From lin-benzoguanines to lin-benzohypoxanthines as ligands for zymomonas mobilis tRNA-guanine transglycosylase: Replacement of protein-ligand hydrogen bonding by importing water clusters

Article abstract of DOI:10.1002/chem.201200809

The foodborne illness shigellosis is caused by Shigella bacteria that secrete the highly cytotoxic Shiga toxin, which is also formed by the closely related enterohemorrhagic Escherichia coli (EHEC). It has been shown that tRNA-guanine transglycosylase (TGT) is essential for the pathogenicity of Shigella flexneri. Herein, the molecular recognition properties of a guanine binding pocket in Zymomonas mobilis TGT are investigated with a series of lin-benzohypoxanthine- and lin-benzoguanine-based inhibitors that bear substituents to occupy either the ribose-33 or the ribose-34 pocket. The three inhibitor scaffolds differ by the substituent at C(6) being H, NH₂, or NH-alkyl. These differences lead to major changes in the inhibition constants, pK_a values, and binding modes. Compared to the lin-benzoguanines, with an exocyclic NH₂ at C(6), the lin-benzohypoxanthines without an exocyclic NH₂ group have a weaker affinity as several ionic protein-ligand hydrogen bonds are lost. X-ray cocrystal structure analysis reveals that a new water cluster is imported into the space vacated by the lacking NH₂ group and by a conformational shift of the side chain of catalytic Asp102. In the presence of an N-alkyl group at C(6) in lin-benzoguanine ligands, this water cluster is largely maintained but replacement of one of the water molecules in the cluster leads to a substantial loss in binding affinity. This study provides new insight into the role of water clusters at enzyme active sites and their challenging substitution by ligand parts, a topic of general interest in contemporary structure-based drug design. Water replacements: A series of lin-benzopurines was evaluated as inhibitors of Zymomonas mobilis tRNA-guanine transglycosylase, an enzyme that was identified as a potential target for the treatment of shigellosis. X-ray cocrystal structures show the import of a new water cluster that replaces lost protein-ligand interactions (see figure), with an overall reduction in binding affinity. Copyright

Full text of DOI:10.1002/chem.201200809

FULL PAPER
DOI: 10.1002/chem.201200809
From lin-Benzoguanines to lin-Benzohypoxanthines as Ligands for
Zymomonas mobilis tRNA–Guanine Transglycosylase: Replacement of
Protein–Ligand Hydrogen Bonding by Importing Water Clusters
Luzi Jakob Barandun,[a] Florian Immekus,[b] Philipp C. Kohler,[a] Sandro Tonazzi,[a]
Bjçrn Wagner,[c] Severin Wendelspiess,[c] Tina Ritschel,[b, d] Andreas Heine,[b]
Manfred Kansy,[c] Gerhard Klebe,*[b] and FranÅois Diederich*[a]
Dedicated to Prof. Dr. Heribert Offermanns on the occasion of his 75th birthday
Abstract: The foodborne illness shigel-
losis is caused by Shigella bacteria that
secrete the highly cytotoxic Shiga
toxin, which is also formed by the
closely related enterohemorrhagic Es-
cherichia coli (EHEC). It has been
shown that tRNA–guanine transglyco-
sylase (TGT) is essential for the patho-
genicity of Shigella flexneri. Herein,
the molecular recognition properties of
a guanine binding pocket in Zymomo-
nas mobilis TGT are investigated with
a series of lin-benzohypoxanthine- and
lin-benzoguanine-based inhibitors that
bear substituents to occupy either the
ribose-33 or the ribose-34 pocket. The
three inhibitor scaffolds differ by the
substituent at C(6) being H, NH2, or
structure analysis reveals that a new
water cluster is imported into the space
vacated by the lacking NH2 group and
by a conformational shift of the side
chain of catalytic Asp102. In the pres-
ence of an N-alkyl group at C(6) in lin-
benzoguanine ligands, this water cluster
is largely maintained but replacement
of one of the water molecules in the
cluster leads to a substantial loss in
binding affinity. This study provides
new insight into the role of water clus-
ters at enzyme active sites and their
challenging substitution by ligand parts,
a topic of general interest in contempo-
rary structure-based drug design.
ꢀ
NH alkyl. These differences lead to
major changes in the inhibition con-
stants, pKa values, and binding modes.
Compared to the lin-benzoguanines,
with an exocyclic NH2 at C(6), the lin-
benzohypoxanthines without an exocy-
clic NH2 group have a weaker affinity
as several ionic protein–ligand hydro-
gen bonds are lost. X-ray cocrystal
Keywords: drug design · hydrogen
bonds · molecular recognition · shi-
gellosis · water clusters
Introduction
In May 2011, the outbreak of the Shiga-toxin-producing bac-
terium enterohemorrhagic Escherichia coli (EHEC) led to
a major health and economic problem in Europe.[1] The
eponymous bacterium for the Shiga toxin—the Shigella bac-
terium—causes the acute inflammatory bowel disease shigel-
losis.[2] It has been shown that the enzyme tRNA–guanine
transglycosylase (TGT, EC 2.4.2.29) is a potential drug
target for the treatment of shigellosis.[3,4] We have intro-
[a] L. J. Barandun,+ Dr. P. C. Kohler, S. Tonazzi, Prof. Dr. F. Diederich
Laboratorium fꢀr Organische Chemie, ETH Zꢀrich
Hçnggerberg, HCI, 8093 Zurich (Switzerland)
Fax : (+41)44-632-1109
E-mail: diederich@org.chem.ethz.ch
[b] F. Immekus,+ Dr. T. Ritschel, Dr. A. Heine, Prof. Dr. G. Klebe
Institut fꢀr Pharmazeutische Chemie
Philipps-Universitꢁt Marburg, Marbacher Weg 6
35032 Marburg (Germany)
duced 6-aminoimidazoACTHNUTRGNEUGN[4,5-g]quinazolin-8(7H)-ones (lin-ben-
zoguanines, Figure 1) as inhibitors of Zymomonas mobilis
TGT.[5,6] The tricyclic, unsubstituted lin-benzopurines, in
which the natural purine is extended by insertion of a ben-
zene ring between the pyrimidine and the imidazole moiet-
ies, were introduced first by Leonard and co-workers in the
1970s.[7] They have, however, found only limited application
in medicinal chemistry.[8–10] Kool and co-workers reported
the synthesis and base-pairing properties of these extended
nucleobases when introduced into oligonucleotides.[11]
In our previous studies on Z. mobilis TGT, we showed
that highly potent inhibitors are obtained by filling either
the ribose-33[12] or ribose-34 pocket[13,14] of the tRNA-bind-
ing site with substituents attached to C(2) or C(4), respec-
Fax : (+49)6421-282-8994
E-mail: klebe@mailer.uni-marburg.de
[c] B. Wagner, S. Wendelspiess, Dr. M. Kansy
Pharma Research Non-Clinical Safety
F. Hoffmann-La Roche AG
4070 Basel (Switzerland)
[d] Dr. T. Ritschel
Current address: Computational Drug Discovery Group
CMBI, Radboud University Nijmegen Medical Centre
P.O. Box 9101, 6500 HB Nijmegen (The Netherlands)
[+] These authors contributed equally to this work.
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/chem.201200809.
Chem. Eur. J. 2012, 00, 0 – 0
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
1
&
ÞÞ
These are not the final page numbers!

Figure 1. Structure and numbering of lin-benzoguanines and lin-benzohy-
poxanthines. Substitution at C(2) or C(4) allows the ribose-33 and ribose-
34 pockets, respectively, to be filled.
tively, of the lin-benzoguanine core (Figure 1). Herein, we
compare lin-benzoguanines with lin-benzohypoxanthines
(imidazoACHTUNGTRENNUNG[4,5-g]quinazolin-8(7H)-ones), which lack the exo-
cyclic NH2 group at C(6). Several comparative studies on
the binding affinity of guanine and their corresponding hy-
poxanthine derivatives have been reported. Both cases are
found in which either the guanine-based compounds have
stronger binding than the hypoxanthine-based ones, or vice
versa.[15,16] On the other hand, 9-deazahypoxanthine- and 9-
deazaguanine-based ligands of the human purine nucleoside
phosphorylase (PNP) were found to be equipotent.[17] These
findings prompted us to investigate and compare the molec-
ular recognition of lin-benzoguanines and lin-benzohypo-
xanthines by Z. mobilis TGT in detail.
Results and Discussion
Ligand design: Z. mobilis TGT catalyzes the guanine/preQ1
base-exchange reaction at the wobble position G34 in the an-
ticodon of bound tRNA with participation of the side chains
of catalytic Asp102 and Asp280.[18] The region of the active
site of interest for our ligand design comprises the guanine/
preQ1 binding pocket and the ribose-33 and ribose-34 pock-
ets occupied by bound tRNA (Figure 2a). lin-Benzoguanines
bound to the nucleobase pocket are intercalated between
Figure 2. Comparison of X-ray cocrystal structures of lin-benzoguanines
and lin-benzohypoxanthines bound to the active site of Z. mobilis TGT.
Color code: Cenzyme gray, O red, N blue. Selected water molecules are
shown as spheres and labeled as W (for residue identifiers, see Table 2SI
in the Supporting Information). Hydrogen bonds are shown as dashed
lines (distances between 2.5 and 3.5 ꢃ). Ligand and water molecules are
well defined by the difference electron density at 3s shown as green
mesh. These characteristics apply to all figure captions unless otherwise
stated. a) Binding mode of lin-benzoguanine 1 (Cligand cyan) in the active
site of TGT (PDB code: 2Z7K[21]), which is indicated as a gray surface.
The water cluster between Asp102 and Asp280 is shown as red spheres.
b) Cocrystal structure of 8a (Cligand green) bound to the active site of
TGT (PDB code: 3S1G). A hydrogen bond between W3 and the water
cluster in the ribose-34 pocket connects the new water cluster lined up by
W1–W6. There is no electron density observable for the side chain of
Gln107. Instead W5 occupies this position. c) Overlay of lin-benzogua-
nine 1 (C cyan, water molecules cyan; PDB code: 2Z7K[21]) and lin-ben-
zohypoxanthine 8a (C green, water molecules green; PDB code: 3S1G)
in the active site of Z. mobilis TGT.
&
2
&
www.chemeurj.org
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 0000, 00, 0 – 0
ÝÝ
These are not the final page numbers!

