within 46 bromodomain-containing proteins (BCPs) in the human proteome
6
which, despite having diverse primary sequences, share a common protein fold and structure. The
KAc residue binds in a well-defined pocket that, in some bromodomains, contains 5 structurally-
conserved water molecules at its base, and possesses a key recognition residue that hydrogen
bonds to the KAc; in canonical bromodomains this is an Asn residue. A sustained effort over recent
years has resulted in small-molecule ligands being identified for an increasing number of these
7-12
bromodomains.
The majority of work has focused on the development of ligands for the
bromodomain and extra terminal domain (BET) family of BCPs, comprising bromodomain-
containing proteins 2-4 (BRD2-4) and the testis-specific BRDT; each of these proteins contains two
adjacent canonical bromodomains. There is also increasing work on the development of ligands for
13-16
the non-BET bromodomains.
The BET bromodomains have emerged as exciting therapeutic
targets with over 20 clinical trials involving BET bromodomain ligands in progress, primarily
14, 17-19
focused on oncology indications.
Here we report further cellular data for our previously
20, 21
reported BET bromodomain ligand OXFBD02 (1),
and subsequent structure-activity
relationship (SAR) studies aimed at optimising the solubility and metabolic stability of this series of
compounds. This work resulted in the development of OXFBD04 (9j), which displays improved
BRD4(1) affinity and substantially enhanced metabolic stability compared to OXFBD02.
3

2
. Results and discussion
A common feature of almost all bromodomain ligands is a component that occupies the KAc-
binding pocket and mimics the interactions formed by KAc with the bromodomain. Work by us
20-25
26-35
and others
has shown that the 3,5-dimethylisoxazole group is a particularly effective KAc
mimic, which has been used as the basis of ligands for the CREBBP bromodomain and the BET
family of BCPs (Fig. 1).
Fig. 1. Chemical structures and in vitro inhibition data of the 3,5-dimethylisoxazole-based BET
2
1, 26, 27, 31, 33, 34
bromodomain ligands 1-5,
the 3,5-dimethylisoxazole-based CREBBP bromodomain
24
36 37
ligand 6, and the structurally unrelated BET bromodomain ligands PFI-1 (7) and (+)-JQ1 (8).
Using a fragment-based approach we previously developed OXFBD02 (1a) and OXFBD03 (1b)
which have IC50 values of 384 nM and 371 nM in an AlphaScreen assay for the first bromodomain
of BRD4 [BRD4(1)], and IC50 values for cytotoxicity of 794 nM and 616 nM in the MV4;11 acute
20, 21
myeloid leukaemia (AML) cell line, respectively.
4

2
.1. Cellular evaluation of OXFBD02 (1)
To determine its activity against a wider panel of cancer cell lines, OXFBD02 (1a) was submitted
38
for testing at a single dose against the NCI-60 human cancer cell line screen. Growth inhibition
(GI) was determined after 48 h treatment at 10 μM, using a sulforhodamine B assay (to indicate
cellular protein content). The compound was subsequently evaluated at five concentrations
between 10 nM and 100 μM to obtain GI50 values, and to indicate the concentration required
(if < 100 μM) for total growth inhibition (TGI) (Table S1). A box plot of the data grouped by cancer
type is shown (Fig. 2).
Fig. 2. Box plot of growth inhibition (GI) by OXFBD02 (1a) in NCI-60 panel. Plot indicates median, range and
quartiles of GI by 1a after 48 h treatment at 10 μM in the NCI-60 DTP Human Tumour Cell Screen, grouped
by cancer type. The number in brackets is the number of cell lines of each type in the NCI-60 panel.
39
(Graphical representation in analogy to Lucas et al.)
Similar to other BET bromodomain ligands, OXFBD02 (1a) was particularly effective at inhibiting
the growth of leukaemia-, breast-, and renal-cancer cell lines. Calculation of the GI50 values (see
Table S1) allow comparison of OXFBD02 (1a) with the NCI-60 data obtained for PFI-1 (7, see Fig.
36
1
), which is a chemically-distinct well-characterised BET bromodomain ligand. PFI-1 (7) and
OXFBD02 (1a) display well correlated patterns of cancer cell toxicity (Pearson’s product-moment
36
correlation coefficient, r = 0.82; Table S1, Fig. S1 and S2). These data suggest that the activity
displayed by both compounds results predominantly from interaction with the BET bromodomains,
rather than other chemotype-specific off-target cellular interactions.
5

Beyond its interactions with KAc residues in histone proteins, BRD4 has been reported to bind
KAc310 of the RelA subunit of NF-κB. The binding of BRD4 leads to recruitment of cyclin-
dependent kinase 9 (CDK9), phosphorylation of RNA polymerase II, and consequent activation of
40-43
NF-κB-dependent gene expression.
Disruption of the RelA-BRD4 interaction with the BET
41
ligand (+)-JQ1 (8) was previously shown to suppress NF-κB-dependent transcription. To
investigate whether OXFBD02 (1a) also disrupts the interaction of BET bromodomains KAc310 of
RelA, and to assess the cellular effects of OXFBD02 (1a) in a functional assay, we employed a
44
luciferase reporter system based on a previously published procedure. This allowed us to
determine the effects of BRD4 bromodomain binding on NF-κB-dependent transcription. (+)-JQ1
(8, see Fig. 1) was employed as a positive control and exerted a potent suppression of NF-κB-
dependent gene expression, whereas the inactive enantiomer (–)-JQ1 did not show significant
effects. Addition of OXFBD02 (1a) led to a concentration-dependent suppression of NF-κB-
dependent gene expression (Fig. 3), indicating that it also disrupts the interaction of the BRD4
bromodomains with KAc310 of RelA. The fact that complete repression of NF-κB-dependent
transcription was not observed with either 1a or (+)-JQ1 suggests that there might be other PPIs
involved in this process.
Fig. 3. Luciferase reporter gene assay allows assessment of the effect of compounds on NF-κB-dependent
gene expression. HeLa cells were transiently transfected with a reporter plasmid containing luciferase and
five κB binding sites. The cells were treated with DMSO as a control (–), or different concentrations (0.2, 1, 5
and 25 µM) of OXFBD02 (1a), (+)-JQ1 (8), or (–)-JQ1 for 12 hours before stimulation with IL-1β (+). The
gene expression without stimulation of IL-1β (–) was also measured for the DMSO-treated control. Results
6

were obtained in quadruplicate and averaged, with error bars signifying standard deviation. OXFBD02 (1a)
shows a concentration-dependent effect on transcription.
2
.2. Designing compounds to investigate the WPF shelf-binding region
Taken together the data above indicate that the series of compounds exemplified by OXFBD02
1a) are potentially useful tools to investigate the function of the BET bromodomains in cellular and
ultimately in vivo settings. With this in mind, we used Metasite 3.1.2 (phase I) and Meteor 2.0.2
phase I and II) to predict the metabolic liabilities of OXFBD02 (1a) (Fig. S3). This analysis
(
(
indicated that the 4-position of the phenyl ring was the most significant liability, presumably through
CYP450-catalysed oxidation. In accordance with this prediction, investigation of the metabolic
stability of OXFBD02 (1a) in human liver microsomes showed that 1a is a relatively high clearance
compound, with CLint of 34.8 µL/min/mg protein and a cellular half-life of 39.8 mins (Table S2).
These data are consistent with work by Sharp et al. who showed that, in a direct comparison, 3,5-
dimethylisoxazole-based compounds showed the highest affinity for the BET bromodomain, but
31
also poor metabolic stability. Therefore, while optimising our compounds we wished to determine
whether poor metabolic stability is an inherent problem with 3,5-dimethylisoxazole-based
compounds, or whether by carefully balancing the properties of the whole molecule this apparent
liability can be overcome.
Given the structure of 1a, and the metabolic liability of the para-position, the region of the molecule
that is most suitable for optimisation is the WPF shelf-binding phenyl ring. It has generally been
observed that binding of a lipophilic moiety to this region confers high affinity for the BET
7
bromodomains. However, the lipophilicity of this group has to be balanced with the overall
properties of the molecule to ensure that the compound is soluble. Solubility Forecast Index
45
(
SFI = clogDpH7.4 + #Ar) is a useful parameter for predicting compound solubility, and is especially
useful for comparing molecules within a series. OXFBD02 (1a) is soluble enough to be used in in
vitro and cellular settings, and has SFI = 6.6, suggesting that new analogues should have an SFI
value of 6.6 or lower to ensure useful solubility. In addition, CLint is hydrophobicity dependent and
therefore reducing compound hydrophobicity might help to improve compound metabolic stability.
46
7

