The structure-activity relationships of L3MBTL3 inhibitors: Flexibility of the dimer interface

Article abstract of DOI:10.1039/c3md00197k

We recently reported the discovery of UNC1215, a potent and selective chemical probe for the L3MBTL3 methyllysine reader domain. In this article, we describe the development of structure-activity relationships (SAR) of a second series of potent L3MBTL3 antagonists which evolved from the structure of the chemical probe UNC1215. These compounds are selective for L3MBTL3 against a panel of methyllysine reader proteins, particularly the related MBT family proteins, L3MBTL1 and MBTD1. A co-crystal structure of L3MBTL3 and one of the most potent compounds suggests that the L3MBTL3 dimer rotates about the dimer interface to accommodate ligand binding.

Full text of DOI:10.1039/c3md00197k

MedChemComm
View Article Online
View Journal | View Issue
CONCISE ARTICLE
The structure–activity relationships of L3MBTL3
inhibitors: flexibility of the dimer interface†
Cite this: Med. Chem. Commun., 2013,
4, 1501
Michelle A. Camerino,a Nan Zhong,b Aiping Dong,b Bradley M. Dickson,a
Lindsey I. James,a Brandi M. Baughman,a Jacqueline L. Norris,a Dmitri B. Kireev,a
William P. Janzen,a Cheryl H. Arrowsmithabcd and Stephen V. Frye
*
a
We recently reported the discovery of UNC1215, a potent and selective chemical probe for the L3MBTL3
methyllysine reader domain. In this article, we describe the development of structure–activity
relationships (SAR) of a second series of potent L3MBTL3 antagonists which evolved from the structure
of the chemical probe UNC1215. These compounds are selective for L3MBTL3 against a panel of
methyllysine reader proteins, particularly the related MBT family proteins, L3MBTL1 and MBTD1. A co-
crystal structure of L3MBTL3 and one of the most potent compounds suggests that the L3MBTL3 dimer
rotates about the dimer interface to accommodate ligand binding.
Received 11th July 2013
Accepted 16th September 2013
DOI: 10.1039/c3md00197k
www.rsc.org/medchemcomm
studies, we have focussed on the recognition of lysine meth-
ylation by MBT domain methyllysine readers, a subclass in the
Introduction
Histone post translational modications such as lysine meth-
ylation play an important role in the function of chromatin by
creating binding sites for reader proteins. This binding event
leads to downstream signalling via the recruitment and stabi-
lization of chromatic template machinery and is an essential
step in the regulation of chromatin and gene expression.1,2
Histone lysine methylation can signal either the activation3,4 or
repression5–7 of gene transcription, depending on the site and
degree of methylation.8 Lysine methylation has been identied
in various positions of the histone tails, mainly on histone 3 at
positions H3K4 (lysine 4), H3K9, H3K27, H3K36, H3K79 and on
histone four at position H4K20.9 The 3-amino group of a lysine
at any of these positions can be mono (me1), di (me2) or tri-
methylated (me3), thus changing the chemical properties of
this residue from a small, relatively “hard” positive charge to a
more diffuse, “so” and polarizable positive charge in the case
of Kme3.10 In most cases, the binding site of this methylated
residue is a hydrophobic ‘aromatic cage’, a conserved struc-
tural motif in almost all reader proteins.11 In our previous
‘Royal family’ of proteins that includes the proteins L3MBTL1,
L3MBTL3, and MBTD1 among others.12,13 While these proteins
have been associated with haematopoiesis14 and cancer
biology,15,16 their exact biological mechanisms still require
elucidation. For this reason, we have endeavoured to identify
small molecule chemical probes that would facilitate further
study and understanding of these proteins.17 To date, we have
reported the discovery of weak small molecule inhibitors of
L3MBTL1 such as UNC669,18,19 and have most recently identi-
ed a potent and selective chemical probe, UNC1215, for the
L3MBTL3 methyllysine reader domain (Fig. 1).20 Interestingly,
the high potency and selectivity of UNC1215 can be explained
by its co-crystal structure with L3MBTL3 which shows that
UNC1215 binds as a 2 : 2 dimer with the protein.20 This was the
rst published evidence of L3MBTL3 dimerization and has
prompted further studies in our group. Here we report a second
series of L3MBTL3 inhibitors, their structure–activity rela-
tionships and their potential unique interactions with the
L3MBTL3 dimer.
aCenter for Integrative Chemical Biology and Drug Discovery, Division of Chemical
Biology and Medicinal Chemistry, University of North Carolina Eshelman School of
Pharmacy, University of North Carolina at Chapel Hill, Chapel Hill, North
Carolina, USA. E-mail: svfrye@email.unc.edu; Fax: +1 919 843 8465; Tel: +1 919
843-5486
bStructural Genomics Consortium, University of Toronto, Ontario, Canada
cDepartment of Medical Biophysics, University of Toronto Ontario, Canada
dPrincess Margaret Cancer Centre, 101 College Street, Toronto, Ontario, Canada, M5G
1L7
† Electronic supplementary information (ESI) available: Experimental procedures
including spectra for all nal compounds, crystallographic methods, in vitro assay
conditions and representative ITC binding curves. See DOI: 10.1039/c3md00197k
Fig. 1 Previously reported inhibitors of L3MBTL1 (UNC669) and L3MBTL3,
(UNC1215).
