Characterization of the binding properties of SIRT2 inhibitors with a N-(3-phenylpropenoyl)-glycine tryptamide backbone

Article abstract of DOI:10.1016/j.bmc.2008.07.059

SIRT2 inhibitors with a N-(3-phenylpropenoyl)-glycine tryptamide backbone were studied. This backbone has been developed in our group, and it is derived from a compound originally found by virtual screening. In addition, compounds with a smaller 3-phenylpropenoic acid tryptamide backbone were also included in the study. Binding modes for the new compounds and the previously reported compounds were analyzed with molecular modelling methods. The approach, which included a combination of molecular dynamics, molecular docking and cluster analysis, showed that certain docking poses were favourable despite the conformational variation in the target protein. The N-(3-phenylpropenoyl)-glycine tryptamide backbone is also a good backbone for SIRT2 inhibitors, and the series of compounds includes several potent SIRT2 inhibitors.

Full text of DOI:10.1016/j.bmc.2008.07.059

Bioorganic & Medicinal Chemistry 16 (2008) 8054–8062
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Characterization of the binding properties of SIRT2 inhibitors with a
N-(3-phenylpropenoyl)-glycine tryptamide backbone
Päivi H. Kiviranta a, , , Heikki S. Salo a, , , Jukka Leppänen a, Valtteri M. Rinne a, Sergiy Kyrylenko b,
Erkki Kuusisto c, Tiina Suuronen d, Antero Salminen d,e, Antti Poso a, Maija Lahtela-Kakkonen a,
Erik A. A. Wallén a,à
*
*
a Department of Pharmaceutical Chemistry, University of Kuopio, PO Box 1627, 70211 Kuopio, Finland
b Department of Biochemistry, University of Kuopio, Finland
c Department of Environmental Health, National Public Health Institute (KTL), PO Box 95, 70701 Kuopio, Finland
d Department of Neurology, Institute of Clinical Medicine, University of Kuopio, Finland
e Department of Neurology, Kuopio University Hospital, PO Box 1777, 70211 Kuopio, Finland
a r t i c l e i n f o
a b s t r a c t
Article history:
SIRT2 inhibitors with a N-(3-phenylpropenoyl)-glycine tryptamide backbone were studied. This back-
bone has been developed in our group, and it is derived from a compound originally found by virtual
screening. In addition, compounds with a smaller 3-phenylpropenoic acid tryptamide backbone were
also included in the study. Binding modes for the new compounds and the previously reported com-
pounds were analyzed with molecular modelling methods. The approach, which included a combination
of molecular dynamics, molecular docking and cluster analysis, showed that certain docking poses were
favourable despite the conformational variation in the target protein. The N-(3-phenylpropenoyl)-glycine
tryptamide backbone is also a good backbone for SIRT2 inhibitors, and the series of compounds includes
several potent SIRT2 inhibitors.
Received 7 April 2008
Accepted 22 July 2008
Available online 24 July 2008
Keywords:
SIRT2
Enzyme inhibitors
Molecular docking
Cluster analysis
Structure–activity relationship
Ó 2008 Elsevier Ltd. All rights reserved.
1. Introduction
role in mammalian development.15 Recent studies have provided
strong evidence that the biological function of SIRT2 can be linked
Human sirtuins (SIRT) consist of seven members, and they be-
long to the class III histone deacetylase protein family.1,2 They
are nicotinamide adenine dinucleotide (NAD+) dependent en-
zymes, which catalyze deacetylation and mono-ADP-ribosyl trans-
ferase reactions.1,3–5 Sirtuins are connected to several cellular
processes in prokaryotic, archae and eukaryotic cells.1,6–8 SIRT2 re-
sides normally in the cytoplasm1,9,10 but can move from the cyto-
plasm to the nucleus during mitosis, where it deacetylates Lys16 of
histone H4.11,12 It seems to be clear that SIRT2 has a role in the con-
trol of cell cycle.13 SIRT2 has also been shown to catalyze the
to glioma tumourigenesis and development of Parkinson’s dis-
ease.16,17 It has also been suggested that SIRT2 inhibitors could
have favourable effects on some physiological changes involved
in Parkinson’s disease.17,18
The NAD+ hydrolysis product, nicotinamide, inhibits the deace-
tylase reaction endogenously. A3 and sirtinol were the first re-
ported potent SIRT2 inhibitors.19 The other known SIRT2
inhibitors are 1,4-bis[2-(4-hydroxy-phenyl)-ethylamino]-anthra-
quinone,20 para-sirtinol,21 EX-527,22 N,N0-bis(2-hydroxybenzylid-
ene)benzene-1,4-diamine23 and cambinol.24 Recently reported
adenosine mimetics,25 suramin anologues26 and phloroglucinol
derivatives27 are also inhibitors of SIRT2. In addition, a potent
and selective SIRT2 inhibitor AGK2 was recently published by
Outeiro et al.17 All reported small-molecule inhibitors inhibit SIRT2
at low micro molar level.
deacetylation of acetylated Lys40 of a-tubulin and colocalize with
the cytoplasmic microtubules together with HDAC6.13,14 The acet-
ylation status of the microtubulin network can probably be altered
by inhibition of SIRT2. It has also been shown that SIRT2 interacts
with the homeobox transcription factor, HOXA10, which plays a
The first SIRT2 inhibitor with a N-(3-phenylpropenoyl)-glycine
tryptamide backbone was compound 1, which was found by virtual
screening.28 In an initial study the size of compound 1 was re-
duced, resulting in a series of N-(3-(4-hydroxyphenyl)-prope-
noyl)-amino acid tryptamides. Compounds 5 and 6 were the
most potent compounds of the series (Fig. 1).29
* Corresponding authors. Tel.: +358 17 163693; fax: +358 17 162456.
E-mail addresses: Paivi.Kiviranta@uku.fi (P.H. Kiviranta), Heikki.Salo@uku.fi (H.S.
Salo).
These authors contributed equally to this work.
Present address: Division of Pharmaceutical Chemistry, Faculty of Pharmacy, PO
à
Box 56, 00014, University of Helsinki, Finland.
0968-0896/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2008.07.059

P. H. Kiviranta et al. / Bioorg. Med. Chem. 16 (2008) 8054–8062
8055
H
N
NH
O
O
H
N
N
H
O
H
N
H N
N
H
N
H
NH
NH
O
O
O
F
O
HO
HO
6
5
F
1
Figure 1. Compound 1 found by using virtual screening and its potent derivatives 5 and 6.28,29
In order to study the N-(3-phenylpropenoyl)-glycine trypta-
mide backbone, a new series of more modified compounds were
synthesized. The modifications also included changes in the N-(3-
phenylpropenoyl)-glycine tryptamide backbone, such as reduction
of the double bond and replacing the indole ring by another aro-
matic ring. The new series of compounds was designed to include
more variations than the earlier reported series of compounds. The
study also included compounds with a shorter 3-phenylpropenoic
acid tryptamide backbone.