30
Based on these observations, and cognisant of work by Gehling et al. we designed a series of
compounds to probe the WPF shelf-binding region of the molecule. In particular we wished to
explore blocking the 4-position of the phenyl ring with metabolically stable functionality to be a
facile way to overcome the issue of metabolic oxidation.
Compounds with both electron-withdrawing and electron-donating substituents at the 2-, 3-, and 4-
positon of the ring (9a-e, g, l, m, o) were designed to probe the optimum electronics and vector for
ring substitution. Compound 9f contains 3,4-dichloro-substitution to determine whether two
substituents are tolerated on the phenyl ring. The 2-, 3-, and 4-pyridyl derivatives (9j, k, p) were
designed as these compounds have favourable SFI values of 5.9, and the 4-chloro-2-pyridyl
derivative (9n) was designed to reduce the basicity of the pyridine nitrogen atom (pK
a
= 2.22 (pK
a
value of the conjugated acid), predicted using ACD/Labs I-Lab 2.0 software [Algorithm Version:
v12.1.0.50374]). We designed compounds with cyclohexyl and cyclopropyl substituents (9h-i) to
determine whether non-aromatic rings are accepted by the WPF shelf. We also wished to probe
whether the secondary hydroxyl group is essential for good BET bromodomain affinity, and so a
series of compounds was designed without this functionality, which also allowed the introduction of
less polar solubilising groups, mainly via reductive amination. In this series, we investigated
whether more polar groups, conveying favourable physicochemical properties to the molecule
(
10a-g), are able to bind the WPF shelf. We were particularly interested in 4,4-difluoropiperidinyl
analogue (10d) as the two geminal fluorine substituents were predicted to lower basicity of the
tertiary amine (pK = 4.67 (conjugated acid of the tertiary amine)), predicted using ACD/Labs I-Lab
.0 software [Algorithm Version: v12.1.0.50374], Table S3), which suggests that the compound will
a
2
exist in a predominantly deprotonated state at physiological pH, allowing binding of the WPF shelf.
We also designed two amide-based compounds (11a-b) to investigate the importance of the
tetrahedral carbon atom proximal to the WPF shelf-binding group.
2
.3. Synthesis
21
Our general synthetic strategy is based on that previously reported for the OXFBD02 (1a). The
synthesis of aldehyde 17 has been optimised (Scheme 1A and Supporting Information) as this is a
8

key intermediate in the synthesis of the secondary alcohols 9a-n, and the amines 10a-d. The
secondary alcohols 9o-p were generated by reaction of the TIPS-protected analogue 18 with
organolithium reagents (Scheme 1A), followed by deprotection.
Scheme 1. Synthesis of the 3,5-dimethylisoxazole-based compounds 9-11. (A) Synthesis of the secondary
alcohols 9a-p (9a: R = 4-chlorophenyl, 9b: R = 2-methoxyphenyl; 9c: R = 3-methoxyphenyl;
9d: R = 4-methoxyphenyl; 9e: R = 4-tolyl; 9f: R = 3,4-dichlorophenyl; 9g: R = 4-fluorophenyl;
9h: R = cyclohexyl; 9i: R = cyclopropyl; 9j: R = 3-pyridyl; 9k: R = 4-pyridyl; 9l: R = 3-fluorophenyl;
9m: R = 4-cyanophenyl; 9n: R = 4-chloro-2-pyridyl, 9o: R = 2-fluorophenyl, 9p: R = 2-pyridyl) and the amines
9

10a-f (10a: R = N-methylpiperazinyl; 10b: R = morpholinyl; 10c: R = N-benzylamino;
10d: R = 4,4-difluoropiperidinyl; 10e: R = piperidinyl; 10f: R = pyrrolidinyl). Reagents and conditions: (a) KF,
L-(+)-tartaric acid, CH
c) MnO , CHCl /EtOAc, reflux, 2 h, 49-78% (n=3), (d) Pd(OAc)
RMgBr, THF, rt, 3-19 h, 51-80%; for 9f: 3,4-dichlorophenylmagnesium bromide, THF, 50 °C, 14 h, 11%; for 9g-i:
Aryl/alkyl bromide, Mg, I , THF, rt to reflux, 30-60 min, then 17, THF, 0 °C to rt, 2-18 h, 72-85%; for
j: 3-bromopyridine, isopropylmagnesium chloride lithium chloride complex, THF, rt, 2 h, then 17, THF, rt, 4 h, 73%;
3 3
CN/THF, rt, 30 min, 59-73% (n =2); (b) BH ·THF, THF, 0 °C then rt, 18 h, 80-96% (n=3);
(
2
3
2
, RuPhos, Na
2
CO , EtOH, 80 °C, 85%; (e) for 9a-e:
3
2
9
for 9k: 4-iodopyridine , isopropylmagnesium chloride, THF, rt, 1 h, then 17, THF, rt to 50 °C, 22 h, 9%; for 9l-n: Aryl
iodide, isopropylmagnesium chloride lithium chloride complex, THF, −10 °C to rt, 2-5 h, then 17, THF, rt to 50 °C, 3 h
to 2 d, 11-23%; (f) for 10a-c: amine, AcOH, EtOH, pH = 4, rt, 20-40 min, then NaBH
0d: 4,4-difluoropiperidine hydrochloride, EtOH, rt, 30 min, then NaBH CN, rt, 19 h, 33%; (g) TIPSCl, imidazole,
DMF, 0 °C to rt, 13 h, 84%; (h) Amine, AcOH, EtOH, rt, 1 h, then NaBH CN, rt, 22 h, then TBAF, THF, 0 °C to rt,
-2 h, 8-33%; (i) Aryl bromide, BuLi, THF, −78 °C, 40 min, then 18, THF, −78 °C to rt, 3-16 h, then TBAF, THF,
°C, 0.5-2 h, 17-89%; (B) Synthesis of 10g. Reagents and conditions: (a) Et SiH, TFA, rt, 15 min, 70%;
3
CN, rt, 17-23 h, 20-69%; for
1
3
3
n
1
0
3
2 2 3
(C) Synthesis of 11a-b. Reagents and conditions: (a) 13, Pd(OAc) , RuPhos, Na CO , EtOH, 90 °C, mw,
1
.5 h, 68%; (b); LiOH, THF, rt, 24 h, 96%; (c) EDC hydrochloride, HOBt hydrate, THF, rt, 20 min, then amine,
5 °C, 3 d, 52-56% (see Experimental Section for more detailed procedures).
5
Reductive amination of 18 with subsequent deprotection furnished the amines 10e-f (Scheme 1A).
The methylene derivative 10g was synthesised by treating OXFBD02 (1a) with TFA and Et SiH to
3
reduce the secondary alcohol (Scheme 1B). To obtain the amides 11a-b we coupled the carboxylic
acid 21 to the appropriate secondary amine using 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide
(EDC) (Scheme 1C). Detailed information for the synthesis of all final compounds (9a-p, 10a-f,
11a-b) as well as the intermediates (13, 15-18, 20-21) can be found in the Experimental Section.
10