Ligands for tRNA–Guanine Transglycosylase
FULL PAPER
the side chains of Met260 and Tyr106 and, similar to the nat-
ural substrates guanine and preQ1, their aminopyrimidinone
ring engages in a complex hydrogen-bonding pattern with
the protein that involves the side chains of Asp102, Asp156,
Gln203, and the backbone NH of Gly230 (Figure 2a; com-
pound 1).[12,13] The lin-benzohypoxanthines lack the exocy-
clic NH2 group at C(6) and, therefore, were expected to lose
hydrogen bonds to the side
The lin-benzohypoxanthines that bear substituents to fill
the ribose-34 pocket were prepared in a similar fashion,
starting from isomer 3. Because the imidazole protecting
group is on the opposite side in this isomer, derivatization at
C(4) is sterically less demanding and therefore facilitated.
Following a literature procedure, benzimidazoles 9a,b were
prepared in eight steps from 3 (Scheme 2).[13] The presence
chains of Asp102 and Asp156,
respectively.
Synthesis: The synthetic route
towards the different new in-
hibitors (Schemes 1–3) follows
a previously reported strategy,
starting from benzimidazole-5-
carboxylic acid (2).[12,13] Esterifi-
cation, nitration, and N-protec-
tion yielded two isomers that
were separated chromatograph-
ically on a 40 g scale. Each
isomer was brominated at C(2)
to furnish the building blocks 3
Scheme 2. Synthesis of lin-benzohypoxanthines 10a,b: a) formamide, 1408C, 18–22 h; b) HCl, MeOH, 658C,
18–24 h; 60–87% (over two steps).
and 4, respectively, which were
both used in the course of the
syntheses (Scheme 1). Substitu-
tion of isomer 4 at C(2) with
different primary amines gave 2-aminobenzimidazoles 5a–d,
which were reduced to the 2,6-diamino derivatives 6a–d.
Cyclization with either formamide or formamidinium ace-
tate led to the formation of the pyrimidinone ring in 7a–d,
which were deprotected to yield lin-benzohypoxanthines
8a–d directing substituents at C(2) into the ribose-33
pocket.
of the substituents at C(4) in 9a,b mandated harsher condi-
tions for the following cyclization step relative to the trans-
formation of C(2)-substituted 6a–d (Scheme 1). Stirring
9a,b at 1408C for several hours in formamide as solvent
provided a mixture of the desired protected lin-benzohypo-
xanthines together with a side product in which the ethyla-
mino group at C(4) was additionally N-formylated. Gratify-
ingly, both the N,N-dimethylsul-
famoyl protecting group and
the undesired N-formyl group
were removed with HCl to fur-
nish target compounds 10a,b.
For the synthesis of the
ꢀ
C(6) N-monoalkylated lin-ben-
zoguanines 11a,b, the ring clo-
sure of an appropriate isatoic
anhydride derivative with S-
methylisothioureas 12a,b, re-
spectively, was envisaged.[19]
Whereas 12a was commercially
available, the required ethyl
derivative 12b was prepared
starting from N,N’-bis-Boc-
protected (Boc=t-butyloxycar-
bonyl) S-methylisothiourea 13,
which was treated with EtI to
afford N-ethylisothiourea 14
(Scheme 3).[20]
with trifluoroacetic acid gave
the desired compound 12b. For
Deprotection
Scheme 1. Synthesis of lin-benzohypoxanthines 8a–d: a) RNH2, EtOAc, or EtOH, 0–258C, 1–7 h; 80–91%;
b) Zn, AcOH/H2O, 25 8C, 20–60 min; 83–99%; c) formamide, 1408C, 18 h, crude; or formamidinium acetate,
EtOH, 808C, 20–22 h; 42–84%; d) HCl, MeOH, 658C, 30 min; 49–99%.
Chem. Eur. J. 2012, 00, 0 – 0
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
&
3
&
ÞÞ
These are not the final page numbers!

F. Diederich, G. Klebe et al.
group of the lin-benzoguanines
reduces the potency even fur-
ther and Ki values are now in
the double-digit micromolar
range. N-Methylated 11a shows
a 480-fold lower binding affinity
relative to the NH2 derivative
19a, and N-ethylated 11b binds
even more weakly by a 700-fold
factor. The origins of these
large differences in binding
strength become apparent in
the discussion of the cocrystal
structures (see below).
Physicochemical properties: We
calculated clogD and clogP
values and conducted measure-
ments to analyze how the re-
moval of the exocyclic NH2
group in lin-benzohypoxan-
thines or its monoalkylation in
lin-benzoguanines affects the
physicochemical properties of
the ligands. Both changes lead
to more positive clogD and
clogP values (calculated using
ACD/Labs software; compare
the 2-methylamino compounds
Scheme 3. Synthesis of the alkylated lin-benzohypoxanthines 11a,b: a) K2CO3, EtI, acetone, 568C, 18 h; 98%;
b) TFA, CH2Cl2, 258C, 45 min; 99%; c) Boc2O, DMAP, Et3N, THF, 258C, 17 h; 98%; d) Zn, AcOH/H2O,
258C, 60 min; 92%; e) LiOH, MeOH/H2O, 80 8C, 2 h; then triphosgene, THF, 0 to 258C, 2.5 h; f) 12a or 12b,
Na2CO3, Me2SO, 808C, 2 h; 18a: 29% (over 3 steps), 18b: 25% (over three steps); g) HCl, MeOH, 658C,
30 min; 11a: 87%, 11b: 86%. Boc2O = dicarbonic acid di-tert-butyl ester, DMAP = 4-(dimethylamino)pyri-
dine, DMF = N,N-dimethylformamide, TFA = trifluoroacetic acid, THF = tetrahydrofuran.
8a
and
11a,b
to
19a;
Table 1),[22,23] although all com-
pounds remain highly hydro-
the cyclization, it was advantageous to protect the amino
group at C(2). The Boc-protected benzimidazole 15 was ob-
tained from 5a and subsequently reduced to give 16. Saponi-
fication yielded the anthranilic acid, which was treated with
triphosgene to furnish isatoic anhydride 17 that was directly
used in the following step. Contrary to the results of Coppo-
la et al.,[19] the cyclization reaction in the absence of Boc
protection gave an inseparable mixture of isomers. Howev-
er, when isatoic anhydride 17 was used, the desired lin-ben-
zoguanines 18a and 18b with alkylated exocyclic NH2
groups at C(6) could be isolated in pure form. Final depro-
tection gave the target molecules 11a and 11b.
philic. Upon the introduction of substituents at C(2) or
C(4), the differences that result from the changes at C(6) in-
creasingly vanish and become masked by the effects from
the added substituents.
Measurements of the parallel artificial membrane permea-
bility (PAMPA) scores[24] were performed to estimate differ-
ences in membrane permeability between the three types of
ligands. The majority of the highly polar compounds do not
penetrate the membrane (Table 1), with only two rated as
“medium” due to accumulation in the membrane. Differen-
ces between the three ligand classes were not observed, as
the highly polar 2-aminoimidazole moiety presumably domi-
nates the behavior of all systems.
Biological activity: Competitive inhibition constants Ki were
obtained at pH 7.3 in 100 mm HEPES buffer at 378C by a ra-
dioactive assay using [8-3H]guanine, as described in the liter-
ature.[6,12,21] The lin-benzoguanine derivatives with an exocy-
clic NH2 group show by far the strongest inhibition (Table 1;
compounds 19a–d and 20a,b) with Ki values in the range of
2–58 nm.[12,13] The removal of the exocyclic NH2 group in the
corresponding lin-benzohypoxanthines 8a–d (with substitu-
ents for the ribose-33 pocket) and 10a,b (with substituents
for the ribose-34 pocket) drastically reduces the affinity by
a factor of 80–680 and yields Ki values in the range near
1 mm. Unexpectedly, monoalkylation at the exocyclic NH2
Classical potentiometric titrations to determine the pKa
values of the ligands were not feasible due to the low solu-
bility of the lin-benzopurines. Instead, measurements were
conducted by parallel capillary electrophoresis.[25] The pKa
values for lin-benzoguanines 19a,b had been previously de-
termined by photometric titration,[12] and the new data were
found to be in good agreement (Table 1SI in the Supporting
Information). The pKa values for the 2-aminoimidazolium
moiety (5.1–5.7) and N(7)H (>10) are similar for all three
ligand classes. In contrast, the removal of the exocyclic NH2
group at C(6) strongly increases the acidity of N(5)H+ and
lowers the pKa from values around 4.0–4.4 in the lin-benzo-
&
4
&
www.chemeurj.org
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 0000, 00, 0 – 0
ÝÝ
These are not the final page numbers!

