Fragment-based ligand design of novel potent inhibitors of tankyrases

Article abstract of DOI:10.1021/jm400211f

Tankyrases constitute potential drug targets for cancer and myelin-degrading diseases. We have applied a structure- and biophysics-driven fragment-based ligand design strategy to discover a novel family of potent inhibitors for human tankyrases. Biophysical screening based on a thermal shift assay identified highly efficient fragments binding in the nicotinamide-binding site, a local hot spot for fragment binding. Evolution of the fragment hit 4-methyl-1,2-dihydroquinolin-2-one (2) along its 7-vector yields dramatic affinity improvements in the first cycle of expansion. A crystal structure of 7-(2-fluorophenyl)-4-methylquinolin-2(1H)-one (11) reveals that the nonplanar compound extends with its fluorine atom into a pocket, which coincides with a region of the active site where structural differences are seen between tankyrases and other poly(ADP-ribose) polymerase (PARP) family members. A further cycle of optimization yielded compounds with affinities and IC ₅₀ values in the low nanomolar range and with good solubility, PARP selectivity, and ligand efficiency.

Full text of DOI:10.1021/jm400211f

Article
Fragment-Based Ligand Design of Novel Potent Inhibitors of
Tankyrases
E. Andreas Larsson,*,†,∥ Anna Jansson,†,∥ Fui Mee Ng,§ Siew Wen Then,§ Resmi Panicker,† Boping Liu,§
Kanda Sangthongpitag,§ Vishal Pendharkar,§ Shi Jing Tai,§ Jeffrey Hill,§ Chen Dan,† Soo Yei Ho,§
Wei Wen Cheong,§ Anders Poulsen,§ Stephanie Blanchard,§ Grace Ruiting Lin,§ Jenefer Alam,§
Thomas H. Keller,§ and Par Nordlund*,†,‡
̈
†School of Biological Sciences, Nanyang Technological University, Lab 07-01, 61 Biopolis Drive (Proteos), Singapore 138673
‡Department of Medical Biochemistry and Biophysics, Karolinska Institute, Scheeles vag 2, SE-171 77 Stockholm, Sweden
̈
§Experimental Therapeutics Centre, 31 Biopolis Way, Nanos Level 3, Singapore 138669
S
* Supporting Information
ABSTRACT: Tankyrases constitute potential drug targets for
cancer and myelin-degrading diseases. We have applied a
structure- and biophysics-driven fragment-based ligand design
strategy to discover a novel family of potent inhibitors for
human tankyrases. Biophysical screening based on a thermal
shift assay identified highly efficient fragments binding in the
nicotinamide-binding site, a local hot spot for fragment
binding. Evolution of the fragment hit 4-methyl-1,2-dihydro-
quinolin-2-one (2) along its 7-vector yields dramatic affinity improvements in the first cycle of expansion. A crystal structure of 7-
(2-fluorophenyl)-4-methylquinolin-2(1H)-one (11) reveals that the nonplanar compound extends with its fluorine atom into a
pocket, which coincides with a region of the active site where structural differences are seen between tankyrases and other
poly(ADP-ribose) polymerase (PARP) family members. A further cycle of optimization yielded compounds with affinities and
IC50 values in the low nanomolar range and with good solubility, PARP selectivity, and ligand efficiency.
perform poly-ADP-ribosylation on its substrates, including
Axin. This stabilizes Axin, and the Axin turnover is shifted,
which leads to a negative regulation of the Wnt/β-catenin
pathway.6 Known inhibitors or inhibitor series of TNKSs such
as 1 (XAV939)6 (Figure 1), 1,1‴,2,2‴-tetrahydrotrispiro-
INTRODUCTION
■
Tankyrase (TNKS) in its two isoforms, TNKS1 and TNKS2, is
a subset of the PARP family of proteins, which constitute
potential drug targets for cancer and myelin-degrading
diseases.1 Functionally, TNKSs as other PARPs are poly-
ADP-ribosylating enzymes that transfer ADP-ribose moieties
from NAD+ to a variety of substrates.2 Poly-ADP-ribosylation
affects the protein substrates in different ways, but most often
by reducing the activity of the modified protein, as in the case
of TNKS modification of TRF1.3 Other members of the PARP
family have been identified as good interference points for drug
discovery,4 perhaps most notably to target DNA repair in
BRCA-driven cancers. TNKSs have been implicated in a diverse
range of functions, such as regulation of telomeric length,
regulation of the Wnt signaling pathway, control of the mitotic
checkpoint, and mediation of insulin-stimulated glucose
uptake.5 TNKS1 and TNKS2 have emerged as potential cancer
targets on the basis of the observation that inhibition of TNKS
acts as a negative regulator for the Wnt/β-catenin pathway in
colon cancer cells by stabilizing Axin.6 Recent studies using
selective inhibitors corroborate the hypothesis that it is the
inhibition of TNKSs and not off-target effects on other PARPs
that is responsible for the observed effect.7 Although these
compounds are not very potent, they prove the point that the
selectivity profile is an important aspect in the development of
TNKS inhibitors. When inhibited, TNKS can no longer
Figure 1. Structure of 1 and fragment hits 2 and 3. The 6- and 7-
vectors are indicated for compound 2.
[indole-3,2′:5′,5″-bis([1,3]dioxane)-2″,3‴-indole]-2,2‴-dione
(JW67), and 4-[4-(4-methoxyphenyl)-5-({[3-(4-methylphen-
yl)-1,2,4-oxadiazol-5-yl]methyl}sulfanyl)-4H-1,2,4-triazol-3-yl]-
pyridine (JW74),8 as well as other compounds,9,10 originated or
were derived from compounds found in cell-based HTS (high-
throughput screening) screens targeting Wnt signaling. TNKS
Received: February 9, 2013
Published: May 14, 2013
© 2013 American Chemical Society
4497
dx.doi.org/10.1021/jm400211f | J. Med. Chem. 2013, 56, 4497−4508

Journal of Medicinal Chemistry
Article
a
Scheme 1. Synthesis of Compounds 8−15
a
Reagents and conditions: (a) toluene, reflux; (b) concentrated H2SO4, 120 °C; (c) Pd(PPh3)4, arylboronic acid, K2CO3, 4:1 dioxane/water, reflux.
a
Scheme 2. Synthetic Route to Compounds 16 and 17
a
Reagents and conditions: HATU, Hunig’s base, dichloromethane.
was only later identified as the molecular target of the
compounds.
and it is perhaps not surprising that this is an affinity hot spot
where most PARP inhibitors are anchored.11 An exception is
the newly discovered class of inhibitors that instead interact
with the D-loop and induce conformational changes.8,12
Fragment Screening. In this study we have used a
fragment-based ligand design (FBLD) strategy to identify novel
inhibitors for TNKS2. Our strategy is based on the extensive
use of biophysical screening to identify leads for crystallography
and further evolution into mature compounds. A thermal shift
assay (TSA)-based strategy using differential scanning
fluorometry (DSF)13 was used for initial fragment screening
of a 500 compound fragment library and hit characterization
and as one of the main metrics for the lead optimization
process. DSF screening was performed at a compound
concentration of 1 mM. Positives were validated to “hits”
internally by checking for dose-dependent response, typically
tested over a concentration range from 5 to 4000 μM. From
our experience, collecting data over a wide range of
concentrations is important to remove oddly behaving
compounds from the hit list. It also assists in revealing other
concentration-dependent effects such as insolubility or
aggregation. This method was used for the characterization of
all compounds in this paper (see Figures S3 and S4, Supporting
Information).
Interpretation of the screening data was complicated by the
fact that the melting profile of TNKS2 did not exhibit a clear
sigmoidal transition, but rather a two/multiple-state transition
(Figure 2c). Assay validation experiments using known TNKS/
PARP inhibitors showed which part of the curve (explicitly the
last transition) was stabilized upon binding. Although the
nonsigmoidal melting behavior does not cause any real trouble
for characterization compounds with large thermal shifts, it
does significantly complicate the use of numerical selection
(and the fitting of Tm) for weakly stabilizing fragments. We
In this study, we specifically targeted the catalytic PARP
domain of TNKS2 directly using a battery of biophysical
methods. Fragment hits, discovered by a thermal shift assay,
were optimized into high-affinity inhibitors of TNKS2. The
expansion of the fragment hits to potent TNKS2 inhibitors was
guided by high-resolution structures of protein−ligand
complexes, and compounds were characterized using both
biophysical and activity assays, and TNKS2/PARP selectivity
was evaluated for representative compounds. Structures of the
more potent late-stage compounds were also used for the
introduction of solubilizing groups in noninterfering positions
to provide compounds with better physical properties.
Synthesis. The synthesis of the TNKS inhibitors is
described in Scheme 1. 3-Bromoaniline was reacted with
ethyl acetoacetate in refluxing toluene, affording the β-
ketoamide (4), which was cyclized in concentrated sulfuric
acid to quinolone-2-one (5). The substituents on C7 of 5 were
introduced via Suzuki coupling (Scheme 1). Similarly, C6-
substituted analogues were prepared from commercially
available 6-bromo-4-methylquinoline-2(1H)-one. Solubilizing
groups were introduced by O-(7-azabenzotriazol-1-yl)-
N,N,N′,N′-tetramethyluronium hexafluorophosphate
(HATU)-mediated amide synthesis (Scheme 2).
RESULTS
■
Binding of NAD+ is a central feature for all PARP proteins as
NAD+ acts as a donor of ADP-ribose moieties onto various
substrates with simultaneous release of nicotinamide. Structur-
ally, the nicotinamide moiety is anchored by three key polar
interactions between the nicotinamide amide group and
residues Gly1032 and Ser1068 in TNKS2 (Figure 2a).
PARPs bind the nicotinamide moiety with high efficiency,
4498
dx.doi.org/10.1021/jm400211f | J. Med. Chem. 2013, 56, 4497−4508