2
.4. In vitro inhibition and SAR studies
A well-validated Amplified Luminescence Proximity Homogeneous Assay (AlphaScreen™) peptide
displacement assay was used to measure the BRD4(1) IC50 values for the synthesised compound,
47
and the data obtained are shown in Table 1.
Table 1. IC50 values, pIC50 values, ligand efficiencies (LE), clogP values, lipophilic ligand efficiencies (LLE),
cLogDPH7.4 values, and solubility forecast index (SFI) for 9a-p, 10a-g, 11a-b, including 1a as a reference
47
compound. Heat map shows relative IC50 values obtained in an AlphaScreen assay. Red indicates low IC50
values, and green indicates high IC50 values. Quoted IC50 values are a mean of three technical repeats.
Ranges in parentheses represent 95% confidence intervals resulting from sigmoidal curve fitting to the
triplicate data.
a
b
b
Compound Substituent
BRD4(1) IC50 (µM)
pIC50
6.42
LE
cLogP
LLE
cLogDpH7.4
SFI
0
.384
1a
9a
9b
9c
9d
c
0.41
2.53
3.89
3.62
6.62
(
0.346-0.420)
0.631
6.20
6.57
6.32
6.23
0.38
0.38
0.37
0.36
3.13
2.45
2.45
2.45
3.07
4.12
3.87
3.78
4.28
3.80
3.80
3.80
7.28
6.80
6.80
6.80
(
0.539-0.739)
0.270
(
0.223-0.327)
0.478
(
0.402-0.570)
0.585
(
0.531-0.644)
0
.296
9
9
9
e
6.53
6.02
6.07
0.40
0.35
0.37
2.99
3.60
2.59
3.54
2.42
3.48
4.32
5.01
4.00
7.32
8.01
7.00
(
0.264-0.332)
0.945
f
(
0.865-1.03)
0.842
g
(
0.697-1.02)
0
.166
9
9
9
h
6.78
6.42
6.78
0.43
0.47
0.43
3.24
1.54
1.04
3.54
4.88
5.74
4.24
3.26
2.96
6.24
5.26
5.96
(
0.158-0.175)
0.377
i
(
0.339-0.420)
0.166
j
(
0.142-0.193)
11

0
.303
9
k
l
6.52
6.10
6.22
0.41
0.37
0.36
1.04
2.59
1.97
5.48
3.51
4.25
2.86
4.00
3.53
5.86
7.00
6.53
(
0.262-0.349)
0.793
9
(
0.685-0.919)
0.604
9m
(
0.536-0.680)
0
.495
9n
9o
9p
6.31
5.85
5.33
0.38
0.36
0.34
1.89
2.59
1.04
4.42
3.26
4.29
3.49
4.00
2.96
6.49
7.00
5.96
(
0.427-0.574)
1.42
(
1.21-1.68)
4.68
(
3.40-6.45)
1
(
.23
1.02-1.47)
.956
0.802-1.14)
.25
1.82-2.77)
10a
10b
10c
5.91
6.02
5.65
0.38
0.40
0.34
0.72
0.97
3.21
5.19
5.05
2.44
1.41
2.34
2.34
3.41
4.34
5.34
0
(
2
(
0
.235
10d
6.63
0.40
1.37
5.26
3.62
5.62
(
0.210-0.263)
4
(
.28
3.39-5.40)
.27
2.43-4.38)
.10
0.990-1.22)
1
1
1
0e
5.37
5.49
5.96
0.36
0.38
0.40
2.55
1.98
4.00
2.82
3.51
1.96
2.22
1.27
4.96
4.22
3.27
7.96
3
(
0f
1
(
0g
0
.722
1
1a
1b
6.14
5.42
0.39
0.34
1.62
0.07
4.52
5.35
3.21
2.38
5.21
4.38
(
0.568-0.918)
3.80
1
(
3.04-4.74)
a
Protein concentration was adjusted to a final assay concentration of 10 nM, peptide concentration was 4 nM
b
(see Supporting Information for detailed assay procedures). cLogP and cLogDpH7.4 values were calculated
c
21
using ACD/Labs I-Lab 2.0 software (Algorithm Version: 5.0.0.184). Values taken from Hewings et al.
Addition of a 4-chloro substituent (9a, Table 1) resulted in a small drop in BRD4(1) affinity (IC50
.631 μM). Addition of a methoxy group was generally well tolerated, with 2-position (9b)
substitution preferred (IC50 = 0.270 μM) over 3- (9c) or 4-position (9d) substitution. The 4-tolyl
=
0
derivative (9e) showed similar affinity to 1a, indicating that lipophilic substituents are favourable for
BRD4(1) binding. Interestingly, fluorine was less well tolerated than other substituents (9g, l, o),
with the 2-fluoro derivative (9o) displaying the lowest BRD4(1) affinity (IC50 = 1.42 μM). The 3-
pyridyl (9j, IC50 = 0.166 μM) and 4-pyridyl (9k, IC50 = 0.303 μM) derivatives showed high BRD4(1)
12

affinity. Intriguingly the 2-pyridyl derivative (9p) showed substantially reduced BRD4(1) affinity (IC50
4.68 μM). Both the cyclopropyl (9i) and cyclohexyl substituents (9h) were well tolerated. The
=
cyclohexyl derivative displayed the joint highest affinity (IC50 = 0.166 μM) for BRD4(1), albeit at the
expense of increased lipophilicity (clogD7.4 = 4.24, LLE = 3.54) compared to the equipotent 3-
pyridyl derivative (9j, clogD7.4 = 2.96, LLE = 5.74). The phenyl derivative in which the secondary
hydroxyl group is removed (10g) showed a significantly reduced BRD4(1) affinity (IC50 = 1.10 μM)
compared to the matched pair of compound 1a. In general, other more polar substituents were
poorly tolerated and displayed higher IC50 values. The piperazine amide derivative (11a) was
relatively well tolerated, whereas the morpholine derivative (11b) was not.
We determined a BRD4(1) dissociation constant (K
titration calorimetry (ITC) (Fig. S4A), which is consistent with the AlphaScreen results. For 1a we
determined a K value of 0.435 μM ± 0.16 μM (Fig. S4B). Selectivity profiling in a BROMOscan
d
) for 9j of 0.247 μM ± 0.08 μM using isothermal
d
assay against 12 phylogenetically diverse bromodomains confirmed that both 1a and 9j potent
BET bromodomain ligands, with additional modest affinity for the CREBBP bromodomain, but no
significant affinity for any of the other of the selected bromodomains investigated (Table S4).
2
.5. Structural studies
To understand the structural basis of our SAR observations we obtained X-ray crystal structures of
the 3-pyridyl derivative (9j; PDB code 6FSY), the cyclopropyl derivative (9i; PDB code 6FT3), and
the difluorinated piperidine derivative (10d; PDB code 6FT4) in complex with BRD4(1). In all cases
the molecules bind to BRD4(1) broadly as expected. The 3,5-dimethylisoxazole acts as the KAc
mimic and forms a hydrogen bond with N140 and a water-mediated hydrogen bond with Y97 (Fig.
4
A). Interestingly in the case of 9j and 9i, only the (R)-enantiomer is observed in the X-ray crystal
structure, despite a racemate being submitted to crystallisation. It should be noted that despite
these two molecules having the same absolute configuration, they have the opposite sense of
stereochemistry due to the priority assignment of the pyridine ring compared to the cyclopropyl
ring.
13

A
B
C
D
Fig. 4A. Overlaid X-ray crystal structures of (R)-OXFBD04 [(R)-9j, PDB code 6FSY, carbon = yellow;
resolution: 1.34 Å] and the diacetylated histone H4-mimicking peptide H41-12KAc5KAc8 (PDB code 3UVW,
carbon = purple) bound to BRD4(1). B. Overlaid X-ray crystal structures of (R)-OXFBD04 [(R)-9j, PDB code
21
6
FSY, carbon = yellow) and (S)-OXFBD02 [(S)-1a, PBD code 4J0S, carbon = orange] bound to BRD4(1),
showing that the molecules have very similar binding modes to BRD4(1). (R)-9j forms additional water-
mediated interactions with D145, which are not possible for 1a. C. Overlaid X-ray crystal structures of (R)-
OXFBD04 [(R)-9j, PDB code 6FSY, carbon = yellow) and the I-BET151 (3, PDB code 3ZYU, carbon =
marine blue) bound to BRD4(1), showing that the pyridine nitrogen does not overlay with that of I-BET151.
D. Overlaid X-ray crystal structures of the cyclopropyl derivative (R)-9i (PDB code 6FT3, carbon = light teal;
1.28 Å) and (R)-OXFBD02 (1a, PBD code 4J0S, carbon = orange) bound to BRD4(1), showing that the
molecules do not overlay precisely when binding to BRD4(1).
Compound 9j overlays precisely with (S)-OXFBD02 (which has the same sense of
stereochemistry; PDB code 4J0S) with the pyridine ring occupying the WPF shelf (Fig. 4B). The
pyridine nitrogen is oriented away from the W81 and forms a water-mediated hydrogen bond with
D145. It is feasible that this additional interaction, which is not present in OXFBD02, is responsible
(at least in part) for the increase in BRD4(1) affinity displayed by 9j. Interestingly, the pyridine
14

