Occupying a flat subpocket in a tRNA-modifying enzyme with ordered or disordered side chains: Favorable or unfavorable for binding?

Article abstract of DOI:10.1016/j.bmc.2016.07.053

Small-molecule ligands binding with partial disorder or enhanced residual mobility are usually assumed as unfavorable with respect to their binding properties. Considering thermodynamics, disorder or residual mobility is entropically favorable and contrib

Full text of DOI:10.1016/j.bmc.2016.07.053

Bioorganic & Medicinal Chemistry xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Occupying a flat subpocket in a tRNA-modifying enzyme with
ordered or disordered side chains: Favorable or unfavorable for
binding?
a,y
b,y
a
b
a
Manuel Neeb , Christoph Hohn , Frederik Rainer Ehrmann , Adrian Härtsch , Andreas Heine ,
b
a,
⇑
François Diederich , Gerhard Klebe
a
Institut für Pharmazeutische Chemie, Philipps-Universität Marburg, Marbacher Weg 6, 35032 Marburg, Germany
Laboratorium für Organische Chemie, ETH Zurich, Vladimir-Prelog-Weg 3, 8093 Zurich, Switzerland
b
a r t i c l e i n f o
a b s t r a c t
Article history:
Small-molecule ligands binding with partial disorder or enhanced residual mobility are usually assumed
as unfavorable with respect to their binding properties. Considering thermodynamics, disorder or resid-
ual mobility is entropically favorable and contributes to the Gibbs energy of binding. In the present study,
we analyzed a series of congeneric ligands inhibiting the tRNA-modifying enzyme TGT. Attached to the
parent lin-benzoguanine scaffold, substituents in position 2 accommodate in a flat solvent-exposed
pocket and exhibit varying degree of residual mobility. This is indicated in the crystal structures by
enhanced B-factors, reduced occupancies, or distributions over split conformers. MD simulations of the
complexes suggest an even larger scatter over several conformational families. Introduction of a terminal
acidic group fixes the substituent by a salt-bridge to an Arg residue. Overall, all substituted derivatives
show the same affinity underpinning that neither order nor disorder is a determinant factor for binding
affinity. The additional salt bridge remains strongly solvent-exposed and thus does not contribute to
affinity. MD suggests temporary fluctuation of this contact.
Received 27 May 2016
Accepted 24 July 2016
Available online xxxx
Keywords:
Drug design
Protein–ligand complexes
Crystallography
Flat-binding pockets
Order-disorder phenomena
Residual mobility
Ó 2016 Elsevier Ltd. All rights reserved.
1
. Introduction
Concepts in rational drug design try to optimize a given ligand
Residual mobility of ligand side chains or attached substituents
is usually observed in flat, solvent-exposed binding pockets where
significant contacts to the neighboring solvent environment are
scaffold to accommodate and interact optimally with a certain tar-
get protein. Our hypotheses and thus the followed design strate-
gies are usually strongly biased by the idea to achieve perfect
complementarity between protein-binding pocket and ligand in
terms of geometry and experienced interaction patterns. We
assume that this strategy leads to energetically favored interac-
tions and a strong binding affinity. By contrast, any residual mobil-
ity or intrinsic flexibility of the bound ligand are intuitively
assumed as unfavorable for binding. However, binding affinity is
a Gibbs energy value and thus composed by an enthalpic and
entropic component. With respect to a favorable entropic compo-
nent to binding, residual mobility or pronounced ligand disorder
can be beneficial as a smaller number of degrees of freedom are
lost by the ligand upon complex formation.
given. In a congeneric series of thermolysin inhibitors with varying
0
P
2
substituents of increasing size and hydrophobicity, we observed
on the one hand a growing influence imposed by the structural
arrangement of the surrounding first and second solvent layer
but on the other hand by an enhanced residual flexibility particu-
1
,2
larly given for larger substituents.
As a second example, we studied the binding of potent inhibi-
tors blocking the function of the tRNA-modifying enzyme tRNA-
guanine transglycosylase (TGT) by attaching hydrophobic sub-
stituents of growing size at the parent lin-benzoguanine scaf-
3
–7
fold.
The target protein performs a complete nucleobase
8
exchange at the wobble position of specific tRNAs. Inhibition of
bacterial TGT may serve as a putative therapeutic concept for drug
development as its function has been linked to the pathogenicity of
9
Shigella, the causative agent of bacterial dysentery. In developing
countries, Shigellosis is a major threat as infections with these bac-
teria are responsible for more than 100,000 lethal cases every
⇑
Corresponding author.
E-mail address: klebe@mailer.uni-marburg.de (G. Klebe).
Equally contributing authors.
10–12
year.
In previous studies, we attached substituents at the 2-
7
y
and 4-position of the parent scaffold and in combination of both.
http://dx.doi.org/10.1016/j.bmc.2016.07.053
968-0896/Ó 2016 Elsevier Ltd. All rights reserved.
0
Please cite this article in press as: Neeb, M.; et al. Bioorg. Med. Chem. (2016), http://dx.doi.org/10.1016/j.bmc.2016.07.053

2
M. Neeb et al. / Bioorg. Med. Chem. xxx (2016) xxx–xxx
The 2-substituents orient into an open, bowl-shaped pocket that
recognizes uracil 33 with its adjacent ribose ring of the natural
1
3
substrate tRNA. In addition, we have developed derivatives,
exhibiting substituents at the 4-position, designed to occupy the
ribose-34 pocket. In the subsequently determined crystal struc-
tures, the position of the 4-substituent could be assigned unequiv-
ocally to the difference electron density, while the substituents,
anchored at C2, indicated in most of the determined crystal struc-
tures enhanced mobility or at least a scatter across multiple con-
formational states to varying extend.4
,7,13,14
This observation
suggests that the 2-substituent, depending on its chemical compo-
sition, experiences pronounced flexibility in the bound state.
Closer inspection of the architecture of the uracil-33 pocket
indicates the exposure of polar and non-polar residues in this bind-
ing cavity. In consequence, lin-benzoguanines with alternative 2-
substituents were synthesized following the design hypothesis to
fix the regarded ligand portion in the uracil-33 pocket by address-
ing the available polar contacts of the protein. In the present study,
we investigated ligands exhibiting an extended phenethyl sub-
stituent in 2-position by molecular dynamics (MD) simulations,
crystal structure analysis, isothermal titration calorimetry (ITC),
and by a radioactive washout-binding assay. At their terminal
end, the phenethyl substituents were decorated with different
polar functional groups, principally capable to establish the desired
contacts with the protein. The consequences of these modifications
are described with respect to the obtained binding poses, dynam-
ics, and energetic properties.
Scheme 1. Chemical formulas of the investigated ligands 1–4.
Table 1
Chemical formulae and binding affinities of the extended 2-amino-lin-benzoguanines
, 6a–d
5
Compound
i
K [nM]
H
10 ± 3a
2 ± 2
5
6
O
a
O
HO
6b
7 ± 3
N
2
2
. Results
6c
21 ± 8
30 ± 7
O
.1. Ligand design
-O
6d
We recently reported on the crystal structures of the parent
a
K
i
value according to Lit. 6.
2-methylamino-lin-benzoguanine (1), the 2-thienylmethyleneamino
(
(
2), 2-piperidylethylenamino- (3), and 2-morpholinoethylenamino
4) derivatives (Scheme 1).1
3,14
the cleavage of the sulfamoyl protection group. One possible expla-
nation for the poor-yielding last step (7–31%) is the observed low
solubility. The lin-benzoguanines as extended version of guanine
are nearly insoluble in most organic solvents (e.g. THF, MeCN, ace-
All crystal structures showed the tricyclic ring system in
ordered state with full occupancy. In the complex crystal structure
with 1, a network of five water molecules is found wrapping
around the terminal methyl group. The two derivatives 3 and 4
with an ethylene linker show the 2-substituent in at least two con-
formations whereas the thienylmethyl substituent in 2 is scattered
over multiple arrangements. Possibly, the ethylene linker is better
suited to place the substituent in ordered fashion. We therefore
selected 5 decorated with a phenethylamino moiety as a starting
point for our ligand design (Table 1). Different to 3 and 4, the phe-
nethyl substituent does not bear an additional protonation site that
may potentially lower the permeation through cell membranes at a
physiological pH value and would thus contribute unsatisfactorily
2 2 2
tone, CH Cl , ethyl acetate) and H O, which results in a difficult
purification process. Canonical nucleobases are known to exhibit
poor solubility due to their ability to form a strong hydrogen bond
network. In the case of lin-benzoguanines, the extension of the ring
system probably further decreases the solubility by strong inter-
molecular
nines either
recrystallization in several solvent systems failed. RP-HPLC with
C18-coded silica and H O/MeCN with either 0.1 vol % NH OAc or
NH OH as solvent system enabled the isolation of pure samples
of the products.
p–p
stacking interactions. Purification of lin-benzogua-
by silica-based normal-phase HPLC or
2
4
4
1
4,15
to bioavailability.