Journal of Medicinal Chemistry
Article
Figure 2. (a) Key active site residues (polar interactions with Gly1032 and Ser1068) for binding of nicotinamide in TNKS2 (PDB 3U9H). (b)
Aligned overlay of the crystal structure of the protein−ligand complex of 2 (magenta) and 3 (yellow) in TNKS2 showing the hydrogen bonding to
Ser1068 and Gly1032 as well as a surface representation of the pocket. Comparison of thermal shift data for TNKS2 (c) and chymotrypsin-treated
TNKS2 (d). The DMSO control is depicted in black and runs in the presence of 2 (here shown at a concentration of 4 mM to better visualize the
stabilization vs the DMSO control) in red.
found it much more reliable to manually inspect all transitions
and to avoid contaminating the hit set with large numbers of
false positives. We also found that adding chymotrypsin to the
DSF experiment greatly simplified the interpretation by
removing the first, nonreporting transition altogether (Figure
2d).
Two compounds, 4-methyl-1,2-dihydroquinolin-2-one (2)
and 4-chloro-1,2-dihydrophthalazin-1-one (3) (Figure 1), gave
melting curves distinct from that of the DMSO control,
stabilizing TNKS2 by 1.1 and 1.3 °C, respectively, at 1 mM.
Further analysis and triage established that these two
compounds also showed a dose-dependent stabilization.
Cleaner looking data with in situ trypsination of TNKS2 (see
the Experimental Section for details) in combination with
concentration-dependent stabilization of Tm were also
collected, further strengthening the validation (Figure 2d).
These compounds were further characterized with a
PARsylation-based activity assay (Table 1) and crystallography
(Figure 2b).
First Round of Expansion. Analysis of the crystal structure
of the two fragments bound to TNKS2 showed hydrogen
bonding to Ser1068 and Gly1032 also seen in nicotinamide
binding and the stacking of the compound to Tyr1071. This
suggested that there were two major areas of fragment 2 that
could be further optimized. One of these is the methyl group of
2 at the 4-position, which is roughly in the same position as the
cyclic thioether moiety of 1, protruding down toward the
catalytic glutamate (Glu1138) (Figure 2b). Substitution at the
4-position is very important, as the analogue lacking the methyl
group (compound 6, Table 1) is inactive in all our assays. A
small set of analogues with small variations/extensions of the
methyl group were prepared and purchased. However, these
compounds did not show any improvements from 2 in TSA or
activity assays (see Table S3, Supporting Information).
The other interesting position for expansion is through the 7-
position of 2, which points toward the extended pocket
responsible for adenosine binding (Figure 2b). It was
immediately apparent from TSA data that compounds extended
in the 7-position induced significantly more stabilization of
TNKS2 than 2. Compound 7 (prepared as a control for the
structure-based expansion hypothesis), which instead extends at
the 6-position, showed no stabilizations in the TSA assay, most
likely since extensions in this vector clash into the protein
(Figure 2b).
Typically, ligand efficiency14 (LE; in this work expressed as
pIC50/HA) is the preferred metric for ranking and prioritization
of hits for follow-up. We decided to optimize a scaffold based
on compound 2 despite the slightly lower LE compared to that
of 3 (0.41 and 0.45, respectively), the main reason being that 2
possesses better geometrical vectors for expansion. In addition,
the binding mode of 2 has less overlap with 1 and hence could
provide more novel compounds.
The expansion strategy was based on Suzuki couplings using
5 and a set of arylboronic acids. Using this strategy, a small set
of compounds were assembled. The most interesting was
4499
dx.doi.org/10.1021/jm400211f | J. Med. Chem. 2013, 56, 4497−4508

Journal of Medicinal Chemistry
Article
Table 1. Biophysical, Structural, and Activity Data on
Fragment Hits and Key Intermediates for TNKS2
Figure 3. (a) Surface representation showing the overall positioning of
compound 11 in the pocket. (b) Crystal structure of the active site of
TNKS2 in complex with 11 showing the nonplanar binding
conformation of the o-fluorophenyl moiety with hydrophobic
interaction with Ile1075, Phe1035, and Tyr1050 as well as the o-
fluoro interaction with the main chain oxygen of Tyr1071.
ortho-substituents. Altered conformational preference can,
when it coincides with the binding conformation, have a
dramatic influence on affinity,15 which we decided to try to
explore by introducing larger ortho-substituents. The crystal
structure also suggests that para-substituents (Figure 3a)
should be well accepted as this vector points toward the exit
of the pocket and hence also a good area to sample polar
features surrounding the exit and to introduce solubilizing
groups if needed.
a
IC50 values are an average of two independent determinations.
b
Curves of the IC50 determination were impacted by the low solubility
c
of the compounds. Differential scanning fluorometry at 1000 μM.
d
e
Differential scanning fluorometry at 125 μM. Surface plasmon
f
resonance, steady-state analysis. Surface plasmon resonance, kinetic
analysis. Ligand efficiency in pIC50/HA, where HA means heavy
atom, i.e., a non-hydrogen atom. Several attempts to soak and
cocrystallize the compounds were made, but no ligand density could
be observed.
g
h
On the basis of this analysis, a set of 15 compounds was
prepared (e.g., compounds 12 and 13). The o-chloro
compound 12 indeed led to larger Tm shifts than the o-fluoro
analogue, and para-substituents were allowed as predicted,
yielding substances with IC50 and Kd in the low nanomolar
range. Interestingly, the activity of compound 13 is several
orders of magnitude lower than that of 12. Since no structural
information is available for 13, we cannot provide a structure-
based understanding of why the difference in activity and
stabilization of 13 and, for example, 8, 11, and 12 is so large. A
possible explanation could be that 13 is unable to displace Wf
and instead the additional bulkiness is only disruptive for the
interaction. It was also apparent that some of the compounds
containing o-chloro substituents behaved poorly and did not
yield interpretable thermal shifts or crystal structures. We
expected this, at least in part, to be due to solubility and or
aggregation problems (Table 2). Hence, optimization of the
para-position was done with these properties in mind, and
some additional compounds (compounds 14−17) were
synthesized at this stage. Ligands synthesized extending on
the para-position only allowed for limited interactions with the
protein and mainly exposed these substituents to the solvent. In
essence, these groups therefore served primarily as solubilizing
groups, with concomitant loss of LE.
compound 11, bearing an o-fluoro substituent (Table 1;
compare to those with p-fluoro (compound 9) and m-fluoro
(compound 10) substituents), invoking large Tm shifts even at
lower ligand concentrations. We subsequently solved the
structure of the protein−ligand complex of this compound,
revealing that the newly added fluorophenyl moiety adopted a
nonflat conformation (torsion of 51°, Figure 3) in which the
fluorine atom had displaced a water molecule present in the
structure of 2 (referred to as Wf below, Figures 2b and 4a). In
addition, the aromatic group is lined up for van der Waals
(vdW) interactions with the hydrophobic side chain of Ile1075
and nonpolar contact to Phe1035, Tyr1050, and Pro1034.
Affinity and Solubility Optimization. Further analysis of
the crystal structures, and crystallographically directed molec-
ular modeling of potential close analogues of 11, suggested
some additional features of the TNKS2 active site pocket that
potentially could be exploited in the next iteration of
compounds. When the fluorine atom in compound 11 displaces
Wf, we conclude that this pocket could potentially harbor even
larger ortho-substituents to maximize interactions with the
carbonyl oxygen of Tyr1071 (Figure 4a). As mentioned above,
11 adopts a nonplanar conformation in the binding pocket. The
conformational preferences of biphenyl torsions are sensitive to
4500
dx.doi.org/10.1021/jm400211f | J. Med. Chem. 2013, 56, 4497−4508