This journal is ª The Royal Society of Chemistry 2013
Med. Chem. Commun., 2013, 4, 1501–1507 | 1501

View Article Online
MedChemComm
Concise Article
The X-ray co-crystal structure of UNC2533 and L3MBTL3
Results and discussion
To investigate the potential binding mode of UNC2533 to
L3MBTL3, the X-ray co-crystal structure of the complex was
In our efforts to develop additional, novel inhibitors of the
methyllysine reader, L3MBTL1, we simulated putative apo-
homodimer conformations of L3MBTL1 and L3MBTL3 using a
recently developed scheme for free energy computations.21 In
each case, we observed stable homodimers with more
compact ligand pockets than those observed in the UNC1215–
L3MBTL3 co-crystal structure.20 Based on this observation,
new small molecules were proposed as possible ligands that
could t within the more compact homodimer pockets while
preserving the dibasic character of UNC1215, which we knew
was required for potent dimer binding. Following Scheme 1,
we synthesised the dibasic compound UNC2533 (1) and
identied it as a potent inhibitor of the L3MBTL3 methyl-
lysine reader domain. In vitro evaluation of UNC2533 in an
AlphaScreenꢀ methylated histone peptide competition
assay22 yielded an IC50 of 62 ꢀ 7.2 nM for L3MBTL3 which is
within three-fold of the chemical probe UNC1215 (IC50 ¼ 24 ꢀ
7.6 nM, Table 1).20‡ The activity of UNC2533 was conrmed by
isothermal titration calorimetry (ITC), to give a Kd of 0.37 mM
which is also within 3-fold of the reported dissociation
constant of UNC1215 (Kd ¼ 0.12 mM).20 Biological screening of
UNC2533 against the other readers in our panel (Table 1)
showed a similar selectivity prole to UNC1215. UNC2533
binds L3MBTL1 weakly (IC50 ¼ 14 ꢀ 0.0 mM) with a similar
affinity to the L3MBTL1 inhibitor UNC66918,19 (IC50 ¼ 21 ꢀ
1.2 mM). The weak binding affinity for L3MBTL1 was also
conrmed by ITC (Kd ¼ 19.0 mM), showing that UNC2533 is 50
times more selective for L3MBTL3 over L3MBTL1. As the
L3MBTL1 monomer is known to bind UNC669 via key inter-
actions between the pyrrolidine and the aromatic reader
pocket in domain 2,18 it is reasonable to suggest that one or
both of the pyrrolidines in UNC2533 is responsible for its
weak L3MBTL1 activity. UNC2533 also displays modest
activity against the other royal family protein MBTD1, as well
as 53BP1, a member of the Tudor domain family. Other
readers from the chromodomain family (CBX7), the Tudor
domain family (UHRF1), and the PHD nger family of zinc-
binding proteins (PHF23 and Jarid1A) did not bind to
UNC2533 (IC50 > 30 mM).
˚
determined and rened to 2.3 A (PDB ID: 4L59). Despite being a
smaller molecule than UNC1215, UNC2533 also binds
L3MBTL3 as a 2 : 2 complex. Each molecule of UNC2533 bridges
the L3MBTL3 dimer interface by interacting with domain 1 of
one monomer and the presumed histone methyllysine binding
site in domain 2 of the other monomer (Fig. 2a). The binding
interaction to domain 2 is primarily mediated by a strong
˚
hydrogen bond (2.6 A) between the carboxylate oxygen of D381
and the pyrrolidine nitrogen of the (pyrrolidinyl)piperidine
moiety (Fig. 2b). Additional hydrophobic contacts and a cation–
p interaction between the positively charged pyrrolidine
nitrogen and the domain 2 aromatic cage strengthen binding.
