Potent inhibitors of tRNA-guanine transglycosylase, an enzyme linked to the pathogenicity of the Shigella bacterium: Charge-assisted hydrogen bonding

Article abstract of DOI:10.1002/anie.200702961

Improving inhibition: tRNA-Guanine transglycosylase (TGT) is a newly recognized target to reduce the pathogenicity of disease-causing Shigella bacteria. A potent family of inhibitors of this enzyme has been developed by structure-based design. Crystallographic data and pK_a, analysis suggest that the aminoimidazole moiety of the central lin-benzoguanine scaffold is protonated and stabilization of the complexes results from charge-assisted hydrogen bonding. (Figure Presented).

Full text of DOI:10.1002/anie.200702961

Communications
DOI: 10.1002/anie.200702961
Medicinal Chemistry
Potent Inhibitors of tRNA-Guanine Transglycosylase, an Enzyme
Linked to the Pathogenicity of the Shigella Bacterium: Charge-
Assisted Hydrogen Bonding**
Simone R. Hörtner, Tina Ritschel, Bernhard Stengl, Christian Kramer, W. Bernd Schweizer,
Björn Wagner, Manfred Kansy, Gerhard Klebe,* and François Diederich*
Dedicated to Professor Dieter Seebach on the occasion of his 70th birthday
tRNA-Guanine transglycosylase (TGT, EC 2.4.2.29) is a
tRNA-modifying enzyme common to nearly all organisms
including humans.[1] In bacteria, TGT catalyzes the exchange
of guanine in position 34 by preQ1, whereas eukaryotic TGT
accelerates the replacement by queuine.[2] This difference in
substrate specificity offers the possibility for selective inhib-
ition of the bacterial enzyme, which has been linked to the
pathogenicity of Shigella, the causative agent of bacillary
dysentery (Shigellosis), a diarrheal disease responsible for
more than a million deaths annually.[3] As substantial
crystallographic and biochemical information about the
bacterial enzyme is available, TGT represents an ideal
target for structure-based drug design to facilitate the
development of selective antibiotics against bacillary dysen-
tery.
natural substrate. This pocket is lined by Val282, Leu68, and
Val45. Lipophilic residues such as phenethyl were therefore
introduced at position 4 of lin-benzoguanine (Figure 1b, see
Scheme 1 for numbering). While the predicted orientation of
the lipophilic vectors into this pocket was confirmed by X-ray
We recently introduced lin-benzoguanine (1)[4,5] as a
nucleobase analogue to fill the pocket occupied by preQ1
(Figure 1, and the Supporting Information).[6] This hetero-
tricyclic skeleton fully maintains the hydrogen-bonding
pattern of the natural substrate (Figure 1a) and shows
mixed competitive–uncompetitive inhibition with respect to
tRNA binding (competitive inhibition constant Kic = 4.1 mm,
uncompetitive inhibition constant Kiu = 7.9 mm; against
Zymomonas (Z.) mobilis TGT).[4] We focused our first
design attempts on the filling of a neighboring shallow
hydrophobic pocket occupied by the ribose 34 moiety of the
[*] S. R. Hörtner, C. Kramer, Dr. W. B. Schweizer, Prof. Dr. F. Diederich
Laboratory of Organic Chemistry, ETH Zürich
Hönggerberg, HCI, 8093 Zürich (Switzerland)
Fax: (+41)44-632-1109
E-mail: diederich@org.chem.ethz.ch
T. Ritschel, Dr. B. Stengl, Prof. Dr. G. Klebe
Institute of Pharmaceutical Chemistry
Philipps-University Marburg
Marbacher Weg 6, 35032 Marburg (Germany)
Fax: (+49)6421-282-8994
Figure 1. a) Binding mode of lin-benzoguanine (1) in the active site of
TGT.[4,5] The two catalytic aspartates Asp280 and Asp102 are solvated
by an unperturbed water cluster (protein data base (PDB) code: 2BBF,
resolution 1.7 Š). Dashed lines indicate hydrogen bonds; distances
between heavy atoms in Š. Ligand skeleton green; C gray; O red;
N blue; S yellow. b) Superimposition of TGT in complex with modified
tRNA (blue backbone, PDB-code 1Q2S, resolution 3.2 Š),[6] 4-phen-
ethyl-substituted lin-benzoguanine (yellow, PDB-code: 1Y5V, resolution
1.58 Š), and inhibitor 8 (green, PDB-code: 2QZR, resolution 1.95 Š).