This aspect is of some importance as we planned to attach in
para-position of the terminal phenyl ring substituents of different
polar and hydrogen-bonding character to interact with the polar
functional groups of amino acid residues exposed to the uracil-33
pocket.
Starting from the commercially available benzimidazole-5-car-
boxylic acid (7, not shown), the synthesis of the regioisomeric 2-
bromobenzimidazoles 8a and 8b (Scheme S1 in the Supplementary
material) followed the sequence esterification, nitration, sulfamoyl
protection, and bromination (Schemes 2 and 3). Both brominated
regioisomers 8a and 8b were used for the synthesis of the lin-ben-
zoguanines 6a–d.
2
.2. Synthesis
Substitution of the 2-bromobenzimidazole-6-carboxylate 8a
with primary ammonium chloride 9a, which was prepared by
acid-catalyzed esterification of the corresponding acid 10 in
methanol, gave the methylated 2-amino-5-nitrobenzimidazole
11 (Scheme 2). The solubility of the bromide in EtOAc was
increased by the addition of methanol, enhancing the velocity of
the reaction. The 5-nitrobenzimidazole 11 was subsequently
reduced to the corresponding amine 12, which showed some
The synthesis route of 2-substituted lin-benzoguanines has
1
5,16
been described by us
and was successfully applied to the
1
7
preparation of 6a–d (Table 1). The last step of the synthesis
requires relatively high temperature (150 °C) to cyclize the sul-
famoyl-protected aminobenzimidazoles with chloroformami-
dinium hydrochloride (Schemes 2 and 3). Simultaneously, this
temperature and the generation of two equivalents of HCl led to
Please cite this article in press as: Neeb, M.; et al. Bioorg. Med. Chem. (2016), http://dx.doi.org/10.1016/j.bmc.2016.07.053

M. Neeb et al. / Bioorg. Med. Chem. xxx (2016) xxx–xxx
3
Scheme 2. Synthesis of elongated lin-benzoguanines 6a,d starting from methyl 2-bromobenzimidazole-6-carboxylate 8a: (a) HCl, MeOH, 1,4-dioxane, reflux, 16 h, 68%; (b)
a, NEt , EtOAc, MeOH, 23 °C, 16 h, 86%; (c) Zn powder, AcOH, H O, 23 °C, 3 h; (d) ClC(NH)NH Cl, (CH SO , 150 °C, 5 h, 13% (from 11); (e) NH OH, 120 °C, 1.5 h, 16%.
9
3
2
3
)
3 2
2
4
Scheme 3. Synthesis of the lin-benzoguanines 6b and 6c starting from methyl 2-bromobenzimidazole-5-carboxylate 8b: (a) MeOH, H
2
SO
, 1-methylimidazole, MeCN, 23 °C, 1 h, 37%; (d) HNPhth, PPh3, DIAD, THF, 23 °C, 16 h, 83%; (e) N 4, EtOH, 90 °C, 16 h, 69%; (f) EtN(iPr)
2
O, 23 °C, 3–5 h; 19a: 50%; 19b: 44%; (h) ClC(NH)NH Cl, (CH SO , 150 °C, 4–5 h, 6b: 7%, 6c: 31%. TBDMS: tert-
4
, 80 °C, 94 h, 90%; (b) LiAlH
4
, THF,
2
3 °C, 16 h, 94%; (c) TBDMSCl, I
3 °C, 16 h, 18a: 49%, 18b: 74%; (g) Zn powder, AcOH, H
2
2
H
2
, EtOAc,
2
2
3
3
)
2
butyldimethylsilyl; DIAD: diisopropyl azodicarboxylate; Phth: phthaloyl.
instability upon light and air exposure, and was directly used with-
out purification for the cyclization with chloroformamidinium
hydrochloride (ClC(NH)NH Cl). At elevated temperatures, lin-ben-
3
of 18c (Scheme 3). The reduction of the nitro derivatives 18a and
18b with zinc powder in AcOH gave the amines 19a and 19b,
which were sensitive to light and air exposure; the higher stability
of 19a allowed its purification, a longer storage time, and a full
characterization. The above-mentioned cyclization with chlorofor-
mamidinium hydrochloride provided the lin-benzoguanines 6b
(7%) and 6c (31%).
zoguanine 6a was prepared in 13% yield over two steps. Treatment
of the ester 6a with aqueous ammonia at 120 °C gave 16% of the
acid 6d after RP-HPLC purification.
The 2-bromobenzimidazole-5-carboxylate 8b was used as the
starting material for the preparation of the lin-benzoguanines 6b
and 6c, which contain the ethanol and nitrile functionalities,
respectively (Scheme 3). Coupling partners were the commercially
available cyanophenethylammonium chloride (9c) and the pro-
tected amino alcohol 9b, which was prepared from 1,4-phenylene-
2.3. Binding affinity
Binding affinities were determined by a radioactive washout
3
Tyr
assay. Thereby, the incorporation of [8- H]guanine into tRNA
1
8
diacetic acid (13) by acid-catalyzed esterification to bis-ester 14,
(ECY2) at position G34 is measured by liquid scintillation counting
at pH 7.3 and 37 °C (see Section 5). Inhibition constants are calcu-
lated by the comparison of the initial velocities of the base-
exchange reaction in absence and presence of the ligand. The
reduction to the diol 15,19 partial silylation to 16, substitution to
the phthalimide 17, and dephthaloylation (Gabriel synthesis).
The coupling of 8b with 9b and 9c was done in the presence of
N,N-diisopropylethylamine in EtOAc at room temperature and led
to a complete consumption of 8b affording 49% of 18a and 74%
20
ligands 1–5 and 6a–d show a binding affinity (K
i
value) in the
low one- to two-digit nanomolar range (Table 1). Compared to
Please cite this article in press as: Neeb, M.; et al. Bioorg. Med. Chem. (2016), http://dx.doi.org/10.1016/j.bmc.2016.07.053

4
M. Neeb et al. / Bioorg. Med. Chem. xxx (2016) xxx–xxx
the unsubstituted reference ligand 1, the phenethyl derivative 5 is
conformation resulting in an analogous binding mode as observed
2
2
about five times more potent and exhibits a binding affinity of
exclusively for 6a (Bsubstituent ¼ 21:0 Aꢀ vs Bcore ¼ 12:3 Aꢀ ). In this
6
about 10 n
M.
Interestingly, the binding affinity of methyl ester
orientation, the substituent refines to 44% occupancy.
6
a is enhanced by a factor of five. In 6b, further decoration of the
We studied a second, independently collected data set using
another crystal of this complex. In the diffraction data taken from
the second crystal, the ligand is visible in the difference electron
density in only the all-trans conformation, however, a slightly
reduced occupancy for the 2-substituent of 84% results from
refinement. Also in this structure enhanced mobility of the 2-sub-
stituent is experienced (Bsubstituent ¼ 27:3 Aꢀ vs Bcore ¼ 13:4 Aꢀ ) and
the position of the hydroxyethyl portion can hardly be assigned
to the density. Further analysis of the difference density does not
allow reliable assignment of the substituent to a second orienta-
tion as in the first data set.
phenyl moiety does not yield a significant change in K
trary, 6c and 6d are slightly decreased in potency.
i
and in con-
2
.4. Crystal structure analysis
For the investigated ligands, crystal structures of the complexes
2
2
have been determined with a resolution of 1.14–1.40 Å. In all
structures, the fully occupied lin-benzoguanine scaffold is well-
defined in the difference electron density (Fig. 1). It is placed into
the guanine-34 recognition pocket undergoing
p-stacking with
the side chains of Tyr106 and Met260 and establishing the same
The 2-substituent of 6c also adopts an all-trans conformation
penetrating into the adjacent ribose-32 subpocket (Fig. 1D). The
entire ligand is fully occupied, nevertheless, its 2-substituent
shows increased temperature factors compared to the parent scaf-
fold (Bsubstituent ¼ 18:3 Aꢀ vs Bcore ¼ 10:2 Aꢀ ). The terminal nitrile
group does not experience specific interactions with any of the
amino acid residues found in the ribose-32 subpocket. Only a sin-
gle water molecule is located in close neighborhood (3.0 Å) to the
nitrile functional group. The side chain of Arg286 has to reshuffle
its orientation and it is shifted out of the subpocket adopting the
previously observed conformation.
hydrogen-bonding interactions to neighboring amino acids as pre-
3
–7
viously described.