The interaction of UNC2533 with domain 1 is mediated by a
˚
second hydrogen bond (2.8 A) between the carboxylate oxygen of
D274 and the C2-linked pyrrolidine nitrogen. Further hydro-
phobic contacts between the pyrrolidine ring and Y301 rein-
force binding in this pocket. Atomic distances for the key
binding interactions in the UNC1215–L3MBTL3 complex20 and
the UNC2533–L3MBTL3 complex were measured using MOE
soware.23 All key contacts between the ligand and the protein
were very similar in distance and strength (ESI Table 5†). This
result was surprising as UNC2533 is signicantly smaller in size
than UNC1215. The intramolecular distance between the two
˚
basic nitrogens in UNC2533 is 10.4 A. The equivalent intra-
˚
molecular distance for the UNC1215 molecule is 13.1 A. This
difference indicates that a change in the conformation of the
L3MBTL3 dimer must have occurred to maintain its bonding
interactions with the smaller inhibitor, UNC2533.
To compare the UNC1215–L3MBTL3 complex with the
UNC2533–L3MBTL3 complex we aligned selected backbone
chains with Kabsch's method24 as implemented in the molec-
ular dynamics program VMD.25 Visualization of the entire 2 : 2
complex shows that the dimer interface has rotated slightly
such that only one of the two monomers in the UNC2533–
L3MBTL3 complex is aligned with the UNC1215–L3MBTL3
complex (Fig. 3). For the “aligned” monomers, there is very little
structural variation in the domain 1 and domain 2 binding
pockets. Upon closer inspection of the domain 2 (methyllysine)
binding pocket we found that all of the key residues are closely
aligned with the exception of some exibility in the side chains
(Fig. 4a). Similarly, the conformation of the domain 1 binding
pocket is also well conserved, with the exception of the E410
side chain which has ipped its orientation due to a steric clash
with the phenyl ring of UNC1215 (Fig. 4b). In contrast, all of the
key residues in the “misaligned” monomers have shied.
Measurement of the distances between the alpha carbons of
residue pairs in the binding site of each crystal structure shows
Scheme 1 Synthetic route to UNC2533 (1). Reagents and conditions: (a) pyr-
rolidine, K2CO3, CH3CN, room temperature, 16 h, 58% yield (b) 4-(pyrrolidin-1-yl)
piperidine, XPhos, Pd2(dba)3, Cs2CO3, dioxane/water, microwave irradiation, 110
ꢁC for 20 min, then 125 ꢁC for 3 h, 12% yield.
˚
a shi of between 2.4 and 4.6 A (Fig. 4c and d, ESI Table 6†). The
same results were observed when the alternate monomer was
aligned using Kabsch's method.24 These ndings reveal that the
L3MBTL3 dimer primarily accommodates binding of the
smaller UNC2533 molecule by rotating the two L3MBTL3
monomers at the dimer interface rather than altering the
‡ It should be noted that the Alphascreenꢀ activities of UNC1215, UNC669 and
UNC1079 are slightly different in various publications due to changes in the
assay format – readers should refer to the supplemental information for the
exact conditions that were used for the assays in this publication.
1502 | Med. Chem. Commun., 2013, 4, 1501–1507
This journal is ª The Royal Society of Chemistry 2013

View Article Online
Concise Article
MedChemComm
Table 1 The activity of UNC2533 (1) compared to control compounds determined from AlphaScreenꢁ and ITC. Compounds screened against L3MBTL1 and MBTD1
were screened at concentrations of up to 300 mM in the AlphaScreenꢁ. All other proteins were screened with a maximal compound concentration of 30 mM
Activity as determined by AlphaScreenꢀ IC50 (mM) and ITC Kd (mM), shows in parentheses [ ]
#
Structure
L3MBTL3
L3MBTL1
MBTD1
53BP1
CBX7
>30
UHRF1
>30
Jarid1A
>30
PHF23
>30
0.062 ꢀ 0.0072
[0.37 ꢀ 0.052]
14 ꢀ 0.0
UNC2533 (1)
53 ꢀ 4
14.5 ꢀ 2.1
[19 ꢀ 1.4]
UNC669
2.8 ꢀ 0.57
21 ꢀ 1.2
170 ꢀ 31
>300
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
UNC1079
8.0 ꢀ 3.0
50 ꢀ 8.8
0.024 ꢀ 0.0076
[0.12 ꢀ 0.11]a
8.9 ꢀ 0.40
UNC1215
78 ꢀ 2.5
13.7 ꢀ 2.0
>30
>30
>30
>30
[9.4 ꢀ 1.7]a
a
Indicates literature values.