Journal of Medicinal Chemistry
Article
a
Table 2. Solubility of Selected Compounds
a
Thermodynamic solubility (see the Experimental Section).
region to the main chain nitrogen of Ile1075 (Figure 4c). The
solubilizing groups are clearly solvent immersed and very
flexible, and it was not possible to model this part of the
molecule into the electron density (Figure 4b).
Characterization of Ligands. All compounds were
characterized using the DSF assay in a dose−response fashion
(Figure 5a). In addition to the benefits mentioned before, this
Figure 4. (a) Overlay of crystal structures of 2 (magenta), 11 (yellow),
and 17 (orange) showing their interactions with the main chain
carbonyl oxygen of Tyr1071. In the structure of the starting fragment
2, Tyr1071 is hydrogen bonded to a water molecule, Wf (H2O−O =
2.9 Å), that is displaced by F in 11 (F−O = 3.4 Å) and Cl in 17 (Cl−
O = 3.0 Å). The Cl atom in the latter compound is distinctly closer to
O(sp2) than the average vdW radius (3.3 Å),16 indicating an attractive
electrostatic interaction between them. This is also supported by the
near linear geometry of the C−Cl···O atoms. (b) Overlay of 16
(purple), 14 (cyan), and 17 (orange). The Fo − Fc electrondensity
map belongs to compound 17 (orange). (c) Crystal structure of 16
(purple) showing water-mediated interactions of the carbonyl oxygen
and residue Ile1075 in addition to the hydrophobic interactions of the
o-chlorophenyl with Tyr1071 and the hydrogen bonds to Ser1068 and
Gly1032.
Figure 5. (a) DSF dose−response data on five compounds. (b)
Comparison of the increase in thermal stability (DSF, blue) vs Kd
(SPR, red) and IC50 (autoPARsylation, green) for a set of compounds
prepared during the optimization. DSF was used for the library screen
and as a primary metric during the optimization. The DSF assay was
found to generally correlate well with IC50 and Kd (SPR).
also allowed us to seamlessly compare ligands at gradually
lower concentrations as the affinity increased. As an example,
fragments 2 and 3 show no significant stabilization at 125 μM,
the concentration where the more high affinity ligands show
clear effects.
A subset of the ligands discussed above was further
characterized with SPR to follow the improvement of affinities
along the compound evolution process. Good general
correlations among SPR-derived Kd, IC50, and responses from
DSF were found (Table 1 and Figure 5b). For compounds for
which SPR affinities were fitted by kinetic analysis, we examined
the on- and off-rates. Overall, improvement in affinity generally
follows the increasingly slower off-rates for the expanded
compounds, although the situation becomes more complicated
for the higher affinity compounds. For example, compounds 16
and 17 have essentially the same observed Kd with SPR (Table
With this group the aggregation/solubility problems were
resolved, as compounds with solubilizing functions in the para-
position were immediately crystallographically productive (e.g.,
14−17) as well as behaved better in assays. Four crystal
structures of protein−ligand complexes of the derivatives were
subsequently solved. Upon analysis of these structures, we
found that the Cl−O(Tyr1071) distance was 3.0 Å, distinctly
shorter than the average Cl−O(sp2) vdW radius of 3.3 Å,
indicative of a electrostatic chlorine−carbonyl oxygen inter-
action (Figure 4a).16 The para-substituents did vector out
toward the exit of the pocket. For two compounds (16 and 17)
bearing p-carboxamido substituents, the different N-attached
alkyl tail clearly protrudes out of the pocket, making polar
interactions with protein-anchored water molecules at the exit
4501
dx.doi.org/10.1021/jm400211f | J. Med. Chem. 2013, 56, 4497−4508

Journal of Medicinal Chemistry
Article
1), but with quite different kinetic balances where the latter
compound shows slower on- and off-rates. (SPR on- and off-
rate data and plots are found in Figure S1 and Table S2,
Supporting Information).
Selectivity. PARPs in general are key interference points in
functional networks, a reason why they have spawned so much
interest as targets for inhibitors. The most elaborate PARP
inhibitors (such as 18 (olaparib)17) exhibit a relatively wide
spectrum PARP inhibition profile. In contrast, our aim was to
create potent inhibitors of TNKSs, if possible with no practical
inhibition of other PARPs.
As compound 11 showed a significant jump in affinity, along
with its novel binding mode, we decided to investigate the
selectivity profile for this compound. We assayed 11 and 18
toward a panel of PARPs (Table 3) and found that 11 in
contrast to 18 showed negligible inhibition at 10 μM.
profile enables therapies targeting TNKS2 without the potential
side effects associated with inhibition of the Wnt pathway.
These novel TNKS2-selective tool compounds can therefore
provide an important chemical tool to study tankyrase biology.
Clearly, more studies on 11 and the more potent compounds
developed in this study are needed to reveal the real potential
for their use, but as mentioned above, as no intertankyrase-
selective inhibitor has been available previously, pure inhibition
effects are not well studied.
DISCUSSION
■
Several families of TNKS inhibitors have been described in the
academic and patent literature;8,9,20 however, so far none of
these families have led to a clinical candidate. FBLD strategies
identify hot spots in the protein for ligand binding and core
fragments which are good starting points for ligand develop-
ment.21 The strategy can therefore generate more efficiently
binding lead compounds as well as novel structural frameworks,
as compared to existing compounds.
Table 3. Inhibition (%) of Selected PARPs at 10 μM for
Compounds 18 and 11
In this work, we have examined whether a fragment strategy
is suitable for human TNKS and if it could be applied to
discover novel compounds, with different properties as
compared to existing TNKS inhibitors. The strategy used was
initially driven by TSA screening, and each step was guided by
X-ray structures of protein−ligand complexes. IC50 measure-
ments were performed when higher affinity ligands were
obtained, and SPR was used retrospectively to characterize
improvements of affinities and ligand efficiency during fragment
evolution. Albeit false negatives in initial TSA screens cannot be
excluded, the strategy appeared efficient and led to the rapid
development of high-affinity ligands for TNKS, with less than
40 compounds synthesized.
compd
PARP1
PARP2
PARP3
PARP6
PARP7
PARP11
11
18
21
97
2
0
31
98
37
84
23
72
99
99
We then generated more detailed data to compare with those
of other inhibitors, particularly 1. Interestingly, 11 is a very
clean TNKS inhibitor, with more than 100-fold selectivity
toward other PARPs (IC50 = 52 nM vs TNKS2). 1 is more
potent than 11 but inhibits PARPs in the middle nanomolar
range6,18 (see Table 4 for a comparison). As shown in Table 1,
a
Table 4. IC50 (μM) and PARP Selectivity of 11 and 1
Most FBLD studies in the literature use NMR, SPR, or X-ray
crystallography for initial fragment screening and evolution. We
generally found the TSA-based data to be reliable for the
TNKS2 system and to correlate well with other biophysical
(SPR) and biochemical (autoPARsylation) data (Table 1 and
Figure 5b).
compd
PARP1
PARP2
PARP3
TNK1
TNK2
PARP6
11
1
>10
0.12
>10
>10
>10
0.86
0.052
0.008
>10
>10
0.046
0.011
a
PARP1, PARP2, PARP3, and PARP6 assays were performed by BPS
Bioscience.
As DSF, in contrast to NMR, requires protein and
compounds at high concentration in the assay, it also, in effect,
reports on compound issues that might hamper successful
crystal soaking or cocrystallization of compounds. In practice, a
major bottleneck of FBLD is the efficient translation of hits
obtained from activity or biophysical assays into crystal
structures. The protein concentration in crystals is very high
(millimolar range), requiring high compound concentrations to
achieve occupancy and interpretable densities of compounds in
the crystal structures. Furthermore, the specific buffer and
precipitant solutions in the soaking or cocrystallization
experiment normally lower the effective compound solubility
(Table 2). Therefore, hits can often be difficult to translate into
structures of protein−ligand complexes. The indicative
correlation for this was, for example, seen for compounds 12
and 8, which appeared to reprecipitate in buffer preparations. In
addition, they also showed low reproducibility in TSA assays
and when soaked with TNKS2 crystals yielded crystal
structures without bound ligand (data not shown). This
problem was solved in the subsequent synthesis cycle where
solubilities of the compounds were improved (Table 2).
As mentioned, we selected to optimize a scaffold based on
compound 2, although it has slightly lower ligand efficiency
than 3. The main reason was that 2 possesses better
geometrical vectors for expansion in the pocket. In addition,
11 was later optimized further into compound 17, which is a 5
times more potent inhibitor of TNKS2. The more potent
compounds prepared in this study (14−17) all show a binding
mode highly similar to that of 11, and except for further
exploiting the halo−O(sp2) interactions with Tyr1071, they
primarily introduce interactions in the exit region and solvent-
immersed solubilizing functions. Because of these similarities, it
is possible that the affinity gains (14−17) may be achievable
with a maintained selectivity profile; however, this needs to be
studied further, but it is clear that these compounds offer a new
class of very PARP selective TNKS inhibitors.
Surprisingly, 11 also showed more than 16-fold selectivity for
TNKS2 over TNKS1 (Table 4). Tankyrase inhibition has
generally been studied as concurrent inhibition of both TNKS1
and TNKS2, and the knowledge of the specific contribution
from the two enzymes is unclear. As there are no previous
potent and intratankyrase-selective inhibitors, the pharmaco-
logical utility of compounds with these properties has not been
explored. TNKS1 and TNKS2 often appear redundant19 in
knockout experiments, and data from Huang et al. indicate that
simultaneous inhibition of TNKS1 and TNKS2 is required to
increase the Axin levels.6 The potential to use 11 and the more
potent analogues as inhibitors of Wnt signaling therefore needs
to be further investigated. On the other hand, this specificity
4502
dx.doi.org/10.1021/jm400211f | J. Med. Chem. 2013, 56, 4497−4508