The difference electron density jF
o
j ꢀ jF j fully defines the bind-
c
2
2
ing mode of 5 (Fig. 1A) including the phenethyl substituent. Never-
theless, in this complex larger B-factors are observed for the 2-
substituent compared to the tricyclic parent core scaffold
2
2
(B
substituent ¼ 34:1 Aꢀ vs Bcore ¼ 17:8 Aꢀ ) (for definition of core and
substituent, see Table S1 in the Supplementary material) indicating
enhanced residual mobility of this moiety. The phenethyl sub-
stituent occupies the uracil-33 pocket adopting an all-trans confor-
mation. Weak hydrophobic interactions are experienced between
the side chains of Ala232, Cys281, Val282, Leu283, and the 2-sub-
stituent which covers in lid-like fashion the hydrophobic residues
in this pocket.
The binding mode of 6d is fully defined in the difference elec-
tron density jF
o
j ꢀ jF
c
2 2
j (Fig. 1E). The C–CH –CH –NH linker adopts
the all-trans conformation pointing into the uracil-33 subpocket.
Under the applied pH conditions (pH 7.8), the terminal carboxylate
group is most likely deprotonated. In this state, it experiences a
bidentate salt-bridge to the guanidinium moiety of Arg286. In
order to establish this contact, the polar group of Arg286 bends
toward the ribose-32 subpocket. Despite of the strong electrostatic
interaction between the carboxylate group of the ligand and the
side chain of Arg286, the 2-substituent exhibits similarly increased
The 2-substituent of 6a refines to a reduced occupancy of 75%
indicating higher flexibility of this moiety compared to that
2
2
observed for 5 (Bsubstituent ¼ 30:7 Aꢀ vs Bcore ¼ 15:1 Aꢀ ). The C–
CH –CH –NH linker adopts an energetically most likely less favor-
able gauche conformation with a torsion angle of ꢀ61.3° (Fig. 1B).
2
2
1
3
Contrary to the gauche conformation of the 2-substituent in 4,
2
the one in 6a adopts a conformation which orients the methyl ben-
zoate moiety out of the uracil-33 pocket facing the backbone of
Val233 and Gly234. Obviously, the extended all-trans conforma-
tion of this substituent is not adopted, as the ester moiety would
interfere with Arg286. Instead, the substituent prefers to fold back
into the described upwards conformation moving the ligand out of
the uracil-33 pocket. Weak van-der-Waals interactions are formed
to the side chain of Ala232. Additionally, a weak hydrogen-bond
contact between the backbone NH group of Gly234 and the car-
bonyl oxygen of the ester group in 6a is formed (3.5 Å). Remark-
ably, this back-folded conformation is observed for the first time
for an attached 2-substituent.
For 6b two data sets collected on different crystals show deviat-
ing results with respect to the substituent’s binding mode (Fig. 1C).
Refinement of the first structure suggests the 2-substituent to be
present in two conformations with a summed overall occupancy
of 100%. With increasing distance from the parent scaffold, the dif-
ference electron density becomes more blurred and accordingly
the assigned temperature factors increase. In consequence, for both
conformers, the substituent is not sufficiently resolved in the dif-
ference electron density due to the residual mobility of that por-
tion of the molecule resulting in a partly defined difference
electron density for the hydroxyethyl moiety. In the first con-
B-factors like the other studied derivatives (Bsubstituent ¼ 22:4 Aꢀ vs
2
B
core ¼ 11:5 Aꢀ ). For comparison, the binding mode of the natural
substrate tRNA is shown in Figure 1F. The bound substrate induces
the same conformation of Arg286 as found in the complexes
TGTꢁ6b and TGTꢁ6c. It opens an additional pocket (highlighted in
light blue) to accommodate the ribose moiety of base at position
32 and forms two charge-assisted hydrogen bonds to the guani-
dinium group of Arg286.
2.5. MD simulations
The crystal structures of 6a–d indicate distinct residual mobil-
ity of the considered substituents directed into the uracil-33
pocket. As indicated by the two data sets recorded for two dis-
tinct crystals of TGTꢁ6b, the observed disorder may also depend
on the crystallization protocol used to grow the crystals. We
therefore performed molecular dynamic simulations using the
2
1
program AMBER under NTP conditions (constant number of par-
ticles, constant pressure of p = 1 bar, constant temperature of
T = 300 K) to assess whether qualitatively the same structural
properties are indicated by the computer simulations as in the
crystalline state. All four complexes of 6a–d were simulated over
a productive simulation time to 100 ns. Subsequently, the ligand
conformations visited along the trajectory were hierarchically
clustered into families with maximal RMSDs of 2 Å using the pro-
gram ptraj (AMBER program suite) and the representative arche-
types were visually inspected. The number of observed
conformation families is listed in Table 2 and the spatial structure
of some family representatives are given in Figure 2. In the Sup-
plementary material, images of all conformer family members
can be found (Fig. S1).
2 2
former, the C–CH –CH –NH linker adopts an all-trans conforma-
tion with an occupancy of 56% and enhanced residual mobility
2
2
(B
substituent ¼ 21:6 Aꢀ vs Bcore ¼ 12:3 Aꢀ ). The adopted orientation is
similar to that observed for 5. Interestingly, the neighboring side
chain of Arg286 is shifted out of the uracil-33 pocket presumably
to create sufficient space to accommodate the hydroxyethyl moi-
ety of the ligand, now extending its binding pose even deeper into
the pocket. In the second conformer, the linker exhibits a gauche
Please cite this article in press as: Neeb, M.; et al. Bioorg. Med. Chem. (2016), http://dx.doi.org/10.1016/j.bmc.2016.07.053

M. Neeb et al. / Bioorg. Med. Chem. xxx (2016) xxx–xxx
5
Figure 1. Binding modes of the analyzed lin-benzoguanines as well as a tRNA analogue. The solvent-accessible protein surface is shown in white. The ligands and interacting
residues are displayed as sticks (nitrogen = blue, oxygen = red, phosphorus = orange). Hydrogen bonds are visualized as dashes. The jF j ꢀ jF j difference electron density is
illustrated at a sigma-level of = 2.5 as green meshes. For clarity, the -stacking residues Tyr106 and Met260 as well as water molecules are not shown. A–E) The tricyclic
o
c
r
p
parent scaffold is well-defined in the difference electron density experiencing hydrogen bonds to residues Asp102, Asp156, Gln203, Gly230, Leu231 and Ala232 within the
guanine-34 recognition site. Depending on its substitution pattern, the phenethyl substituent in 2-position adopts either an orientation within the uracil-33 pocket or is
rotated out of this position now facing the backbone of Val233 and Gly234. Only 6d (E) forms additional polar interactions to the guanidinium head group of Arg286. While in
the complex structures TGTꢁ5 (A), TGTꢁ6a (B) and TGTꢁ6d (E) the side chain of Arg286 closes the preceding pocket at position 32, 6b (C) and 6c (D) bind to the site by shifting
the side chain of Arg286 apart. For TGTꢁ6b (C), the results collected at two independent crystals are shown. In one crystal structure (ligand C-atoms dark blue), an expended
conformation with an occupancy of 84% of the substituent was found. In the second crystal (ligand C-atoms light blue), a scatter over two orientations (extended and back-
folded) with occupancies of 56% and 44% could be refined. (F) Binding of the natural substrate tRNA (PDB ID 1Q2R) induces the same conformation of Arg286 as found in the
complexes TGTꢁ6b and TGTꢁ6c opening the pocket accommodating the ribose moiety of base at position 32 (light blue). The ribose moiety at position 32 interacts with the
side chain of Arg286 via two charge-assisted hydrogen bonds.