nitrogen does not overlay with that of I-BET151 (3) when bound to BRD4(1) (PDB code 3ZYU)
Fig. 4C), and the nitrogen atom in I-BET151 does not appear to form any interactions with
(
BRD4(1). The phenol moiety of 9j forms a hydrogen bond to the ZA-channel water molecule in the
same manner as OXFBD02 (1a). The cyclopropyl derivative (9i) does not overlay so precisely with
(R)-OXFBD02, with the hydroxyl group oxygen atoms displaced by 1.4 Å. The cyclopropyl group
occupies the WPF shelf and is oriented towards W81. However, analysis of the B-factors for the
ligand indicates that the cyclopropyl ring is the most flexible component of the molecule, and that in
general 9i is more flexible than 9j, perhaps reflecting the higher affinity of the latter for BRD4(1).
The SAR of the pyridyl derivatives (9j, k, p) is particularly intriguing given the significant difference
in BRD4(1) resulting from moving the nitrogen atom one position around the ring. We hypothesised
that the decreased potency of 2-pyridyl derivative 9p could be affected by the formation of an
intramolecular H-bond between the nitrogen lone pair of the 2-pyridyl substituent and the
secondary hydroxyl group. This interaction cannot form in the 3-pyridyl derivative 9j. We conducted
1
H NMR experiments to determine whether we could detect formation of an internal hydrogen bond
in 9p in solution. To assess the presence and strength of a solution-phase hydrogen bond we
1
observed the change in chemical shift of the given hydrogen atom when the H NMR solvent is
48
changed from CDCl
hydrogen bonds typically show a Δppm CDCl
predominantly dictated by the surrounding solvent. Hydrogen atoms that are involved in a
hydrogen bond typically show Δppm CDCl → D -DMSO < 1 ppm, as their environment is mainly
3
6
to D -DMSO. The chemical shift of hydrogen atoms that are not engaged in
3
→ D -DMSO = 2-4 ppm, as their environment is
6
3
6
dictated by the intramolecular interaction, and hence less affected by the surrounding solvent. This
technique is especially powerful when combined with structural studies, as it allows comparison
between intramolecular hydrogen bonds formed in solution phase and those present when a ligand
is bound to a protein.
In D
DMSO = +4.2 ppm) consistent with this group being solvent exposed. In contrast, only a small
change was seen for the secondary alcohol (Δppm CDCl → D -DMSO = +0.8 ppm), which supports
the idea that this group is engaged in an intramolecular hydrogen bond (Fig. 5A). We hypothesised
6
-DMSO, the signal for the phenolic hydroxyl group of 9p shifts downfield (Δppm CDCl
3 6
→D -
3
6
15

that the 4-chloro-2-pyridyl derivative (9n) would have a less basic nitrogen atom due to the
inductive electron withdrawing effects of the 4-chloro substituent, and that this would result in a
weaker internal H-bond. Consistent with this prediction, we observed an increased Δppm
CDCl
bond (Δppm CDCl
intermediate BRD4(1) affinity (IC50 = 495 nM) between 9j and 9p.
3 6
→D -DMSO of +1.6 ppm for the secondary alcohol, indicating a weaker internal hydrogen
3
→D -DMSO = +4.7 ppm for the phenol of 9n). This compound shows
6
1
Fig. 5. Studies to rationalise the observed structure-activity relationship. A) H NMR spectra of 9p dissolved
in different mixtures of CDCl and D -DMSO displayed in a range from 5.0-10.0 ppm. (B) Chemical shifts
of -CHOH protons of pyridyl analogues 9p, 9n, 9j, and 1a, which was used as a control for a compound that
is not able to form an intramolecular H-bond, plotted against D -DMSO concentration. All experiments were
3
6
6
performed at a compound concentration of 2 mg/mL. See Figure S5-8 for additional information. (C)
Molecular dynamics studies for the enantiomers of 9p and 9j bound to BRD4(1). Representative plots
showing the moving average of the dihedral angle between the bonds shown in red during three 50 ns MD
simulations. A graph summarising the results of all randomised MD simulations that have been performed
can be found in Figure S9-10.
16

We proposed that the internal hydrogen bond present in 9p would result in the molecule adopting a
solution state conformation that was unfavourable for protein binding. To investigate this idea, we
conducted molecular dynamics (MD) simulations to predict the conformation that both enantiomers
of 9p would adopt when bound to BRD4(1). A 50 ns simulation indicates that (S)-9p adopts a very
similar conformation to (R)-OXBDF02, and that in this conformation the internal hydrogen bond is
intact (Fig. 6A). In this conformation the pyridine nitrogen atom is facing towards the solvent and
away from W81, which is also the case in the X-ray crystal structure of 9j, and an X-ray crystal
structure of I-BET151 (3) which contains a 2-pyridyl substituent (Fig. 4C).
A
B
Fig. 6A. A representative image of a 45 ns MD simulation of (S)-9p overlaid with the X-ray crystal structures
of (R)-OXFBD04 [(R)-9j, PDB code 6FSY, carbon = yellow] bound to BRD4(1). The predicted internal
hydrogen-bond is present in (S)-9p. B. A representative image of a 50 ns MD simulation of (R)-9p overlaid
with the X-ray crystal structures of (R)-OXFBD04 [(R)-9j, PDB code 6FSY, carbon = yellow] bound to
BRD4(1). The simulation indicates that it is favourable for the pyridine nitrogen atom of (S)-9p to orient away
from W8, requiring the internal hydrogen bond to be broken.
We reasoned that the orientation of the nitrogen atom away from the hydrophobic W81 residue is
likely favourable. However, while the (S)-enantiomer of 9p can adopt this orientation and maintain
the internal hydrogen bond, the (R)-enantiomer would not be able to bind to BRD4(1) and maintain
the internal hydrogen bond. This was shown to be the case in a 50 ns MD simulation, where the
pyridine nitrogen is oriented away from W81, and consequently the internal hydrogen bond is
broken (Fig. 6B). This observation provides an explanation for the low BRD4(1) affinity displayed
by racemic 9p, as only half the concentration of the ligand can bind to BRD4(1) with the internal
hydrogen bond intact. In the opposite enantiomer a large enthalpic penalty to break the hydrogen
17

bond must first be overcome before the ligand can bind to BRD4(1). This observation predicts that
S)-9p should have a much higher affinity for BRD4(1) than (R)-9p. Work to investigate this is
(
ongoing but is beyond the scope of these studies.
Compounds 10a-f were designed to investigate whether more polar groups would be tolerated on
the WPF shelf. As might be expected, most of these compounds show low BRD4(1) affinity, in line
with the idea that a lipophilic group is preferred in this region. The notable exception in this series
is the geminal difluorinated piperidine (10d), which has an IC50 value of 235 nM for BRD4(1). We
attribute this to the electron-withdrawing effects of the fluorine atoms resulting in reduced basicity
that supresses amine protonation, allowing the piperidine to interact better with the lipophilic WPF
shelf. An X-ray crystal structure of 10d bound to BRD4(1) (PDB code 6FT4; Fig. 7A) confirms that
the difluoropiperidine moiety does reside on the WPF shelf, as predicted. We calculated the pK
a
values of the amine analogues (Table S3) and plotted the predicted appearance of the non-ionic
state under assay conditions against the detected IC50 values (Fig. 7B). These data suggest that
positively charged amines do not bind well to the WPF shelf, with a correlation (linear regression
2
R = 0.8162) between BRD4(1) affinity and the pK
a
of the conjugated acid of the tertiary amine
observed.
18