For the TGT-tRNA complex, the U33-preQ1-U35 fragment bound to the
active site is shown. The 4-phenethyl substituent orients towards the
ribose 34 binding pocket; two preferred conformations are shown. In
contrast, the naphthylmethylamino substituent of 8 points into the
ribose 33 binding pocket. Figures were prepared using PyMOL.[17]
E-mail: klebe@mailer.uni-marburg.de
B. Wagner, Dr. M. Kansy
Pharmaceuticals Division, Discovery Chemistry
F. Hoffmann-La Roche AG, 4070 Basel (Switzerland)
[**] Work at Zürich was supported by the ETH Research Council and the
Fonds der Chemischen Industrie, work at Marburg by the Deutsche
Forschungsgemeinschaft.
Supporting information for this article is available on the WWW
under http://www.angewandte.org or from the author.
8266
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2007, 46, 8266 –8269

Angewandte
Chemie
crystallography, binding efficacy of the inhibitors, whose
mechanism of action was fully competitive, remained in the
micromolar range (e.g. Ki = 1 mm for the phenethyl-substi-
tuted ligand). An extended crystallographic investigation[5]
subsequently revealed that the lipophilic vector introduced
into position 4 of the heterocyclic scaffold disrupted a highly
conserved water network solvating the two catalytic aspar-
tates Asp280 and Asp102 (Figure 1a). Apparently, energetic
gains from the ligandꢀs complementary occupation of the
prominent shallow hydrophobic pocket were fully compen-
sated by energetic costs arising from the clearly unfavorable
disruption of the water network and violation of the solvation
requirements of the two catalytic Asp side chains.
The synthesis of the representative 2-benzylamino-sub-
stituted lin-benzoguanine 5 is shown in Scheme 1 (for the
synthesis and characterization of the other new ligands, see
the Supporting Information). Esterification of commercially
To avoid this serious problem and to enhance binding
efficacy, we subsequently concentrated our design towards
the introduction of liphophilic vectors at position 2 of the lin-
benzoguanine skeleton. Molecular modeling using MOLOC[7]
suggested these substituents to point towards the pocket
accommodated by the ribose 33 residue of the substrate
(rather than ribose 34). It was therefore expected to leave the
solvation water network of the two Asp side chains intact.
Based on the modeling considerations regarding space and
environmental polarity of the ribose 33 site, ligands 2–12 with
various alkylamino- and arylalkylamino substituents were
designed (Table 1). Compounds 2–4 with small residues were
prepared as initial controls to estimate experimentally the
gain in binding free enthalpy resulting from penetration and
increasing occupation of the ribose 33 pocket by larger
residues. The introduction of morpholine moieties in 11 and
12 was intended to give full water solubility of the notoriously
poorly soluble lin-benzoguanine derivatives, which is required
for planned cell-based assays.
Scheme 1. Synthesis of ligand 5. a) SOCl2, MeOH, 658C, 92%;
b) HNO3/H2SO4, 508C, 66%; c) Me2NSO2Cl, Et3N, toluene, 1118C,
35% (a), 33% (b); d) 1. LiN(TMS)2, THF, ꢁ788C; 2. CBr4, THF,
ꢁ788C, 75% (a), 60% (b); e) BnNH2, 08C, 92%; f) Zn, AcOH, H2O,
258C, 94%; g) dimethyl sulfone, chloroformamidinium chloride,
1508C, 20%.
available benzimidazole-5-carboxylic acid (13) afforded
methyl ester 14, which was subjected to nitration, resulting
in compound 15. Subsequent protection of the imidazole
moiety with the N,N-dimethylaminosulfonyl group furnished
the two regioisomers 16a (protected at N1) and 16b
(protected at N3) which were separated by column chroma-
tography and identified by X-ray crystal-structure analysis of
the latter (Supporting Information). Bromination of regio-
isomer 16a at position 2 led to intermediate 17a, which
underwent nucleophilic aromatic substitution with benzyl-
amine to yield 18. Reduction of the nitro group gave amine
19, and following cyclization with chlorformamidinium chlo-
ride in dimethyl sulfone yielded the desired lin-benzoguanine
derivative 5. As a result of the hydrochloric acid liberated
during this reaction, deprotection of the imidazole moiety was
achieved concurrently with the cyclization reaction.