Please cite this article in press as: Neeb, M.; et al. Bioorg. Med. Chem. (2016), http://dx.doi.org/10.1016/j.bmc.2016.07.053

6
M. Neeb et al. / Bioorg. Med. Chem. xxx (2016) xxx–xxx
Table 2
Occupancy of the different conformer families in percentage found as clusters along the MD trajectory of 6a–d (calculated by ptraj after
hierarchical clustering)
a
Conformer clusters approaching the conformers in the crystal structures most closely are highlighted in gray.
Conformers with a relative occupancy of more than 10% are highlighted in bold.
b
c
Conformer approaching closest the 84% populated conformer in the first and 56% populated in the second X-ray structure.
d
Conformer approaching closest the 44% populated conformer in the second X-ray structure.
All simulations found among others a binding pose of the 2-
tion crystal structure, assignment of an additional conformer
appears appropriate if it is indicated by additional electron density
and its occupancy refines to at least 15–20%.
substituent approaching the one observed in the crystal structure
within an RMSD between 0.4 Å and 1.6 Å. Nevertheless, in all cases
the MD simulations suggest an additional number of possible con-
formers. The substituent conformation of 6a matches in the second
most populated conformer well with the geometry (RMSD = 0.4 Å)
found in the crystal structure, the most populated one agrees bet-
ter with a trans orientation of the substituent. For the other ligands
The enhanced mobility of the 2-substituents involves in case of
6a and 6b a back folding of this part of the ligand combined with a
partial abandonment of the pocket. In one data set, collected on
another independent crystal of TGTꢁ6b, a second conformation
could be attributed to the difference electron density. Molecular
dynamics simulations also assign enhanced mobility to this part
of the ligand. They suggest an even larger scatter over multiple
configurations extending those observed in the crystal structures.
Many of these, however, show a population well below 15–20%.
In a crystallographic diffraction experiment, an additional con-
former will only be indicated in the density if a periodic arrange-
ment from unit cell to unit cell is given.
6
b, 6c, and 6d, a conformer close to the crystal structure is also
observed, however, to a minor fraction along the trajectory. Also
here, the occurrence of additional conformers with gauche orienta-
tion is indicated, in agreement to the disorder indicated in the first
data set collected from a crystal of TGTꢁ6b.
2
.6. Thermodynamic analysis
As the binding affinity is not strongly altered across the ligands
showing deviating residual mobility, this property cannot be detri-
mental to binding. This underlines the thermodynamic considera-
tion that residual ligand mobility takes a beneficial contribution to
the entropic portion of the Gibbs binding energy. The number of
conformational clusters found for the studied ligands by MD sim-
ulations differs, however; in all cases, the conformer observed in
the crystal is also traversed along the trajectory with a similar
geometry. In the packing environment of a crystal structure, mole-
cules are accommodated under restricted conditions. The local
crystal environment limits the available space to evolve the
dynamic properties of a protein-ligand complex. In case of trypsin,
To reveal a more detailed picture of the thermodynamic proper-
ties of ligand binding, we performed ITC titration experiments. The
novel data for 5 and 6d are listed in Table 3 together with those of
1
–4, reported previously.13
3
. Discussion
In the present study, we investigated the conformational prop-
erties and the residual mobility of extended substituents oriented
into a flat bowl-shaped binding pocket that recognizes the nucle-
obase uracil-33 with its attached ribose sugar ring of the natural
tRNA substrate. Crystal structure analysis suggests residual mobil-
ity of the placed substituents in the protein bound state. This resid-
ual dynamic property is indicated by enhanced B-factors attributed
to the atoms of the substituents. Not in all cases could the popula-
tion of the 2-substituent be refined to 100%, thus we assume that
minor populated conformers are present that scatter over several
configurations. In the crystallographic refinement of a high-resolu-
we experienced
a remarkable selection of deviating crystal
2
3
forms. Remarkably, ligand binding was only successful in a par-
ticular crystal form if the ligand could establish sufficient residual
mobility in the bound state. It accommodated only successfully in
the crystal packing if enough unoccupied space next to the binding
pocket was accessible so that the ligand could experience the
residual mobility necessary for its binding. We even observed
Please cite this article in press as: Neeb, M.; et al. Bioorg. Med. Chem. (2016), http://dx.doi.org/10.1016/j.bmc.2016.07.053

M. Neeb et al. / Bioorg. Med. Chem. xxx (2016) xxx–xxx
7
crystal.26 The applied cooling protocol can be crucial for the results
Binding poses
Occurrence
RMSD
2
7
observed in the determined crystal structure. Systematic studies
have shown that molecular motions such as the adjustment of
side-chain rotamers will find sufficient time to equilibrate accord-
ing to the conformational distribution given at low temperature.26
However, adjustment involving larger movements and transforma-
tions such as the flipping of entire ring systems or larger rearrange-
ments of ligand portions may be too slow to adjust with flash
cooling, thus the situation at ambient in solution is probably
reflected in the solid state. Supposedly, the time needed to cool
below the glass transition temperature of water in the solvent
channels spanning through a protein crystal is crucial for the fixing
of molecular motions. Finally, as mentioned above, periodicity in
the arrangement across the crystal must be given to detect a
molecular portion in the diffraction pattern, whereas complete
random distribution will remain unresolved.
1
9.0%
6.4%
5.6 Å
0.4 Å
6
a
b
1
4
6.8%
.3%
4.9 Å
0.4 Å
6
4
For 6b, we could determine two independent crystal structures
and the conformational multiplicity of the substituent differs
between the two structures. The back-folded conformation is only
observed in one crystal, whereas the extended conformer is found
in both determinations. Presumably, the latter conformer corre-
sponds to the energetically more stable situation. In case the
adjustment of this conformer into the most stable configuration
is slow, a dependence on the cooling rate of the crystals below
the glass transition temperature of water could be given. This
can lead to deviating populations of conformational states in the
measured crystal structures. It remains speculative that in the
crystal showing only one predominant conformer, the ligands
had more time to properly equilibrate upon cooling.
7
6.1%
.5%
3.5 Å
1.6 Å
6
c
2
1
7.3%
.1%
1.7 Å
1.0 Å
6d
6
Facing the MD trajectory with the different crystal structures
suggests that residual mobility is significantly reduced in the solid
state (even though populations beyond 15–20% are difficult to rec-
ognize in the diffraction pattern). Nonetheless, scatter over multi-
ple states and enhanced B-factors serve as a qualitative indicator
for residual mobility of the ligand in its bound state.
Figure 2. Results of the hierarchical clustering of the frames observed along the
trajectory of the 100 ns MD simulations. The conformation observed in the crystal
structure (carbon atoms yellow) is superimposed with the most populated
conformer (carbon atoms light green) and the conformer approximating most
closely the crystal structure (carbon atoms light blue) together with the occupancy
in percentage of the simulation time and the RMSD with respect to the geometry in
the crystal structure. RMSD values (in Å) were calculated using fconv after
Most likely, the cooling protocol, the quality and size of the
crystal, as well as the rate by which the experimentalist performs
the plunging of the crystal into liquid nitrogen and the thickness of
cold-gas layer formed above the liquid cryogen (the transfer takes
several milliseconds where major part of the crystal cooling
already happens) take an impact on the resulting picture in the
2
2
alignment on the tricyclic scaffold without hydrogens.
decomposition of the crystals if ligands requiring binding with
residual mobility were soaked into crystals not offering a large
enough volume. Vice versa, the conformationally restricted ligands
did not populate in this crystal form. In another study, we evalu-
ated in total 17 data sets of a complex of aldose reductase with
2
7
crystal structure. Even the liquid cryogen used for cooling can
influence the results, as different cryo media (e.g. nitrogen, ethane,
helium) exhibit different heat capacities and heat conductivities to
absorb and conduct heat through the crystal upon flash freezing. As
these parameters are difficult to control by the experimentalist in a
systematic fashion, a reproducible protocol will hardly be possible
to accomplish. However, for the interpretation of crystal structure
data this aspect has to be taken into account. In case of aldose
reductase, we observed a peptide bond flip in the refined struc-
tures of a protein-ligand complex, recorded for data sets taken
2
4
zopolrestat. Crystals were obtained under varying soaking and
co-crystallization conditions. Depending on the applied protocol,
a flip of a peptide bond was observed, accompanied by a rupture
of an H-bond formed to the bound ligand.