In amide formation reactions the carboxylic acid was activated
either as an anhydride using ethyl chloroformate or pivaloyl chlo-
ride, as an acid chloride using oxalyl chloride or thionyl chloride, or
as an N-hydroxysuccinimide ester using HOSu and DCC. The latter
procedure was used for activation of the carboxylic acid terminal of
L
-alanine to avoid racemization.33 More detailed synthetic proce-
dures are described in Supplementary material.
2.2. In vitro assay for SIRT2 activity
Molecular modelling methods were used to characterize the
binding and interaction properties of all compounds (also the pre-
viously reported N-(3-(4-hydroxyphenyl)-propenoyl)-amino acid
tryptamides). So far only a little is known about the binding prop-
erties of SIRT2 inhibitors. Currently there is only one experimental
crystal structure of SIRT2 publicly available.30 As the only pub-
lished structure is in apo-form it does not provide any experimen-
tal data on the binding sites and binding modes of small-molecule
regulators. In order to efficiently design and screen new inhibitors
more information on the interaction properties of inhibitors is ur-
gently needed. A cavity adjacent to the active site of the protein has
been hypothesized to serve as a binding site for some known sir-
tuin inhibitors.17,31,32 As there is no precise information available
concerning the inhibitor binding site, the area used in the search
for interactions was defined to be quite large. The search was con-
centrated on the active site and the cavity formations around the
catalytic site. The area included also the narrow, water filled, chan-
nel leading from the active site to the other site of the protein.
The inhibitory activities were tested in a Fluor de Lys fluores-
cence-based assay. The assay was slightly modified from method
described in the BioMol product sheet.
2.3. Computational procedures
The interaction possibilities between the area around the active
site and the inhibitors were investigated with a molecular model-
ling approach, which includes a combination of molecular dynam-
ics and molecular docking. Molecular dynamics simulations were
performed in order to characterize possible conformational rear-
rangements in the protein structure, which could have favourable
effects on ligand binding. In most docking programs the target pro-
tein is treated as a rigid structure. However, changes in the recep-
tor structure may have drastic effects on ligand binding. Using
multiple MD (molecular dynamics) snapshot conformations as tar-
get structures in the docking calculations, conformational flexibil-
ity of the protein could be incorporated in the binding mode
predictions. This kind of approach has successfully been used for
including receptor flexibility into docking calculations.34,35 The tar-
get protein sets were created taking the probabilities of different
protein conformations into account. The distribution of different
conformational states was analyzed with cluster analysis. Based
on the cluster data, cluster representatives were chosen keeping
the number of cluster representatives relative to the sizes of the
clusters. This way the probability distribution of the conformations
in the target protein sets was similar to the distribution seen in the
MD simulations.
2. Methods
2.1. Synthetic chemistry
The new compounds 10–27 are based on a similar chemical
backbone as the earlier reported compounds 1–9, and they were
synthesized by similar methods as described previously.29 A few
simplified compounds without the amino acid in the middle of
the backbone were also synthesized (compounds 28–30). General
structures of the new compounds are described in Figure 2.
O
R
R
5
4
H
N
R1= H, OH or F
R2= H, OH, OCH3 or NO2
R3= H or OCH3
R4= H or CH3
NH
H
N
O
N
N
Ar
H
NH
O
N
H
O
O
R5= H or CH3
R
R
1
3
R
1
R
R1
2
R
1 = H or a benzyloxy group
R1= H or OH
Compounds 10-22,25-27
Compounds 23-24
Compounds 28-30
Figure 2. The general structures of new compounds 10–30.

8056
P. H. Kiviranta et al. / Bioorg. Med. Chem. 16 (2008) 8054–8062
3.2. Binding mode prediction
3. Results and discussion
3.1. Inhibitory activities
The binding possibilities of the reported compounds were
investigated by analyzing the docking poses of the most potent
compound belonging to the series, compound 6, and then compar-
ing the docking poses of the other compounds with these results.
The docking poses obtained from the docking calculations were
analyzed with cluster analysis in order to distinguish docking
poses, which occur frequently in the docking calculations (Fig. 3).
The clustering of the binding modes was performed based on root
mean square deviation (RMSD) values calculated on heavy atoms
of the ligands. Based on the cluster data the possible interactions
and calculated scoring values of different binding modes were
analyzed.
The docking results obtained using the protein conformations
from the three individual molecular dynamics runs, MD run 1,
MD run 2 and MD run 3 (referred later as MD1, MD2, MD3a and
MD3b, see details in Section 5), were analyzed separately. The
docking poses seen in these dockings were compared to the poses
obtained for compound 6 in the docking runs performed using the
X-ray crystal conformation of the protein as a target structure. Of
the 15 docking runs performed on each ligand MD snapshot pair,
three best-ranked solutions were chosen for the set of conforma-
tions, which was subjected to cluster analysis with a hierarchical
average-linkage cluster method. The cluster analysis was done in
order to see which docking poses were ranked high when sets of
distinct target protein structures were used in the dockings.
Visual inspection of the cluster data of the docking results of
compound 6 revealed that the docking poses clearly were divided
into distinct binding modes. In order to determine the frequency of
different binding modes a cut-off value for the cluster formation
had to be defined. Choosing the right cut-off level is crucial step
in the analysis of the data obtained with hierarchical cluster meth-
ods as it determines the sizes and the homogeneities of the formed
clusters.36 The cut-off level of the clustering was first determined
by applying the Kelley–Gardner–Sutcliffe (KGS) function.37 How-
ever, visual inspection of the cluster members showed that KGS
led in this case to too high cut-off levels. Members belonging to
the same cluster were observed to have substructures located in
distinct adjacent binding pockets on the protein surface. Thus,
the cut-off level was reduced based on visual inspection of the
cluster members, and a cut-off value of 3.2 Å was used in the
analysis.
The distribution of the cluster sizes showed that certain binding
modes occurred clearly more frequently than others. Interestingly,
the most common docking pose for 6 was similar in all of the four
individual dockings performed on the MD snapshots and, also, in
the dockings performed on the crystal structure. Despite the differ-
ences in the MD simulation length, protonation of His187 and the
used scoring function of the different docking runs, a similar bind-
ing mode was found to be the most common docking solution for
the compound.
The binding mode (referred later as binding mode A), which
was seen as the most frequent docking pose in all of the individual
docking series, had the para-hydroxyphenyl group of the ligand lo-
cated in a narrow channel leading from the active site to the oppo-
site site of the protein. The phenyl ring is surrounded by the
residues Ile93, Pro94, Leu138 and Leu103. The indole ring is situ-
ated in a hydrophobic cavity formed between residues His187
and Phe119, which is the area where the acetylated lysine of the
substrate is shown to bind.38 Due to variation in the side-chain
conformations in this protein area (especially His187), the biggest
clusters included also poses, which had the indole ring oriented
slightly towards the cavity close to Phe96, Phe119, and Tyr104.