Journal of Medicinal Chemistry
Article
the binding mode of 2 was less overlapping with 1 and hence
could provide more novel compounds.
6b). By reducing the conformational flexibility in solution, the
entropic penalty upon binding to the protein is reduced. This is
similar to a concept used in the design of protein tyrosine
phosphatase-1B inhibitors25 and inhibitors reviewed for other
proteins.15
It is interesting to note that compound 3, a fragment not
selected for expansion, is identical to the nicotinamide
anchoring moiety of 18 (PDB 3U9Y),12 and the binding
conformation of this part of 18 and 3 can be completely
superimposed (see the Supporting Information). 18 is primarily
a PARP1−4 inhibitor (Table 4) and shows little interaction and
stabilization with TNKS.22 18 is further extended using the
vector occupied by a chlorine atom in 3, an extension not
structurally compatible (without perturbing the protein) in the
crystal structure of 3. As expected, the binding of 18 in TNKS2
invokes a conformational change in the D-loop, required to
accommodate the rest of the compound. Although this shows
that there is some degree of plasticity of the TNKS2 pocket, it
also shows the risks of disturbing these residues, as this
potentially is the reason for the lower affinity 18 shows toward
TNKS2.
The compounds reported in this work are unique to
previously reported TNKS and PARP inhibitors but have
some similarities in binding mode. All earlier reported
compounds, where TNKS structures are available for
protein−ligand complexes, have their major interactions in
the nicotinamide pocket, with the exception of 3-(4-
methoxyphenyl)-5-({[5-methyl-4-(4-methylphenyl)-4H-1,2,4-
triazol-3-yl]sulfanyl}methyl)-1,2,4-oxadiazole8 (PDB 3UDD)
and 4-(1,3,3a,4,7,7a-hexahydro-1,3-dioxo-4,7-methano-2H-iso-
indol-2-yl)-N-8-quinolinyl-benzamide12 (IWR1, PDB 3UA9),
which have major interactions in the adenine pocket with minor
interactions in the nicotinamide pocket. One distinguishing
feature of previous compounds binding in the nicotinamide-
binding pocket is that they expanded further down into the
pocket (toward Glu1138) as compared to the compounds of
the present work, for example, 1 (Figure 7), but also 6(5H)-
In contrast, the X-ray structures reveal that the orientation of
the 2-hydroxy-4-methylquinoline scaffold is highly conserved
and acts as an anchor all the way from compound 2, the original
fragment hit, to the more leadlike compounds 11 and 17
(Figure 6). This is in line with what other investigators have
Figure 6. Overlay of the bound conformations extracted from
protein−ligand crystal structures of the original fragment hit 2
(magenta), 11 (yellow), 15 (green), 16 (purple), and 17 (orange)
showing that (a) the binding mode of the 2-hydroxyquinoline scaffold
is conserved throughout the expansion of the fragment and (b) the
biphenyl C−C−C−C torsion angles of the bound conformers are
close to 51°.
Figure 7. Comparison of the binding modes of 1 (PDB 3KR8) and 17
to TNKS2 based on crystal structures. Schematic (a) and molecular
surface representation (b) overlay of the compounds interacting with
the pocket. 1 is depicted in green and a representative of our series
(17) in orange. To further illustrate the differences and similarities in
binding modes, a molecular surface was created using the same colors.
reported for other fragment expansions previously23 and lends
support to the initial fragment 2 performing high-quality
interactions. Also, the protein conformation is very conserved
during fragment evolution, suggesting limited loss of binding
energy due to protein conformational entropic costs upon
binding.
The final compounds have several interesting properties that
differentiate them from previous TNKS inhibitors. For
example, substituents in ortho-positions can in addition to
sampling direct interactions (Cl−carbonyl O(sp2), water
displacement in this case) also contribute significantly to
affinity by altering the conformational preference of the
compound by shifting the biaryl torsion from 30° to 40° for
non-ortho-substituted derivatives vs 55° for ortho-substituted
derivatives.24 In effect, the ligands are preorganized in the
binding conformation, and the observed biaryl torsion for the
binding conformation in the crystal structures of all ortho-
substituted compounds prepared is very close to 51° (Figure
phenanthridinone. This part of the pocket is highly conserved
in both the TNKS and PARP families, and interactions in this
pocket are a characteristic of several of the compounds with
lower specificity between TNKS and PARP (primarily PARP1−
4). The pocket contains the glutamate (Glu1138 in TNKS2)
conserved in all family members that possess PARP activity,
and our analysis suggests that none of the known compounds
interacting in this pocket of TNKS (including our compound
2) have managed to establish polar interactions with this
carboxylate moiety.
Of the previously reported compounds, 1 stands out as a
potential high-quality close-to-lead candidate, while other
compounds suffer from either limited affinity or unsatisfactory
LE. Although the compounds and binding modes in general are
very different, the central interactions in the nicotinamide-
binding pocket have similarities in 1 and our series; both make
4503
dx.doi.org/10.1021/jm400211f | J. Med. Chem. 2013, 56, 4497−4508

Journal of Medicinal Chemistry
Article
a π-bonded interaction with a protein-bound water (Figure 7).
In our series this is by the second ring of the quinolinone
moiety, while for 1 it is accomplished by the (trifluoromethyl)-
aryl ring. As discussed above, the cyclic thioether moiety of 1
makes interactions in the Glu1138 pocket including the residue
Phe1061 (3.58 Å), while this space is occupied by a water
molecule in, for example, the structure with compound 15
(CH3−H2O = 3.1 Å, H2O−Glu1138 = 3.1 Å, H2O−Phe1061 =
2.8 Å). Although 1 fits nicely in this pocket, it does not make
polar interactions with Glu1138. The most striking differences
are seen at the other end of the compound, where in our series
a chlorine atom explores a novel interaction with Tyr1071,
substituting a water molecule present in other inhibitor
structures. This is enabled by the nonplanar geometry (torsion
is near 51°) between the hydroxyquinoline moiety and the o-
halo-substituted ring seen in all our structures.
Superposition of TNKS2 in complex with compound 11
onto PARP2 (PDB 3KJD) reveals that the amino acids
responsible for the interactions around the 2-hydroxy-4-
methylquinoline part of the compound are well conserved.
However, a patch of much less conserved hydrophobic residues
is apparent in the o-chloro-binding site: Pro1034, Phe1035, and
the aromatic ring of Tyr1050 in the D-loop (Figure 8). The D-
nicotinamide-pocket-anchored fragments which were grown to
exploit a novel chloro−O(sp2) interaction with the main chain
carbonyl oxygen of Tyr1071. In the final rounds of chemistry,
additional efforts were made to improve the physicochemical
properties of the compounds in the series and a set of
solubility-improving modifications were introduced to generate
compounds that have high affinity, good physicochemical
properties, and a unique selectivity profile.
EXPERIMENTAL SECTION
■
Protein Expression and Purification. The TNKS2 PARP
domain (amino acids 947−1162) was produced in Escherichia coli
and purified using Ni chromatography as previously described.18 This
protein construct was used for biophysical characterization (DSF and
SPR) and X-ray crystallography.
Fragment Screening and Biophysics. Compound Library. A
500 compound fragment library was acquired from Maybridge (RO3
500 library 2009 Maybridge (Trevillett, Tintagel, Cornwall, U.K.)).
Differental Scanning Flourometry. A DSF assay was used to probe
compound effects on TNKS2 thermal stability. The protein was
diluted to 0.2 mg/mL in buffer containing (1:1000, 5×) SYPRO
Orange (Invitrogen). We added 0.5 μL of either compound solution
or DMSO control, making the final volume 25 μL in all wells. Melt
analysis was performed on an ICycler IQ (Bio-Rad) real-time PCR
instrument with filters set up for detection of SYPRO Orange (λexcitation
= 490 nm, λemission = 575 nm) fluorescence. The plates were covered in
optically clear tape before initiation of the experiment, where a
temperature interval of 25−80 °C was scanned at a ramp rate of
approximately 2 °C/min.
For fragment screening, the compounds were added from stocks in
50 mM DMSO, giving an end compound concentration of 1 mM and
a DMSO concentration of 2%. The compounds were plated in a layout
of 80 fragments and 16 DMSO controls for each 96-well plate.
Analysis of the screen data was primarily done by ocular inspection of
individual curves, and fitting Tm to sigmoidal transitions where
applicable was performed using an Excel script.26 For dose−response
data and DSF assay on expansion compounds made at various
concentrations, dilution of the compounds was done in 100% DMSO
at 50× assay concentration. For data generated with in situ digestion of
TNKS2, trypsin, 1:100 (m/m) vs TNKS2, was added to the protein
buffer dye mixture just prior to the DSF experiment.
Figure 8. Binding of compound 11 (yellow) and compound 17
(orange) to TNKS2 (blue) compared to PARP2 (gray) showing the
differences/similarities in substrate-binding area between the proteins.
The amino acids in TNKS2 are marked in gray and those in PARP2 in
orange.
DSF in Situ Digestion of TNKS2. To try to simplify the
interpretation of the thermal shift data, we decided to investigate
the use of in situ trypsination in combination with DSF. In situ
trypsination is commonly used in crystallography as a way to promote
crystallization,27 presumably by removing less ordered and well-folded
contaminants. It has previously been described to work well on
TNKS2 for crystallization purposes.18 Chymotrypsin (1/100 TNKS2
mass equivalents) was added to the protein buffer dye mixture just
prior to the DSF experiment. This provided much cleaner curves,
which substantially simplified the interpretation and fitting of the
weakly shifting original fragment hits.
Surface Plasmon Resonance. All measurements were performed
using a Biacore T200 (GE Healthcare) using a running buffer of 20
mM PBS, pH 7.4, 0.05% Tween20, and 2% DMSO. TNK2 was
immobilized using a capture-coupling protocol very similarly to what
has previously been described.28 Briefly, TNK2−6HIS (20 μg/mL) in
20 mM PBS buffer, pH 7.4, was introduced with a flow rate of 10 μL/
min to a Ni2+ and NHS doubly preactivated NTA chip (NTA-S
Biacore BR-1005-32) with a response target of 5000 RU. This
procedure led to a stable and robust surface of about 4600 RU, which
was used for further characterization.
loop in TNKS is shorter than in PARP1−4.20,22 In addition, in
the structure of TNKS2, the loop containing Ile1075 closes
down much more onto the substrate-binding site than the
equivalent amino acids in PARP2.
There are also some hydrophilic residues from the α-helix-5
from the regulatory domain in the PARP2 structure near the
NAD+ cleft opening (this domain is not present in TNKS2). In
the superposition of the proteins, these hydrophilic amino acids
reach into the binding site and clash with the compounds in the
TNKS2 structures.
Overall, this region of the binding cavity in TNKS2 contains
more hydrophobic residues than in PARP2 and other members
of the PARP family.20,22 This could also describe the selectivity
of compound 11 as an explicit TNKS inhibitor.
CONCLUSION
■
In conclusion, a new class of potent TNKS inhibitors has been
developed using a TSA-driven fragment-based approach where
major expansions were guided by crystal structures and the
activity improvments were monitored throughout the process.
Key compounds were further characterized using SPR to
establish that TSA and IC50 values correlated with an
improvement of Kd. Chemical expansion was started from
For the kinetic experiments, the T200 standard method LMW
kinetic was used with a contact time 60 s and a dissociation time of
3000 s at a flow rate of 30 μL/min. For more details on sample
preparation and the experimental layout, see the Supporting
Information. Analysis of the SPR data was performed using Biacore
T200 evaluation software and Scrubber2.29
4504
dx.doi.org/10.1021/jm400211f | J. Med. Chem. 2013, 56, 4497−4508