Ligands for tRNA–Guanine Transglycosylase
FULL PAPER
Table 1. Inhibition constants Ki, clogD (calculated logarithmic distribution constant for n-octanol/water at pH 7.4) and clogP (calculated logarithmic
partition constant for n-octanol/water) values, PAMPA scores, and pKa values for lin-benzohypoxanthines and lin-benzoguanines.
clogD[a]
(pH 7.4)
PAMPA score
Pe [10ꢀ6 cmsꢀ1
pKa measurements
N(7)H
X
Ki [nm]
G
clogP[a]
]
category
N(3)H+
N(5)H+
19a
8a
11a
11b
NH2
H
NHMe
NHEt
58ꢁ36[12]
ꢀ0.33
0.18
0.18
ꢀ0.05
ꢀ0.19
0.47
0.00[b]
0.00
0.00
medium
low
low
>10
>10
n.d.[c]
5.7
5.6
5.6
4.4
1.8
4.2
6500ꢁ2900
28050ꢁ11970
40830ꢁ4270
0.72
0.97
0.05
low
19b
8b
NH2
H
35ꢁ9[2]
1.09
1.76
1.41
1.27
0.00[b]
0.03
medium
low
>10
>10
5.5
5.1
4.0
2.0
2900
19c
8c
NH2
H
6ꢁ6[12]
ꢀ0.89
ꢀ0.22
ꢀ0.45
ꢀ0.59
0.00
0.05
low
low
4100
19d
8d
NH2
H
10ꢁ3[12]
1.76
2.37
2.04
1.90
3700
20a
10a
NH2
H
2ꢁ1[13]
ꢀ0.87
ꢀ0.57
1.78
1.65
0.00
0.04
low
low
740ꢁ170
20b
10b
NH2
H
4ꢁ2[13]
ꢀ0.23
ꢀ0.10
2.31
2.17
0.09
0.00
low
low
1100ꢁ370
n.d.[c]
5.3
n.d.[c]
[a] Calculated using ACD/Labs.[23] [b] Accumulation in membrane, hence the compound is categorized as medium. [c] n.d.=could not be determined
with the applied method.
guanines to 2.0 and below in the lin-benzohypoxanthines
(Table 1). This may have a large impact on the binding
mode as it is assumed that N(5) is protonated in the active
site to undergo hydrogen bonding to the presumably depro-
tonated side chain of Asp102.[21,26] The lin-benzohypoxan-
thines have to pay a greater energetic penalty for protona-
tion at N(5) relative to the lin-benzoguanines. The data indi-
cate that the parent tricyclic scaffolds are predominantly in
their unprotonated form under assay conditions (pH 7.3), al-
though Poisson–Boltzmann calculations for lin-benzogua-
nines suggest an increase of the pKa value of the 2-amino-
imidazolium moiety upon binding in the enzyme pocket.[21]
tion of the 2-aminoimidazole core of the ligand and the ad-
ditional Glu235, which polarizes the addressed backbone
carbonyl even further.[12] The tricyclic aromatic skeleton is
sandwiched between Tyr106 and Met260, thus undergoing
favorable p-stacking interactions. As reported for the coc-
rystal structures with lin-benzoguanines, the binding of lin-
benzohypoxanthines induces a cis-peptide bond conforma-
tion between Val262 and Gly263 (not shown).[21]
For the unsubstituted ligand 8a, a dataset with a resolution
of 1.82 ꢃ showed a well-defined electron density for the
ligand and the surrounding water molecules within the
active site (Figure 2b; for hydrogen-bonding lengths, see Fig-
ure 1SI in the Supporting Information). The comparison of
this structure with that previously published for lin-benzo-
guanine 1 (PDB code: 2Z7K[21]) revealed that removal of
the exocyclic amino functionality at C(6) leads to a rotation
of the side chain of Asp102 within the guanine subpocket,
away from the bound ligand (Figure 2c). Instead, it now un-
dergoes hydrogen bonding to the side chain NH2 of Asn70
and the backbone NH of Thr71. This interaction is already
known from apo structures of TGT (Figure 2SI in the Sup-
porting Information)[27] and from cocrystal structures with li-
gands bound to the guanine subpocket that lack a hydrogen-
bond donor in a fitting distance to the carboxyl functionality
of Asp102 (e.g., PDB code: 1F3E;[4] Figure 3SI in the Sup-
porting Information). The emerging space at the bottom of
X-ray cocrystal structures of lin-benzohypoxanthines: For
the four ligands 8a–c and 10a, crystal structures in complex
with TGT could be obtained with a maximum resolution be-
tween 1.53 and 1.82 ꢃ (Table 2). The tricyclic aromatic ring
system of 2-amino-lin-benzohypoxanthines (Figure 2b) binds
in a way that is similar to the closely related 2-amino-lin-
benzoguanines (Figure 2a) in the guanine subpocket. In all
structures, the tricyclic scaffold is fixed by means of hydro-
gen bonds to the side chains of Asp156 and Gln203, the
backbone NH of Gly230, and the backbone C=O groups of
Leu231 and Ala232. The interaction with the backbone
C=O of Leu231 can be assumed to be rather strong due to
the charge assistance provoked by the most likely protona-
Chem. Eur. J. 2012, 00, 0 – 0
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
&
5
&
ÞÞ
These are not the final page numbers!

F. Diederich, G. Klebe et al.
Table 2. X-ray cocrystal structures: data collection and refinement statistics.
8a
8b
8c
10a
11a
11b
PDB code
no. crystal used
l [ꢃ]
3S1G
1
0.91841
C2
3GEV
1
0.91841
C2
3V0Y
1
0.91841
C2
3SM0
1
0.91841
C2
3RR4
1
0.91841
C2
3TLL
1
0.91481
C2
space group
a [ꢃ]
90.8
91.2
91.0
91.0
90.7
91.2
b [ꢃ]
64.8
64.7
64.9
64.9
64.8
64.8
c [ꢃ]
70.6
70.3
70.6
71.2
70.5
70.3
b [8]
95.9
95.9
95.9
96.4
96.0
96.0
resolution range [ꢃ]
unique reflections
R(I)sym [%][b]
completeness [%]
redundancy
30–1.82
35920 (1807)[a]
5.8 (24.1)[a]
97.9 (98.2)[a]
2.7 (2.2)[a]
17.9 (3.6)[a]
30–1.82
34666
30–1.53
54813 (2380)[a]
6.3 (43.8)[a]
89.3 (77.1)[a]
2.6 (1.6)[a]
16.0 (1.9)[a]
10–1.53
51600
30–1.65
45274 (2008)[a]
4.7 (29.9)[a]
95.9 (81.8)[a]
2.5 (2.2)[a]
19.0 (2.6)[a]
30–1.65
43009
25–1.57
54458 (2425)[a]
5.6 (27.0)[a]
93.8 (83.9)[a]
2.7 (2.2)[a]
16.2 (2.7)[a]
25-1.57
51761
30–1.68
45107 (2115)[a]
5.9 (28.7)[a]
96.3 (90.9)[a]
2.9 (2.3)[a]
19.5 (2.1)[a]
30–1.68
42831
30–1.37
81623 (3991)[a]
3.1 (13.1)[a]
95 (94.5)[a]
2.7 (2.7)[a]
25.2 (8.1)[a]
30–1.37
76018
I/s (I)
resolution range [ꢃ]
reflections used in refinement
final R values
Rfree (Fo; Fo >4sFo)[c]
Rwork (Fo; Fo >4sFo)[d]
no. of atoms (non-hydrogen)
protein atoms
21.4 (20.0)
16.3 (15.3)
20.9 (19.1)
17.4 (16.0)
21.8 (20.5)
17.4 (16.4)
20.3 (18.8)
16.4 (15.4)
20.0 (19.0)
15.9 (15.0)
16.8 (16.3)
12.6 (12.4)
2794
249
16
2723
207
16
2734
246
16
2763
291
25
2864
290
18
2873
367
19
water molecules
ligand atoms
RMSD, angle [8]
RMSD, bond [ꢃ]
Ramachandran plot[e]
most favored regions [%]
additionally allowed regions [%]
generously allowed regions [%]
mean B factors [ꢃ2]
protein atoms
1.9
0.007
2.3
0.010
2.0
0.008
2.2
0.009
2.0
0.008
2.1
0.013
95.0
4.6
0.4
96.3
3.4
0.3
95.7
4.0
0.3
94.5
4.9
0.6
94.9
4.8
0.3
96.2
3.5
0.3
15.1
24.5
12.7
13.7
22.3
15.6
19.4
28.7
23.8
17.6
28.6
15.0
15.8
25.7
14.8
14.0
28.3
12.4
water molecules
ligand atoms
[a] Values in parentheses are statistics for the highest resolution shell. [b] R(I)sym =[ꢀhꢀi jIi(h)ꢀhI(h)ij/ꢀhꢀiIi(h)]ꢄ100, in which hI(h)i is the mean of the
I(h) observation of reflection h. [c] Rwork =ꢀhkl jFoꢀFc j/ꢀhkl jFo j. [d] Rfree was calculated as shown for Rwork but on refinement-excluded 5% of data.
[e] Calculated with Procheck.[38]
the guanine subpocket provoked by the Asp102 rotation is
filled with water molecules, which form a well-defined six-
membered network (W1–W6) (Figure 2b; for residue identi-
fiers, see Table 2SI in the Supporting Information; for hy-
drogen-bonding lengths, see Figure 1SI in the Supporting In-
formation).[28] W1 solvates Asp156, whereas W3 interacts
through a hydrogen bond with N(5) of the ligand. A contact
to the rotated Asp102 is built by means of W5 and W6.
Water W3 is also connected to the five-membered water
cluster, which is well conserved between Asp102 and
Asp280 and has already attracted attention in preceding
studies[14] (Figure 2a and c). For the side chain of Gln107
(not shown), which flanks the guanine subpocket, no elec-
tron density is visible at the applied contour level. It seems
to be solvent-exposed without directed interactions.
For the complexes of ligands 8b and 8c, datasets with res-
olutions of 1.53 and 1.65 ꢃ, respectively, were obtained
(Figure 3; for hydrogen-bonding lengths see Figures 4SI and
5SI in the Supporting Information). Their binding mode is
analogous to the one observed for 8a. Asp102 is rotated
away from the bound ligand, and the newly formed space is
filled with five well-defined water molecules that are equally
positioned to W1–W4 and W6 in the complex of 8a (Fig-
ure 3b). The water network described above is apparently
highly conserved in the crystal structures. The position of
W5 is different due to a change in the network pattern.
Gln107 adopts a position facing the guanine subpocket in
the structures with 8b,c and establishes a directed hydrogen
bond to W6 (Figure 6SI in the Supporting Information). In
the complex with 8c, the side chain C=O of Gln107 almost
adopts the position of W5. A similar situation is found in
the cocrystal structure with 8b, but Gln107 is slightly tilted
relative to 8c, and the side chain O-position shifted by
about 0.4 ꢃ. Nevertheless, the Gln107 side chain interacts
with W6 directly in 8c as well. In the two complexes 8b and
8c, W3 again interacts with N(5) of the ligand. The water
positions W1, W2, W4, and W6 also occur in the apo struc-
ture (PDB code: 1P0D;[4] Figure 2SI in the Supporting Infor-
mation). In the unoccupied active site, Gln107 shows a differ-
ent conformation from 8b,c (Figure 6SI in the Supporting
Information). Nevertheless, in this conformation we also ob-
served a hydrogen bond from the terminal carboxamide
C=O to W6 (2.5 ꢃ). Similar to the reported cocrystal struc-
tures of C(2)-substituted lin-benzoguanines, there is no rea-
sonable electron density seen for the C(2) side chains in 8b
and 8c; hence they can be assumed to be highly flexible or
&
6
&
www.chemeurj.org
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 0000, 00, 0 – 0
ÝÝ
These are not the final page numbers!