The binding mode A offers several hydrogen bonding possibili-
ties for 6. The backbone carbonyl group of Glu137 could serve as a
The structures of the compounds, their inhibitory activities at
200 lM and the IC50 values for the most potent compounds are
presented in Tables 1 and 2. The inhibitory activities are compared
to the results of the earlier published compounds 1–9.29 From the
presented 21 new compounds, five compounds gave an inhibition
of over 50% at 200
were determined.
lM (11, 12, 15, 27 and 29), and their IC50 values
The earlier reported compounds showed that the central amino
acids -aminoisobutyric acid and -alanine gave equipotent com-
a
L
pounds, whereas the central amino acid glycine resulted in less po-
tent compounds. All of these central amino acids were also used in
the new series.
In the first group of new compounds 10–20 the substitution
of phenyl ring was varied and the adjacent double bond was re-
duced. Compounds 11, 12 and 15 gave over 50% inhibition at
200
phenyl ring gave inhibitions of 49.5% and 59.6% at 200
IC50 values of 100 M and 109 M, respectively. Compound 15
with the para-fluoro substituent on the phenyl ring gave an inhi-
bition of 61.3% at 200 M and an IC50 of 105 M. These com-
lM. Compounds 11 and 12 without the substituent on the
lM and
l
l
l
l
pounds are not as potent as the corresponding earlier reported
para-hydroxy substituted compounds 5 and 6, but more potent
than the corresponding meta-hydroxy and meta- or ortho-meth-
oxy substituted compounds 10, 13 and 14. Also a meta-nitro
substituted compound 16 gave an equipotent inhibitory activity
with compound 13.
Replacing the central amino acid alanine by glycine gave 17 and
18, which gave the lower inhibitory activities, 22.7% and 24.2% at
200 lM, respectively. Reduction of the adjacent double bond of
the phenyl group resulted in compounds 19 and 20. Increased flex-
ibility decreased the inhibitory activity. These two compounds
were clearly less potent than the corresponding compounds 11
and 5 with the double bond, respectively.
The second group of compounds constitute 21 and 22 where the
indole ring has been replaced by a 3-pyridyl and a phenyl ring.
These compounds gave lower inhibitory activities as compared to
the corresponding indole ring containing compounds 5 and 12.
The 3-pyridyl ring reduced the inhibitory activity strongly,
whereas the phenyl ring only lowered the inhibitory activity
slightly.
The effect of a more rigid central amino acid proline was
studied with compounds 23 and 24. Compound 23 has an inter-
mediate inhibitory activity as compared to compounds 12 and
17 with the central amino acids L-alanine and glycine, respec-
tively. The third group of compounds constitute 24–27, which
have a benzyloxy group in 5-position of the indole ring. The ben-
zyloxy group did not have a significant effect on the inhibitory
activity. Also a similar effect was observed with a methoxy
group in the 5-position of the indole ring in compound 18. How-
ever, compound 27 with a reduced double bond in combination
with a benzyloxy group in the 5-position of the indole ring gave
a good inhibitory activity, with an inhibition of 50.2% at 200
lM
and an IC50 value of 86 M. It was even slightly more potent
l
than the corresponding compound with a double bond 25 and
the other reduced compounds without the benzyloxy substitu-
ents 19 and 20.
The fourth group of compounds 28–30 with the shorter 3-phe-
nylpropenoic acid tryptamide backbone are presented in Table 2.
Interestingly, this group of relatively small compounds gave good
inhibitory activities as compared to compounds discussed above.
The best compound 29 gave an inhibition of 58.0% at 200
and an IC50 of 173 M.
lM
l

P. H. Kiviranta et al. / Bioorg. Med. Chem. 16 (2008) 8054–8062
8057
Table 1 (continued)
Table 1
SIRT2 inhibitors with N-(3-phenylpropenoyl)-glycine tryptamide backbone (95%
confidence intervals for IC50 given in parentheses)
Compound Structure
Inhibition
200
SD,a
IC50
(lM)
l
M
SIRT2b
% SIRT2
Compound Structure
Inhibition
200
% SIRT2
IC50
lM
SD,a
(lM)
SIRT2b
H
N
O
N
H
NH
11
49.5 0.7
100 (71–
141)
H
N
O
O
H
N
N
H
129
77.3 4.8
51 (27–
75)
NH
O
O
O
F
H
N
H
O
F
N
H
12
NH
NH
NH
59.6 11.3
109 (49–
243)
H
N
O
O
H
N
229
79.1 1.4
63 (41–
96)
N
NH
NH
H
O
O
O
H
N
O
N
H
O
O
13
45.9 15.5
35.0 8.5
61.3 0.5
—
—
H
N
O
N
H
329
32.6 18.2
—
O
H
N
O
N
H
14
O
O
O
O
O
H
N
N
H
429
55.4 3.3
99 (66–
150)
NH
NH
NH
NH
O
O
O
H
N
O
N
NH
15
105 (53–
209)
H
O
F
H
N
N
H
529
83.3 3.7
50 (23–
109)
O
H
N
N
H
HO
NH
O
16
42.1 15.3
—
—
NO2
H
N
H
O
N
H
629
88.6 0.8
47 (28–
79)
O
H
N
N
H
17
NH
NH
22.7 11.9
HO
O
O
H
N
H
O
H
N
N
729
8.5 6.0
44.3 10.6
67.5 1.1
—
—
H
N
H
O
18
24.2 3.7
—
—
HO
O
O
O
H
N
O
H
N
N
H
829
NH
NH
O
O
N
H
19
NH
12.5 1.3
O
HO
H
N
O
O
H
N
N
H
929
80 (53–
120)
N
H
NH
N
20
19.5 1.2
—
O
O
HO
F
HO
O
H
H
N
N
N
N
H
NH
H
21
2.0 2.8
—
O
10
46.7 0.1
—
HO
H
O
(continued on next page)

8058
P. H. Kiviranta et al. / Bioorg. Med. Chem. 16 (2008) 8054–8062
Table 1 (continued)
hydrogen bond acceptor for the hydroxyl group attached to the
phenyl ring. There are several other hydrogen bond possibilities
for a hydroxyl group deeper in the channel. The cut-off value in
the clustering allowed some variation in the position of the para-
hydroxyphenyl among the members of the biggest cluster. Deeper
in the pocket backbone amines of Gly141, Gln142 and Phe143
could act as hydrogen bond donors and the backbone carbonyl
group of Tyr139 as an acceptor for a hydroxyl group. The binding
mode A allows the amide bond of the compound, located closer
to the para-hydroxyphenyl group, to form a hydrogen bond either
with the backbone amine of Ile169 or with the carboxylate group
of Asp170. Both orientations for the amide bond were seen among
the solutions of these clusters.