Journal of Medicinal Chemistry
Article
Crystallography. TNKS2 crystals were obtained from chymotryp-
sin-cleaved TNK at a concentration of 10.5 mg/mL in hanging drops
at 4 °C in 0.1 M Tris−HCl, pH 8.5, 15−20% PEG3350, and 0.2 M
lithium sulfate as previously described.18 The crystals were then
transferred to a solution containing a 0.5 mM (5% DMSO)
concentration of the ligand of interest and soaked for about 2 h up
to overnight before they were flash frozen in liquid nitrogen. Glycerol
(20%) was used as a cryoprotectant.
Diffraction data for the complexes were collected to resolutions
between 1.9 and 2.4 Å on the BL13B1 beamline (ADSC Quantum-315
CCD detector) at the National Synchrotron Radiation Research
Center (NSRRC), Taiwan, ROC, as well as on the MX1 beamline at
the Australian Synchrotron, Victoria, Australia.
(d, J = 7.60 Hz, 1 H), 7.32−7.23 (m, 2 H), 3.58 (s, 2 H), 2.19 (s, 3 H);
MS (ESI) m/z [C10H10BrNO2 + H]+ 257.
7-Bromo-4-methyl-1,2-dihydroquinolin-2-one (5). A mixture of 4
(6.3 g, 24.6 mmol) and concentrated sulfuric acid (30 mL) was heated
to 120 °C and stirred for 2 h. Upon completion of reaction, the
reaction mixture was cooled and poured into ice. The resulting
precipitate was filtered and washed with water and ether. The crude
material was then recrystallized from methanol to give 5 as a colorless
1
solid (4 g, yield 68%): H NMR (400 MHz, DMSO-d6) δ (ppm)
11.65 (br s, 1 H), 7.65 (d, J = 8.53 Hz, 1 H), 7.48 (d, J = 2.01 Hz, 1
H), 7.35 (dd, J = 8.53, 1.88 Hz, 1 H), 6.43 (s, 1 H), 2.40 (d, J = 0.88
Hz, 3 H); MS (ESI) m/z [C10H8BrNO + H]+ 239.
General Synthetic Route for Compounds 7−15. A solution of 5
(0.21 mmol) in 4:1 dioxane/water (2.1 mL) was added to the
corresponding boronic acid (0.42 mmol). Potassium carbonate (0.53
mmol) was added to the reaction mixture, followed by tetrakis-
(triphenylphosphine)palladium (0.02 mmol). The reaction mixture
was heated to reflux and allowed to stir overnight. Upon completion,
the reaction mixture was cooled to room temperature, and the solvent
was removed by rotary evaporation. The residue was poured into
water and extracted with dichloromethane. The organic layer was dried
over sodium sulfate and filtered, and the solvent was removed by
rotary evaporation. The crude material was then recrystallized from
ethyl acetate/dichloromethane/hexane to give a colorless solid.
4-Methyl-6-phenyl-1,2-dihydroquinolin-2-one (7). 7 was prepared
according to the general procedure using commercially available 6-
bromo-4-methyl-1,2-dihydroquinolin-2-one and 4,4,5,5-tetramethyl-2-
phenyl-1,3,2-dioxaborolane as reactants. The crude material was
recrystallized from ethyl acetate/dichloromethane/hexane to give 7
The crystals belonged to space group C2221 (unit cell parameters a
= 94 Å, b = 94 Å, and c = 116 Å) with 2 molecules/ASU or P212121
(unit cell parameters a = 67 Å, b = 67 Å, and c = 118 Å) with 4
molecules/ASU.
At the Australian synchrotron, Blu-Ice30 was used for data collection
and processing. All the other data were processed using HKL200031
and mosflm,32 and the structures were solved with molecular
replacement using Phaser33 from the CCP4 program suite, with the
apo structure (PDB 3KR7) as a model.34 Model building was done
using COOT35 and refinement using Refmac.36 The ligand CIF
libraries were generated using Corina,37 and the ligand, Zn ion, and
some sulfate ions were fitted into the electron density using COOT
and refined further.
Details on data processing and refinement statistics are given in the
Supporting Information, Table S2.
Pymol38 and MOE2011.1039 were used for constructing the
structure figures.
1
as a colorless solid (yield 81%, mp 266−268 °C): H NMR (400
Chemistry. General Procedures. All reagents were purchased from
commercial sources and used as received. Reaction progress was
monitored by TLC using Merck silica gel 60 F254 on glass plates with
detection by UV at 254 nM. LC−MS analysis was carried out with a
Shimadzu LC-20AD and LCMS-2020. The column used was a
Phenomenex Kinetex 2.6 μm, 50 × 2.10 mm). Proton nuclear
magnetic resonance (1H NMR) spectra were obtained using a Bruker
Ultrashield 400 PLUS/R system, operating at 400 MHz. All resonance
bands were referenced to tetramethylsilane (internal standard).
Splitting patterns are indicated as follows: s, singlet; d, doublet; t,
triplet; m, multiplet; br, broad peak. The compounds’ purities were
≥95% determined by a VARIAN ProStar HPLC instrument. Melting
points were determined in Pyrex capillary tubes using a StuartnSMP30
melting point apparatus.
Thermodynamic Solubility Study. A 6 μL volume of 50 mM
DMSO stock from the stock plate is added to the reaction deep well
plate containing 600 μL of pH 4.0 pION buffer, mixed, and incubated
for 18 h. The plate is sealed well during the incubation process. The
test compound concentration is 500 μM. At the end of the incubation
period, 100 μL of sample from the storage plate is vacuum filtered
using a filter plate. This step wets the filters, and the filtrate is
discarded. Another 200 μL of the sample from the deep well plate is
vacuum filtered using the same filter block but a clean filter plate. A 75
μL volume of the filtrate from the filter plate is transferred to a UV
sample plate. A 75 μL volume of 1-propanol is added to this UV plate.
The solution is mixed, and the spectrum is read using the UV
spectrophotometer (Spectramax-Molecular Devices). The analysis is
carried out by using pION μSOL EXPLORER software, version 3.3.
Synthesis. N-(3-Bromophenyl)-3-oxobutanamide (4). A solution
of ethyl acetoacetate (10 g, 58.1 mmol) in dry toluene (58 mL) was
added to 3-bromoaniline (12.1 g, 93 mmol) in a dropwise manner in a
round-bottomed flask, and the reaction mixture was refluxed
overnight. Upon completion of reaction, the reaction mixture was
cooled and quenched with sodium carbonate solution. The aqueous
fraction was extracted with dichloromethane; the organics were dried
over sodium sulfate and concentrated in vacuo to give a brown residue.
Hexane was added to the residue, and the flask was cooled with ice to
aid precipitation of the product. A light brown precipitate of 4 (6.3 g,
yield 42%) was obtained after filtration via a Buchner funnel: 1H NMR
(400 MHz, DMSO-d6) δ (ppm) 10.26 (br s, 1 H), 7.96 (s, 1 H), 7.46
MHz, chloroform-d) δ (ppm) 11.45 (br s, 1 H), 7.86 (d, J = 1.60 Hz, 1
H), 7.76 (d, J = 1.60 Hz, 1 H), 7.74 (d, J = 1.6 Hz, 2 H), 7.68−7.46
(m, 3 H), 7.43 (d, J = 8.4 Hz, 1 H), 7.0 (s, 1 H), 2.52 (s, 3 H); MS
(ESI) m/z [C16H13NO + H]+ 236.
4-Methyl-7-phenyl-1,2-dihydroquinolin-2-one (8). 8 was prepared
according to the general procedure using 5 and commercially available
4,4,5,5-tetramethyl-2-phenyl-1,3,2-dioxaborolane as reactants. The
crude material was recrystallized from ethyl acetate/dichloro-
methane/hexane to give 8 as a colorless solid (yield 47%, mp 290−
1
292 °C): H NMR (400 MHz, chloroform-d) δ (ppm) 10.38 (br s, 1
H), 7.74 (d, J = 8.4 Hz, 1 H), 7.65 (d, J = 7.60 Hz, 2 H), 7.55−7.43
(m, 5 H), 6.56 (s, 1 H), 2.52 (s, 3 H); MS (ESI) m/z [C16H13NO +
H]+ 236.
7-(4-Fluorophenyl)-4-methyl-1,2-dihydroquinolin-2-one (9). 9
was prepared according to the general procedure using 5 and
commercially available 2-(4-fluorophenyl)-4,4,5,5-tetramethyl-1,3,2-
dioxaborolane as reactants. The crude material was recrystallized
from dichloromethane/methanol/hexane to give 9 as a colorless solid
1
(yield 34%, mp 255−257 °C): H NMR (400 MHz, chloroform-d) δ
(ppm) 10.51 (br s, 1 H), 7.73 (d, J = 8.0 Hz, 1 H), 7.63 (dd, J = 3.2,
6.4 Hz, 2 H), 7.48 (d, J = 8.4 Hz, 1 H), 7.44 (s, 1H), 7.18 (m, 2 H),
6.57 (s, 1 H), 2.53 (s, 3 H); MS (ESI) m/z [C16H12FNO + H]+ 254.
7-(3-Fluorophenyl)-4-methyl-1,2-dihydroquinolin-2-one (10). 10
was prepared according to the general procedure using 5 and
commercially available (3-fluorophenyl)boronic acid as reactants. The
crude material was recrystallized from dichloromethane/methanol/
hexane to give 10 as a colorless solid (yield 20%, mp 264−266 °C): 1H
NMR (400 MHz, chloroform-d) δ (ppm) 10.51 (br s, 1 H), 7.75 (d, J
= 8.4 Hz, 1 H), 7.48−7.27 (m, 5 H), 7.13−7.09 (m, 1 H), 6.58 (s, 1
H), 2.53 (s, 3 H); MS (ESI) m/z [C16H12FNO + H]+ 254.
7-(2-Fluorophenyl)-4-methyl-1,2-dihydroquinolin-2-one (11). 11
was prepared according to the general procedure using 5 and
commercially available (2-fluorophenyl)boronic acid as reactants. The
crude material was recrystallized from dichloromethane/methanol/
hexane to give 11 as a colorless solid (yield 20%, mp 285−287 °C): 1H
NMR (400 MHz, chloroform-d) δ (ppm) 9.77 (br s, 1 H), 7.74 (d, J =
8.4 Hz, 1 H), 7.50−7.39 (m, 4 H), 7.27−7.17 (m, 2 H), 6.57 (s, 1 H),
2.52 (s, 3 H); MS (ESI) m/z [C16H12FNO + H]+ 254.
7-(2-Chlorophenyl)-4-methyl-1,2-dihydroquinolin-2-one (12). 12
was prepared according to the general procedure using 5 and
4505
dx.doi.org/10.1021/jm400211f | J. Med. Chem. 2013, 56, 4497−4508