Fig. 2 (a) X-ray crystal structure of UNC2533 (1) in complex with L3MBTL3 as a 2 : 2 complex (PDB ID: 4L59). UNC2533 is shown in cyan and the two monomers of
L3MBTL3 are colored in green and pink. (b) The binding site of UNC2533 with L3MBTL3 at the dimer interface showing the key hydrogen bonding interactions of the
ligand with residues D381 and D274.
conformation of the protein backbone in the ligand binding explored modications to the (pyrrolidinyl)piperidine, the pyr-
site. This exibility at the dimer interface could be a mechanism rolidine, and the C2-linker. Synthesis was performed following
that endogenous L3MBTL3 uses to read different histone the methods exemplied in Schemes 1 or 2 and as described in
sequences. Whether this mechanism is truly important to the the ESI.† Briey, compounds were synthesised from the
biological function of L3MBTL3 is the subject of ongoing commercially available 1-bromo-4-(2-bromoethyl)benzene via
studies.
an alkylation reaction followed by cross coupling (Scheme 1).
For Buchwald couplings, the reaction conditions were improved
by employing a RuPhos/RuPhos pre-catalyst system, with NaOt-
Bu as the base and tetrahydrofuran as the solvent.26 These
conditions were more suitable for the coupling of cyclic amines
affording the desired products in superior yields (18–98%)
within a shorter reaction time. Under microwave irradiation,
complete conversion to the product was oen observed by
Structure–activity relationships
As the central core of the UNC2533 molecule is structurally
distinct from UNC1215, we further investigated how changes in
this structure would affect its activity and selectivity for
L3MBTL3. We rst synthesised a series of compounds that
This journal is ª The Royal Society of Chemistry 2013
Med. Chem. Commun., 2013, 4, 1501–1507 | 1503

View Article Online
MedChemComm
Concise Article
Scheme 2 Reagents and conditions: (a) pyrrolidine, TBTU, NEt3, DMF, room
temperature, 17 h (b) 4-(pyrrolidin-1-yl)piperidine, RuPhos, RuPhos pre-catalyst,
NaOt-Bu, THF, microwave irradiation, 120 ꢁC, 10 min (c) LiAlH4, THF, reflux, 18 h.
were used as positive controls for L3MBTL3 and L3MBTL1,
respectively. UNC1079 20 was used as a less active control as it
binds L3MBTL3 poorly (IC50 ¼ 8.0 ꢀ 3.0 mM). As we were most
interested in the selectivity of the compounds for L3MBTL3
against the other MBT family proteins L3MBTL1 and MBTD1,
compounds evaluated at these proteins were tested at concen-
trations of up to 300 mM. All other assays were carried out with a
maximal compound concentration of 30 mM. While Alpha-
Screenꢀ is a suitable method for the high-throughput screening
of compounds, an orthogonal assay, such as isothermal titra-
tion calorimetry (ITC) was used to conrm SAR for a represen-
tative subset of compounds.
Fig. 3 Alignment of the UNC2533–L3MBTL3 crystal structure (PDB ID: 4L59,
cyan) with the UNC1215–L3MBTL3 crystal structure (PDB ID: 4FL6, grey) using the
Kabasch method implemented in VMD. The monomer on the right is aligned
while the one on the left is misaligned.
We initially modied the (pyrrolidinyl)piperidine moiety to a
slightly larger 1,40-bipiperidine in compound 2 (Table 2). As
desired, the L3MBTL1 activity of this compound decreased to 50
ꢀ 3.2 mM. Unfortunately this small change also resulted in a 9-
fold drop in activity for L3MBTL3 which was mirrored in the ITC
by a 7-fold loss in activity (Kd ¼ 2.4 mM). This indicates that
binding to the L3MBTL3 dimer is primarily mediated by the
interaction of the (pyrrolidinyl)piperidine with the second
methyllysine reader domain.
Next, we set out to synthesise compounds of different sizes
by modifying the length and position of the 2-carbon linker.
Shortening the linker to a single methylene, in compound 3
resulted in a 4 to 5-fold drop in L3MBTL3 activity (IC50 ¼ 0.35 ꢀ
0.11 mM, Kd ¼ 1.4 mM), which could be due to a weaker inter-
action of this compound with either or both of the domain 1
and domain 2 binding sites of the L3MBTL3 dimer. Length-
ening the linker to a 3-carbon chain in compound 4 maintained
L3MBTL3 activity in the Alphascreenꢀ and is consistent with
the ability of a larger molecule like UNC1215 to be accommo-
dated in the binding site. Compound 4, however, was less active
against L3MBTL3 by ITC, which gave a Kd of 1.84 mM. Moving
the 2-carbon chain meta to the piperidine in (compound 5) did
not result in a signicantly different IC50 (0.12 ꢀ 0.023 mM)
however the ITC showed much weaker binding (Kd ¼ 2.7 mM)
Fig. 4 Superimposition of the UNC1215–L3MBTL3 complex (PDB ID: 4FL6, grey
carbons) and the UNC2533–L3MBTL3 complex (PDB ID: 4L59, cyan carbons) (a)
view of the domain 2 (methyllysine) binding pocket showing residues in the
aligned monomer (b) view of the domain 1 binding pocket showing residues in
the aligned monomer (c) view of the domain 2 (methyllysine) binding pocket
showing the shifted residues in the misaligned monomer (d) view of the domain 1
(methyllysine) binding pocket showing the shifted residues in the misaligned
monomer.