A
B
Fig. 7A. Overlaid X-ray crystal structures of 10b (PDB code 6FT4, carbon = yellow; resolution: 1.34 Å) and
21
(
S)-OXFBD02 (1a, PBD code 4J0S, carbon = orange) bound to BRD4(1). The inductive electron-
withdrawing effect of the fluorine atoms reduces the basicity of the piperidine nitrogen sufficiently for it to
remain unprotonated and bind to the WPF shelf. B. BRD4(1) IC50 values of amines (10a-f) plotted against
the predicted appearance of the non-ionic state under assay conditions (pH = 7.6). Linear regression gives a
2
R value of 0.8162, indicating a good correlation between reduced basicity and increased BRD4(1) affinity.
a
Calculated using ACD/Labs I-Lab 2.0 software (Algorithm Version: v5.0.0.184).
While the cyclopropyl-derived 9i shows the highest LE of the series, the 3-pyridyl derivative 9j and
the cyclohexyl analogue 9h have the highest BRD4(1) affinity. However, while 9h has a LLE of
3
.54, 9j shows a LLE of 5.74, indicating that this BRD4(1) ligand has an optimised balance of
affinity and physicochemical properties.
2
.6. Metabolism and cellular studies
The metabolic stability of our most promising compounds (9j, 10d) was tested in a human
microsomal stability assay using OXFBD02 (1a) as a reference. The cyclohexyl derivative 9h was
excluded from these studies as its more lipophilic nature limits its solubility and increases the
likelihood of a poor metabolic profile. OXFBD02 (1a) and the 4,4-difluoropiperidinyl derivative (10d)
exhibited similar metabolic half-lives, with t1/2 = 39.8 min and t1/2 = 27.0 min, respectively.
Compounds 1a and 10d displayed CLint values of 34.8 and 51.4 µL/min/mg protein and are
49
consequently considered to be medium and high clearance compounds, respectively. In contrast,
the 3-pyridyl analogue (9j) displayed a significantly longer metabolic half-life of 388 min (Table 2).
With an intrinsic clearance (CLint) of 3.57 µL/min/mg protein, 9j is therefore classified as a low
19

clearance compound. These properties indicate that 9j is the optimal ligand in this series for
progression to use in cellular and potentially in vivo studies. However, it should be noted that the
microsomal stability assay only provides insights into phase I metabolism. More comprehensive
metabolic characterisation, including phase II conjugation reactions, will be included in future
studies.
Table 2. Metabolic stability data of 1a, 9j, and 10d including the intrinsic clearance (CLint ± standard error)
and the apparent half-life (t1/2) detected by means of a microsomal stability assay using human liver
microsomes. Dextromethorphan and verapamil were used as medium and high clearance controls,
respectively. Compounds were tested at a concentration of 3 µM.
Compound
CLint (µL/min/mg protein)
34.8 ± 3.76
3.57 ± 2.21
t
1/2 (min)
n
5
5
5
5
3
OXFBD02 (1a)
39.8
388
27.0
54.5
7.20
9j
10d
51.4 ± 2.6
dextromethorphan 25.4 ± 4.23
verapamil 192 ± 11.3
To rationalise the selection of compounds used for further cellular studies we have predicted the
cell permeability of all final compounds presented herein. All compounds were predicted to have
-
1
an absorption rate constant (K ) between 0.053 – 0.058 min which suggests excellent cellular
a
uptake (Table S5). As there is no significant difference in permeability among the synthesised
compounds we decided to further study the compound 9j displaying to most promising results in
our previous studies in a cellular setup. In a cell growth assay using A498 (renal), HT-29 (colon),
and MCF7 (breast) cancer cell lines, 9j showed low micromolar activity (Table 3 and Fig. S11).
OXFBD02 was used as a positive control, and the data for OXFBD02 are in line with that obtained
in the NCI-60 screen. To address the mechanism of cytotoxicity, we probed the effect of 9j and 1a
on MYC suppression in MCF7 cells. We show that 9j exerts a potent and time-dependent effect of
MYC suppression similar to 1a and (+)-JQ1, which was used as a positive control (Fig. 8). These
data indicate that OXFBD04 (9j) shows similar cellular efficacy to OXFBD02 (1a), which combined
with its enhanced metabolic stability makes OXFBD04 (9j) a useful tool compound for studying the
function of the BET bromodomains.
20

Table 3. GI50 data for OXFBD02 (1a) and OXFBD04 (9j). A498 (renal), HT-29 (colon), and MCF7 (breast)
cancer cell lines were used. Values quoted are a mean of three repeats. The standard deviation is shown.
Compound
GI50 (μM) A498
1.58 ± 1.35
4.88 ± 1.54
GI50 (μM) HT-29
5.79 ± 0.84
4.40 ± 0.81
GI50 (μM) MCF7
1.25 ± 0.47
1.40 ± 0.60
OXFBD02 (1a)
OXFBD04 (9j)
Fig. 8. Inhibition of BET bromodomains by means of 3,5-dimethylisoxazole-based ligands OXFDD02 (1a) or
OXFBD04 (9j) induces MYC suppression in MCF7 breast cancer cells. Representative western blot
detection of c-myc after treatment with OXFBD02 (1a), OXFBD04 (9j), or (+)-JQ1 as a reference.
Compounds were tested at a concentration of 10 µM.
3
. Conclusion
In conclusion, we have shown that the BET bromodomain ligand OXFBD02 (1a) displays a
promising profile in the NCI-60 panel of cancer cell lines, but a short metabolic half-life of 40
minutes in human liver microsomes. To optimise the metabolic stability of this compound series,
we investigated the SAR of WPF-binding groups, with a view to optimising the overall compound
properties. In the pyridyl series we showed that an intramolecular hydrogen bond detrimentally
affects the affinity of 2-pyridyl derivative (9p), by holding the (R)-enantiomer in a conformation that
disfavours BRD4(1) binding. However, the 3-pyridyl derivative, OXFDB04 (9j), displays higher
BRD4(1) affinity than the parent compound and an LLE value of 5.74, indicating that it is a
BRD4(1) ligand with an optimised balance of affinity and physicochemical properties. The addition
of the pyridine ring led to an increased metabolic half-life of 6.5 hours in human liver microsomes.
These data indicate that the 3,5-dimethylisoxazole group is not inherently metabolically labile, and
optimisation of the overall compound properties can lead to compounds with useful metabolic
stabilities. In cancer cell lines, OXFDB04 (9j) showed similar effects on cytotoxicity and MYC
suppression compared to the parent compound OXFBD02 (1a). Consequently, we report
21

OXFDB04 (9j) as an improved tool compound to study BRD4 in in vitro and potentially in vivo
settings.
22

4
. Experimental
4
.1. General chemistry experimental details
1
H NMR spectra were recorded on Bruker AVIII HD 400 (400 MHz) or Bruker AVII 500 (500 MHz).
part per million (ppm) relative to the solvent reference peak as internal
deuterium lock. The multiplicity of each signal is indicated by: s (singlet), d (doublet), t (triplet), dd
doublet of doublets), q (quartet), sp (septet) or m (multiplet). Identical proton coupling constants (J) are
averaged in each spectrum and are reported to the nearest 0.1 Hz. Coupling constants were determined
Chemical shifts are reported as δ
H
50
(
13
using Bruker TopSpin software. C NMR spectra were recorded on Bruker AVIII HD 400 (101 MHz) or
Bruker AVII 500 (126 MHz). Chemical shifts are reported as δ part per million (ppm) relative to the
C
solvent reference peak as internal deuterium lock. Coupling constants (J) are quoted in Hz and are
recorded to the nearest 1 Hz. Identical coupling constants (J) are averaged in each spectrum and
reported to the nearest 1 Hz. The coupling constants are determined by analysis using Bruker TopSpin
19
software. F NMR spectra were recorded on a Bruker AVII 500 (470 MHz) using a broadband proton
decoupling pulse sequence and deuterium internal lock. The chemical shift data for each signal are given
11
as δ
F
in units of parts per million (ppm). B NMR spectra were recorded on a Bruker DRX500 (160 MHz).
The chemical shift data for each signal are given as δ
B
in units of parts per million (ppm). Coupling
constants (J) are quoted in Hz and are recorded to the nearest 1 Hz. Identical coupling constants (J) are
averaged in each spectrum and reported to the nearest 1 Hz. The coupling constants are determined by
analysis using Bruker TopSpin software. Note: in all isoxazole-containing compounds, positions on the
central aromatic ring are numbered first, the isoxazole ring with primes (´) and any additional rings with
double primes (´´) etc. Low-resolution mass spectra (LRMS) using electron spray ionisation were
recorded on a Micromass LCT Premier spectrometer. Electrospray Ionisation (ES) High-Resolution Mass
Spectrometry (HRMS) spectra were acquired on either a Bruker MicroTOF spectrometer or a Thermo
Exactive mass spectrometer, equipped with Agilent 1100 liquid chromatography systems for flow
injection analysis, from solutions of MeOH, H O or MeCN as stated. MicroTOF data were processed
2
using Bruker Hystar software, while Exactive data were analysed using Thermo Xcalibur software.
Melting points on crystallised samples were determined using either a) a Leica Galen III hot stage
microscope or b) a Griffin capillary tube melting point apparatus and are uncorrected. The solvents of
23