Ligands for tRNA–Guanine Transglycosylase
FULL PAPER
scattered over multiple conformations.[21] These side chains
are most likely directed towards the flat and solvent-ex-
posed ribose-33 pocket. As a driving force behind the rota-
tion of Gln107 in the complexes of 8b and 8c, an influence
of the flexible C(2) side chain on the active site solvation
pattern is imaginable.
A different orientation of Asp102 and, in consequence,
a varied water network can be found in the complex of
TGT with 10a (Figure 3c). The parent scaffold contains an
extended substituent on C(4) of the tricycle that points into
the ribose-34 pocket. In analogy to preceding studies, the
ethylamine linker is used to occupy the polar ribose-34
pocket, thereby displacing the water cluster around the most
likely negatively charged Asp102 and Asp280.[13,14] The ter-
minal cyclopentyl ring undergoes a hydrophobic interaction
with Val45 and Leu68 that form the apolar bottom of the
ribose-34 pocket. In contrast to the complexes with 8a–c,
the side chain of Asp102 cannot be solvated by an interac-
tion to Asn70 and Thr71. The C(4) substituent of the ligand
hinders the complete rotation of Asp102, hence it stays in
a tilted intermediate position. However, this small rotation
also opens space in the guanine subpocket, which is filled
with water molecules that build a network (W1–W3, W5,
and W6). W1, W2, W5, and W6 are at the conserved posi-
tions as can be seen in the structures with 8a–c. Because of
the tilted position of Asp102, the former water molecule W4
is displaced and W3 now occupies a shifted position and un-
dergoes hydrogen-bonding interactions with Asp102, W2,
W5, and N(5) of the ligand (for hydrogen-bonding lengths,
see Figure 7SI in the Supporting Information).
ꢀ
X-ray cocrystal structures of C(6) N-alkylated lin-benzo-
ꢀ
guanines: For the complexes with C(6) N-alkylated lin-ben-
zoguanines 11a and 11b, datasets with resolutions of 1.68
and 1.37 ꢃ, respectively, were obtained (Figure 4). The
ꢀ
adaptation of the guanine subpocket to the binding of C(6)
N-alkylated lin-benzoguanines matches very well with the
described binding of lin-benzohypoxanthines. Due to the
ꢀ
spatial requirements of the C(6) N-alkyl moieties, the hy-
Figure 3. X-ray cocrystal structures of lin-benzohypoxanthines 8a–c and
10a bound to the active site of Z. mobilis TGT. a) Cocrystal structure of
8b (Cligand magenta) bound to the active site of TGT (PDB code: 3GEV).
There is no reasonable electron density observable for the N-ethylthio-
phene side chain, which was already found to be highly disordered in an
earlier cocrystal structure of 19b bound to TGT.[21] Gln107 forms with its
side chain C=O
a bridging hydrogen bond between W4 and W6
(d(OGln107···OW4)=2.9 ꢃ and d(OGln107···OW6)=2.7 ꢃ). After detailed
visual inspection of the electron density maps and the B values, the nitro-
gen/oxygen orientation of the Gln107 carboxamide side chain has been
rotated by 1808 and deviates from the coordinate file deposited in the
PDB. b) Superposition of 8a (Cligand green), 8b (Cligand magenta), and 8c
(Cligand yellow) in complex with TGT. Illustrated are the ligand, the dis-
cussed water molecules (shown in 8a: green, 8b: magenta, and 8c:
yellow), and the residues involved in the water network. Water positions
W1–W4 and W6 are highly conserved in the three structures. The bridg-
ing position of W5 in the complex of 8a is adopted by the side chain
C=O of Gln107 in the complexes of 8b,c (Figure 6SI in the Supporting
Information). c) Cocrystal structure of 10a bound to the active site of
TGT (PDB code: 3SM0).
Chem. Eur. J. 2012, 00, 0 – 0
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
&
7
&
ÞÞ
These are not the final page numbers!

F. Diederich, G. Klebe et al.
drogen bonds to Asp102 are also not formed in these com-
plexes. Hence, Asp102 also rotates out of the ligand binding
pocket here and interacts with Asn70 and Thr71. Water mol-
ecules fill the cavity beneath the ligand but form a different
pattern than the complexes with 8a–c and 10a (see the Sup-
porting Information). In the cocrystal structure with 11a,
three water molecules (W1–W3) take the above-mentioned
enabled space (Figure 4a). W1 is located at a supposedly en-
ergetically less favorable position surrounded by the hydro-
phobic parts of the side chains of Asp102, Ile201, and
Met153 (Figure 8SI in the Supporting Information) and only
forms hydrogen bonds to the protein with the side chain
OH and the backbone NH of Ser103. W1 also interacts with
W2, which forms hydrogen bonds to Tyr258 (not shown)
and W3. Hydrogen bonds are established between W3 and
N(5) of the ligand and one of the well-conserved water mol-
ecules between Asp102 and Asp280. Extending through
three following water molecules (W4–W6), this network
also finds its endpoint in an interaction with the side chain
ꢀ
of Asp102. The C(6) N-alkyl moieties of 11a and 11b both
adopt the s-cis conformation in the cocrystal structures. An
s-trans orientation is sterically hindered as it would create
a consequent clash if the side chain of Asp156 will not move
out of space as Asp102 does in all the studied complexes.
Likely, the movement of Asp102 is also more favored as this
movement is expected to occur during catalysis.[18] Further-
more, the s-cis conformation is also the preferred one in so-
lution, as revealed by DFT calculations at the B3LYP/6-
311GACTHUNRTGNEUNG(d,p) level of theory (Figure 9SI in the Supporting In-
formation). When comparing the cocrystal structures of 11a
and 11b, well-conserved positions for the water molecules
W2–W6 can be observed. W1 is displaced by the additional
carbon of the N-ethyl residue of 11b, which in one of the
two observable orientations (occupancy 36%) points into
the hydrophobic area of the guanine subpocket around
Ile201, Met153, and Ser103 (Figure 4b; see also Figure 10SI
in the Supporting Information). This leaves both the NH
and OH groups of Ser103 unsolvated. The orientation of the
terminal carbon of the N-ethyl moiety additionally produces
steric clashes with the side chain of Met260 that can in con-
sequence only be partly observed in the highly conserved
conformation (Figure 10SI in the Supporting Information).
In the second orientation (64% occupancy), the terminal
carbon of the N-ethyl group of 11b occurs in a nearly 1808
rotated orientation and is unfavorably directed towards the
backbone NH of Tyr106. Only when the terminal carbon of
the N-ethyl moiety adopts the latter conformation does the
side chain of Met260 assume its highly conserved position
with a corresponding occupancy.
ꢀ
Figure 4. X-ray cocrystal structures of Z. mobilis TGT with C(6) N-mon-
Origin of the differences in biological activity between lin-
benzoguanines, their N-monoalkylated derivatives, and lin-
benzohypoxanthines: Taken together, the results from the
biological and physicochemical assays and from biostructure
analysis allow a detailed insight into the molecular recogni-
tion of lin-benzoguanines and lin-benzohypoxanthines at the
active site of TGT. The lin-benzohypoxanthines are 80–680
times less potent than the corresponding lin-benzoguanines
(Table 1). The pKa measurements (Table 1) suggest that
oalkylated lin-benzoguanines 11a,b. a) Cocrystal structure of 11a bound
to the active site of TGT (PDB code: 3RR4). For clarity, the residues
Met153 and Ile201 that surround W1 are not shown (Figure 8SI in the
Supporting Information). Tyr258, hidden behind the ligand, is also not
displayed. b) Cocrystal structure of 11b bound to the active site of TGT
(PDB code: 3TLL). Two conformations of the N-ethyl moiety are ob-
served, one with an occupancy of 36%, with the terminal methyl group
pointing towards the hydrophobic cavity around Ser103, and a second,
with an occupancy of 64%, in which the terminal methyl group is nearly
flipped by 1808, thus facing the tricyclic ligand scaffold.
&
8
&
www.chemeurj.org
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 0000, 00, 0 – 0
ÝÝ
These are not the final page numbers!