LCMS aer 10 minutes of heating. If the di-bromo starting indicating that substitution at this position is not well tolerated.
material was unavailable, compounds were synthesised from The shorter methylene linker in the meta position (6) had
the acid via an initial amide coupling. A Buchwald coupling similar activity to 5.
followed by reduction of the amide with LiAlH4 gave the nal
dibasic compounds (Scheme 2.).
Condent that the piperidine moiety acts simply as a linker
group in both UNC2533 and UNC1215, we modied the
The biological activity of all compounds was primarily eval- piperidine to a 2-carbon aliphatic chain in compound 7. This
uated against our reader panel using the AlphaScreenꢀ assay.22 change resulted in a 20-fold loss in potency against L3MBTL3
Previously reported compounds UNC1215 20 and UNC669 18,19 such that its activity was similar to the monobasic compound
1504 | Med. Chem. Commun., 2013, 4, 1501–1507
This journal is ª The Royal Society of Chemistry 2013

View Article Online
Concise Article
MedChemComm
Table 2 Effects of changing the (pyrrolidinyl)piperidine or the ethyl-pyrrolidine
Activity as determined by AlphaScreenꢀ IC50 (mM) and ITC Kd (mM), shows in parentheses [ ]
#
Structure
L3MBTL3
L3MBTL1
MBTD1
53BP1
CBX7
UHRF1
Jarid1A
PHF23
0.062 ꢀ 0.0072
[0.37 ꢀ 0.052]
14 ꢀ 0.0
UNC2533 (1)
53 ꢀ 4
14.5 ꢀ 2.1
>30
>30
>30
>30
>30
>30
>30
>30
[19 ꢀ 1.4]
0.58 ꢀ 0.027
2
50 ꢀ 3.2
170 ꢀ 39
>30
[2.4 ꢀ 0.13]
0.35 ꢀ 0.11
[1.4 ꢀ 0.13]
3
4
57 ꢀ 0.17
157 ꢀ 38
135 ꢀ 33
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
0.070 ꢀ 0.013
[1.8 ꢀ 0.38]
10 ꢀ 2.8
30 ꢀ 2.6
0.12 ꢀ 0.051
[2.7 ꢀ 0.33]
5
261 ꢀ 34
>30
>30
>30
>30
>30
6
7
0.12 ꢀ 0.023
1.3 ꢀ 0.87
45 ꢀ 4.9
51 ꢀ 4.0
78 ꢀ 17
94 ꢀ 11
19.4 ꢀ 2.9
>30
>30
>30
>30
>30
>30
>30
>30
>30
2.4 ꢀ 1.0
[38 ꢀ 2.2]
8
>300
>300
>30
>30
>30
>30
>30
9
2.4 ꢀ 0.28
88 ꢀ 10
32 ꢀ 2.1
>300
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
10
0.32 ꢀ 0.11
149 ꢀ 8.5
58 ꢀ 1.9
13 ꢀ 1.4
90 ꢀ 1.1
11
[44 ꢀ 4.9]
[48, n ¼ 1]
UNC669, and is potentially due to the inability of this smaller efficiently. As expected, complete removal of the ethyl pyrroli-
compound to engage both pockets of the dimer. To determine dine in compound 11, resulted in over a 200 fold loss in
if two basic amines are critical to maintaining L3MBTL3 L3MBTL3 activity (IC50 ¼ 13 ꢀ 1.4 mM), indicating that while
activity, we synthesised the corresponding amide, compound 8, this fragment still binds, both amines are required for high
and found that this modication decreased L3MBTL3 activity L3MBTL3 potency. In comparison with UNC2533, the activity of
by more than 30-fold. This loss in activity was also evident by 11 at L3MBTL1 was lower (IC50 ¼ 90 ꢀ 1.1 mM), indicating that
ITC, where only weak binding was observed (Kd ¼ 38 mM). The the ethyl pyrrolidine is somewhat responsible for the observed
same trend was observed when comparing compounds 6 and 9 L3MBTL1 binding of UNC2533.