The docking solutions of binding mode A had obtained fairly
high scoring values by the scoring functions. In the dockings per-
formed on the crystal structure and in the docking series of
MD3b the largest cluster included the best-ranked solution. The
clusters of the binding mode A in MD1, MD2 and MD3a were all
ranked among the best clusters when the ranking was made based
on the best-scored individual found in the cluster.
The best-ranked docking poses on each MD docking series of 6
had the para-hydroxyphenyl structure placed in the channel lead-
ing through the protein. In the best-ranked solution of MD1 and
MD2 the indole ring is located in a pocket formed by residues
Ala85, Ser88, Gly92, Ile93, Pro94, Asp95, Phe96 and Arg97. The
best-ranked docking conformations in MD1 and MD2 were very
close to each other. In MD3a the para-hydroxyphenyl group is
placed slightly different than in MD1 and MD2. The indole ring is
under the flexible loop, but has a slightly different orientation as
in MD2, allowing the indole nitrogen to form a hydrogen bond with
Asp170. In MD1 and MD2 the amide nitrogens are able to form a
hydrogen bond to the side chain of Asp170. Also a hydrogen bond
between the other amide nitrogen and the backbone carbonyl of
Ile169 could be possible. In MD3a similar hydrogen bond forma-
tion between the amide structures and the residues cannot be
seen. This probably causes some difference in the location of the
indole ring. The binding mode seen in the highest-scored solutions
in MD1, MD2, and MD3a is later referred as binding mode B. As
stated earlier the highest-scored docking pose of MD3b was the
most often occurring binding mode A. The highest-scored confor-
mation placed the para-hydroxyphenyl group in a position allow-
ing the hydroxyl group to form hydrogen bonds with the side-
chain carbonyl group of Gln142, and the indole-side amide bond
carbonyl formed a hydrogen bond with Ile169.
Compound Structure
Inhibition
200
SD,a
% SIRT2
IC50
(lM)
lM
SIRT2b
O
H
N
H
O
N
H
22
29.6 1.8
—
H
N
N
N
23
24
36.6 8.4
29.8 3.2
43.7 2.7
32.9 2.3
—
—
—
—
NH
NH
O
O
O
O
H
N
O
O
H
N
H
O
N
H
NH
25
O
O
O
H
N
N
H
NH
NH
O
26
O
H
N
H
O
N
H
27
50.2 1.1
86
(34–216)
O
a
SD, standard deviation.
IC50 were determined for compounds which had over 50% inhibition at 200 lM
b
for SIRT2.
Although the biggest clusters in all MD runs were similar there
were some variations in the sizes and binding modes of the other
clusters. In the docking performed using the protein conformations
of MD1 run two distinct binding modes clearly occurred more fre-
quently than the others. The two most populated clusters were al-
most similar in size. The docking poses in the second biggest
cluster differ significantly from the conformations in the binding
modes A and B. In these docking poses the para-hydroxyphenyl
ring is located in the hydrophobic cavity between His187 and
Phe119. The indole ring is located in a pocket formed by residues
Ala85, Ser88, Gly92, Ile93, Pro94, Asp95, Phe96 and Arg97. The fact
that this binding mode is the only frequently occurring binding
mode, which does not involve entering the narrow channel, makes
this binding mode interesting. This binding mode is later referred
as binding mode C. The binding mode of the third most populated
cluster in MD1 was binding mode B as the para-hydroxyphenyl
structure is buried in the narrow channel and the indole ring is lo-
cated in the pocket near residues Ala85, Ser88, Gly92, Ile93, Pro94,
Asp95, Phe96 and Arg97.
Table 2
SIRT2 inhibitors with 3-phenylpropenoic acid tryptamide backbone (95% confidence
intervals for IC50 given in parentheses)
Compound Structure
Inhibition
200
IC50
(
l
M)
lM
SD,a % SIRT2
SIRT2b
NH
O
N
H
28
44.3 4.1
—
HO
NH
NH
O
N
H
29
58.0 11.1
173 (108–
276)
O
N
H
30
44.1 7.4
—
a
As the first two clusters in MD1 were practically equally sized,
in MD2 the most populated cluster is significantly larger in size
than the other clusters. In MD2 86% of the docking poses had the
SD, standard deviation.
b
IC50 were determined for compounds which had over 50% inhibition at 200
for SIRT2.
lM

P. H. Kiviranta et al. / Bioorg. Med. Chem. 16 (2008) 8054–8062
8059
Figure 3. Dendrograms representing the cluster data of compound 6 in docking series MD1, MD2, MD3a and MD3b. The green rectangle illustrates the clusters of binding
mode A, the red binding mode B and the blue binding mode C.
para-hydroxyphenyl group located in the narrow channel. In the
case of MD1 only 52% of the docking poses had the para-hydroxy-
phenyl group in this area. In the MD2 dockings binding poses sim-
ilar to the ones in cluster 2 of MD1 were extremely rare. This could
be explained by frequency of the different conformational forms of
the active site residues in the used target protein set. The simula-
tion length of MD1 was significantly shorter than in MD2 and MD3.
Because of this the distribution of conformational states could dif-
fer from that of the longer simulations. In MD2, MD3a and MD3b
the target protein sets were created based on cluster analysis
and, in these sets, the significance of the frequent conformational
states is therefore emphasized. This might, respectively, reduce
the significance of the protein conformations, which would favour
docking poses similar to the cluster 2 in MD1, if these conforma-
tions do not occur frequently in the simulation. One apparent dif-
ference between the conformations of target proteins of MD1 and
MD2 could be seen in the conformational states of the side chain of
His187. In the shorter molecular dynamics run of MD1 the confor-
mation of His187 remains close to the conformation seen in the
crystal structure. However, in the long 10 ns, MD2 run, the confor-
mation of His187 changes dramatically before 1 ns and the confor-
mation remains similar throughout the rest of the simulation.
In the MD3a most frequent docking pose was similar to the
most common binding mode in the MD2 docking calculations. In
the protein structures of MD3 the distribution of the conforma-
tional states of His187 varies from the one of MD2. His187 could
be found more often in the conformation close to the conformation
seen in the crystal structure than in the structures of MD2. Also in
the case of MD3a over 80% of the docking poses belonged to clus-
ters, which had the para-hydroxyphenyl group in the channel.
sition of the indole ring varied. In order to include a frequently
occurring binding mode, which does not bind into the channel, into
further analysis the second biggest cluster of MD1 was chosen
(binding mode C). Although the MD1 differs from the MD2,
MD3a and MD3b in the length of the simulation and in the ap-
proach used in the protein conformation selection, as one of the
frequent binding modes in these calculations was significantly dif-
ferent than the frequent docking poses in other MDs, it was neces-
sary to include this binding mode for further analysis.