crystallisation are shown in parentheses. Infrared spectra were obtained from thin films using a diamond
attenuated total reflectance module. The spectra were recorded on a Bruker Tensor 27 spectrometer.
-1
Absorption maxima (νmax) are reported in wavenumbers (cm ) and are classified as broad (br), strong (s),
medium (m) or weak (w). Analytical HPLC was carried out on a PerkinElmer Flexar system with a Binary
LC Pump and UV/VIS LC Detector. For determination of compound purity following methods were
®
applied. Method 1 (M1): a Dionex Acclaim 120 column (C18, 5 μm, 120 Å, 4.6 × 150 mm) was used and
the solvents employed were A = 0.1% (v/v) solution of formic acid in 95% H
solution of formic acid in 95% MeCN/5% H O, and the gradient (A:B). A 10-minute linear gradient of 0-
00% B was run with a flow rate of 1 mL/min and detection at 254 nm. Samples were injected in DMSO,
2
O/5% MeCN; B = 0.1% (v/v)
2
1
®
MeOH, DMSO/MeOH or DMSO/CHCl
3
. Method 2 (M2): a Dionex Acclaim 120 column (C18, 5 μm, 120
2
Å, 4.6 × 150 mm) was used and the solvents employed were A = H O; B = MeCN. Linear gradient
conditions (0−10 min, linear increase from 5% to 95% of B; 10−15 min, B = 95%) with a flow rate of 1.5
mL/min and detection at 254 nm. Samples were injected in DMSO, MeOH, DMSO/MeOH or
®
DMSO/CHCl
used and the solvents employed were A = H
increase from 5% to 95% of B; 10−15 min, B = 95%) with a flow rate of 1.5 mL/min and detection at 220
nm. Samples were injected in DMSO, MeOH, DMSO/MeOH or DMSO/CHCl . All compounds that were
3
. Method 3 (M3): a Dionex Acclaim 120 column (C18, 5 μm, 120 Å, 4.6 × 150 mm) was
2
O; B = MeCN. Linear gradient conditions (0−10 min, linear
3
subjected to biological evaluation had purity of ≥95% determined by HPLC and LCMS analysis.
Anhydrous solvents were obtained under the following conditions: anhydrous DMF, anhydrous MeOH
and anhydrous EtOH were purchased from Sigma-Aldrich UK in SureSeal™ bottles and used without
further purification; anhydrous THF and CH
argon or nitrogen atmosphere; where stated, THF was distilled from sodium and benzophenone and
pyridine was distilled from CaH . Chemicals were purchased from Acros, Sigma-Aldrich, Alfa Aesar,
Fisher, Apollo Scientific or Fluorochem. Where appropriate and if not stated otherwise, all non-aqueous
2
Cl were dried over activated 3 Å molecular sieves under an
2
2
reactions were performed in a flame-dried flask under an inert atmosphere of nitrogen or argon. MnO
2
was activated at 250 °C overnight prior to use. Organolithium reagents were titrated against
51
diphenylacetic acid, and commercially-available organomagnesium compounds were titrated against
52
salicaldehyde phenylhydrazone. Where stated, NEt was dried with KOH and distilled onto KOH pellets.
3
24

Isolute® SCX-2 cartridges for cation exchange were purchased from Biotage UK and were used
according to manufacturers’ protocols.
4
4
.2. Synthetic procedures
21
.2.1. 3-(3,5-Dimethyl-1,2-oxazol-4-yl)-5-[hydroxyl(phenyl)methyl]phenol (1a)
21
Following the procedure of Hewings et al., to a solution of 17 (2.63 g, 12.1 mmol, 1.0 eq) in
anhydrous THF at 0 °C was added dropwise a 1 M solution of phenylmagnesium bromide (34.0 mL,
3
4.0 mmol, 2.8 eq) in THF. The reaction solution was warmed to rt and stirred for 4 h. After this
time, the volatile components were removed in vacuo. The resulting solid was crystallised from hot
MeCN and the mother liquor was purified by silica gel chromatography, eluting with 1% AcOH and
EtOAc in petroleum ether (gradient elution 15 → 40%). The combined solids were crystallised from
hot MeCN yielding 1a (1.43 g, 40%) as an off-white crystalline solid: R
f
(50% EtOAc/petroleum
ether) 0.38; mp 187-190 °C (MeCN) [lit. mp 187-188 °C ]; H NMR (400 MHz, D -acetone) δ 8.44
1H, br s, C(1)OH), 7.49-7.43 (2H, m, C(2’’)H & C(6’’)H), 7.34-7.28 (2H, m, C(3’’)H and C(5’’)H),
21
1
6
(
7
.24-7.19 (1H, m, C(4’’)H), 6.94-6.91 (1H, m, C(6)H), 6.91-6.88 (1H, m, C(4)H), 6.67 (1H, dd, J =
.9, 1.9 Hz, C(2)H), 5.83-5.79 (1H, m, CHOH), 4.90 (1H, d, J = 3.9 Hz, CHOH), 2.36 (3H, s,
1
+
+
C(5’)CH
3
), 2.18 (3H, s, C(3’)CH ); LRMS m/z (ES ) 296 ([M + H] , 100%). Data are in good
3
21
agreement with literature values.
4
.2.2. 3-(3,5-Dimethyl-1,2-oxazol-4-yl)-5-[(4-chlorophenyl)(hydroxy)methyl]phenol (9a)
To a solution of 17 (96 mg, 442 µmol, 1.0 eq) in anhydrous THF (2 mL) was added a 1 M solution
of 4-chlorophenylmagnesium bromide (1.00 mL, 216 mg, 1.00 mmol, 2.3 eq) in Et O at rt, and the
reaction solution was stirred for 19 h. The reaction was quenched with a saturated aqueous
solution of NH Cl (15 mL), and the aqueous phase was extracted with EtOAc (3 × 15 mL). The
combined organic layers were washed with H O (45 mL), passed through an anhydrous frit and
2
4
2
concentrated in vacuo. Purification by silica gel chromatography, eluting with EtOAc in
cyclohexane (gradient elution 5 → 100%) followed by mass-directed autopurification (0.1% formic
acid, gradient elution MeCN/H
2
O 30 → 85%) afforded the desired product 9a (74 mg, 51%) as a
clear and colourless oil that solidified under vacuum to give a colourless amorphous solid: R
f
(50%
25