Ligands for tRNA–Guanine Transglycosylase
FULL PAPER
N(5) (pKa values 1.8–2.0) of the lin-benzohypoxanthines
most probably is not protonated in this part of the scaffold,
in contrast to the lin-benzoguanines, with corresponding pKa
values 4.0–4.4. Therefore, interaction of the ligand with the
carboxylate side chain of Asp102 is unfavorable, which
causes this side chain to rotate away from the ligand into
a conformation similar to the one adopted in the apoen-
zyme. However, the largest part of the loss in binding
energy, when changing to the lin-benzohypoxanthine ligands,
most probably originates from the loss of the two short ionic
hibitors of catechol-O-methyltransferase.[32] Furthermore,
the backbone NH and the OH group of Ser103 become
fully desolvated, which should also reduce the overall bind-
ing affinity.
Conclusion
Here we have compared the properties of three series of li-
ꢀ
gands for Z. mobilis TGT: lin-benzoguanines, their C(6) N-
ꢀ
hydrogen bonds, which the exocyclic C(6) NH2 groups of
alkylated derivatives, and lin-benzohypoxanthines. Their
physical properties are dominated by the highly polar, basic
2-aminoimidazole moiety, which is the origin of low parti-
tioning and distribution constants and poor membrane per-
meability. A large difference in the measured pKa value for
the N(5)H+ moiety, which shifts from approximately 4 in
the lin-benzoguanines to around 2 in the lin-benzohypoxan-
thines may have a strong influence on the binding affinity to
the protein. Whereas lin-benzoguanines bind with Ki values
in the single- to double-digit nanomolar range, as a result of
ionic hydrogen bonding of the aminopyrimidinone ring to
two Asp side chains, lin-benzohypoxanthines only show af-
finities in the single-digit micromolar range or slightly
the lin-benzoguanines form with the carboxylates of Asp102
(e.g., in the complex of 1: d
(N···O)=2.8 ꢃ). Also, a weaker interaction with the OH of
Ser103 (d(N···O)=3.7 ꢃ) is lost. The import of a well-con-
ACHTUNGTNER(NUNG N···O)=2.8 ꢃ) and Asp156 (d-
ACHTUNGTRENNUNG
ACHTUNGTRENNUNG
served water cluster, which consists of 5 to 6 molecules that
form a network among each other and a series of hydrogen
bonds to the protein, does not compensate for the loss of
ꢀ
the hydrogen bonds from the C(6) NH2 to the anionic side
chains of Asp102 and Asp156.
ꢀ
ꢀ
The bulk of the C(6) N-methyl and C(6) N-ethyl groups
in ligands 11a and 11b prevents an interaction of the ligand
with the side chain of Asp102, which turns away to adopt
the conformation seen in the apoenzyme and in the com-
plexes of lin-benzohypoxanthines 8a–c. Compared to lin-
ꢀ
below. The C(6) N-alkylated lin-benzoguanines are even
weaker ligands with Ki values in the double-digit micromo-
lar range. The origin of the different binding affinities was
investigated by X-ray cocrystal structure analysis. The tricy-
clic scaffold of all three ligand classes adopts the same bind-
ing geometry in the guanine recognition pocket. For the lin-
ꢀ
benzoguanines with unsubstituted C(6) NH2 group, the two
hydrogen bonds to the carboxylate of Asp102 are lost, as
well as the weaker interaction with the OH group of Ser103.
In the complex of 11a, a water cluster W1–W6 is imported
into the space vacated by the conformational change of
Asp102. However, this cluster is energetically less favorable
than the cluster in the structures of lin-benzohypoxanthines
8a–c, as some of the water molecules are forced by the
ꢀ
benzohypoxanthines that lack the C(6) NH2, an ionic hy-
drogen bond to the carboxylate of Asp156, which remains in
a conserved orientation, is lost. Furthermore, the carboxyl-
ate of Asp102 no longer binds to the ligand and changes its
conformation to adopt a position seen in the apoenzyme,
which is supposedly important in the catalytic mechanism.
This conformational change opens up a space that is filled
by a conserved water cluster. Although the water cluster un-
dergoes several favorable interactions with the protein, they
are not sufficient to compensate for the loss of ligand–pro-
tein ionic hydrogen bonding, and the overall binding affinity
is strongly reduced. The structural data show that the bound
ꢀ
C(6) N-methyl group into less favorable, more hydrophobic
environments. In particular, W1 loses the ionic hydrogen
bond to the side chain of Asp156, which is seen in the com-
plexes of 8a–c (Figure 8SI in the Supporting Information),
and the water network chain to N(5) through W2 and W3 is
ꢀ
ruptured. As
a result, C(6) N-methylated 11a (Ki =
28’050 nm) has a 480-fold lower binding affinity than the
NH2 derivative 19a. The stability of the complex of N-ethy-
lated 11b (Ki =40’830 nm) is even further reduced by
a factor of 700-fold relative to 19a. First of all, the desolva-
tion of the ethyl derivative requires more energy, particular-
ly as the ethyl group does not find a favorable hydrophobic
environment in the protein. Second, the ethyl substituent
displaces W1 from the water cluster. Displacement of water
molecules from an environment of uncharged residues ap-
pears to be related to a negligible effect in the free energy;
however, large compensating contributions in enthalpy and
entropy are observed.[29,30] If displacement from an environ-
ment with charged residues is attempted, an unfavorable
free energy contribution can also be observed.[31] The latter
is also given for the complexes studied here, because W1 is
involved in interactions to the charged residue Asp156. Nev-
ertheless, favorable replacements of water molecules by N-
alkyl moieties have also been reported, for example, for in-
ꢀ
C(6) N-alkyl lin-benzoguanines maintain the double ionic
hydrogen bonding to the side chain of Asp156. However,
ꢀ
for steric reasons, the bulky C(6) N-alkyl moiety prevents
the interaction with the side chain of Asp102, which again
switches to the orientation seen in the apoenzyme. The
opened space is again filled by a water cluster, which never-
theless interacts in a less favorable way with the protein and
the ligand. In particular, the cocrystal structure of the N-
ethyl derivative 11b shows a fully desolvated Ser103 residue
along with the loss of a water contact to the adjacent
charged Asp156 residue, which presumably causes the poor-
est binding affinity in the entire series. Together with our
previous work on the substitution of the water cluster be-
tween the two side chains of catalytic Asp102 and
Asp280,[12,21] this investigation documents that much remains
to be learned about the energetics of such water clusters in
Chem. Eur. J. 2012, 00, 0 – 0
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
&
9
&
ÞÞ
These are not the final page numbers!