and is rationalised by the crystal structure of UNC2533, which
With the knowledge that a two carbon linker is optimal for
shows that a strong ionic interaction occurs between the activity, we continued by modifying the aromatic core and the
positively charged nitrogen of the ethyl-pyrrolidine and the ethyl-pyrrolidine of UNC2533 (Table 3). Addition of a chloro ortho
carboxylate of D274 in domain 1. To decrease linker exibility, to the 2-carbon linker in compound 22 was tolerated, however no
the 2-carbon chain was replaced with the more rigid piperidine signicant increase in potency was observed. Hypothesizing that
in symmetrical compound 10. As a result, the L3MBTL3 affinity we may be able to increase compound selectivity by designing a
was weakened about 5-fold (0.32 ꢀ 0.11 mM). This loss in covalent inhibitor, we synthesised the electrophilic 3-cyano and
activity could be explained by the UNC2533 crystal structure 3-keto compounds 12 and 13 that could be susceptible to
which shows that it is necessary for the ethylene linker to adopt nucleophilic attack by the cysteine thiol present in the domain 1
a gauche conformation in-order to engage the carboxylate of or domain 2 binding pockets. Both compounds were evaluated in
D274. A atter geometry may not be able to achieve this the same AlphaScreenꢀ assay. Aer an incubation time of 30
This journal is ª The Royal Society of Chemistry 2013
Med. Chem. Commun., 2013, 4, 1501–1507 | 1505

View Article Online
MedChemComm
Concise Article
Table 3 Effects of substituting the pyrrolidine moiety and core aromatic
Activity as determined by AlphaScreenꢀ IC50 (mM) and ITC Kd (mM), shows in parentheses [ ]
#
Structure
L3MBTL3
L3MBTL1
14 ꢀ 0.0
128 ꢀ 15
MBTD1
53 ꢀ 4
>300
53BP1
14.5 ꢀ 2.1
>30
CBX7
>30
UHRF1
>30
Jarid1A
>30
PHF23
>30
UNC2533 (1)
0.062 ꢀ 0.0072
0.36 ꢀ 0.039
12
>30
>30
>30
>30
13
0.39 ꢀ 0.24
44 ꢀ 5.5
166 ꢀ 34
18.4 ꢀ 6.5
>30
>30
>30
>30
14
15
16
17
18
19
20
21
0.048 ꢀ 0.013
1.0 ꢀ 0.55
86 ꢀ 12
>300
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
>30
152 ꢀ 28
>300
0.059 ꢀ 0.035
[0.40 ꢀ 0.17]
60 ꢀ 9.9
>300
[30, n ¼ 1]
0.10 ꢀ 0.038
63 ꢀ 2.1
249 ꢀ 2.8
290 ꢀ 9.2
>300
0.080 ꢀ 0.041
[0.61 ꢀ 0.13]
50 ꢀ 5.5
[3.3 ꢀ 0.15]
0.85 ꢀ 0.091
0.65 ꢀ 0.095
0.27 ꢀ 0.005
125 ꢀ 14
92 ꢀ 14
65 ꢀ 9.9
>300
247 ꢀ 17
22
0.037 ꢀ 0.0017
21 ꢀ 9.7
176 ꢀ 71
21.0 ꢀ 3.5
>30
>30
>30
>30
minutes, both modications displayed a 5-fold drop in potency the nearby Y301. A loss in activity was also observed for the
(Table 3), suggesting that covalent inhibition of L3MBTL3 seems spirocyclic compounds 20 and 21 which may undergo similar
unlikely and that substitution at the 3-position of the pyrrolidine electronic clashes with the binding site.
ring is sub-optimal for ligand binding. As no signicant
improvement over UNC2533 was observed, these compounds compound 14 (versus UNC2533). Interestingly, this compound
were not pursued further. was also more selective than UNC1215 against L3MBTL1 and
Decreasing the ring size to an azetidine gave the equipotent
Further modications to the ethyl-pyrrolidine by increasing MBTD1 in our Alphascreenꢀ assay. Further modication to this
the size of the pyrrolidine ring to either an azepane (17) or moiety by introduction of a diuoro in the 3-position of the
piperidine (18) did not signicantly change the L3MBTL3 azetidine ring in 15 dropped the L3MBTL3 activity to 1 mM.