3.3. SAR
The effect of structural modifications of the reported com-
pounds on the binding properties was analyzed by investigating
the docking poses resembling the binding modes A, B and C of
compound 6. The ability of the used scoring functions to rank the
compounds according to the biological activity data was also
tested.
Accurate prediction of the binding affinity seems to be an extre-
mely difficult task for the scoring functions of the available docking
programs. This problem has been a widely acknowledged weak-
point in the most currently available docking applications.39,40 As
anticipated, the scoring function was not able to create a correct
rank-ordering for the compounds. When the ranking and scoring
was based on the scoring value of the highest-scored solution on
each MD run, no correlation could be seen between the scoring val-
ues and the inhibition percentages. There was no difference be-
tween the performance of the two distinct scoring functions,
Goldscore and Chemscore used in MD3a and MD3b. Neither was
there difference between MD1, MD2, MD3a and MD3b. In order
to test the effect of the structural variation in the used target struc-
tures mean fitness scores were calculated for each compound using
the three best-ranked docking solution on each ligand–protein
pair. The rationale behind this idea is in the fact that as the target
protein structures are chosen based on the occurrence frequency in
the MD simulation, compounds, which get higher scores on com-
mon target conformations, would get higher mean scores. How-
ever, in this case this approach neither gave any correlation
between the calculated scores and the biological data. The fact that
Docking poses similar to binding mode
observed.
C were only rarely
Based on the cluster analysis of the docking results of MD1,
MD2, MD3a and MD3b the most frequently occurring binding
modes were chosen for further analysis (binding mode A). The
binding mode, which had obtained the highest score in MD1,
MD2 and MD3a, was also chosen for further characterization
(binding mode B). In both of these binding modes the para-
hydroxyphenyl group was located in the channel, and only the po-

8060
P. H. Kiviranta et al. / Bioorg. Med. Chem. 16 (2008) 8054–8062
the activity range of the compounds in this study is quite small, the
correct rank-scoring and activity prediction would require high
accuracy from the scoring function.
ring could, however, make such interactions stronger. Compound
21 had obtained low scoring values in both scoring functions and
in all individual docking series. A clear difference could not be seen
in the binding modes of these compounds.
The results of cluster analysis of the docking poses of the
other docked compounds were in general terms similar to the
results of compound 6. In the MD1 the most common binding
pose for almost all of the compounds is the binding mode in
which the phenyl ring is located in the hydrophobic cavity be-
tween Phe119 and His187, and the indole part of the compound
is located under the loop structure of Pro94, Asp95 and Phe96
(binding mode C). In the dockings of MD2 the most frequently
occurring binding pose for most compounds resembled the bind-
ing mode A of 6. Some variation could be seen in the orientation
of the indole ring as the indole ring could be seen oriented
slightly towards the cavity close to Phe96, Phe119 and Tyr104.
The most frequent binding mode for some compounds, for
example, compounds 8, 28, 29 and 30, had the para-hydroxy-
phenyl in the channel, but the indole ring is located under the
loop structure formed by residues Ala85, Ser88, Gly92, Ile93,
Pro94, Asp95, Phe96 and Arg97 resembling the binding mode
B. Also in MD3a the highest-populated clusters of most com-
pounds were similar to the binding mode A. The biggest clusters
of compounds 8, 28, 29 and 30 had a binding conformation sim-
ilar to the binding mode B. In MD3b the highest-populated clus-
ters included a wider range of different binding modes.
Addition of a benzyloxy group to the 5-position of the indole
ring of compounds 12 and 17 resulting in compounds 25 and 26
had no effect on the inhibitory activity. Compounds 12 and 25
had inhibitions in the range of 59.6–43.7% at 200
pounds 17 and 26 had inhibitions in the range 22.7–32.9% at
200 M. As increased molecular structure offers more interactions
lM, and com-
l
upon binding and thus higher inhibitory activity, the fact that the
inhibitory activity does not increase indicate that there are no
favourable interaction partners for the benzyloxy group in the
binding site. However, both of the used scoring functions clearly
overestimated the binding affinity of these structures. The larger
molecular structure led to higher scoring values in all of the indi-
vidual docking series, MD1, MD2, MD3a and MD3b. The results
of the cluster analysis also indicate that single favourable interac-
tion site for the benzyloxy group was not found in the docking
runs. A wider variety of docking poses were created in the docking
process for these larger compounds, which could be seen as smal-
ler cluster sizes.
In addition, the relatively small compounds 28–30 were almost
equipotent. The inhibitory activities of 28, 29 and 30 were 44.3%,
58.0% and 44.1% at 200
lM, respectively, and the IC50 value of 29
The differently substituted phenyl groups showed that a hydro-
xyl group in the para-position of the phenyl ring had a positive ef-
fect on the inhibitory activity. Compounds 5, 6 and 9 had an
was 173 M. In MD2 and MD3a the most frequent binding mode
l
for these compounds had the para-hydroxyphenyl group placed
in the narrow channel, and the docking poses were similar to bind-
ing mode B. In MD1 the compounds 29 and 30 had often docking
poses resembling binding mode C as for 28 the binding mode B oc-
curred most frequently. In MD3b a wider variety of different dock-
ing poses were observed. The para-hydroxy substituent on phenyl
ring did not seem to be as important for the inhibitory activity
among these compounds as in the case of the bigger compounds.
Removal of the para-hydroxy substituent of 28 did not lower the
inhibitory activity. Neither did the removal of the double bond in
the propenoyl side-chain lower the inhibitory activity. These dif-
ferences could raise the question whether the binding of these
smaller compounds differs from the binding of the other com-
pounds. However, these modelling results do not explain this,
and further studies would be needed.
inhibition in the range 67.5–88.6% at 200 lM. In binding modes
A and B hydrogen bonding possibilities for a hydroxyl group can
be found at the end of the narrow channel. Compounds without
a hydroxyl group in this position cannot form these interactions.
In these binding modes any of the substituents, which compounds
10, 13, 14 and 16 have in the meta- and ortho-positions, do not of-
fer any hydrogen bond interaction possibilities. Neither does the
binding mode offer the hydrogen bond possibility for compound
4 that has a methoxy substituent in para-position. However, com-
pound 4 gave an inhibition of 55.4% at 200
lM and IC50 was
99
l
M.29 Binding mode C also offers a hydrogen bond possibility
for a hydroxyl group in the para-position. The same as in the bind-
ing modes A and B, the binding mode C does not offer additional
hydrogen bonding interactions for structures with other substitu-
ents or substituents in other positions.