–1
1
EtOAc/cyclohexane) 0.35; νmax (thin film)/cm : 3457 (br) (O–H), 3015 (br), 2970 (w), 2949 (br); H
NMR (400 MHz, CD OD) δ 7.42-7.36 (2H, m, C(2’’)H & C(6’’)H), 7.34-7.28 (2H, m, C(3’’)H &
C(5’’)H), 6.88-6.83 (1H, m, C(6)H), 6.82-6.77 (1H, m, C(4)H), 6.64 (1H, dd, J = 2.2, 1.6 Hz, C(2)H),
3
1
3
5
1
1
7
.75 (1H, s, CHOH), 2.36 (3H, s, C(5’)CH
3
), 2.21 (3H, s, C(3’)CH
3
); C NMR (101 MHz, CD OD) δ
3
66.9 (C(5’)), 160.0 (C(3’)), 159.0 (C(1)), 148.0 (C(5)), 144.7 (C(4’’)), 134.0 (C(1’’)), 132.6 (C(3)),
29.4 (C(3’’) & C(5’’)), 129.3 (C(2’’) & C(6’’)), 119.7 (C(4)), 117.9 (C(4’)), 115.9 (C(2)), 114.0 (C(6)),
+
3
5.9 (CHOH), 11.5 (C(5’)CH ), 10.8 (C(3’)CH
3
); HRMS m/z (ES ) Found: 352.0715 & 354.0698.
+
3
35
+
37
+
−
18
C H16ClNNaO
requires [M( Cl)] & [M( Cl)] 352.0711 & 354.0683; LRMS m/z (ES ) 328
35
–
37
–
(
[M( Cl) – H] , 100%), 330 ([M( Cl) – H , 39%]; HPLC RT = 11.55 min, purity 95.5% (M1).
4
.2.3. 3-(3,5-Dimethyl-1,2-oxazol-4-yl)-5-[hydroxy(2-methyoxyphenyl)methyl]phenol (9b)
To a solution of 17 (94 mg, 433 µmol, 1.0 eq) in anhydrous THF (2 mL) was added a 1 M solution
-methoxyphenylmagnesium bromide (1.00 mL, 211 mg, 1.00 mmol, 2.3 eq) in THF at rt, and the
reaction solution was stirred for 3 h. The reaction was quenched with a saturated aqueous solution
of NH Cl (15 mL), and the aqueous phase was extracted with EtOAc (3 × 15 mL). The combined
organic layers were washed with H O (45 mL), passed through an anhydrous frit and concentrated
in vacuo. Purification by silica gel chromatography, eluting with EtOAc in cyclohexane (gradient
elution 10 → 60%) afforded 9b (113 mg, 80%) as a colourless solid: R (50% EtOAc/cyclohexane)
2
4
2
f
–1
0
.36; νmax (thin film)/cm : 3196 (br) (O–H), 2993 (w), 2837 (w), 1642 (m) 1596 (s); mp 167-169 °C
1
(
EtOAc); H NMR (400 MHz, CD
3
OD) δ 7.49 (1H, dd, J = 7.5, 1.5 Hz, C(6’’)H), 7.25 (1H, ddd, J =
7
6
2
.8, 7.8, 1.5 Hz, C(3’’) H), 7.01-6.92 (2H, m, C(4’’)H & C(5’’)H), 6.88-6.84 (1H, m, C(6)H), 6.80-
.76 (1H, m, C(4)H), 6.60 (1H, dd, J = 1.9, 1.9 Hz, C(2)H), 6.12 (1H, s, CHOH), 3.81 (3H, s, OCH
3
),
13
.37 (3H, s, C(5’)CH
3
), 2.21 (3H, s, C(3’)CH
3
); C NMR (101 MHz, CD OD) δ 166.7 (C(5’)), 160.0
3
(C(3’)), 158.8 (C(1)), 157.7 (C(2’’)), 148.1 (C(5)), 133.9 (C(1’’)), 132.0 (C(3)), 129.5 (C(3’’)), 127.9
(C(6’’)), 121.6 (C(4’’) or C(5’’)), 119.9 (C(4)), 118.0 (C(4’)), 115.4 (C(2)), 114.2 (C(6)), 111.7 (C(5’’)
+
or C(4’’)), 70.5 (CHOH), 55.9 (OCH
3
), 11.4 (C(5’)CH
3
), 10.7 (C(3’)CH
3
); HRMS m/z (ES ) Found:
+
4
+
+
3
26.1386. C19H20NO
requires M 326.1387; LCMS (formic acid) RT = 0.88, [M + H] = 326; HPLC
RT = 10.95 min, purity 98.0% (M1).
26

4
.2.4. 3-(3,5-Dimethyl-1,2-oxazol-4-yl)-5-[hydroxy(3-methyoxyphenyl)methyl]phenol (9c)
To a solution of 17 (103 mg, 474 µmol, 1.0 eq) in anhydrous THF (2 mL) was added a 1 M solution
of 3-methoxyphenylmagnesium bromide (1.00 mL, 211 mg, 1.00 mmol, 2.1 eq) in THF at rt, and
the reaction solution was stirred for 5 h. The reaction was quenched with a saturated aqueous
solution of NH
4
Cl (15 mL), and the aqueous phase was extracted with EtOAc (3 × 15 mL). The
combined organic layers were washed with H
2
O (45 mL), passed through an anhydrous frit and
concentrated in vacuo. Purification by silica gel chromatography, eluting with EtOAc in
cyclohexane (gradient elution 10 → 60%) afforded 9c (106 mg, 68%) as a colourless solid: R (50%
f
–1
EtOAc/cyclohexane) 0.34; νmax (thin film)/cm : 3176 (br) (O–H), 2933 (w), 1640 (w), 1596 (w); mp
1
9
4-96 °C (EtOAc); H NMR (400 MHz, CD
3
OD) δ 7.24 (1H, dd, J = 7.8, 7.8 Hz, C(2’’)H), 7.05-6.93
(2H, m, C(4’’)H & C(5’’)H or C(4’’)H & C(6’’)H or C(5’’)H & C(6’’)H), 6.89-6.84 (1H, m, C(6)H), 6.84-
6
3
1
1
.76 (2H, m, C(4)H & C(4’’)H, C(5’’)H or C(6’’)H), 6.67-6.59 (1H, m, C(2)H), 5.72 (1H, s, CHOH),
13
.78 (3H, s, OCH
3
), 2.73 (3H, s, C(5’)CH
3
), 2.22 (3H, s, C(3’)CH
3
); C NMR (101 MHz, CD OD) δ
3
66.8 (C(5’)), 161.2 (C(3’’)), 160.0 (C(3’)), 158.9 (C(1)), 148.3 (C(1’’)), 147.4 (C(5)), 132.4 (C(3)),
30.3 (C(2’’)), 120.0 (C(4’’), C(5’’) or C(6’’)), 119.8 (C(4’’), C(5’’) or C(6’’)), 118.0 (C(4’)), 115.7
(
C(2)), 114.1 (C(6)), 113.7 (C(4)), 113.2 (C(4’’), C(5’’) or C(6’’)), 76.6 (CHOH), 55.7 (OCH
3
), 11.4
+
+
4
+
(
C(5’)CH
3
), 10.7 (C(3’)CH
3
); HRMS m/z (ES ) Found: 326.1386. C19
H20NO
requires M 326.1387;
+
LCMS (formic acid) RT = 0.87, [M + H] = 326; HPLC RT = 10.84 min, purity 98.5% (M1).
4
.2.5. 3-(3,5-Dimethyl-1,2-oxazol-4-yl)-5-[hydroxy(4-methyoxyphenyl)methyl]phenol (9d)
To a solution of 17 (102 mg, 470 µmol, 1.0 eq) in anhydrous THF (2 mL) was added a 0.5 M
solution of 4-methoxyphenylmagnesium bromide (2.00 mL, 211 mg, 1.00 mmol, 2.1 eq) in THF at
rt, and the reaction solution was stirred for 3 h. The reaction was quenched with a saturated
aqueous solution of NH
4
Cl (15 mL), and the aqueous phase was extracted with EtOAc (3 × 15 mL).
The combined organic layers were washed with H
2
O (45 mL), passed through an anhydrous frit
and concentrated in vacuo. Purification by silica gel chromatography, eluting with EtOAc in
cyclohexane (gradient elution 10 → 60%) afforded 9d (104 mg, 68%) as a clear and colourless oil.
Precipitation from Et
2
O and hexane afforded a colourless amorphous solid: R
f
(50%
–1
1
EtOAc/cyclohexane) 0.44; νmax (thin film)/cm : 3016 (w) (O–H), 2970 (w), 1739 (s); H NMR (400
27