activity. The L3MBTL3 activity of 18 was conrmed by ITC Calculation of the pKa of the azetidine nitrogen in both
yielding a Kd of 0.61 mM for L3MBTL3. Surprisingly, the Kd of 18 compounds revealed that addition of the diuoro decreases the
for L3MBTL1 was more potent than expected based on the pKa from 10.3 to 7.1 and consequently weakens the ionic
Alphascreenꢀ results (Kd ¼ 3.3 mM) and is currently under interaction of the pyrrolidine N with D274. The hydrolysed
further investigation. Introduction of a hydrogen bond acceptor spirocyclic, compound 16 had a similar activity and selectivity
to the six-membered ring was originally hypothesised to have prole to the unsubstituted azetidine (14) suggesting that the
favourable interactions with the tyrosine hydroxyl located at the hydroxyl groups are not making new favourable interactions but
back of the domain 1 binding pocket however the inclusion of are also not interfering with binding. ITC of 16 conrmed the
oxygen in the morpholino compound 19 was not tolerated. The activity of this compound for L3MBTL3, yielding a Kd of 0.40
activity for this compound was weakened by more than an order mM. The selectivity against L3MBTL1 was also conrmed by ITC
of magnitude when compared to the alicyclic compound. This to be 75 fold (Kd ¼ 30 mM).
may be due to the decreased basicity of the morpholino
nitrogen compared with the pyrrolidine nitrogen. From the
crystal structure of UNC2533, it is also reasonable to suggest
Conclusions
that this may be the result of an unfavourable electronic clash In summary, we have reported the discovery of a second series
between the electronegative oxygen atom and the pi electrons of of L3MBTL3 inhibitors based on our original compound
1506 | Med. Chem. Commun., 2013, 4, 1501–1507
This journal is ª The Royal Society of Chemistry 2013

View Article Online
Concise Article
MedChemComm
UNC2533. These compounds are structurally more compact
than the chemical probe UNC1215 and maintain nanomolar
L3MBTL3 potency. In addition, these compounds have
profound selectivity over the Royal family proteins L3MBTL1
and MBTD1 as well as a broad range of other methyllysine
readers exemplied in our AlphaScreenꢀ assay. Through
structure activity relationship studies, we deduced that a
dibasic amine is necessary for compound potency as both
6 D. M. Stewart, J. Li and J. Wong, Mol. Cell. Biol., 2005, 25,
2525–2538.
7 Y. B. Schwartz and V. Pirrotta, Nat. Rev. Genet., 2007, 8, 9–22.
8 Y. Zhang and D. Reinberg, Genes Dev., 2001, 15, 2343–2360.
9 G. Schotta, R. Sengupta, S. Kubicek, S. Malin, M. Kauer,
E. Callen, A. Celeste, M. Pagani, S. Opravil, I. A. De La Rosa-
Velazquez, A. Espejo, M. T. Bedford, A. Nussenzweig,
M.BusslingerandT. Jenuwein,Genes Dev., 2008, 22, 2048–2061.
functionalities take part in key cation–pi binding interactions 10 N. Zacharias and D. A. Dougherty, Trends Pharmacol. Sci.,
with the L3MBTL3 dimer. Modications to the (pyrrolidinyl)- 2002, 23, 281–287.
piperidine showed that ligand binding is primarily mediated by 11 M. Yun, J. Wu, J. L. Workman and B. Li, Cell Res., 2011, 21,
the interaction of UNC2533 with the domain 2 reader pocket 564–578.
and that there is little room for variation in this portion of the 12 S. Maurer-stroh, N. J. Dickens, L. Hughes-davies,
molecule. In addition, we found that decreasing the ring size of
the ethyl-pyrrolidine moiety to an ethyl-azetidine was tolerated,
T. Kouzarides, F. Eisenhaber and C. P. Ponting, Trends
Biochem. Sci., 2003, 28, 69–74.
giving the equipotent compound 14. Most importantly, X-ray 13 J. Kim, J. Daniel, A. Espejo, A. Lake, M. Krishna, L. Xia,
crystallography of the UNC2533-L3MBTL3 complex showed that Y. Zhang and M. T. Bedford, EMBO Rep., 2006, 7, 397–403.
UNC2533 binds the L3MBTL3 dimer as a 2 : 2 complex and is 14 F. R. Zahir, S. Langlois, K. Gall, P. Eydoux, M. A. Marra and
accommodated by rotation of the L3MBTL3 dimer interface.
J. M. Friedman, Am. J. Med. Genet., Part A, 2009, 149, 1257–
This is the rst evidence that demonstrates the dynamic nature
1262.
of the L3MBTL3 dimer and its ability to change its conforma- 15 N. Gurvich, F. Perna, A. Farina, F. Voza, S. Menendez,
tion upon ligand binding. This has implications for the design
of further L3MBTL3 inhibitors and possibly histone binding.
J. Hurwitz and S. D. Nimer, Proc. Natl. Acad. Sci. U. S. A.,
2010, 107, 22552–22557.
16 J. Qin, D. Van Buren, H. S. Huang, L. Zhong, R. Mostoslavsky,
S. Akbarian and H. Hock, J. Biol. Chem., 2010, 285, 27767–
27775.