Changes made to the central amino acids located between the
amide structures had various effects on the inhibitory activity.
Changing the alanine residue of 6 to glycine residue of 8 or proline
Compound 28 is also a known natural product isolated from
kernels of maize, safflower seed and Ravensara anisata Danguy,
among others.41–43
4. Conclusion
residue of 23 lowered the inhibition to 40% at 200
lM. However,
changing the alanine residue of 6 or 12 to -aminoisobutyric acid
a
The most potent new compounds were 11, 12, 15, 27 and 29
in 5 or 11 had no effect on the inhibitory activity. Explanations for
the effect of such small changes could not be easily found based on
the binding modes. The proline residue could restrict forming low
energy conformations, which would allow favourable interactions
with the protein. In the case of the glycine residue the inhibitory
activity could be affected by the increased conformational flexibil-
ity. As the flexibility increases the binding process has larger neg-
ative effect on the entropic term of the binding free energy as in
the case of more rigid compounds.
with IC50 values in the range 86–173
earlier reported compounds with the same backbone 5 and 6 with
IC50 values in the range 47–50 M, these new compounds did not
lead to an improvement of the inhibitory activity. However, the
new compounds clearly indicate that the N-(3-phenylpropenoyl)-
glycine tryptamide backbone is a good backbone for the design
of SIRT2 inhibitors. One-third of the synthesized compounds have
lM. As compared to the best
l
over 50% inhibition at the concentration of 200 lM. In addition, the
study revealed that a series of compounds with a smaller 3-phe-
nylpropenoic acid tryptamide backbone also were good SIRT2
inhibitors.
Molecular modelling methods were used for analyzing possible
binding modes for all the compounds. Molecular docking calcula-
tions were performed using several sets of protein conformations
obtained from molecular dynamics simulations. The used sets of
protein conformations were created using cluster analysis and
choosing cluster representatives based on the relative sizes of the
Compounds 11 and 12 gave higher inhibitions of 49.5% and
59.6% at 200
dole ring replaced by other aryl groups. The inhibitory activities
of 21 and 22 were 2.0% and 29.6% at 200 M, respectively. This
lM than similar compounds 21 and 22 with the in-
l
clearly indicates that the indole ring is forming important interac-
tions in the binding site. The interactions of the indole ring in all of
the binding modes A, B and C are mainly hydrophobic and aro-
matic interactions. Similar interactions are also possible for com-
pounds 21 and 22. The larger molecular structure of the indole

P. H. Kiviranta et al. / Bioorg. Med. Chem. 16 (2008) 8054–8062
8061
clusters. Cluster analysis was also used for the investigation of the
binding poses obtained in the docking runs. Based on the cluster
data of the docking poses three distinct binding modes clearly
stand out from the results. The binding mode A was found to be
a common binding mode in all of the MD dockings and with most
of the compounds. The binding mode occurred frequently despite
the difference in the His187 protonation state and the used scoring
function. The binding mode B stands out as this binding mode
seemed to obtain high scoring values in MD1, MD2 and MD3a.
The third binding mode, which clearly stands out from the results,
is the binding mode C. This binding mode was frequent in MD1 for
most of the compounds but was only rarely seen in MD2, MD3a
and MD3b.
The study gave important information about how the com-
pounds interact with SIRT2. Reasonable binding modes were found
for these compounds in the area, which has been earlier postulated
as a binding site for sirtuin inhibitors. However, prediction of a
binding mode without experimental structural data is an extre-
mely challenging task. In this study binding modes, which had ob-
tained high scoring values despite the variation in the target
structure, were studied. In future we are planning to perform
extensive free energy calculations for evaluating the binding affin-
ity of the compounds in the binding modes, which were observed
in this study. These free energy calculations may provide further
validation for the binding mode predictions.
snapshots from the trajectory at a regular 5 ps time interval
(MD1). The protein conformations from the 10 ns side-chain MD
runs (MD2 and MD3) were obtained based on a RMSD cluster anal-
ysis of the active site side-chain conformations. The clusterings
were performed with the g_cluster program implemented in GRO-
MACS using the gromos cluster method.50 The clustering included
10,000 snapshots from the 10 ns MD trajectory, and was made
using RMSD value of 1.0 Å as a cut-off distance defining the cluster
size. The pool of conformations was reduced to 95 conformations
in MD2 and 92 in MD3 keeping the proportions of cluster represen-
tatives and the size of the original clusters constant. All of the con-
formations were minimized in GROMACS with steepest descent.
Prior to using the conformations as docking targets all hydrogens
were added to the protein structures with the BIOPOLYMER mod-
ule of Sybyl 7.3.3.51
5.2.2. Molecular docking
The 3D ligand structures of the inhibitors used in the docking
were built with the sketch module of Sybyl 7.3.3. The compounds
were minimized with Powell minimization method using a conver-
gence criterion of 0.05 kcal/molÅ until the convergence was
reached. MMFF94 charges were assigned to the atoms of the ligand
compounds. The docking calculations were performed with version
3.2 of Gold docking software.52 The search area for the docking
program was defined by choosing residues within 15 Å from
Ile169 to represent the active site of the protein and by applying
the cavity detection method of Gold. For each ligand 15 docking
runs were performed, and the docking poses were scored with
Goldscore scoring function in MD1, MD2 and MD3a and with
Chemscore scoring function in MD3b. Three binding poses for each
ligand–protein pair were chosen for further analysis based on the
obtained scoring values. The binding poses were clustered within
each ligand, and frequently occurring docking modes were de-
tected. The selected binding modes were clustered with a hierar-
chical average-linkage clustering method based on atom pair
RMSD values.53 The cluster height for the hierarchical clustering
was chosen based on KGS function37 and visual inspection of the
formed clusters.
5. Experimental
5.1. In vitro assay for SIRT2 activity
The Fluor de Lys fluorescence assays were modified from meth-
od described in the BioMol product sheet. Briefly, assay was carried
out using Fluor de Lys-acetylated 250
lM SIRT2-peptide substrate
(KI-179), 500
l
M NAD+ (N6522, Sigma), recombinant GST-SIRT2-
enzyme and SIRT2-assay buffer (HDAC assay buffer, KI-143, sup-
plemented with 1 mg/mL BSA, A3803, Sigma). GST-SIRT2-enzyme
was produced as described recently.20 The buffer, SIRT2-enzyme,
NAD+ and DMSO/compounds in DMSO (2.5
lL in 50 lL total vol-
ume of reaction mixture; DMSO from Sigma, D2650) for testing
were preincubated for 5 min at room temperature. The reaction
was started by adding the substrate, and the reaction mixture
was incubated for 3 h at 37 °C. After that Fluor de Lys-developer
Acknowledgments
We thank Prof. Jouko Vepsäläinen for guidance in NMR, Ph.D.