MHz, CD
3
OD) δ 7.34-7.26 (2H, m, C(3’’)H & C(5’’)H), 6.91-6.84 (3H, m, C(2’’)H, C(6’’)H & C(6)H),
), 2.35
OD) δ 166.8 (C(5’)), 160.4
6
.81-6.78 (1H, m, C(4)H), 6.65-6.60 (1H, m C(2)H), 5.72 (1H, s, CHOH), 3.74 (3H, s, OCH
3
13
(
(
(
(
3H, s, C(5’)CH
3
), 2.20 (3H, s, C(3’)CH
3
); C NMR (126 MHz, CD
3
C(4’’)), 160.0 (C(3’)), 158.9 (C(1)), 148.6 (C(5) or C(1’’)), 137.9 (C(1’’) or C(5)), 132.3 (C(3)), 129.0
C(3’’) & C(5’’)), 119.6 (C(4)), 118.0 (C(4’)), 115.5 (C(2)), 114.7 (C(2’’) & C(6’’)), 113.9 (C(6)), 76.2
−
CHOH), 55.7 (OCH
3
), 11.4 (C(5’)CH
3
), 10.7 (C(3’)CH
3
); HRMS m/z (ES ) Found: 324.1243.
−
4
−
−
–
19
C H18NO
requires M , 324.1241; LRMS m/z (ES ) 324 ([M – H] , 100%); HPLC RT = 10.74 min,
purity 97.1% (M1).
4
.2.6. 3-(3,5-Dimethyl-1,2-oxazol-4-yl)-5-[hydroxy(4-methylphenyl)methyl]phenol (9e)
To a solution of 17 (93 mg, 428 µmol, 1.0 eq) in anhydrous THF (2 mL) was added a 1 M solution
of 4-tolylmagnesium bromide (1.00 mL, 195 mg, 1.00 mmol, 2.3 eq) in THF at rt, and the reaction
solution was stirred for 1 h. To aid solubility, additional THF (2 mL) was added and the reaction
solution was stirred at rt for 17 h. The reaction was quenched with a saturated aqueous solution of
NH
organic layers were washed with H
in vacuo. Purification by silica gel chromatography, eluting with acetone in petroleum ether
gradient elution 10 → 50%) afforded 9e (75 mg, 56%) as a clear and colourless oil. Precipitation
from Et O and hexane gave a colourless, amorphous solid: R (20% EtOAc/cyclohexane) 0.08; νmax
thin film)/cm : 3280 (br) (O–H), 1596 (s), 1421 (s); H NMR (400 MHz, CD OD) δ 7.27 (2H, d, J =
.9 Hz, C(3’’)H & C(5’’)H), 7.12 (2H, d, J = 7.9 Hz, C(2’’)H & C(6’’)H), 6.90-6.85 (1H, m, C(6)H),
.82-6.77 (1H, m, C(4)H), 6.65-6.60 (1H, m, C(2)H), 5.72 (1H, s, CHOH), 2.33 (3H, s, C(5’)CH ),
OD) δ 166.4 (C(5’)), 159.9
4
Cl (15 mL), and the aqueous phase was extracted with EtOAc (3 × 15 mL). The combined
2
O (45 mL), passed through an anhydrous frit and concentrated
(
2
f
–1
1
(
3
7
6
2
3
13
.28 (3H, s, C(4’’)CH
3
), 2.19 (3H, s, C(3’)CH
3
); C NMR (126 MHz, CD
3
(C(3’)), 158.5 (C(1)), 148.1 (C(5)), 142.4 (C(1’’)), 137.7 (C(4’’)), 131.9 (C(3)), 129.5 (C(2’’) & C(6’’)),
1
27.3 (C(3’’) & C(5’’)), 119.3 (C(4)), 117.6 (C(4’)), 115.2 (C(2)), 113.6 (C(6)), 76.1 (CHOH), 20.7
+
+
3
+
(
C(4’’)), 11.1 (C(5’)), 10.4 (C(3’)); HRMS m/z (ES ) Found: 310.1433. C19
H20NO
requires M
+
3
10.1438; LRMS (formic acid) RT = 0.92, [M + H] = 310; HPLC RT = 11.29 min, purity 98.5% (M1).
28

4
.2.7. 3-(3,5-Dimethyl-1,2-oxazol-4-yl)-5-[(3,4-dichlorophenyl)(hydroxy)methyl]phenol (9f)
To a solution of 17 (110 mg, 506 µmol, 1.0 eq) in anhydrous THF (2 mL) was added a 0.5 M
solution of 3,4-dichlorophenylmagnesium bromide (2.50 mL, 313 mg, 1.25 mmol, 2.5 eq) in THF at
rt, and the reaction solution was stirred for 7 h, then was heated at 50 °C for 15 h. Additional 3,4-
dichlorophenylmagnesium bromide (1.00 mL, 125 mg, 500 µmol, 1.0 eq) was added and the
reaction solution was stirred at 50 °C for 7 h. The reaction was quenched with a saturated aqueous
solution of NH
4
Cl (15 mL), and the aqueous phase was extracted with EtOAc (3 × 15 mL). The
combined organic layers were washed with H
2
O (45 mL), passed through an anhydrous frit and
concentrated in vacuo. The residue was resuspended in MeOH, filtered and the filtrate was
concentrated in vacuo. Purification by silica gel chromatography, eluting with EtOAc in
cylcohexane (gradient elution 5 → 55%) followed by mass-directed autopurification (0.1% formic
acid, gradient elution MeCN/H
EtOAc/cyclohexane) 0.49; νmax (thin film)/cm : 3198 (br) (O–H), 1637 (m), 1596 (s); H NMR
400 MHz, CD OD) δ 7.58 (1H, d, J = 1.8 Hz, C(2’’)H), 7.47 (1H, d, J = 8.3 Hz, C(5’’)H), 7.31 (1H,
dd, J = 8.3, 1.8 Hz, C(6’’)H), 6.86-6.82 (1H, m, C(6)H), 6.82-6.78 (1H, m, C(4)H), 6.69-6.63 (1H, m,
2
O 30 → 85%) afforded 9f (20 mg, 11%) as a colourless solid: R (50%
f
–1
1
(
3
13
C(2)H), 5.74 (1H, s, CHOH), 2.38 (3H, s, C(5’)CH
3
), 2.23 (3H, s, C(3’)CH ); C NMR (101 MHz,
3
CD OD) δ 166.9 (C(5’)), 160.0 (C(3’)), 159.2 (C(1)), 147.5 (C(3’’)), 146.9 (C(4’’)), 133.2 (C(3)),
3
1
1
3
32.7 (C(5)), 131.9 (C(1’’)), 131.4 (C(5’’)), 129.6 (C(2’’)), 127.5 (C(6’’)), 119.7 (C(4)), 117.9 (C(4’)),
+
16.2 (C(2)), 113.9 (C(6)), 75.3 (CHOH), 11.5 (C(5’)CH
3
), 10.7 (C(3’)CH
3
); HRMS m/z (ES ) Found:
+
3
+
+
2
86.0322, C18H15Cl NNaO
requires M 386.0321; LCMS (formic acid) RT = 1.05, [M + H] 364;
HPLC RT = 13.52 min, purity 99.2% (M1).
4
.2.8. 3-(3,5-Dimethyl-1,2-oxazol-4-yl)-5-[(4-fluorophenyl)(hydroxyl)methyl]phenol (9g)
Magnesium turnings (63 mg, 2.49 mmol, 6.0 eq) and a crystal of iodine were added to anhydrous
THF (5 mL) and the mixture was stirred. 1-Bromo-4-fluorobenzene (314 µL, 500 mg,
2
.85 mmol, 6.6 eq) in anhydrous THF (5 mL) were added to the dropping funnel. The solution was
added dropwise at rt, and the funnel was rinsed with further anhydrous THF (0.5 mL). The reaction
mixture was heated under reflux with stirring for 1 h. After this time, the Mg turnings had dissolved
and the solution became cloudy. The solution was cooled to 0 °C and an ice-cooled solution of 17
29