Journal of Medicinal Chemistry
Article
commercially available (2-chlorophenyl)boronic acid as reactants. The
crude material was recrystallized from dichloromethane/methanol/
hexane to give 12 as a colorless solid (yield 49%, mp 275−277 °C): 1H
NMR (400 MHz, chloroform-d) δ (ppm) 10.10 (br s, 1H), 7.73 (d, J
= 8.28 Hz, 1 H), 7.48−7.54 (m, 1 H), 7.31−7.40 (m, 4 H), 7.29 (s, 1
H), 6.58 (s, 1 H), 2.54 (s, 3 H); MS (ESI) m/z [C16H12ClNO + H]+
270.
7-(2-Methoxyphenyl)-4-methyl-1,2-dihydroquinolin-2-one (13).
13 was prepared according to the general procedure using 5 and
commercially available (2-methylphenyl)boronic acid as reactants. The
crude material was recrystallized from dichloromethane/methanol/
hexane to give 13 as a colorless solid (yield 27%, mp 270−273 °C): 1H
NMR (400 MHz, chloroform-d) δ (ppm) 11.50 (br s, 1H), 7.71 (d, J
= 8.4 Hz, 1 H), 7.42−7.32 (m, 4 H), 7.09−7.00 (m, 2 H), 6.53 (s, 1
H), 3.84 (s, 3 H), 2.51 (s, 3 H); MS (ESI) m/z [C17H15NO2 + H]+
266.
H), 3.28 (s, 3 H), 2.46 (s, 3 H); MS (ESI) m/z [C20H19ClN2O3+ H]+
371.
Biological Assays. TNKS2 DNA Cloning and Protein Production.
The gene encoding the PARP domain of TNKS2 (amino acids 934−
1166) was synthesized (GenScript, Piscataway, NJ) with EcoRI and
SalI sites at the 5′ and 3′ ends of the gene to allow in-frame subcloning
into the expression vector pGEX-6P-1.
Gene expression in BL21(DE3) cells (Merck, Germany) was
performed in the presence of 0.5 mM IPTG at 18 °C for 18 h. The
recombinant protein was purified by immobilized affinity chromatog-
raphy (IMAC) on the Profinia protein purification system (Bio-Rad,
Hercules, CA). The protein was eluted in elution buffer containing 20
mM glutathione. The purified protein was concentrated using a 10 mL
concentrator with cellulose membrane, 10 kDa NMWL. By successive
30 min centrifugation, the glutathione buffer was exchanged for 0.1 M
Tris buffer, pH 8.0. The protein was concentrated to about 1.1 mg/
mL. The amount of purified protein obtained from a 1 L induction was
8.9 mg. The resultant protein has a purity of ∼91% as determined by
the 2100 bioanalyzer (Agilent, Santa Clara, CA).
The auto-PARsylation reactions were carried out in 40 μL volumes
in the presence of the compound (concentration varying from 0.006 to
100 μM, 2.5% DMSO), 20 nM GST-TNKS2, and 250 μM NAD+
(Sigma-Aldrich). The reactions were incubated at room temperature
for 2 h and then quenched by adding 10 μL of 20% formic acid. Then
100 μL of acetonitrile was added, and the samples were centrifuged for
30 min at 3500 rpm and 4 °C. The supernatant was transferred to a
new plate and subjected to the LC/MS analysis.
7-(4-Amino-2-chlorophenyl)-4-methylquinolin-2(1H)-one (14).
14 was prepared according to the general procedure using
commercially available 6-bromo-4-methyl-1,2-dihydroquinolin-2-one
and 3-chloro-4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)aniline as
reactants. The crude material was recrystallized from ethyl acetate/
dichloromethane/hexane to give 14 as a colorless solid (yield 33%, mp
1
264−266 °C): H NMR (400 MHz, chloroform-d) δ (ppm) 9.71 (br
s, 1 H), 7.69 (d, J = 8.4 Hz, 1 H), 7.76 (dd, J = 8.4 Hz, 1.60 Hz, 1 H),
7.31 (d, J = 1.6 Hz, 2 H), 7.22 (s, 1 H), 7.16 (d, J = 8.4 Hz, 1 H), 6.81
(s, J = 2.4 Hz, 1 H), 6.65 (dd, J = 8 Hz, 2.4 Hz, 1 H), 6.54 (s, 1 H),
2.51 (s, 3 H); MS (ESI) m/z [C16H13NO + H]+ 285.
Analytical Method. HPLC−MS/MS analysis was carried out using
an Agilent 1200RRLC series HPLC system consisting of a binary
pump, vacuum degasser, and column switching valve with an
autosampler equipped with a 40 μL sample loop interfaced with a
TSQ quantum ultra triple-stage quadrupole tandem mass spectrometer
(Thermo Scientific). Both systems were controlled by either XCalibur
or LCQuan software. Nicotinamide was separated onto a Phenom-
enexKinetexHILIC column (50 × 2.1 mm, 2.6 μM) using mobile
phases (mobile phase A, 0.1% formic acid in Milli-Q water with 10
mM ammonium formate; mobile phase B, 0.1% formic acid in
acetonitrile). The isocratic LC method was run at 0.4 mL/min with
70% mobile phase B with a run time of 2 min.
The mass spectrometer was operated with positive ion detection for
nicotinamide. The selected reaction monitoring (SRM) transition for
nicotinamide was m/z 123 → 79.7. The vaporization temperature was
maintained at 400 °C, and a voltage of 4 kV was applied to the sprayer
needle. Nitrogen gas was used as the sheath gas and auxiliary gas. The
argon gas was used for the collision energy. The detection and relative
quantification of analytes were performed using the SRM mode.
Data Processing/Analysis. The chromatogram data were acquired
by LCQuan software. The raw data from each sample were processed
by a processing method for each batch of sequences. The
quantification of an individual component as a peak area was done
in arbitrary units. The area of the peak corresponding to nicotinamide
on the MS chromatogram was plotted against the log of compound
concentrations. The experimental data were analyzed by nonlinear
regression (GraphPad Prism).
3-Chloro-4-(4-methyl-2-oxo-1,2-dihydroquinolin-7-yl)benzoic
Acid (15). 15 was prepared according to the general procedure using 5
and commercially available 3-chloro-5-(tetramethyl-1,3,2-dioxaboro-
lan-2-yl)benzoic acid as reactants. Upon completion, the reaction was
filtered through Celite and washed with methanol. The filtrate was
acidified to pH 2, extracted with isopropyl alcohol/dichloromethane
(1:7), dried over magnesium sulfate, and evaporated to dryness in
vacuo. The crude material was recrystallized from dichloromethane/
methanol/hexane. 15 was obtained as a colorless solid (yield 33%, mp
> 370 °C): 1H NMR (400 MHz, D2O) δ (ppm) 11.68 (s, 1H), 8.02 (s,
1 H), 7.94 (d, J = 7.6 Hz, 1 H), 7.81(d, J = 8 Hz, 1 H), 7.51 (d, J = 7.6
Hz, 1 H), 7.38 (d, J = 1.6 Hz), 7.26 (d, J = 7.6 Hz, 1H), 6.45 (s, 1 H),
2.47 (s, 3 H); MS (ESI) m/z [C17H12ClNO3 + H]+ 314.
General Synthesis Route for Compounds 16 and 17. A solution of
15 (0.064 mmol) in N,N-dimethylformamide (1 mL) was added to
N,N-diisopropylethylamine (0.03 mL) and 2-(1H-7-azabenzotriazol-1-
yl)-1,1,3,3-tetramethyluronium hexafluorophosphate methanaminium
(0.14 mmol) and the resulting solution stirred for 10 min before
addition of the corresponding amine (0.127 mmol). The solution was
left to stir at room temperature for 2 h. Upon completion, the reaction
was extracted with dichloromethane, dried over magnesium sulfate,
and evaporated to dryness in vacuo. The crude material was purified
by recrystallization with methanol to afford a white solid.
3-Chloro-4-(4-methyl-2-oxo-1,2-dihydroquinolin-7-yl)-N-[2-(mor-
pholin-4-yl)ethyl]benzamide (16). 16 was prepared according to the
general procedure using 15 and commercially available 4-(2-
aminoethyl)morpholine as reactants. The crude material was purified
by recrystallization with methanol to afford 16 as a white solid (yield
Computational Tools. Instant JChem 5.9.0 was used for structure
database management, search, and prediction (ChemAxon, 2012,
http://www.chemaxon.com).
The protein−ligand crystal structures of both initial fragment hits
and the expansions were analyzed and visualized using Lead-IT v2.1
(2012, www.Biosolveit.com/LeadIT) and MOE (Molecular Operating
Environment), version 2011.10 (Chemical Computing Group,
Montreal, Canada, http://www.chemcomp.com/software.htm).
1
12%, mp 165−169 °C): H NMR (400 MHz, DMSO-d6) δ (ppm)
11.68 (s, 1 H), 8.62 (t, J = 5.3 Hz, 1 H), 8.04 (d, J = 1.5 Hz, 1 H), 7.90
(d, J = 8.1 Hz, 1H), 7.81 (d, J = 8.4 Hz, 1 H), 7.56 (d, J = 8.0 Hz, 1
H), 7.36 (s, 1 H), 2.43 (m, 4 H), 7.27 (dd, J = 1.6, 8.0 Hz, 1 H), 6.46
(s, 1 H), 3.58 (m, 5 H), 3.42 (m, 3 H), 2.47 (s, 3 H); MS (ESI) m/z
[C23H24ClN3O3+ H]+ 426.
3-Chloro-N-(2-methoxyethyl)-4-(4-methyl-2-oxo-1,2-dihydroqui-
nolin-7-yl)benzamide (17). 17 was prepared according to the general
procedure using 15 and commercially available 2-methoxyethylamine
as reactants. The crude material was purified by recrystallization with
methanol to afford 17 as a white solid (yield 21%, mp 257−158 °C):
1H NMR (400 MHz, DMSO-d6) δ (ppm) 11.68 (s, 1 H), 8.74 (t, J =
5.1 Hz, 1 H), 8.06 (d, J = 1.6 Hz, 1 H), 7.93 (dd, J = 1.6, 8.0 Hz, 1 H),
7.81 (d, J = 8.2 Hz, 1 H), 7.55 (d, J = 8.0 Hz, 1 H), 7.37 (d, J = 1.4 Hz,
1 H), 7.27 (dd, J = 1.6, 8.3 Hz, 1 H), 6.46 (s, 1 H), 3.47−3.48 (m, 4
ASSOCIATED CONTENT
■
S
* Supporting Information
Table with SPR data for on- and off-rates, table of
crystallographic details for all X-ray structures, figure showing
the superpositioning of compounds 3 and 18 binding to
TNKS2, list of less successful modifications of 2 and related
4506
dx.doi.org/10.1021/jm400211f | J. Med. Chem. 2013, 56, 4497−4508