Anu J. Tervo for the successful virtual screening and Tiina Koivunen
and Marko Koskivuori for their skilful technical assistance, and the
Graduate School of Drug Discovery, the National Technology
Agency of Finland, the Academy of Finland (Grant No. 108649 to
E.A.A.W., Grant No. 106313 to J.L. and Grant No. 111453 to A.P.),
Association of Finnish Chemical Societies, the Gustav Komppa Fund
of the Kordelin Foundation and the Finnish Cultural Foundation for
financial support.
(KI-176) plus 2 mM nicotinamide in 50 lL were added, and incuba-
tion was continued for 45 min at 37 °C. Fluorescence readings were
obtained using the VictorTM 1420 Multilabel Counter (Wallac, Fin-
land) with excitation wavelength 360 nm and emission 460 nm.
Each experiment was repeated at least three times.
5.2. Computational details
5.2.1. Molecular dynamics
The molecular dynamics runs were performed using the GRO-
MACS 3.3.1 software package44 with GROMOS9645 force field.
The monomer B of the crystal structure of SIRT2 (pdb code: 1J8F)
was used as the starting geometry for the simulations. SPC water
model,46 PME electrostatics,47,48 Berendsen pressure and tempera-
ture coupling49 and periodic boundary conditions were applied in
the simulations. After initial minimizations and equilibration runs
three distinct MD runs were performed. A 400 ps and two separate
10 ns MD runs were performed holding the backbone atoms fixed.
The two 10 ns fixed backbone simulations differed in the proton-
ation state of His187. A set of protein conformations was chosen
from the fixed-backbone MD simulations to be used as target
structures in molecular docking calculations. Eighty-one protein
conformations were obtained from the 400 ps MD run by taking
Supplementary data
Detailed experimental procedures for the syntheses, NMR spec-
tra, ESI-MS results and elemental analysis data for the new com-
pounds, and in vitro assay for SIRT2 activity are available.
Supplementary material associated with this article can be found
in the online version. Supplementary data associated with this arti-
cle can be found, in the online version, at doi:10.1016/
j.bmc.2008.07.059.
References and notes
1. Sauve, A. A.; Wolberger, C.; Schramm, V. L.; Boeke, J. D. Annu. Rev. Biochem.
2006, 75, 435.

8062
P. H. Kiviranta et al. / Bioorg. Med. Chem. 16 (2008) 8054–8062
2. Smith, J. S.; Brachmann, C. B.; Celic, I.; Kenna, M. A.; Muhammad, S.; Starai, V. J.;
Avalos, J. L.; Escalante-Semerena, J. C.; Grubmeyer, C.; Wolberger, C.; Boeke, J.
D. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 6658.
3. Tanner, K. G.; Landry, J.; Sternglanz, R.; Denu, J. M. Proc. Natl. Acad. Sci. U.S.A.
2000, 97, 14178.
28. Tervo, A. J.; Suuronen, T.; Kyrylenko, S.; Kuusisto, E.; Kiviranta, P. H.; Salminen,
A.; Leppanen, J.; Poso, A. J. Med. Chem. 2006, 49, 7239.
29. Kiviranta, P. H.; Leppanen, J.; Rinne, V. M.; Suuronen, T.; Kyrylenko, O.;
Kyrylenko, S.; Kuusisto, E.; Tervo, A. J.; Jarvinen, T.; Salminen, A.; Poso, A.;
Wallen, E. A. Bioorg. Med. Chem. Lett. 2007, 17, 2448.
4. Smith, B. C.; Denu, J. M. Biochemistry 2006, 45, 272.
5. Michan, S.; Sinclair, D. Biochem. J. 2007, 404, 1.
6. Brachmann, C. B.; Sherman, J. M.; Devine, S. E.; Cameron, E. E.; Pillus, L.; Boeke,
J. D. Genes Dev. 1995, 9, 2888.
7. Guarente, L. Genes Dev. 2000, 14, 1021.
30. Finnin, M. S.; Donigian, J. R.; Pavletich, N. P. Nat. Struct. Biol. 2001, 8, 621.
31. Huhtiniemi, T.; Wittekindt, C.; Laitinen, T.; Leppanen, J.; Salminen, A.; Poso, A.;
Lahtela-Kakkonen, M. J. Comput. Aided Mol. Des. 2006, 20, 589.
32. Neugebauer, R. C.; Uchiechowska, U.; Meier, R.; Hruby, H.; Valkov, V.; Verdin,
E.; Sippl, W.; Jung, M. J. Med. Chem. 2008, 51, 1203.
8. Lin, S. J.; Ford, E.; Haigis, M.; Liszt, G.; Guarente, L. Genes Dev. 2004, 18, 12.
9. Frye, R. A. Biochem. Biophys. Res. Commun. 2000, 273, 793.
10. Afshar, G.; Murnane, J. P. Gene 1999, 234, 161.
11. Vaquero, A.; Scher, M. B.; Lee, D. H.; Sutton, A.; Cheng, H. L.; Alt, F. W.; Serrano,
L.; Sternglanz, R.; Reinberg, D. Genes Dev. 2006.
12. Inoue, T.; Hiratsuka, M.; Osaki, M.; Yamada, H.; Kishimoto, I.;
Yamaguchi, S.; Nakano, S.; Katoh, M.; Ito, H.; Oshimura, M. Oncogene
2007, 26, 945.
13. North, B. J.; Marshall, B. L.; Borra, M. T.; Denu, J. M.; Verdin, E. Mol. Cell 2003, 11,
437.
33. Garcia-Lopez, M. T.; Gonzalez-Muniz, R.; Molinero, M. T.; Naranjo, J. R.; Del Rio,
J. J. Med. Chem. 1987, 30, 1658.
34. Lin, J. H.; Perryman, A. L.; Schames, J. R.; McCammon, J. A. J. Am. Chem. Soc.
2002, 124, 5632.
35. Lin, J. H.; Perryman, A. L.; Schames, J. R.; McCammon, J. A. Biopolymers 2003, 68,
47.
36. Bottegoni, G.; Cavalli, A.; Recanatini, M. J. Chem. Inf. Model. 2006, 46, 852.
37. Kelley, L. A.; Gardner, S. P.; Sutcliffe, M. J. Protein Eng. 1997, 10, 737.
38. Avalos, J. L.; Celic, I.; Muhammad, S.; Cosgrove, M. S.; Boeke, J. D.; Wolberger, C.
Mol. Cell 2002, 10, 523.
14. Hubbert, C.; Guardiola, A.; Shao, R.; Kawaguchi, Y.; Ito, A.; Nixon, A.; Yoshida,
M.; Wang, X. F.; Yao, T. P. Nature 2002, 417, 455.