Protein crystals are usually studied under cryo conditions at
1
00 K to limit radiation damage.25 Unfortunately, it is difficult to
define a temperature corresponding to the situation seen in a fro-
zen crystal structure. The flash freezing of a crystal will clearly take
an impact on residual molecular motions and the probability
distribution of putative conformers actually seen in the studied
2
4
from multiple crystals. The amount by which this flip had taken
place was dependent on the protocol how the crystals were grown
and manipulated prior to measurement.
Table 3
Different terminal functional groups were attached to the para
position of the phenyl ring at the 2-substituent to provide this
ligand portion with a facility to undergo directional interactions
with the protein. In case of the attached carboxylate group, the
ligand vector departing from the 2-position recruits the guani-
dinium group of Arg286 to form a salt-bridge and it adopts an
ordered conformation. As the assay data show, the formed sol-
vent-exposed salt-bridge contributes hardly anything to the bind-
ing affinity of the ligand. ITC confirms this observation. A virtually
unchanged thermodynamic profile is observed for 5 and 6d. This
example underlines the fact that a salt-bridge, formed between
ligand and protein, can only enhance binding if it forms in a deeply
Raw data of the investigated ligands measured at 25 °C in Hepes buffer
Ligand
K
d
[n
M
]
D
[
G0
kJ molꢀ1]
D
H
obs
ꢀT
D
S0
ꢀ1
[kJ mol ]a
ꢀ1
[kJ mol ]
1
2
3
4
5
52.4 ± 6.9
ꢀ41.6 ± 0.3
ꢀ74.8 ± 1.8
ꢀ50.9 ± 0.6
ꢀ66.7 ± 0.5
ꢀ78.3 ± 1.2
ꢀ45.3 ± 1.8
ꢀ47.0 ± 1.4
33.2 ± 1.9b
11.2 ± 0.6
24.1 ± 0.7
35.7 ± 1.3
3.9 ± 1.9
111.6 ± 11.8 ꢀ39.7 ± 0.3
34.3 ± 6.5
35.0 ± 6.9
56.6 ± 14.3
29.8 ± 8.9
ꢀ42.6 ± 0.5
ꢀ42.6 ± 0.5
ꢀ41.4 ± 0.6
ꢀ43.0 ± 0.8
6
d
4.0 ± 1.6
a
S0 was calculated according to the Gibbs–Helmholtz equation.
Errors were estimated by means of standard deviation. The error for ꢀT
ꢀ
TD
b
S0 was
D
calculated according to error propagation.
Please cite this article in press as: Neeb, M.; et al. Bioorg. Med. Chem. (2016), http://dx.doi.org/10.1016/j.bmc.2016.07.053

8
M. Neeb et al. / Bioorg. Med. Chem. xxx (2016) xxx–xxx
buried binding pocket exhibiting low dielectric conditions and
excluding access of solvent molecules. In a previous study, we
could show that similar salt-bridges, formed at the opposing end
of the lin-benzoguanine scaffold, which becomes entirely buried
was started by the 1:1 addition of a similarly prepared solution
Tyr
containing E. coli tRNA (ECY2; 3
l
M
) and a mixture of guanine
and radioactively labelled [8- H]guanine (20 ). 15 L aliquots
were removed from this mixture (76 L) every hour and pipetted
3
lM
l
l
1
3
in the pocket, make a huge enthalpic contribution to binding.
on Whatman GC-F glass microfiber filters. The reaction was imme-
The molecular dynamics simulations even indicate that the salt-
bridge found in the crystal structure is not stable along the trajec-
tory and forms only temporarily.
diately quenched in 10% (w/v) trichloroacetic acid solution at 0 °C.
In order to separate the tRNA from excess [8- H]guanine, not incor-
3
porated into tRNA, the glass microfiber filters were washed twice
in a 5% (w/v) TCA solution for 10 min, followed by an additional
washing step using technical grade ethanol over 20 min. The
labelled tRNA was captured in the filters during the described
steps. Afterwards, the filters were dried at 60 °C for at least
4
. Conclusions
Protein-ligand binding occurs overwhelmingly in surface-
3
0 min. The resulting count rate was obtained after the addition
exposed depressions and cavities on a protein. The available pock-
ets are of different shape, burial, and physicochemical composition.
The affinity of ligand binding to such pockets depends on the
strength of the interactions established in these pockets, the
change in the dynamic properties and thus in the degrees of free-
dom of the system, and the modulation of the local solvent
structure.
TM
of 4 mL Rotiszent to each filter by liquid scintillation counting.
i
K values were determined using the method described by Dixon
3
0
at least in duplicate. Thereby, a Michaelis–Menten constant of
ꢀ1
0
.9
l
molꢁL was used for data evaluation.
5
.2. Synthesis
Usually in deep and highly buried binding pockets, major con-
tributions to binding affinity are experienced. Nonetheless, also
accommodation of ligand portions to flat solvent-exposed pockets
can be essential to optimize affinity and particularly selectivity.
However, how much affinity can be gained by filling such flat pock-
ets, and should the binding pose of a ligand be restricted to one ori-
entation or is a residual mobility with a scatter over multiple
orientations beneficial for binding? Considering conformational
entropy and the restriction to a small number of degrees of free-
dom, ligand disorder can be of advantage for binding.
The individual synthesis protocols and the analytical character-
izations are given in the Supplementary material.
5
.3. Isothermal titration calorimetry
A Microcal iTC200 microcalorimeter system (Malvern) was used
to perform the ITC measurements. The protein was dissolved in the
experimental buffer (50 m Hepes, 200 m NaCl and 0.037%
Tween 20, pH 7.8) to a final concentration of 10 containing
% DMSO. Due to their low solubility, the ligands were first dis-
M
M
lM
3
In the present study, we investigated the properties of a series
of 2-substituted lin-benzoguanine inhibitors binding to the tRNA-
modifying enzyme TGT. The attached substituents were designed
to fill the flat bowl-shaped uracil-33 binding pocket of the protein.
Crystal structure analysis revealed either significantly enhanced
residual mobility of the attached substituents in the pocket indi-
cated by larger B-factors or a distribution over multiple orienta-
tions is found. Molecular dynamics simulations show an even
larger scatter over several clusters of deviating side chain conform-
ers. In the crystal environment, molecular motion is restricted and
the flash-cooling process of the crystals freezes or at least reduces
mobility to a certain amount, which is difficult to quantify.
Nonetheless, some configurations experienced along the computed
MD trajectory agree with geometries similarly found in the solid
state. Most likely, the flash-cooling protocol takes impact on the
final scatter seen in the crystal structures. The interpretation of
the properties of a particular binding pose seen in a crystal struc-
ture must consider this fact. Finally, the attachment of a functional
group competent to establish a salt-bridge in the flat solvent-
exposed pocket helps to restrict the bound ligand in a fixed orien-
tation; however, the formation of a solvent exposed salt-bridge has
no detectable contribution to binding affinity.
solved in 100% DMSO and then diluted in the buffer solution to a
final DMSO concentration of 3%. The ligand concentration in the
syringe was adjusted to 200
lM with the same experimental buffer.
All ITC experiments were run at 25 °C after a stable baseline had
been achieved. The reference cell was filled with filtered deminer-
alized water. The initial delay before the injections were started
and the spacing between each injection were adjusted to 180 s.