Fully competitive inhibitory behavior of all the ligands in
the base exchange reaction (G34!preQ1) was confirmed by
trapping experiments (see the Supporting Information) as
previously described[5] and the inhibition constants Ki derived
from kinetic measurements.[4] In the trapping experiments,
TGT is incubated with an excess of tRNA and the inhibitors
are added to validate whether they block tRNA binding in a
fully competitive way. In a subsequent sodium dodecyl sulfate
(SDS)-polyacrylamide gel electrophoresis (PAGE) analysis,
bands for the TGT–tRNA complex are not observed if the
inhibitor shows fully competitive behavior. Gratifyingly, all
ligands showed fully competitive behavior with binding
affinities in the nanomolar range, with the best ones (10, 11)
featuring single-digit nanomolar inhibitory constants
(Table 1), which are unprecedented activities for inhibitors
of TGT enzymes. In addition, the introduction of the
morpholino substituent led to the desired free solubility of
the best ligand 11 in aqueous buffers.
Table 1: 2-Substituted lin-benzoguanine derivatives prepared for the inhibition
of TGT.
Compd.
R
Ki [nm]
Compd.
R
Ki [nm]
35ꢀ6
1
H
4100ꢀ1000 7[a]
2
Me
1600ꢀ400
8[a]
55ꢀ11
3
4
NH2
77ꢀ12
58ꢀ36
9
27ꢀ12
10ꢀ3
10[a]
5[a]
6[a]
70ꢀ1
35ꢀ9
11[b]
12[b]
6ꢀ6
40ꢀ18
[a] Isolated and used as the bis(trifluoroacetate) salt (TFA=CF3COOH).
[b] Isolated and used as the tris(trifluoroacetate) salt.
Angew. Chem. Int. Ed. 2007, 46, 8266 –8269
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 8267

Communications
The X-ray crystal structures of 2, 3, 6, 8, 10, and 11 as
related to the move of the substituent from the 4- to the 2-
position of the heterocyclic platform. This change avoids
disturbance of the crucial water network between Asp102 and
Asp280. The resolution of the X-ray crystal structure allows
location of two crystal water molecules solvating the catalytic
residues (Figure 2).
binary complexes with Z. mobilis TGT were solved. Herein
we only discuss in detail the structure of the bound naphthyl
derivative 8, solved with a resolution of 1.95 Š (Figure 2).[8]
Measured values for the logarithmic distribution coeffi-
[9]
cient logD(octanol/water) at pH 7.4 (Table 2) clearly show that
the increased binding affinity is not at all due to greater
lipophilicity resulting from the introduction of the substitu-
ents at position 2 of the central scaffold.
Table 2: Binding affinities (inhibition constants Ki [nm]), pKa values, and
logD values of selected inhibitors of TGT.
Compd.
Ki [nm]
pKa
logD
1
2
3
4
6
4100ꢀ1000[4]
1600ꢀ400
77ꢀ12
5.2
5.4
6.3
6.2
5.8
5.9
6.7
ꢁ0.05
0.11
ꢁ0.33
ꢁ0.07
1.49
3.29
ꢁ0.31
58ꢀ36
35ꢀ9
8
11
55ꢀ11
6ꢀ6
A slight increase (factor 2.5) in binding affinity is seen
upon expanding the parent lin-benzoguanine 1, (Kic = 4.1 mm)
by a methyl group in 2-position to give 2 (Ki = 1.6 mm). This
methyl group forms additional hydrophobic contacts with a
small niche next to the side-chain methyl group of Ala232 and
the backbone of Gly261.