Journal of Medicinal Chemistry
Article
Bouwmeester, T.; Porter, J. A.; Bauer, A.; Cong, F. Tankyrase
inhibition stabilizes axin and antagonizes Wnt signalling. Nature 2009,
461, 614−620.
(7) Waaler, J.; Machon, O.; Tumova, L.; Dinh, H.; Korinek, V.;
Wilson, S. R.; Paulsen, J. E.; Pedersen, N. M.; Eide, T. J.; Machonova,
O.; Gradl, D.; Voronkov, A.; von Kries, J. P.; Krauss, S. A novel
tankyrase inhibitor decreases canonical Wnt signaling in colon
carcinoma cells and reduces tumor growth in conditional APC mutant
mice. Cancer Res. 2012, 72, 2822−2832.
compounds, dose−response data on fragment hits 2 and 3, and
table on methods used for PARP1 activity assays. This material
is available free of charge via the Internet at http://pubs.acs.org.
Accession Codes
The crystal structures of compounds 2, 3, 11, 14, 15, 16, and
17 in complex with TNKS2 have been deposited with the
RCSB Protein Data Bank under the accession codes 4JIZ,
3W5I, 41UE, 4J21, 4J22, 4J3L, 4J3M.
(8) Shultz, M. D.; Kirby, C. A.; Stams, T.; Chin, D. N.; Blank, J.;
Charlat, O.; Cheng, H.; Cheung, A.; Cong, F.; Feng, Y.; Fortin, P. D.;
Hood, T.; Tyagi, V.; Xu, M.; Zhang, B.; Shao, W. [1,2,4]Triazol-3-
ylsulfanylmethyl)-3-phenyl-[1,2,4]oxadiazoles: antagonists of the Wnt
pathway that inhibit tankyrases 1 and 2 via novel adenosine pocket
binding. J. Med. Chem. 2012, 55, 1127−1136.
AUTHOR INFORMATION
Corresponding Author
*E-mail: andreas.larsson@ntu.edu.sg, +65 65869756 (E.A.L.);p
nordlund@ntu.edu.sg, +65 65869693(P.N.).
■
Author Contributions
(9) Chen, B.; Dodge, M. E.; Tang, W.; Lu, J.; Ma, Z.; Fan, C.-W.;
Wei, S.; Hao, W.; Kilgore, J.; Williams, N. S.; Roth, M. G.; Amatruda, J.
F.; Chen, C.; Lum, L. Small molecule-mediated disruption of Wnt-
dependent signaling in tissue regeneration and cancer. Nat. Chem. Biol.
2009, 5, 100−107.
∥E.A.L. and A.J. contributed equally to this work.
Notes
The authors declare no competing financial interest.
(10) Peterson, R. T. Drug discovery: propping up a destructive
regime. Nature 2009, 461, 599−600.
ACKNOWLEDGMENTS
■
Portions of this research were carried out at the National
Synchrotron Radiation Research Center, a national user facility
supported by the National Science Council of Taiwan, Republic
of China. The Synchrotron Radiation Protein Crystallography
Facility is supported by the National Research Program for
Genomic Medicine. This research was also undertaken on the
MX1 beamline at the Australian Synchrotron, Victoria,
Australia. We thank the beamline staff at both the synchrotrons
for all their help during data collection. This work was
supported by funding from a principal investigator start-up
grant from Nanyang Technological University, the National
Research Foundation (Grant NRF-CRP4-2008-02), and the
Biomedical Research Council.
(11) Ferraris, D. V. Evolution of poly(ADP-ribose) polymerase-1
(PARP-1) inhibitors. From concept to clinic. J. Med. Chem. 2010, 53,
4561−4584.
(12) Narwal, M.; Venkannagari, H.; Lehtio, L. Structural basis of
̈
selective inhibition of human tankyrases. J. Med. Chem. 2012, 55,
1360−1367.
(13) Niesen, F. H.; Berglund, H.; Vedadi, M. The use of differential
scanning fluorimetry to detect ligand interactions that promote protein
stability. Nat. Protoc. 2007, 2, 2212−2221.
(14) Abad-Zapatero, C.; Metz, J. T. Ligand efficiency indices as
guideposts for drug discovery. Drug Discovery Today 2005, 10, 464−
469.
(15) Leung, C. S.; Leung, S. S. F.; Tirado-Rives, J.; Jorgensen, W. L.
Methyl effects on protein-ligand binding. J. Med. Chem. 2012, 55,
4489−4500.
(16) Lommerse, J. P. M.; Stone, A. J.; Taylor, R.; Allen, F. H. The
nature and geometry of intermolecular interactions between halogens
and oxygen or nitrogen. J. Am. Chem. Soc. 1996, 118, 3108−3116.
(17) Menear, K. A.; Adcock, C.; Boulter, R.; Cockcroft, X. L.;
Copsey, L.; Cranston, A.; Dillon, K. J.; Drzewiecki, J.; Garman, S.;
Gomez, S.; Javaid, H.; Kerrigan, F.; Knights, C.; Lau, A.; Loh, V. M.,
Jr.; Matthews, I. T.; Moore, S.; O’Connor, M. J.; Smith, G. C.; Martin,
N. M. 4-[3-(4-Cyclopropanecarbonylpiperazine-1-carbonyl)-4-fluoro-
benzyl]-2H-phthalazin-1-one: a novel bioavailable inhibitor of poly-
(ADP-ribose) polymerase-1. J. Med. Chem. 2008, 51, 6581−6591.
ABBREVATIONS USED
■
FBLD, fragment-based ligand design; SPR, surface plasmon
resonance; DSF, differential scanning fluorometry; NAD+,
nicotinamide adenine dinucleotide; PARP, poly(ADP-ribose)
polymerase; TNKS, tankyrase (TRF1-interacting, ankyrin-
related ADP-ribose polymerase)
REFERENCES
■
(1) Fancy, S. P. J.; Harrington, E. P.; Yuen, T. J.; Silbereis, J. C.;
Zhao, C.; Baranzini, S. E.; Bruce, C. C.; Otero, J. J.; Huang, E. J.;
Nusse, R.; Franklin, R. J. M.; Rowitch, D. H. Axin2 as regulatory and
therapeutic target in newborn brain injury and remyelination. Nat.
Neurosci. 2011, 14, 1009−1016.
(18) Karlberg, T.; Markova, N.; Johansson, I.; Hammarstrom, M.;
̈
Schutz, P.; Weigelt, J.; Schuler, H. Structural basis for the interaction
̈
̈
between tankyrase-2 and a potent Wnt-signaling inhibitor. J. Med.
Chem. 2010, 53, 5352−5355.
(19) Chiang, Y. J.; Hsiao, S. J.; Yver, D.; Cushman, S. W.; Tessarollo,
L.; Smith, S.; Hodes, R. J. Tankyrase 1 and tankyrase 2 are essential
but redundant for mouse embryonic development. PLoS One 2008, 3,
e2639.
(2) D’Amours, D.; Desnoyers, S.; D’Silva, I.; Poirier, G. G.
Poly(ADP-ribosyl)ation reactions in the regulation of nuclear
functions. Biochem. J. 1999, 342 (Part2), 249−268.
(3) Smith, S.; Giriat, I.; Schmitt, A.; de Lange, T. Tankyrase, a
poly(ADP-ribose) polymerase at human telomeres. Science 1998, 282,
1484−1487.
(20) Gunaydin, H.; Gu, Y.; Huang, X. Novel binding mode of a
potent and selective tankyrase inhibitor. PLoS One 2012, 7, e33740.
(21) Murray, C. W.; Rees, D. C. The rise of fragment-based drug
discovery. Nat. Chem. 2009, 1, 187−192.
(4) Rouleau, M.; Patel, A.; Hendzel, M. J.; Kaufmann, S. H.; Poirier,
G. G. PARP inhibition: PARP1 and beyond. Nat. Rev. Cancer 2010, 10,
293−301.
(22) Wahlberg, E.; Karlberg, T.; Kouznetsova, E.; Markova, N.;
̈
Macchiarulo, A.; Thorsell, A.-G.; Pol, E.; Frostell, Å.; Ekblad, T.; Oncu,
̈
(5) Riffell, J. L.; Lord, C. J.; Ashworth, A. Tankyrase-targeted
therapeutics: expanding opportunities in the PARP family. Nat. Rev.
Drug Discovery 2012, 11, 923−936.
(6) Huang, S.-M. A.; Mishina, Y. M.; Liu, S.; Cheung, A.; Stegmeier,
F.; Michaud, G. A.; Charlat, O.; Wiellette, E.; Zhang, Y.; Wiessner, S.;
Hild, M.; Shi, X.; Wilson, C. J.; Mickanin, C.; Myer, V.; Fazal, A.;
Tomlinson, R.; Serluca, F.; Shao, W.; Cheng, H.; Shultz, M.; Rau, C.;
Schirle, M.; Schlegl, J.; Ghidelli, S.; Fawell, S.; Lu, C.; Curtis, D.;
Kirschner, M. W.; Lengauer, C.; Finan, P. M.; Tallarico, J. A.;
D.; Kull, B.; Robertson, G. M.; Pellicciari, R.; Schuler, H.; Weigelt, J.
̈
Family-wide chemical profiling and structural analysis of PARP and
tankyrase inhibitors. Nat. Biotechnol. 2012, 30, 283−288.
(23) Murray, C. W.; Verdonk, M. L.; Rees, D. C. Experiences in
fragment-based drug discovery. Trends Pharmacol. Sci. 2012, 33, 224−
232.
(24) Brameld, K. A.; Kuhn, B.; Reuter, D. C.; Stahl, M. Small
molecule conformational preferences derived from crystal structure
4507
dx.doi.org/10.1021/jm400211f | J. Med. Chem. 2013, 56, 4497−4508