15. Bae, N. S.; Swanson, M. J.; Vassilev, A.; Howard, B. H. J. Biochem. (Tokyo) 2004,
135, 695.
16. Hiratsuka, M.; Inoue, T.; Toda, T.; Kimura, N.; Shirayoshi, Y.; Kamitani, H.;
Watanabe, T.; Ohama, E.; Tahimic, C. G.; Kurimasa, A.; Oshimura, M. Biochem.
Biophys. Res. Commun. 2003, 309, 558.
17. Outeiro, T. F.; Kontopoulos, E.; Altmann, S. M.; Kufareva, I.; Strathearn, K. E.;
Amore, A. M.; Volk, C. B.; Maxwell, M. M.; Rochet, J. C.; McLean, P. J.; Young, A. B.;
Abagyan, R.; Feany, M. B.; Hyman, B. T.; Kazantsev, A. G. Science 2007, 317,
516.
18. Garske, A. L.; Smith, B. C.; Denu, J. M. ACS Chem. Biol. 2007, 2, 529.
19. Grozinger, C. M.; Chao, E. D.; Blackwell, H. E.; Moazed, D.; Schreiber, S. L. J. Biol.
Chem. 2001, 276, 38837.
20. Tervo, A. J.; Kyrylenko, S.; Niskanen, P.; Salminen, A.; Leppanen, J.; Nyronen, T.
H.; Jarvinen, T.; Poso, A. J. Med. Chem. 2004, 47, 6292.
21. Mai, A.; Massa, S.; Lavu, S.; Pezzi, R.; Simeoni, S.; Ragno, R.; Mariotti, F. R.;
Chiani, F.; Camilloni, G.; Sinclair, D. A. J. Med. Chem. 2005, 48, 7789.
22. Napper, A. D.; Hixon, J.; McDonagh, T.; Keavey, K.; Pons, J. F.; Barker, J.; Yau, W.
T.; Amouzegh, P.; Flegg, A.; Hamelin, E.; Thomas, R. J.; Kates, M.; Jones, S.;
Navia, M. A.; Saunders, J. O.; DiStefano, P. S.; Curtis, R. J. Med. Chem. 2005, 48,
8045.
23. Kiviranta, P. H.; Leppanen, J.; Kyrylenko, S.; Salo, H. S.; Lahtela-Kakkonen, M.;
Tervo, A. J.; Wittekindt, C.; Suuronen, T.; Kuusisto, E.; Jarvinen, T.; Salminen, A.;
Poso, A.; Wallen, E. A. J. Med. Chem. 2006, 49, 7907.
39. Warren, G. L.; Andrews, C. W.; Capelli, A. M.; Clarke, B.; LaLonde, J.; Lambert, M.
H.; Lindvall, M.; Nevins, N.; Semus, S. F.; Senger, S.; Tedesco, G.; Wall, I. D.;
Woolven, J. M.; Peishoff, C. E.; Head, M. S. J. Med. Chem. 2006, 49, 5912.
40. Leach, A. R.; Shoichet, B. K.; Peishoff, C. E. J. Med. Chem. 2006, 49, 5851.
41. Andrianaivoravelona, J. O.; Terreaux, C.; Sahpaz, S.; Rasolondramanitra, J.;
Hostettmann, K. Phytochemistry 1999, 52, 1145.
42. Sato, H.; Kawgishi, H.; Nishimura, T.; Yoneyama, S.; Yoshimoto, Y.; Sakamura,
S.; Furusaki, A.; Katsuragi, S.; Matsumoto, T. Agric. Biol. Chem. 1985, 49, 2969.
43. Takii, T.; Kawashima, S.; Chiba, T.; Hayashi, H.; Hayashi, M.; Hiroma, H.;
Kimura, H.; Inukai, Y.; Shibata, Y.; Nagatsu, A.; Sakakibara, J.; Oomoto, Y.;
Hirose, K.; Onozaki, K. Int. Immunopharmacol. 2003, 3, 273.
44. Van Der Spoel, D.; Lindahl, E.; Hess, B.; Groenhof, G.; Mark, A. E.; Berendsen, H.
J. J. Comput. Chem. 2005, 26, 1701.
45. van Gunsteren, W. F; Billeter, S. R.; Eising, A. A.; Hünenberger, P. H.; Krüger, P.;
Mark, A. E.; Scott, W. R. P.; Tironi, I. G. In Biomolecular Simulation: The
GROMOS96 Manual and User Guide, Hochschulverlag AG an der ETH Zürich:
Zürich, Switzerland, 1996.
46. Berendsen, H. J. C.; Postma, J. P. M.; van Gunsteren, W. F.; Hermans, J. In
Intermolecular Forces; Pullman, B., Ed.; D. Reidel Publishing Company
Dordrecht, 1981; pp 331–342.
47. Darden, T.; York, D.; Pedersen, L. J. Chem. Phys. 1993, 98, 10089.
48. Essmann, U.; Perera, L.; Berkowitz, M. L.; Darden, T.; Lee, H.; Pedersen, L. G. J.
Chem. Phys. 1995, 103, 8577.
49. Berendsen, H. J. C.; Postma, J. P. M.; DiNola, A.; Haak, J. R. J. Chem. Phys. 1984, 81,
3684.
24. Heltweg, B.; Gatbonton, T.; Schuler, A. D.; Posakony, J.; Li, H.; Goehle, S.;
Kollipara, R.; DePinho, R. A.; Gu, Y.; Simon, J. A.; Bedalov, A. Cancer Res. 2006,
66, 4368.
25. Trapp, J.; Jochum, A.; Meier, R.; Saunders, L.; Marshall, B.; Kunick, C.; Verdin, E.;
Goekjian, P.; Sippl, W.; Jung, M. J. Med. Chem. 2006, 49, 7307.
26. Trapp, J.; Meier, R.; Hongwiset, D.; Kassack, M. U.; Sippl, W.; Jung, M.
ChemMedChem 2007, 2, 1419.
50. Daura, X.; Gademann, K.; Jaun, B.; Seebach, J. D.; van Gunsteren, W. F.; Mark, A.
E. Angew. Chem., Int. Ed. 1999, 38, 236.
51. Sybyl 7.3, Tripos International, 1699 South Hanley Rd., St. Louis, Missouri,
63144, USA.
52. Jones, G.; Willett, P.; Glen, R. C.; Leach, A. R.; Taylor, R. J. Mol. Biol. 1997, 267,
727.
53. In R Development Core Team (2005). R: A language and environment for
statistical computing. R Foundation for Statistical Computing, Vienna, Austria,
2008. http://www.R-project.org.
27. Gey, C.; Kyrylenko, S.; Hennig, L.; Nguyen, L. H.; Buttner, A.; Pham, H. D.;
Giannis, A. Angew. Chem., Int. Ed. 2007, 46, 5219.