A larger increase in binding affinity, down to the single-
digit nanomolar range, is observed for the compounds bearing
amino substituents in position 2. The crystal structures show
an additional hydrogen bond between the exocyclic amino
group and the carbonyl group of Ala232 (d(N···O) = 2.8 Š in
the complex with 3 (not shown) and 3.6 Š with 8 (Figure 2)).
More importantly however, this exocyclic amino group
increases the basicity of the imidazole ring. The now
guanidine-like 2-aminoimidazol group is protonated, thereby
bearing positive charge and enabling strong charge-assisted
hydrogen bonding.
In the structure of the complex of TGT and preQ1, which
was obtained by co-crystallization, the carbonyl group of the
peptide bond between Leu231 and Ala232 is oriented
towards the exocyclic basic aminomethyl group of the
modified nucleobase (Figure 2b).[10] At the same time, the
Figure 2. a) Crystal structure of naphthyl-substituted ligand 8 bound to
TGT, resolution 1.95 Š. The difference electron density of the com-
pound is shown together with the hydrogen-bonding contacts between
the protein and the ligand. Ligand skeleton orange; C gray; O red;
N blue; S yellow. b) Superimposition of ligand 8 (gray) and preQ1
(yellow) bound to the active site of TGT. Distances between heavy
atoms in Š.
The lin-benzoguanine scaffold is well-defined, but the naph-
thyl substituent is not detected in the difference electron
density. It is likely that this part of the ligand is scattered over
multiple conformations that point towards the ribose 33
binding pocket. Similarly, the substituents in 6, 10, and 11
bound to TGT are not detected in the difference electron
density maps. This situation indicates that strong directional
interactions have not been established in the environment of
the ribose 33 binding pocket, which would lock the substitu-
ents in an energetically favorable conformation. Nevertheless,
binding free enthalpy should be gained by desolvation of the
2-substituents and the uracil 33 binding pocket. Gains in
enthalpy resulting from interactions with the pocket might
not be large; on the other hand a substantially unfavorable
entropic compensation should also not occur given the
apparent residual mobility of these substituents in bound
state.
ꢁ
N H group of this peptide bond interacts with the side chain
of Glu235 and stabilizes this orientation. At physiological
pH value, the aminomethyl group should be protonated
whereas the side chain of Glu235 should be deprotonated.
This way, a strong charge-assisted hydrogen-bonding network
with two ionic hydrogen bonds (d(OꢁE235···HNAla232) = 2.7 Š
=
(distance
O
between
heavy
atoms)
and
d(C
Leu231···HN+preqQ1) = 2.9 Š) is formed (Figure 2).[11] In gen-
eral, charge-assistance is thought to contribute significantly to
binding affinity.[12] We therefore hypothesized that the 2-
aminoimidazole ring in the most potent of our ligands could
also be protonated, thereby establishing a similar charge-
assisted hydrogen-bonding network. In other words, an ionic
Clearly, the significant increase in affinity of 2-substituted,
compared to 4-substituted lin-benzoguanines (all of which
had activities in the micromolar inhibitory range[4]), can be
8268 www.angewandte.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2007, 46, 8266 –8269

Angewandte
Chemie
rather than a neutral hydrogen bond would form between N1
Keywords: drug design · glycosylases · hydrogen bonds ·
inhibitors · shigellosis · TGT
.
=
of the aminoimidazolium ring and C OLeu231
.
To verify this assumption, the pKa values of some
representative ligands were determined by photometric
titration.[9,13] Table 2 shows the relevant pKa values assigned
to the protonation of the imidazole rings in the ligands; the
entire set of pKa data for each molecule can be found in the
the Supporting Information. The pKa value is indeed sub-
stantially increased upon introduction of an amino substituent
in position 2: it changes from lin-benzoguanine (1, pKa = 5.2)
and its methyl derivative 2 (5.4) to substantially higher values
between 5.9 (6) and 6.7 (11). Hence, under the conditions of
the binding assay at pH 7.3, the 2-amino-substituted imida-
zole moiety of the ligands is more readily protonated.