The first injection contained 0.3–0.5
lowed by 14–24 injections of 1.0–1.8
000 rpm was adjusted. Raw data were collected as released heat
per time.
lL of the ligand solution fol-
lL. A stirring speed of
1
To analyze the raw data using the Origin 7.0 software, the base-
line and integration limits were adjusted manually. After integrating
the area under the peaks, the first data point was removed due to a
3
1
reduced accuracy. The influence of the heat of dilution was cor-
rected considering heat contributions collected after saturation of
0
the protein. K
d
(dissociation constant) as well as
DH
(enthalpy of
binding) were extracted by applying a single-site binding model as
0
provided by the manufacturer. Subsequently, ꢀT
DS was calculated
according to the Gibbs–Helmholtz equation. For the current protein-
ligand system, we have detected a buffer dependence of the heat sig-
1
4
nal owing to a protonation linkage. Since the groups involved in
the change of protonation state are remote from the attached acidic
group in 6d, data collected in one bufferis sufficient to determine the
relative differences between the ligands reported in Table 3. The val-
ues should not be used for comparisons on absolute scale.
5
5
. Experimental section
.1. Enzyme assay
The enzyme kinetic characterization of TGT is based on a
5.4. X-ray data
method described by Grädler et al.28 and Stengl et al. Due to
low solubility, the ligands were dissolved in 100% DMSO and sub-
sequently diluted to the desired concentration with 5% DMSO with
the assay buffer. The protein dissolved in the same buffer was
added to the various ligand samples with a final concentration of
29
5.4.1. mobilis TGT crystallization and ligand soaking
Z. mobilis TGT was overexpressed and purified as described else-
3
2,33
where.
Crystals were grown using the sitting drop vapor diffu-
sion method at 291 K in the presence of the inhibitor. The protein
ꢀ1
9
nM
and incubated for 10 min at 37 °C. Additionally, a reference
solution was adjusted to 12 mgꢁmL by dilution with high salt
buffer (10 m Tris, 2 m NaCl, 1 m EDTA, pH 7.8) and incubated
with the inhibitor previously dissolved in 100% DMSO at a final
without the ligand only containing DMSO and buffer was incu-
bated under the same conditions. Subsequently, base exchange
M
M
Please cite this article in press as: Neeb, M.; et al. Bioorg. Med. Chem. (2016), http://dx.doi.org/10.1016/j.bmc.2016.07.053

M. Neeb et al. / Bioorg. Med. Chem. xxx (2016) xxx–xxx
9
Table 4
Data collection, processing, and refinement statistics for the crystallographically investigated TGT-ligand-complexes
Crystal data
PDB ID
TGTꢁ5
TGTꢁ6a
TGTꢁ6b
TGTꢁ6c
TGTꢁ6d
4Q8T
4Q8U
5JXQ
4Q8V
4Q8W
(
A) Data collection and processing
Collection site
No. crystals used
k [Å]
BESSY 14.2
1
0.91841
BESSY 14.2
1
0.91841
PETRA P14
1
0.97627
BESSY 14.2
1
0.91841
PETRA P14
1
0.97627
Space group
C2
C2
C2
C2
C2
Unit cell parameters
a [Å]
b [Å]
c [Å]
b [°]
90.2
64.6
71.0
93.2
89.8
64.5
70.9
93.2
90.2
64.8
70.3
96.2
91.2
65.1
70.8
96.3
89.8
64.7
70.7
96.3
(
B) Diffraction dataa
Resolution range [Å]
30–1.40
30–1.31
(1.33–1.31)
96126
(4340)
4.8 (47.4)
99.1
80–1.20
(1.24–1.20)
123477
(11058)
2.9 (49.2)
98.3
50–1.40
(1.48–1.40)
79794
(12619)
6.2 (48.3)
97.8
80–1.14
(1.19–1.14)
134549
(15089)
2.8 (30.9)
92.0
(
1.42–1.40)
80321
3968)
Unique reflections
(
R(I)sym [%]b
Completeness [%]
5.8 (48.9)
99.9
(
98.3)
(89.9)
(94.4)
(95.8)
(80.0)
Redundancy
4.2 (3.8)
22.7 (2.4)
17.1
3.3 (2.9)
21.8 (2.0)
15.8
3.3 (2.8)
19.8 (2.2)
12.8
2.9 (2.9)
11.0 (2.0)
12.1
5.7 (5.0)
27.8 (4.7)
11.6
I/
Wilson B-factor [Å ]
Matthews Coefficient [Å /Da]
r (I)
2
3
2.4
2.4
2.4
2.4
2.4
(
C) Refinement
PHENIX version
1.8.4_1496
29.8–1.40
1949
1.8.4_1496
18.1–1.31
4815
1.8.4_1496
69.9–1.20
6174
1.8.4_1496
21.8–1.40
3990
1.8.4_1496
70.3–1.14
6728
Resolution range [Å]
Reflections used for Rfree
Reflections used for Rwork
78372
91306
117303
75804
127821
Final R valuesa
R
R
free [%]c
work [%]
15.4
13.8
16.2
14.0
15.7
13.7
16.3
13.1
15.0
13.4
d
No. of atoms (non-hydrogen)
Protein atoms
Water molecules
Ligand atoms
2899
311
24
2892
330
28
2941
351
27
2913
384
26
2927
376
27
RMSD, angle [°]
1.1
1.1
1.1
1.1
1.1
RMSD, bond [Å]
0.007
0.007
0.005
0.007
0.007
Ramachandran plote
Most favoured regions [%]
Additionally allowed regions [%]
Generously allowed regions [%]
95.3
4.4
0.3
94.9
4.7
0.3
94.6
5.1
0.3
95.0
4.7
0.3
94.9
4.8
0.3
2
Mean B-factors [Å ]
Protein atoms
Water molecules
Ligand atoms
20.7
33.8
22.6
18.8
32.8
21.2
16.6
31.7
18.6
14.6
29.4
13.0
15.1
28.0
15.5
a
b
c
Values in parentheses are statistics for the highest resolution shell.
P P
P P
RðIÞ
¼ ½
jI
i
ðhÞ ꢀ IðhÞj=
iI
i
ðhÞꢂ ꢃ 100, in which IðhÞ is the mean of the IðhÞ observation of reflection h.
sym
h
i
h
P
P
R
R
work
¼
jF
o
ꢀ F
c
j=
jF
o
j.
hkl
hkl
d
e
free was calculated as shown for Rwork but on refinement-excluded 2.4–5% of data.
36
Statistics from PROCHECK.
concentration up to 1.5 mm depending on the solubility of the cor-
responding ligand. This solution was mixed with 1.5 L reservoir
solution (100 m MES, pH 5.5, 10% (v/v) DMSO, 11–13% (w/v)
PEG 8000) to a 3 L droplet. The reservoir contained 1.0 mL of
5.5. Data collection
l
M
For data collection, the soaked crystals were transferred into a
cryoprotectant solution consisting of 50 m MES, pH 5.5, 300 m
NaCl, 0.5 m DTT, 2% (v/v) DMSO, 4% (w/v) PEG 8000, 30% (v/v)
l
M
M
the above-mentioned solution. Within one week, crystals showing
M
an appropriate size for data collection were obtained.
glycerol for 20 s followed by immediately flash-freezing in liquid
nitrogen. Data sets TGTꢁ5, TGTꢁ6a and TGTꢁ6c were collected at
the BESSY II (Helmholtz-Zentrum, Berlin, Germany) beamline
14.2 at a wavelength of k = 0.91841 Å using a Rayonix MX225
CCD detector. Complex structures for TGTꢁ6b and TGTꢁ6d were col-
lected at PETRA III (EMBL, Hamburg, Germany) beamline P14 at a
wavelength of k = 0.97627 Å using a PILATUS 6 M-F detector. To
minimize radiation damage, all data sets have been collected at
cryo-conditions (100 K).
Crystals grown in the presence of ligand 5 did not show an
appropriate size or form sufficient for diffraction experiments.