Acknowledgements
We thank the SGC for providing the constructs and/or plasmids 17 S. V. Frye, Nat. Chem. Biol., 2010, 6, 159–161.
for L3MBTL1, L3MBTL3, MBTD1, UHRF1 and CBX7 and Greg 18 J. M. Herold, T. J. Wigle, J. L. Norris, R. Lam, V. K. Korboukh,
Wang for providing PHF23 and JARID1 proteins. The research
described here was supported by the National Institute of General
Medical Sciences, US National Institutes of Health (NIH, grant
R01GM100919), the Carolina Partnership and the University
C. Gao, L. A. Ingerman, D. B. Kireev, G. Senisterra,
M. Vedadi, A. Tripathy, P. J. Brown, C. H. Arrowsmith,
J. Jin, W. P. Janzen and S. V. Frye, J. Med. Chem., 2011, 54,
2504–2511.
Cancer Research Fund, University of North Carolina at Chapel 19 J. M. Herold, L. I. James, V. K. Korboukh, C. Gao, K. E. Coil,
Hill, the Ontario Research Fund (grant ORF-GL2), and the
Structural Genomics Consortium which is a registered charity
(number 1097737) that receives funds from AbbVie, Boehringer
D. J. Bua, J. L. Norris, D. Kireev, P. J. Brown, J. Jin,
W. P. Janzen, O. Gozani and S. Frye, MedChemComm, 2012,
3, 45–51.
Ingelheim, Canada Foundation for Innovation, the Canadian 20 L. I. James, D. Barsyte-Lovejoy, N. Zhong, L. Krichevsky,
Institutes for Health Research (CIHR), Genome Canada through
the Ontario Genomics Institute [OGI-055], GlaxoSmithKline,
Janssen, Lilly Canada, the Novartis Research Foundation, the
Ontario Ministry of Economic Development and Innovation,
Pzer, Takeda, and the Wellcome Trust [092809/Z/10/Z]. C.H.A.
holds a Canada Research Chair in Structural Genomics.
V. K. Korboukh, J. M. Herold, C. J. MacNevin, J. L. Norris,
C. A. Sagum, W. Tempel, E. Marcon, H. Guo, C. Gao,
X. P. Huang, S. Duan, A. Emili, J. F. Greenblatt,
D. B. Kireev, J. Jin, W. P. Janzen, P. J. Brown,
M. T. Bedford, C. H. Arrowsmith and S. V. Frye, Nat. Chem.
Biol., 2013, 9, 184–191.
21 B. M. Dickson, H. Huang and C. B. Post, J. Phys. Chem. B,
2012, 116, 11046–11055.
22 T. J. Wigle, J. M. Herold, G. A. Senisterra, M. Vedadi,
D. B. Kireev, C. H. Arrowsmith, S. V. Frye and
W. P. Janzen, J. Biomol. Screening, 2010, 15, 62–71.
Notes and references
1 B. D. Strahl and C. D. Allis, Nature, 2000, 403, 41–45.
2 T. Jenuwein and C. D. Allis, Science, 2001, 293, 1074–1080.
3 B. E. Bernstein, E. L. Humphrey, R. L. Erlich, R. Schneider, 23 Molecular Operating Environment (MOE), 2012.10,
P. Bouman, J. S. Liu, T. Kouzarides and S. L. Schreiber,
Proc. Natl. Acad. Sci. U. S. A., 2002, 99, 8695–8700.
Chemical Computing Group Inc., 1010 Sherbooke St. West,
Suite #910, Montreal, QC, Canada, H3A 2R7, 2012.
4 H. Santos-Rosa, R. Schneider, A. J. Bannister, J. Sherriff, 24 W. Kabsch, Acta Crystallogr., Sect. A: Found. Crystallogr.,
B. E. Bernstein, N. C. Emre, S. L. Schreiber, J. Mellor and
T. Kouzarides, Nature, 2002, 419, 407–411.
5 G. Schotta, A. Ebert, V. Krauss, A. Fischer, J. Hoffmann,
1976, 32, 922–923.
25 W. Humphrey, A. Dalke and K. Schulten, J. Mol. Graphics,
1996, 14, 33–38.
S. Rea, T. Jenuwein, R. Dorn and G. Reuter, EMBO J., 2002, 26 D. Maiti, B. P. Fors, J. L. Henderson, Y. Nakamura and
21, 1121–1131.
S. L. Buchwald, Chem. Sci., 2011, 2, 57–68.
This journal is ª The Royal Society of Chemistry 2013
Med. Chem. Commun., 2013, 4, 1501–1507 | 1507