F. Diederich, G. Klebe et al.
1566 (s), 1454 (s), 1367 (w), 1289 (m), 1218 (w), 1093 (w), 913 (w), 851
(w), 786 (w), 695 cmꢀ1 (m); HR-MALDI-MS: m/z: calcd (%) for
C14H12N5OS+: 298.0757; found: 299.0787 (19), 298.0755 (100) [M+H]+.
protein–ligand complexes and of the substitution of individ-
ual water molecules in these clusters by ligand parts. It is
clear from this work that such an enhancement in the
knowledge will have to rely on accurate, high-resolution X-
ray cocrystal structure analysis such as that reported in this
contribution. Small changes in the ligand may not only lead
to losses/gains in protein–ligand hydrogen bonds but might
also profoundly affect the solvation and the formation of
contiguously connected water networks that mediate inter-
actions in such a complex, which makes it even more diffi-
cult to assign accurate energetic quantities to lost/formed
hydrogen bonds.
2-{[2-(4-Morpholinyl)ethyl]amino}-1,7-dihydro-8H-imidazoACTHNUTRGNEUGN[4,5-g]quina-
zolin-8-one (8c): According to GP 2b, starting from 7c (28 mg,
0.07 mmol). Flash chromatography (MCI gel; H2O/MeCN 100:0 to
80:20), followed by lyophilization, yielded 8c (20 mg, 99%) as a white
powder. M.p. >1608C (decomp); 1H NMR (300 MHz, CD3OD): d=2.56
(t, J=4.5 Hz, 4H; N
ACHTUNGTRENNUNG
(t, J=6.3 Hz, 2H; CH2NH), 3.71 (brt, J=4.5 Hz, 4H; OAHCTUNGTRENNUNG
ꢀ
ꢀ
ꢀ
1H; H C(4)), 7.95 (s, 1H; H C(6)), 7.96 ppm (s, 1H; H C(9));
13C NMR (100 MHz, D2O+1 drop TFA): d=37.49 (CH2NH), 52.24 (2C;
NACHTNURTGENN(UG CH2)2), 54.84 (CH2CH2NH), 63.65 (2C; OCATHUNGTRNEN(UNG CH2)2), 102.79 (C(9)),
109.49 (C(4)), 117.03 (C(8a)), 131.13 (C(4a)), 135.09 (C(3a)), 136.32
(C(9a)), 147.46 (C(6)), 152.68 (C(2)), 160.24 ppm (C(8)); IR (ATR): n˜ =
3238 (brw), 1626 (s), 1610 (s), 1570 (m), 1462 (s), 1381 (w), 1297 (s), 1205
(w), 1111 (s), 1003 (m), 911 (w), 866 (m), 786 cmꢀ1 (m); HR-MALDI-
MS: m/z: calcd (%) for C15H19N6O2+: 315.1564; found: 316.1598 (15),
315.1562 (100) [M+H]+.
Experimental Section
Materials and general methods: Compounds 3, 4, 5a, 6a,[12] and 9a,b[13]
were prepared as described in the literature. In the following, the experi-
mental details for the syntheses of compounds 8a–d, 10a,b, and 11a,b
are described. All other synthetic details and experimental data, NMR
spectra, description of the pKa measurements and PAMPA score, and the
biological assay can be found in the Supporting Information. Crystallo-
graphic pictures were prepared using PyMol.[33]
2-[(2-Phenylethyl)amino]-1,7-dihydro-8H-imidazoACTHNUGTRNEUNG[4,5-g]quinazolin-8-one
(8d): According to GP 2b, starting from 7d (32 mg, 0.077 mmol). Flash
chromatography (MCI gel; H2O/MeCN 100:0 to 80:20), followed by lyo-
philization, yielded 8d (26 mg, 89%) as a white solid. M.p. >2508C;
1H NMR (300 MHz, CD3OD): d=3.06 (t, J=7.2 Hz, 2H; CH2CH2NH),
ꢀ
3.77 (t, J=7.2 Hz, 2H; CH2NH), 7.17–7.23 (m, 1H; H C(4’)), 7.28–7.34
ꢀ
ꢀ
ꢀ
(m, 4H; H C(2’,3’,5’,6’)), 7.65 (s, 1H; H C(4)), 8.13 (s, 1H; H C(6)),
8.92 ppm (s, 1H; H C(9)); 13C NMR (100 MHz, (CD3)2SO): d=34.44
(CH2CH2NH), 44.26 (CH2NH), 106.37 (C(9)), 107.02 (C(4)), 117.76
ꢀ
General procedure 1 (GP 1) for the cyclization to the lin-benzohypoxan-
thines with formamide: A solution of the benzimidazole in anhydrous
formamide was heated at 1408C for 18–22 h under Ar and evaporated by
bulb-to-bulb distillation (1 mbar, 1408C).
(C(8a)), 126.47 (C(4’)), 128.36 (2 C; C
N
ACHTUNGTRENNUNG(3’,5’)),
130.07 (C(4a)), 135.78 (C(1’)), 138.12 (2 C; C
AHCTUNGTRENNUNG
151.94 (C(2)), 160.14 ppm (C(8)); IR (ATR): n˜ =3213 (w), 2988 (w), 2901
(w), 1705 (s), 1668 (s), 1657 (s), 1486 (m), 1319 (m), 1298 (m), 1047 (w),
902 (w), 873 (m), 752 cmꢀ1 (m); HR-MALDI-MS: m/z: calcd (%) for
C17H16N5O+: 306.1349; found: 307.1387 (14), 306.1349 (100) [M+H]+.
General procedure 2 (GP 2) for the cleavage of the protecting group(s):
Method a: The protected lin-benzopurine (1.0 equiv) was heated in con-
centrated aqueous HCl solution/MeOH 1:2 (2.0–6.0 mL) at 658C for 18–
24 h, neutralized (pH 6–7) with saturated aqueous NaHCO3 solution, and
evaporated. The residue was suspended in MeOH, filtered, purified by
preparative HPLC (Phenomenex, 50ꢄ21.1 mm, Gemini 5 mm, C18,
110 A, AXIA; flow rate 15 mLminꢀ1, H2O +0.1 vol% HCOOH/MeCN
100:0 for 10 min, 100:0 to 80:20 within 40 min), and lyophilized to yield
the free amine. Method b: The protected lin-benzohypoxanthine
(1.0 equiv) was heated in 2m aqueous HCl/MeOH 1:1 at 658C for 30 min
and then evaporated. The residue was dissolved in H2O and neutralized
by addition of saturated aqueous NaHCO3 solution. The white precipi-
tate was removed by filtration and washed with H2O.
4-{2-[(Cyclopentylmethyl)amino]ethyl}-2-(methylamino)-1,7-dihydro-8H-
imidazoACTHUNRTGNEUNG[4,5-g]quinazolin-8-one (10a): According to GP 1, starting from
9a (58 mg, 0.13 mmol) in anhydrous formamide (4.0 mL). The crude
product was directly used according to GP 2a to yield 10a (26 mg, 60%
over two steps) as a white foam. M.p. >3158C (decomp); 1H NMR
ꢀ
(600 MHz, (CD3)2SO): d=1.23 (dt, J=13.3, 7.6 Hz, 2H; H CACHTUNGTRENNUNG(2’,5’)),
ꢀ
ꢀ
1.46–1.54 (m, 2H; H CACTHNUGTRENNUG(3’,4’)), 1.57–1.62 (m, 2H; H CACHTUNGTNER(NUGN 3’,4’)), 1.77 (dt,
ꢀ
ꢀ
J=13.3, 7.1 Hz, 2H; H CACTHNUTRGNE(UNG 2’,5’)), 2.16 (q, J=7.5 Hz, 1H; H C(1’)), 2.93
ꢀ
(d, J=7.2 Hz, 2H; CH2 C(1’)), 2.96 (s, 3H; NMe), 3.17 (t, J=7.1 Hz,
2H; CH2 C(4)), 3.54 (t, J=7.1 Hz, 2H; CH2CH2 C(4)), 7.60 (brs, 1H;
ꢀ
ꢀ
2-(Methylamino)-1,7-dihydro-8H-imidazo
G
(8a):
ꢀ
ꢀ
NH), 7.70 (s, 1H; H C(9)), 7.95 (s, 1H; H C(6)), 8.39 ppm (s, 2H;
According to GP 1, starting from 6a (50 mg, 0.153 mmol) in anhydrous
formamide (2.0 mL). The resulting crude 7a was directly used according
to GP 2a to yield 8a (16 mg, 49% over two steps) as a white solid. M.p.
>2458C (decomp); 1H NMR (300 MHz, 1m NaOD in D2O): d=2.83 (s,
2NH); 13C NMR (150 MHz, (CD3)2SO): d=23.27 (CH2 C(4)), 25.08
ꢀ
(2C; C
(3’,4’)), 29.40 (NMe), 30.62 (2C; C
(2’,5’)), 37.27 (C(1’)), 47.67
ꢀ
ꢀ
(CH2 C(1’)), 52.14 (CH2CH2 C(4)), 104.62 (C(4)), 115.86 (C(9)), 116.33
(C(8a)), 138.09 (C(3a)), 141.65 (C(9a)), 145.75 (C(4a)), 160.03 (C(6)),
161.90 (C(2)), 165.97 ppm (C(8)); IR (ATR): n˜ =3352 (m), 3143 (m),
3016 (m), 2911 (m), 2869 (m), 2803 (m), 1699 (s), 1664 (s), 1600 (s), 1449
(s), 1291 (m), 1226 (m), 1190 (m), 1158 (m), 1097 (m), 1023 (m), 984 (m),
896 (m), 826 (m), 802 (m), 760 (m), 689 cmꢀ1 (m); HR-MALDI-MS: m/z:
calcd (%) for C18H25N6O+: 341.2084; found: 342.2129 (13), 341.2088
(100) [M+H]+.
ꢀ
ꢀ
3H; NMe), 7.18 (s, 1H; H C(4)), 7.64 (s, 1H; H C(6)), 8.00 ppm (s, 1H;
H C(9)); 13C NMR (100 MHz, 1m NaOD in D2O): d=29.23 (NMe),
ꢀ
103.90 (C
(4 or 9)), 104.46 (C
(4 or 9)), 113.74 (C(8a)), 144.23 (C(9a)),
146.39 (C(3a)), 152.39 (C(4a)), 153.10 (C(6)), 170.93 (C(2)), 173.34 ppm
(C(8)); IR (ATR): n˜ =3540–2500 (brm), 1678 (s), 1649 (s), 1625 (s), 1568
(m), 1463 (m), 1375 (m), 1349 (m), 1279 (m), 1224 (m), 1197 (m), 1155
(m), 1099 (m), 985 (m), 920 (m), 841 (m), 783 (m), 694 cmꢀ1 (m); HR-
MALDI-MS: m/z: calcd (%) for C10H10N5O+: 216.0880; found: 216.0878
(100) [M+H]+.
4-{2-[(Cyclohexylmethyl)amino]ethyl}-2-(methylamino)-1,7-dihydro-8H-
imidazoACTHUNRTGNEUNG[4,5-g]quinazolin-8-one (10b): According to GP 1, starting from
9b (70 mg, 0.15 mmol) in anhydrous formamide (4.0 mL). The crude
product was directly used according to GP 2a to yield 10b (35 mg, 66%)
2-[(Thien-2-ylmethyl)amino]-1,7-dihydro-8H-imidazo
one (8b): According to GP 2a, starting from 7b (53 mg, 0.13 mmol) to
yield 8b (36 mg, 92%) as white solid. M.p. >1988C (decomp);
1H NMR (300 MHz, D2O/TFA 95:5): d=4.39 (s, 2H; CH2), 6.52 (dd, J=
ACHTUNGERTN[NUNG 4,5-g]quinazolin-8-
as
a
white foam. M.p. >2988C (decomp); 1H NMR (600 MHz,
a
ꢀ
(CD3)2SO): d=0.93 (qd, J=11.9, 2.4 Hz, 2H; Hax
C
ꢀ
N
ꢀ
ꢀ
ꢀ
12.3, 3.2 Hz, 1H; Hax C(4’)), 1.16–1.24 (m, 2H; Hax
5.1, 3.6 Hz, 1H; H C(4’)), 6.66 (dd, J=3.6, 1.2 Hz, 1H; H C(3’)), 6.89
ꢀ
ꢀ
ꢀ
ꢀ
(m, 4H; H C(1’), Heq
CACTHNGUTERN(UNG 2’,3’,5’)), 1.75 (brd, J=13.2 Hz, 2H;
(2’,6’)), 2.79 (d, J=6.8 Hz, 2H; CH2 C(1’)), 2.96 (s, 3H; NMe),
3.13 (t, J=7.3 Hz, 2H; CH2 C(4)), 3.51 (t, J=7.3 Hz, 2H; CH2CH2
C(4)), 7.60 (brs, 1H; NH), 7.69 (s, 1H; H C(9)), 7.94 (s, 1H; H C(6)),
(dd, J=5.1, 1.2 Hz, 1H; H C(5’)), 7.20 (s, 1H; H C(4)), 7.68 (s, 1H;
H C(6)), 8.65 ppm (s, 1H; H C(9)); 13C NMR (100 MHz, D2O/TFA
95:5): d=41.13 (CH2), 101.37 (C(9)), 108.83 (C(4)), 116.09 (C(8a)),
ꢀ
ꢀ
ꢀ
ꢀ
Heq
C
G
ꢀ
ꢀ
125.97 (C(5’)), 126.67 (CACTHNUGRTNENUG(3’ or 4’)), 126.80 (CACHUTNGTREN(NUNG 3’ or 4’)), 130.86 (C(4a)),
133.26 (C(2’)), 135.96 (C(6)), 136.54 (C(9a)), 147.00 (C(3a)), 151.65
(C(2)), 158.96 ppm (C(8)); IR (ATR): n˜ = 3024 (brw), 1620 (s), 1594 (s),
13
8.42 ppm (s, 2H; NH); C NMR (150 MHz, (CD3)2SO): d=22.92 (CH2
C(4)), 25.02 (2C;
CU
(3’,5’)), 25.61 (NMe), 28.82 (C(4’)), 30.05 (2C;
&
10
&
www.chemeurj.org
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 0000, 00, 0 – 0
ÝÝ
These are not the final page numbers!