Journal of Medicinal Chemistry
Article
data. A medicinal chemistry focused analysis. J. Chem. Inf. Model. 2008,
48, 1−24.
(25) Black, E.; Breed, J.; Breeze, A. L.; Embrey, K.; Garcia, R.; Gero,
T. W.; Godfrey, L.; Kenny, P. W.; Morley, A. D.; Minshull, C. A.;
Pannifer, A. D.; Read, J.; Rees, A.; Russell, D. J.; Toader, D.; Tucker, J.
Structure-based design of protein tyrosine phosphatase-1B inhibitors.
Bioorg. Med. Chem. Lett. 2005, 15, 2503−2507.
(26) The script is based on the script prepared by Frank Niesen,
SGC Oxford (ftp://ftp.sgc.ox.ac.uk/pub/biophysics), adapted to
integrate XLFIT (http://www.idbs.com/products-and-services/
desktop-and-oem/xlfit-curve-fitting/). Accessed 1 June 2013.
(27) Dong, A.; Xu, X.; Edwards, A. M.; Chang, C.; Chruszcz, M.;
Cuff, M.; Cymborowski, M.; Di Leo, R.; Egorova, O.; Evdokimova, E.;
Filippova, E.; Gu, J.; Guthrie, J.; Ignatchenko, A.; Joachimiak, A.;
Klostermann, N.; Kim, Y.; Korniyenko, Y.; Minor, W.; Que, Q.;
Savchenko, A.; Skarina, T.; Tan, K.; Yakunin, A.; Yee, A.; Yim, V.;
Zhang, R.; Zheng, H.; Akutsu, M.; Arrowsmith, C.; Avvakumov, G. V.;
Bochkarev, A.; Dahlgren, L. G.; Dhe-Paganon, S.; Dimov, S.;
Dombrovski, L.; Finerty, P., Jr.; Flodin, S.; Flores, A.; Graslund, S.;
Hammerstrom, M.; Herman, M. D.; Hong, B. S.; Hui, R.; Johansson,
I.; Liu, Y.; Nilsson, M.; Nedyalkova, L.; Nordlund, P.; Nyman, T.; Min,
J.; Ouyang, H.; Park, H. W.; Qi, C.; Rabeh, W.; Shen, L.; Shen, Y.;
Sukumard, D.; Tempel, W.; Tong, Y.; Tresagues, L.; Vedadi, M.;
Walker, J. R.; Weigelt, J.; Welin, M.; Wu, H.; Xiao, T.; Zeng, H.; Zhu,
H. In situ proteolysis for protein crystallization and structure
determination. Nat. Methods 2007, 4, 1019−1021.
(28) Kimple, A. J.; Muller, R. E.; Siderovski, D. P.; Willard, F. S. A
capture coupling method for the covalent immobilization of
hexahistidine tagged proteins for surface plasmon resonance. Methods
Mol. Biol. 2010, 627, 91−100.
(29) Scrubber2. http://www.biologic.com.au/. Accessed 1 June 2013.
(30) McPhillips, T. M.; McPhillips, S. E.; Chiu, H. J.; Cohen, A. E.;
Deacon, A. M.; Ellis, P. J.; Garman, E.; Gonzalez, A.; Sauter, N. K.;
Phizackerley, R. P.; Soltis, S. M.; Kuhn, P. Blu-Ice and the Distributed
Control System: software for data acquisition and instrument control
at macromolecular crystallography beamlines. J. Synchrotron Radiat.
2002, 9, 401−406.
(31) Otwinowski, Z.; Minor, W. Processing of X-ray diffraction data
collected in oscillation mode. In Macromolecular Crystallography, Part
A; Carter, C. W., Jr., Sweet, R. M., Eds.; Methods in Enzymology, Vol.
276; Academic Press: New York, 1997; pp 307−326 .
(32) Leslie, A. G. W.; Powell, H. R. Evolving Methods for
Macromolecular Crystallography; Springer Verlag: Dordrecht, The
Netherlands, 2007; Vol. 245, pp 41−51.
(33) McCoy, A. J.; Grosse-Kunstleve, R. W.; Adams, P. D.; Winn, M.
D.; Storoni, L. C.; Read, R. J. Phaser crystallographic software. J. Appl.
Crystallogr. 2007, 40, 658−674.
(34) Winn, M. D.; Ballard, C. C.; Cowtan, K. D.; Dodson, E. J.;
Emsley, P.; Evans, P. R.; Keegan, R. M.; Krissinel, E. B.; Leslie, A. G.;
McCoy, A.; McNicholas, S. J.; Murshudov, G. N.; Pannu, N. S.;
Potterton, E. A.; Powell, H. R.; Read, R. J.; Vagin, A.; Wilson, K. S.
Overview of the CCP4 suite and current developments. Acta
Crystallogr., D :Biol. Crystallogr. 2011, 67, 235−242.
(35) Emsley, P.; Cowtan, K. Coot: model-building tools for
molecular graphics. Acta Crystallogr., D: Biol. Crystallogr. 2004, 60,
2126−2132.
(36) Murshudov, G. N.; Vagin, A. A.; Dodson, E. J. Refinement of
macromolecular structures by the maximum-likelihood method. Acta
Crystallogr., D: Biol. Crystallogr. 1997, 53, 240−255.
(37) The 3D structure generator CORINA is available from
Molecular Networks GmbH, Erlangen, Germany (http://www.
molecular-networks.com). Accessed 1 June 2013.
(38) Schrodinger, LLC. The PyMOL Molecular Graphics System,
version 1.3r1; Portland, OR, 2010.
(39) Molecular Operating Environment (MOE), 2011.10; Chemical
Computing Group Inc.: Montreal, Quebec, Canada, 2011.
4508
dx.doi.org/10.1021/jm400211f | J. Med. Chem. 2013, 56, 4497−4508