Thus, a soaking protocol was applied. For this purpose, crystals
were grown in the absence of the ligand according to the previ-
ously described protocol. Instead of mixing the protein solution
with the inhibitor before crystallization, pre-grown apo crystals
were transferred to a 3
the stock solution containing the ligand to a final concentration of
. The droplet was sealed against 1.0 mL reservoir solution and
soaked into the crystal overnight.
lL droplet of reservoir solution mixed with
1
m
M
All TGT crystals corresponded the monoclinic space group C2
containing one monomer in the asymmetric unit. Data sets TGTꢁ5
Please cite this article in press as: Neeb, M.; et al. Bioorg. Med. Chem. (2016), http://dx.doi.org/10.1016/j.bmc.2016.07.053

1
0
M. Neeb et al. / Bioorg. Med. Chem. xxx (2016) xxx–xxx
and TGTꢁ6a were processed and scaled with the HKL2000 pack-
molecules were extracted. Missing amino acids as well as the most
probable conformer of missing side chains were added. In case of
more than one visible conformation, the highest occupied one
was kept during the simulation. Protonation states of histidines
were inspected visually and set as HID (hydrogen at d-position,
3
4
age. Data processing and scaling for TGTꢁ6b, TGTꢁ6c, and TGTꢁ6d
35
were performed with XDS and XSCALE, respectively. Cell dimen-
sions, data collection, and processing statistics are given in Table 4.
5
.6. Structure determination and refinement
HIS#: 90, 257, 319, 332) or HIE (hydrogen at
3, 145, 333). The zinc ion was mimicked by four massless dummy
atoms, each with a charge of +0.5 using the CaDA approach by
Pang. Cysteine residues involved in the binding of the zinc ion
were set as CYM (CYS#: 308, 310, 313), histidines as HIN
(HIS#:349).
e-position, HIS#:
7
The coordinates of the TGT apo structure 1PUD served as a start-
4
6
ing model for molecular replacement using the program Phaser MR
of the ccp4 program suite.37 Structures were refined using the pro-
gram Phenix (phenix.refine 1.8.4_1496)38 starting with a first cycle
of simulated annealing using default parameters. Further refine-
ment cycles comprised the coordinates (xyz), occupancy, and indi-
vidual B-factor refinement as well as applying metal restraints for
the zinc ion. The individual atomic displacement parameter (ADP)
weights were refined for all structures. The temperature factors of
all structures were refined anisotropically. The calculation of the
Due to the required computational time, the calculations were
based on only one TGT monomer. Parameters for the ligands were
generated with the program antechamber using the general amber
4
7
force field (gaff), its charges were calculated via bond charge cor-
4
8,49
rection (bcc).
Addition of hydrogen atoms to the protein, neu-
tralization of the system by adding two sodium ions, and solvation
of the complex in a TIP3P water box was done with the tleap.50
After a minimization of the water box comprising 100 steps and
the whole system comprising 500 steps performed with a general-
ized Born solvent model, all following simulations included peri-
odic boundary conditions, the Particle Mesh Ewald procedure
Rfree value comprised a 2.4–5% fraction of the data.
For all structures, amino acid side chains were fitted according
to their
r
A-weighted 2jF
o
j ꢀ jF
c
j and jF
o
j ꢀ jF j electron density
c
3
9
obtained in the program Coot. After the initial refinement cycles,
the zinc ion as well as water and glycerol molecules were imple-
mented in the model. For adding water molecules, the option ‘up-
date waters’ included in Phenix was used after increasing the
hydrogen-bond length threshold for the solvent-model and sol-
vent–solvent contacts to 2.3 Å. The inserted molecules were visu-
ally inspected afterwards. Ligand restraints were generated by
the CSD based gradeWebServer [http://grade.globalphasing.org]
in case of 6b, 6c, and 6d. Ligands were built and minimized using
the program MOE [MOE 2012.10] and geometric restraints calcu-
lated subsequently by the program Monomer Library Sketcher in
case of 5 and 6a. Multiple protein-residue conformations were
assigned in case a reasonable electron density was observed and
were kept during refinement if the side chain with the lowest
occupancy showed a value of at least 20%. In case of 6a and 6b,
the occupancy of the substituent was refined due to elevated
B-factors observed for this portion. Ramachandran plots have been
5
1
52
53
(PME) and the SHAKE algorithm using the ff99SB force field
with a cut-off of 10 Å. Thereby, the system is heated up to 300 K
stepwise (0 Kꢁꢁꢁ150 Kꢁꢁꢁ225 Kꢁꢁꢁ300 K) over a period of 150 ps fixing
ꢀ1
ꢀ2
the TGT monomer with weak restraints (25 kcal ꢁ mol ꢁ Aꢀ ). Sub-
sequently, the pressure is adjusted to 1 bar over a time scale of
50 ps followed by a productive simulation for 10 to 100 ns using
2 fs time steps under NPT conditions. The trajectory derived under
these conditions was further analyzed with the program ptraj,
whereby every second frame was included into the analyses.
4
0
Acknowledgements
We are grateful to the beamline staff at BESSY II (Helmholtz-
Zentrum Berlin) and Petra in Hamburg (Desy, EMBL) for provid-
ing outstanding support during data collection. We also thank
the Helmholtz-Zentrum Berlin for the travel support. We are
grateful to the Chemical Computing Group (Montreal, Canada)
for making the MOE program available to us. The usage of the
AMBER program suite is kindly acknowledged. The present study
was supported by the ERC Advanced Grant no. 268145-DrugPro-
filBind awarded to G.K. and a grant from the ETH Research
Council presented to F.D.
36
generated with the program PROCHECK. For the analysis of tem-
perature factors the program Moleman has been used.41 The burial
of molecular ligand portions in the protein binding pocket has been
computed using the program MS.42
5
.7. Accession codes
The atomic coordinates and structure factors of the TGT com-
plex with ligands 5, 6a–d have been deposited in the RCSB Protein
Data Bank (PDB) with accession codes: 4Q8T, 4Q8U, 5JXQ, 4Q8V,
Supplementary data
4
Q8W.
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.bmc.2016.07.053.
5
.8. Docking
References and notes
Docking was performed using GOLD Suite v5.1 [The Cambridge
Crystallographic Data Centre: Cambridge, U.K.], the default values
1. Krimmer, S. G.; Betz, M.; Heine, A.; Klebe, G. ChemMedChem 2014, 9, 833.
2. Biela, A.; Nasief, N. N.; Betz, M.; Heine, A.; Hangauer, D.; Klebe, G. Angew. Chem.,
Int. Ed. 2013, 52, 1822.
43
for the genetic algorithm (GA) and the scoring function ChemS-
4
4
core. For each ligand, 30 GA runs were performed with 100,000
operations. As a protein model, the coordinates of the crystal struc-
ture 2QZR were used. The binding site included all atoms within a
distance of 10 Å to the sulfur of Met260. Ligands were drawn using
the MOE builder.
3.
4.
5.
6.
7.
8.
Stengl, B.; Meyer, E. A.; Heine, A.; Brenk, R.; Diederich, F.; Klebe, G. J. Mol. Biol.
2007, 370, 492.
Kohler, P. C.; Ritschel, T.; Schweizer, W. B.; Klebe, G.; Diederich, F. Chem. Eur. J.
2009, 15, 10809.
Ritschel, T.; Kohler, P. C.; Neudert, G.; Heine, A.; Diederich, F.; Klebe, G.
ChemMedChem 2009, 4, 2012.
Ritschel, T.; Hoertner, S.; Heine, A.; Diederich, F.; Klebe, G. ChemBioChem 2009,
1
0, 716.
5
.9. Molecular dynamic simulation
Barandun, L. J.; Immekus, F.; Kohler, P. C.; Ritschel, T.; Heine, A.; Orlando, P.;
Klebe, G.; Diederich, F. Acta Crystallogr., Sect. D 2013, D69, 1798.
Xie, W.; Liu, X.; Huang, R. H. Nat. Struct. Biol. 2003, 10, 781.
Molecular dynamic (MD) simulations were performed with the
program AMBER11.4 Docking results of the corresponding ligands
5
9. Kotloff, K. L.; Winickoff, J. P.; Ivanoff, B.; Clemens, J. D.; Swerdlow, D. L.;
Sansonetti, P. J.; Adak, G. K.; Levine, M. M. WHO Bull. 1999, 77, 651.
served as starting coordinates. Prior to the simulation, the pdb-file
had to be modified: All crystallographically determined water
10. Durand, J. M.; Okada, N.; Tobe, T.; Watarai, M.; Fukuda, I.; Suzuki, T.; Nakata,
N.; Komatsu, K.; Yoshikawa, M.; Sasakawa, C. J. Bacteriol. 1994, 176, 4627.