Ligands for tRNA–Guanine Transglycosylase
FULL PAPER
sets, further refinement using SHELXL-97 was performed.[36] For each
refinement step, at least 20 cycles of conjugate gradient minimization
were made with default restraints on bond lengths, angles, and B values.
A total of 5% of all data were used for Rfree calculation. Intermittent
model building was performed using COOT.[37] The ligand, water, and
glycerol molecules were placed into the sigma-A weighted difference
electron density and included in further refinement cycles. Riding hydro-
gen atoms were added for the protein in a final refinement cycle without
using additional parameters. Model analysis was performed using PRO-
CHECK.[38] Coordinate files were deposited in the PDB with the follow-
ing access codes: 3S1G, 3GEV, 3V0Y, 3SM0, 3RR4, 3TLL.
ꢀ
ꢀ
C
E
(C(9)), 115.31 (C(4)), 116.00 (C(8a)), 137.30 (C(3a)), 141.00 (C(9a)),
144.68 (C(4a)), 159.44 (C(6)), 161.33 (C(2)), 165.53 ppm (C(8)); IR
(ATR): n˜ =3366 (m), 3008 (m), 2924 (m), 2853 (m), 2808 (m), 1689 (s),
1661 (s), 1598 (s), 1445 (s), 1291 (m), 1225 (m), 1188 (w), 1157 (w), 1102
(w), 983 (m), 889 (m), 843 (m), 762 (m), 689 cmꢀ1 (m); HR-MALDI-MS:
m/z: calcd (%) for C19H27N6O+: 355.2241; found: 356.2283 (23), 355.2247
(100) [M+H]+.
2,6-Bis(methylamino)-1,7-dihydro-8H-imidazoACTHNUTRGNEUG[N 4,5-g]quinazolin-8-one
(11a): According to GP 2a, starting from 18a (49 mg, 0.109 mmol) and
yielding 11a (23 mg, 87%) as a white solid. M.p. >3048C (decomp);
1H NMR (600 MHz, (CD3)2SO): d=2.82 (d, J=4.2 Hz, 3H; NMe), 2.89
(d, J=3.0 Hz, 3H; NMe), 6.29 (brs, 1H; NH), 6.82 (brs, 1H; NH), 6.91
(s, 1H; H C(4)), 7.55 ppm (s, 1H; H C(9)); 13C NMR (150 MHz,
(CD3)2SO): d=27.39 (NMe), 28.87 (NMe), 104.95 (C(9)), 110.16 (C(4)),
145.70 (2C;
CACHTUNGTRENNUNG(3a,9a)), 150.25 (C(2)), 158.52 (C(6)), 163.65 (2C;
Acknowledgements
CACHTUNGTRENNUNG(4a,8a)), 166.84 ppm (C(8)); IR (ATR): n˜ =3253 (w, sh), 3042 (w), 2932
(w), 1682 (m), 1672 (m), 1597 (s), 1570 (m), 1536 (m), 1450 (m), 1411 (s),
1350 (m), 1330 (m), 1283 (m), 1227 (m), 1207 (m), 1152 (m), 1118 (m),
979 (m), 871 (m), 864 (m), 782 (m), 627 cmꢀ1 (m); HR-MALDI-MS: m/z:
calcd (%) for C11H12N6NaO+: 267.0970; found: 267.0970 (99) [M+Na]+;
m/z: calcd (%) for C11H13N6O+: 245.1145; found 245.1151 (100) [M+H]+.
We are grateful to the ETH Research Council, F. Hoffmann-La Roche
(Basel), and Chugai Pharmaceuticals for the generous support of the
work done at ETH. The research at the University of Marburg was sup-
ported by a grant from the Deutsche Forschungsgesellschaft (KL1204/13-
1). We thank Daniel Zimmerli (Roche, Basel) for advice regarding
HPLC, Dr. Bruno Bernet (ETH) for proofreading the Experimental Sec-
tion, Michael Seet (ETH) for proofreading the manuscript, and Pablo
Rivera-Fuentes (ETH) for DFT calculations. We acknowledge the sup-
port from the beamline staff at BESSY II in Berlin and a travel grant
from the Helmholtz-Zentrum fꢀr Materialien und Energie in Berlin.
6-(Ethylamino)-2-(methylamino)-1,7-dihydro-8H-imidazoACTHNUTRGNE[UNG 4,5-g]quinazo-
lin-8-one (11b): According to GP 2a, starting from 18b (124 mg,
0.266 mmol) and yielding 11b (59 mg, 86%) as a white solid. M.p.
>2698C (decomp); 1H NMR (300 MHz, 1m aqueous NaOD solution):
d=1.07 (t, J=7.7 Hz, 3H; NCH2CH3), 2.82 (s, 3H; NMe), 3.22 (q, J=
ꢀ
7.7 Hz, 2H; NCH2CH3), 6.95 (s, 1H; H C(4)), 7.54 ppm (s, 1H;
ꢀ
H C(9)); 13C NMR (150 MHz, (CD3)2SO): d=14.84 (NCH2CH3), 28.95
[1] C. Frank, M. S. Faber, M. Askar, H. Bernard, A. Fruth, A. Gilsdorf,
M. Hçhle, H. Karch, G. Krause, R. Prager, A. Spode, K. Stark, D.
Werber, Euro. Surveill. 2011, 16, pii=19878. Available online:
http://www.eurosurveillance.org/ViewArticle.aspx?ArticleId=19878.
[2] P. J. Sansonetti, FEMS Microbiol. Rev. 2001, 25, 3–14.
[3] J. M. B. Durand, B. Dagberg, B. E. Uhlin, G. R. Bjçrk, Mol. Micro-
biol. 2000, 35, 924–935.
[4] U. Grꢁdler, H.-D. Gerber, D. M. Goodenough-Lashua, G. A. Garcia,
R. Ficner, K. Reuter, M. T. Stubbs, G. Klebe, J. Mol. Biol. 2001, 306,
455–467.
(NMe), 34.96 (CH2CH3), 105.20 (C(9)), 110.18 (C(4)), 145.85 (2C;
C
(3a,9a)), 149.10 (C(2)), 158.66 (C(6)), 163.38 (2C;
CACTHNGUTERN(UNG 4a,8a)),
167.32 ppm (C(8)); IR (ATR): n˜ =3448 (w), 3249 (w, sh), 3103 (w, sh),
3030 (w), 2973 (w), 2920 (w), 2878 (w), 1672 (w), 1592 (s), 1567 (m), 1515
(m), 1472 (m), 1432 (m), 1409 (s), 1365 (m), 1340 (m), 1321 (m), 1282
(m), 1228 (m), 1186 (m), 1164 (m), 1123 (m), 1106 (m), 1017 (w), 982
(w), 858 (m), 844 (m), 780 (m), 679 (m), 657 (m), 621 cmꢀ1 (m); HR-
MALDI-MS: m/z: calcd (%) for C12H14N6NaO+: 281.1127; found:
281.1118 (87) [M+Na]+; m/z: calcd (%) for C12H15N6O+: 259.1302;
found: 259.1298 (100) [M+H]+.
[5] B. Stengl, E. A. Meyer, A. Heine, R. Brenk, F. Diederich, G. Klebe,
J. Mol. Biol. 2007, 370, 492–511.
[6] E. A. Meyer, N. Donati, M. Guillot, W. B. Schweizer, F. Diederich,
B. Stengl, R. Brenk, K. Reuter, G. Klebe, Helv. Chim. Acta 2006, 89,
573–597.
Z. mobilis TGT crystallization: Protein was cloned, overexpressed, and
purified as described in detail elsewhere.[34] Crystals of TGT appropriate
for data collection were obtained by using the hanging-drop, vapor-diffu-
sion method at 288 K. TGT was cocrystallized with inhibitors. A protein
solution (12 mgmLꢀ1 TGT, 10 mm Tris-HCl (Tris=tris(hydroxymethyl)-
aminomethane), pH 7.8, 2m NaCl, 1 mm EDTA (ethylenediaminetetra-
acetic acid), 15% (v/v) Me2SO) was incubated with 1.5 mm inhibitor. A
total of 2 mL of this solution was mixed with 2 mL reservoir solution
(100 mm MES (2-(N-morpholino)ethanesulfonic acid), pH 5.5, 1 mm DTT
(dithiothreitol), 10% (v/v) Me2SO, 13% (w/v) polyethylene glycol (PEG)
8000). After three weeks of crystal growth, crystals reached dimensions
around 0.7ꢄ0.7ꢄ0.2 mm3. Instead of cocrystallization, compounds 8b,c
were soaked at a final concentration of 5 mm for one day into wild-type
crystals.
[7] For an overview, see: a) N. J. Leonard, A. G. Morrice, M. A. Spre-
cker, J. Org. Chem. 1975, 40, 356–363; b) G. E. Keyser, N. J. Leo-
´
nard, J. Org. Chem. 1976, 41, 3529–3532; c) N. J. Leonard, F. Kaz-
mierczak, A. Rykowski, J. Org. Chem. 1987, 52, 2933–2935.
[8] For lin-benzoadenine-based inhibitors of the S-adenosyl-l-homocys-
ˇ
´
teine (SAH) hydrolase, see: a) H. Dvorꢅkovꢅ, A. Holy, M. Masojꢆd-
kovꢅ, Collect. Czech. Chem. Commun. 1988, 53, 1779–1794; b) V. P.
Rajappan, S. W. Schneller, Tetrahedron 2001, 57, 9049–9053.
[9] For lin-benzoadenine-based inhibitors of the epidermal growth
factor receptor (EGFR), see: a) G. W. Rewcastle, B. D. Palmer, A. J.
Bridges, H. D. H. Showalter, L. Sun, J. Nelson, A. McMichael, A. J.
Kraker, D. W. Fry, W. A. Denny, J. Med. Chem. 1996, 39, 918–928;
b) Y. B. Lee, C. H. Ahn, Patent US 2005/0187231A1, 2005.
X-ray data collection: Crystals were transferred for 10 s into a cryoprotec-
tant-containing buffer (50 mm MES, pH 5.5, 300 mm NaCl, 0.5 mm DTT,
2% (v/v) Me2SO, 4% (w/v) PEG 8000, 30% (v/v) glycerol). Afterwards,
crystals were flash-frozen in liquid nitrogen. All datasets were collected
under cryo conditions (100 K) at the BESSY-PSF Beamline 14.2 in
Berlin at a wavelength of l=0.91841 ꢃ. A Rayonix MX225 CCD detec-
tor was used for data collection. TGT crystals exhibited a monoclinic
space group C2 with one monomer per asymmetric unit. Data processing
and scaling was carried out using HKL2000. Unit-cell dimensions, data
collection, and processing statistics are given in Table 2.
[10] For 6-arylimidazoACTHNUTRGNE[NUG 4,5-g]quinazolin-8(7H)-ones as inhibitors of phos-
phodiesterase type 5 (PDE5), see: a) J. E. Macor, D. P. Rotella,
H. N. Weller, III, D. W. Cushman, J. P. Yevich, Patent US 6087368,
2000; b) D. P. Rotella, Z. Sun, Y. Zhu, J. Krupinski, R. Pongrac, L.
Seliger, D. Normandin, J. E. Macor, J. Med. Chem. 2000, 43, 5037–
5043.
[11] For an overview, see: a) H. Liu, J. Gao, L. Maynard, Y. D. Saito,
E. T. Kool, J. Am. Chem. Soc. 2004, 126, 1102–1109; b) H. Liu, S. R.
Lynch, E. T. Kool, J. Am. Chem. Soc. 2004, 126, 6900–6905; c) J.
Gao, H. Liu, E. T. Kool, J. Am. Chem. Soc. 2004, 126, 11826–11831;
d) H. Liu, J. Gao, E. T. Kool, J. Org. Chem. 2005, 70, 639–647;
e) A. R. Hernꢅndez, E. T. Kool, Org. Lett. 2011, 13, 676–679;
Structure determination and refinement: For all complexes, the CNS pro-
gram package was used to perform the initial rigid-body refinement and
a cycle of conjugate gradient energy minimization, simulated annealing,
and B-factor refinement.[35] The coordinates of the TGT apo structure
1P0D were used as starting model. Due to the high resolution of all data-
Chem. Eur. J. 2012, 00, 0 – 0
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
&
11
&
ÞÞ
These are not the final page numbers!