Furthermore, in the polar environment of the binding
pocket an additional pKa shift is expected stabilizing the
charged state. To support such considerations we performed
Poisson–Boltzmann calculations using our recently intro-
duced PEOE-PB charge model for protein–ligand com-
plexes.[14–16] These calculations provide good evidence that
the actual pKa values of the bound amino-substituted ligands
are shifted by 1–2 pKa units (logarithmic scale) to higher
values, so that the ligands should be fully protonated once
accommodated in the active site (for details on pKa calcu-
lations see the Supporting Information). We therefore
[1] C. Romier, J. E. W. Meyer, D. Suck, FEBS Lett. 1997, 416, 93 –
98.
[2] D. Iwata-Reuyl, Bioorg. Chem. 2003, 31, 24 – 43.
[3] N. Okada, C. Sasakawa, T. Tobe, M. Yamada, S. Nagai, K. A.
Talukder, K. Komatsu, S. Kanegasaki, M. Yoshikawa, Mol.
Microbiol. 1991, 5, 187 – 195.
[4] E. A. Meyer, N. Donati, M. Guillot, W. B. Schweizer, F. Die-
derich, B. Stengl, R. Brenk, K. Reuter, G. Klebe, Helv. Chim.
Acta 2006, 89, 573 – 597.
[5] B. Stengl, E. A. Meyer, A. Heine, R. Brenk, F. Diederich, G.
Klebe, J. Mol. Biol. 2007, 370, 492 – 511.
[6] W. Xie, X. Liu, R. H. Huang, Nat. Struct. Biol. 2003, 10, 781 – 788.
[7] P. R. Gerber, K. Müller, J. Comput.-Aided Mol. Des. 1995, 9,
251 – 268.
[8] B. Stengl, Doctoral Dissertation, Philipps-Universität, Marburg,
2006.
[9] J. Olsen, P. Seiler, B. Wagner, H. Fischer, T. Tschopp, U. Obst-
Sander, D. W. Banner, M. Kansy, K. Müller, F. Diederich, Org.
Biomol. Chem. 2004, 2, 1339 – 1352.
[10] R. Brenk, M. T. Stubbs, A. Heine, K. Reuter, G. Klebe,
ChemBioChem 2003, 4, 100 – 111.
[11] B. Stengl, K. Reuter, G. Klebe, ChemBioChem 2005, 6, 1066 –
1077.
[12] S. H. Reich, M. A. Fuhry, D. Nguyen, M. J. Pino, K. M. Welsh, S.
Webber, C. A. Janson, S. R. Jordan, D. A. Matthews, W. W.
Smith, J. Med. Chem. 1992, 35, 847 – 858.
[13] For a recent review on tuning and predicting amine basicities,
see: M. Morgenthaler, E. Schweizer, A. Hoffmann-Röder, F.
Benini, R. E. Martin, G. Jaeschke, B. Wagner, H. Fischer, S.
Bendels, D. Zimmerli, J. Schneider, F. Diederich, M. Kansy, K.
Müller, ChemMedChem 2007, 2, 1100 – 1115.
[14] P. Czodrowski, I. Dramburg, C. A. Sotriffer, G. Klebe, Proteins
Struct. Funct. Bioinf. 2006, 65, 424 – 437.
[15] P. Czodrowski, C. A. Sotriffer, G. Klebe, J. Mol. Biol. 2007, 367,
1347 – 1356.
[16] P. Czodrowski, C. A. Sotriffer, G. Klebe, J. Chem. Inf. Model.
2007, 47, 1590 – 1598.
[17] W. DeLano, 2006, www.pymol.org.
=
propose that the hydrogen bond between N1 of 8 and C
OLeu231 (d = 3.0 Š) has ionic character (Figure 2).
In summary, we have reported the synthesis and biological
evaluation of the most potent small-molecule inhibitors of
TGT known to date. Strong evidence is provided that the 2-
aminoimidazole moieties in the most effective ligands are
protonated and that complexation is enhanced by charge-
assisted hydrogen bonding between protein and ligand as
found for the complexation with the natural substrate preQ1.
The synthesis of these inhibitors with single-figure nanomolar
inhibition constants is an important step towards the develop-
ment of new, specific drugs against Shigellosis.
Received: July 3, 2007
Published online: September 27, 2007
Angew. Chem. Int. Ed. 2007, 46, 8266 –8269
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 8269