Please cite this article in press as: Neeb, M.; et al. Bioorg. Med. Chem. (2016), http://dx.doi.org/10.1016/j.bmc.2016.07.053

M. Neeb et al. / Bioorg. Med. Chem. xxx (2016) xxx–xxx
11
1
1
1
1
1
1. Levine, M. M.; Kotloff, K. L.; Barry, E. M.; Pasetti, M. F.; Sztein, M. B. Nat. Rev.
30. Dixon, M. Biochem. J. 1953, 55, 170.
31. Mizoue, L. S.; Tellinghuisen, J. Anal. Biochem. 2004, 326, 125.
32. Romier, C.; Ficner, R.; Reuter, K.; Suck, D. Proteins: Struct., Funct., Bioinf. 1996,
24, 516.
33. Reuter, K.; Ficner, R. J. Bacteriol. 1995, 177, 5284.
34. Otwinowski, Z.; Minor, W. Methods Enzymol. 1997, 276, 307.
35. Kabsch, W. Acta Crystallogr., Sect. D 2010, D66, 125.
36. Laskowski, R. A.; Macarthur, M. W.; Moss, D. S.; Thornton, J. M. J. Appl.
Crystallogr. 1993, 26, 283.
Microbiol. 2007, 5, 540.
2. Gu, B.; Cao, Y.; Pan, S.; Zhuang, L.; Yu, R.; Peng, Z.; Qian, H.; Wei, Y.; Zhao, L.;
Liu, G.; Tong, M. Int. J. Antimicrob. Agents 2012, 40, 9.
3. Neeb, M.; Betz, M.; Heine, A.; Barandun, L. J.; Hohn, C.; Diederich, F.; Klebe, G. J.
Med. Chem. 2014, 57, 5566.
4. Neeb, M.; Czodrowski, P.; Heine, A.; Barandun, L. J.; Hohn, C.; Diederich, F.;
Klebe, G. J. Med. Chem. 2014, 57, 5554.
5. Barandun, L. J.; Immekus, F.; Kohler, P. C.; Tonazzi, S.; Wagner, B.;
Wendelspiess, S.; Ritschel, T.; Heine, A.; Kansy, M.; Klebe, G.; Diederich, F.
Chem. Eur. J. 2012, 18, 9246.
6. Hörtner, S. R.; Ritschel, T.; Stengl, B.; Kramer, C.; Schweizer, W. B.; Wagner, B.;
Kansy, M.; Klebe, G.; Diederich, F. Angew. Chem. 2007, 119, 8414; Angew. Chem.,
Int. Ed. 2007, 46, 8266.
37. McCoy, A. Acta Crystallogr., Sect. D 2007, D63, 32.
38. Adams, P. D.; Afonine, P. V.; Bunkoczi, G.; Chen, V. B.; Davis, I. W.; Echols, N.;
Headd, J. J.; Hung, L. W.; Kapral, G. J.; Grosse-Kunstleve, R. W.; McCoy, A. J.;
Moriarty, N. W.; Oeffner, R.; Read, R. J.; Richardson, D. C.; Richardson, J. S.;
Terwilliger, T. C.; Zwart, P. H. Acta Crystallogr., Sect. D 2010, D66, 213.
39. Emsley, P.; Cowtan, K. Acta Crystallogr., Sect. D 2004, D60, 2126.
40. 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. Acta Crystallogr., Sect D 2011, 67, 235.
1
1
1
1
7. B. Zacharie, C. Penney, L. Gagnon, B. Grioux, L. Geerts, S. Abbott, WO2008/
1
31547, (2008).
8. S. Hachiya, M. Oku, H. Mukai, T. Shin, K. Matsuura, R. Seo, T. Kamikubo, Y.
Terada, M. Sanagi, K. Yoshihara, T. Takahashi, WO2006/123725, (2006).
9. Desai, D.; Madhunapantula, S. V.; Gowdahalli, K.; Sharma, A.;
Chandagaludoreswamy, R.; El-Bayoumy, K.; Robertson, G. P.; Amin, S. Bioorg.
Med. Chem. Lett. 2010, 20, 2038.
41. Kleywegt, G. J.; Zou, J. Y.; Kjeldgaard, M.; Jones, T. A.; Around, O. International
Tables for Crystallography; Kluwer Academic Publishers: Dordrecht, 2001.
42. Connolly, M. L. J. Appl. Crystallogr. 1983, 16, 548.
2
0. Bartoszewicz, A.; Kalek, M.; Nilsson, J.; Hiresova, R.; Stawinski, J. Synlett 2008,
3
7.
43. Jones, G.; Willett, P.; Glen, R. C.; Leach, A. R.; Taylor, R. J. Mol. Biol. 1997, 267,
727.
44. Eldridge, M. D.; Murray, C. W.; Auton, T. R.; Paolini, G. V.; Mee, R. P. J. Comput.-
Aided Mol. Des. 1997, 11, 425.
45. Case, D. A.; Cheatham, T. E.; Darden, T.; Gohlke, H.; Luo, R.; Merz, K. M.;
Onufriev, A.; Simmerling, C.; Wang, B.; Woods, R. J. J. Comput. Chem. 2005, 26,
1668.
2
1. Case, D. A.; Babin, V.; Berryman, J. T.; Betz, R. M.; Cai, Q.; Cerutti, D. S.;
Cheatham, T. E.; Darden, T. A.; Duke, R. E.; Gohlke, H.; Goetz, A. W.; Gusarov, S.;
Homeyer, N.; Janowski, P.; Kaus, J.; Kolossvary, I.; Kovalenko, A.; Lee, T. S.;
LeGrand, S.; Luchko, T.; Luo, B.; Wu, X.; Kollman, P. A. Amber 14; University of
California: San Francisco, 2014.
2
2
2. Neudert, G.; Klebe, G. Bioinformatics 2011, 27, 1021.
3. Brandt, T.; Holzmann, N.; Muley, L.; Khayat, M.; Wegscheid-Gerlach, C.; Baum,
B.; Heine, A.; Hangauer, D.; Klebe, G. J. Mol. Biol. 2011, 405, 1170.
4. Steuber, H.; Zentgraf, M.; Gerlach, C.; Sotriffer, C. A.; Heine, A.; Klebe, G. J. Mol.
Biol. 2006, 363, 174.
5. Garman, E. F. Acta Crystalogr., Sect. D 2010, D66, 339.
6. Fraser, J. S.; van den Bedem, H.; Samelson, A. J.; Lang, P. T.; Holton, J. M.; Echols,
N.; Alber, T. Proc. Natl. Acad. Sci. 2011, 108, 16247.
46. Pang, Y.-P. Proteins: Struct., Funct., Bioinf. 2001, 45, 183.
47. Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case, D. A. J. Comput. Chem.
2004, 25, 1157.
48. Jakalian, A.; Bush, B. L.; Jack, D. B.; Bayly, C. I. J. Comput. Chem. 2000, 21, 132.
49. Jakalian, A.; Jack, D. B.; Bayly, C. I. J. Comput. Chem. 2002, 23, 1623.
50. Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. J.
Chem. Phys. 1983, 79, 926.
2
2
2
2
2
2
7. Warkentin, M.; Berejnov, V.; Husseini, N. S.; Thorne, R. E. J. Appl. Crystallogr.
51. Darden, T.; York, D.; Pedersen, L. J. Chem. Phys. 1993, 98, 10089.
52. Ryckaert, J. P.; Ciccotti, G.; Berendsen, H. J. C. J. Comput. Phys. 1977, 23, 327.
53. Hornak, V.; Abel, R.; Okur, A.; Strockbine, B.; Roitberg, A.; Simmerling, C.
Proteins: Struct., Funct., Bioinform. 2006, 65, 712.
2
006, 39, 805.
8. Grädler, U.; Gerber, H.-D.; Goodenough-Lashua, D. M.; Garcia, G. A.; Ficner, R.;
Reuter, K.; Stubbs, M. T.; Klebe, G. J. Mol. Biol. 2001, 306, 455.
9. Stengl, B.; Reuter, K.; Klebe, G. ChemBioChem 2005, 6, 1926.
Please cite this article in press as: Neeb, M.; et al. Bioorg. Med. Chem. (2016), http://dx.doi.org/10.1016/j.bmc.2016.07.053