F. Diederich, G. Klebe et al.
f) A. T. Krueger, L. W. Peterson, J. Chelliserry, D. J. Kleinbaum,
E. T. Kool, J. Am. Chem. Soc. 2011, 133, 18447–18451.
[23] ACD/Labs, Version 12.01, Advanced Chemistry Development, Inc.,
Toronto, 2009.
[12] S. R. Hçrtner, T. Ritschel, B. Stengl, C. Kramer, W. B. Schweizer, B.
Wagner, M. Kansy, G. Klebe, F. Diederich, Angew. Chem. 2007, 119,
8414–8417; Angew. Chem. Int. Ed. 2007, 46, 8266–8269.
[13] P. C. Kohler, T. Ritschel, W. B. Schweizer, G. Klebe, F. Diederich,
Chem. Eur. J. 2009, 15, 10809–10817.
[24] For reviews of PAMPA scores, see: a) S. Bendels, O. Tsinman, B.
Wagner, D. Lipp, I. Parrilla, M. Kansy, A. Avdeef, Pharm. Res. 2006,
23, 2525–2535; b) A. Avdeef, S. Bendels, L. Di, B. Faller, M. Kansy,
K. Sugano, Y. Yamauchi, J. Pharm. Sci. 2007, 96, 2893–2909.
[25] X. Gong, M. Figus, J. Plewa, D. A. Levorse, L. Zhou, C. J. Welch,
Chromatographia 2008, 68, 219–225.
[26] B. Stengl, K. Reuter, G. Klebe, ChemBioChem 2005, 6, 1926–1939.
[27] R. Brenk, M. T. Stubbs, A. Heine, K. Reuter, G. Klebe, ChemBio-
Chem 2003, 4, 1066–1077.
[28] For reviews on water clusters, see: a) J. E. Ladbury, Chem. Biol.
1996, 3, 973–980; b) R. Ludwig, Angew. Chem. 2001, 113, 1856–
1876; Angew. Chem. Int. Ed. 2001, 40, 1808–1827; c) S. W. Homans,
Drug Discovery Today 2007, 12, 534–539; d) P. Ball, Chem. Rev.
2008, 108, 74–108.
[29] T. Petrova, H. Steuber, I. Hazemann, A. Cousido-Siah, A. Mitschler,
R. Chung, M. Oka, G. Klebe, O. El-Kabbani, A. Joachimiak, A.
Podjarny, J. Med. Chem. 2005, 48, 5659–5665.
[14] T. Ritschel, P. C. Kohler, A. Heine, F. Diederich, G. Klebe, Chem-
MedChem 2009, 4, 2012–2023.
[15] For some examples on the preference for guanine-based over hypo-
xanthine-based ligands, see: a) R. Seifert, U. Gether, K. Wenzel-Sei-
fert, B. K. Kobilka, Mol. Pharmacol. 1999, 56, 348–358; b) M. J.
Barnes, N. Cooper, R. J. Davenport, H. J. Dyke, F. P. Galleway,
F. C. A. Galvin, L. Gowers, A. F. Haughan, C. Lowe, J. W. G. Meiss-
ner, J. G. Montana, T. Morgan, C. L. Picken, R. J. Watson, Bioorg.
Med. Chem. Lett. 2001, 11, 1081–1083; c) S. Kohgo, K. Yamada, K.
Kitano, Y. Iwai, S. Sakata, N. Ashida, H. Hayakawa, D. Nameki, E.
Kodama, M. Matsuoka, H. Mitsuya, H. Ohrui, Nucleosides Nucleo-
tides Nucleic Acids 2004, 23, 671–690; d) A. B. Eldrup, C. R. Aller-
son, C. F. Bennett, S. Bera, B. Bhat, N. Bhat, M. R. Bosserman, J.
Brooks, C. Burlein, S. S. Caroll, P. D. Cook, K. L. Getty, M. Mac-
Coss, D. R. McMasters, D. B. Olsen, T. P. Prakash, M. Prhavc, Q.
Song, J. E. Tomassini, J. Xia, J. Med. Chem. 2004, 47, 2283–2295.
[16] For some examples on the preference for hypoxanthine-based over
guanine-based ligands, see: a) Q. Chao, L. Deng, H. Shih, L. M.
Leoni, D. Genini, D. A. Carson, H. B. Cottam, J. Med. Chem. 1999,
42, 3860–3873; b) Y. Chong, H. Choo, Y. Choi, J. Mathew, R. F.
Schinazi, C. K. Chu, J. Med. Chem. 2002, 45, 4888–4898; c) M. Bret-
ner, S. Schalinski, P. Borowski, T. Kulikowski, Nucleosides Nucleo-
tides Nucleic Acids 2003, 22, 1531–1533.
[17] For some examples of PNP inhibitors, see: a) S. Niwas, P. Chand,
V. P. Pathak, J. A. Montgomery, J. Med. Chem. 1994, 37, 2477–2480;
b) P. E. Morris, Jr., A. J. Elliott, S. P. Walton, C. H. Williams, J. A.
Montgomery, Nucleosides Nucleotides Nucleic Acids 2000, 19, 379–
404; c) E. A. Taylor, K. Clinch, P. M. Kelly, L. Li, G. B. Evans, P. C.
Tyler, V. L. Schramm, J. Am. Chem. Soc. 2007, 129, 6984–6985.
[18] W. Xie, X. Liu, R. H. Huang, Nat. Struct. Biol. 2003, 10, 781–788.
[19] G. M. Coppola, G. E. Hardtmann, O. R. Pfister, J. Org. Chem. 1976,
41, 825–831.
[30] A. Biela, F. Sielaff, F. Terwesten, A. Heine, T. Steinmetzer, G.
Klebe, J. Med. Chem. DOI: 10.1021/jm300337q.
[31] H. Steuber, A. Heine, G. Klebe, J. Mol. Biol. 2007, 368, 618–638.
[32] M. Ellermann, R. Jakob-Roetne, C. Lerner, E. Borroni, D. Schlatter,
D. Roth, A. Ehler, M. G. Rudolph, F. Diederich, Angew. Chem.
2009, 121, 9256–9260; Angew. Chem. Int. Ed. 2009, 48, 9092–9096.
[33] The PyMOL Molecular Graphics System, Version 1.3; Schrçdinger,
Inc., New York, 2010.
[34] For a detailed description of Z. mobilis TGT crystallization, see:
a) C. Romier, R. Ficner, K. Reuter, D. Suck, Proteins 1996, 24, 516–
519; b) C. Romier, K. Reuter, D. Suck, R. Ficner, EMBO J. 1996,
15, 2850–2857; c) K. Reuter, R. Ficner, J. Bacteriol. 1995, 177,
5284–5288.
[35] A. T. Brꢀnger, P. D. Adams, G. M. Clore, W. L. DeLano, P. Gros,
R. W. Grosse-Kunstleve, J.-S. Jiang, J. Kuszewski, M. Nigles, N. S.
Pannu, R. J. Read, L. M. Rice, T. Simonson, G. L. Warren, Acta
Crystallogr. Sect. D 1998, 54, 905–921.
[36] G. M. Sheldrick, T. R. Schneider, Methods Enzymol. 1997, 277, 319–
343.
[37] P. Emsley, K. Cowtan, Acta Crystallogr. Sect. D 2004, 60, 2126–2132.
[38] R. A. Laskowski, J. Appl. Crystallogr. 1993, 26, 283–291.
[20] N. Pluym, A. Brennauer, M. Keller, R. Ziemek, N. Pop, G. Bern-
hardt, A. Buschauer, ChemMedChem 2011, 6, 1727–1738.
[21] T. Ritschel, S. Hçrtner, A. Heine, F. Diederich, G. Klebe, ChemBio-
Chem 2009, 10, 716–727.
Received: March 10, 2012
[22] A. Leo, C. Hansch, D. Elkins, Chem. Rev. 1971, 71, 525–616.
Published online: && &&, 0000
&
12
&
www.chemeurj.org
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 0000, 00, 0 – 0
ÝÝ
These are not the final page numbers!

Ligands for tRNA–Guanine Transglycosylase
FULL PAPER
Water replacements: A series of lin-
benzopurines was evaluated as inhibi-
tors of Zymomonas mobilis tRNA–
guanine transglycosylase, an enzyme
that was identified as a potential target
for the treatment of shigellosis. X-ray
cocrystal structures show the import of
a new water cluster that replaces lost
protein–ligand interactions (see
figure), with an overall reduction in
binding affinity.
Drug Design
L. J. Barandun, F. Immekus,
P. C. Kohler, S. Tonazzi, B. Wagner,
S. Wendelspiess, T. Ritschel, A. Heine,
M. Kansy, G. Klebe,*
F. Diederich* .................. &&&& — &&&&
From lin-Benzoguanines to lin-Benzo-
hypoxanthines as Ligands for Zymo-
monas mobilis tRNA–Guanine Trans-
glycosylase: Replacement of Protein–
Ligand Hydrogen Bonding by
Importing Water Clusters
Chem. Eur. J. 2012, 00, 0 – 0
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
&
13
&
ÞÞ
These are not the final page numbers!