Inhibitors for human glutaminyl cyclase by structure based design and bioisosteric replacement

Article abstract of DOI:10.1021/jm900969p

The inhibition of human glutaminyl cyclase (hQC) has come into focus as a new potential approach for the treatment of Alzheimer's disease. The hallmark of this principle is the prevention of the formation of Aβ _{3,11(pE)-40,42}, as these Aβ-speci

Full text of DOI:10.1021/jm900969p

J. Med. Chem. 2009, 52, 7069–7080 7069
DOI: 10.1021/jm900969p
Inhibitors for Human Glutaminyl Cyclase by Structure Based Design and Bioisosteric Replacement
†
†
†
§
¥
¥
‡
€
Mirko Buchholz, Antje Hamann, Susanne Aust, Wolfgang Brandt, Livia Bohme, Torsten Hoffmann, Stephan Schilling,
Hans-Ulrich Demuth,†,‡,¥ and Ulrich Heiser*,†
†Department of Medicinal Chemistry and ‡Department of Enzymology and ¥Department of Preclinical Pharmacology, Probiodrug AG,
Weinbergweg 22, 06120 Halle, Germany, and §Department of Bioorganic Chemistry, Leibniz Institute of Plant Biochemistry, Weinberg 3,
D-06120 Halle, Germany
Received July 1, 2009
The inhibition of human glutaminyl cyclase (hQC) has come into focus as a new potential approach for
the treatment of Alzheimer’s disease. The hallmark of this principle is the prevention of the formation of
Aβ3,11(pE)-40,42, as these Aβ-species were shown to be of elevated neurotoxicity and likely to act as a
seeding core leading to an accelerated formation of Aβ-oligomers and fibrils. Starting from 1-(3-(1H-
imidazol-1-yl)propyl)-3-(3,4-dimethoxyphenyl)thiourea, bioisosteric replacements led to the develop-
ment of new classes of inhibitors. The optimization of the metal-binding group was achieved by
homology modeling and afforded a first insight into the probable binding mode of the inhibitors in the
hQC active site. The efficacy assessment of the hQC inhibitors was performed in cell culture, directly
monitoring the inhibition of Aβ3,11(pE)-40,42 formation.
Introduction
The N-terminal modification of peptides by the generation
of a pGlua moiety from the corresponding glutamine pre-
cursors is an important process in the maturation of bioactive
peptides1-3 which is catalyzed by glutaminyl cyclases (QC).4
QC activity is commonly found in nature.5-8 In the case of the
mammalian QC activity, the catalytic process is found to be
zinc-dependent9 whereas bacterial and plant QC catalysis
differs with regard to mechanism and the structure of
the catalyzing protein.10,11 The protein structure of human
QC is closely related to bacterial aminopeptidases and other
members of the zinc hydrolase family.12,13 In contrast to
structurally related aminopeptidases, mammalian QCs pos-
sess only one zinc ion in the active site adopting a tetrahedral
geometry.9,13,14
The protein sequence is highly conserved throughout the
different mammalian species, thereby having an identity
of 80-90%.15 The recently identified isoforms of the human
and murine enzymes have a high sequence homology and
show a close substrate specificity compared to the human
protein.16
Human QC activity is highly abundant in secretory and
neuronal tissue;17 it is preferentially found in brain regions
such as hippocampus and hypothalamus.7,17 Recent develop-
ments have related QC activity to the processing of the
β-amyloid peptides (Aβ) in the course of Alzheimer’s disease
(AD).18-20 It was shown that QC is probably the responsible
enzyme for the formation of N-terminal pGlu residues after
the liberation of the N-terminus of glutamate at position 3 of
the Aβ-peptide, thereby exhibiting a glutamate cyclase (EC)
Figure 1. First potent hQC inhibitors.26
activity.21 The formation of this modified Aβ species is likely
to be a crucial event in the initiation and the progress of the
disease, as these pGlu containing peptides exhibit an elevated
neurotoxicity and a tendency to aggregate.22-24
The application of inhibitors of QC as a new strategy for
the treatment of AD has proven to be successful in different
transgenic animal models.19,25 A reduction of the amount
of pGlu peptides as well as the total plaque load was achieved,
leading to a significant improvement of learning and
memory.19
Recently, we have presented the first generation of potent
QC inhibitors (Figure 1).26 The compounds were developed
without information of the target protein structure, postulat-
ing the contact of the N-alkylimidazole with the catalytic zinc
ion to be crucial for the inhibitory potency.
In this manuscript we describe further development of QC
inhibitors by means of a structure-based approach utilizing a
homology model of the enzyme. Furthermore, new active
compound classes could be identified by the bioisosteric
replacement of the unfavorable thiourea moiety.27-31 The
efficacy of the compounds regarding their ability to suppress
pGlu formation at the N-terminus of the Aβ-peptide was
shown in a cell-based model.
*To whom correspondence should be addressed. Phone: þ49 (0) 345
5559908. Fax: þ49 (0) 345 5559901. E-mail: ulrich.heiser@probiodrug.de.
aAbbreviations: hQC, human glutaminyl cyclase; pGlu, pyro-gluta-
myl.
r
2009 American Chemical Society
Published on Web 10/28/2009
pubs.acs.org/jmc

7070 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 22
Buchholz et al.
Scheme 1a
a Reagents and conditions: (a) trityl chloride, NEt3, DMF, 12 h, room temp; (b) 2-(3-bromopropyl)isoindoline-1,3-dione, MeCN, 12 h, reflux;
(c) TFA (5%), MeOH, 3 h, reflux; (d) hydrazine, EtOH, 8 h, reflux, then 4 N HCl, 4 h, reflux.
Scheme 2a
a Reagents and conditions: (a) 2-(3-bromopropyl)isoindoline-1,3-dione, DMF, 8 h, 90 ꢀC; (b) trityl chloride, dichloromethane, 1.5 h 0 ꢀC to room
temp; (c) chromatographic separation; (d) 7, hydrazine, EtOH, 8 h, reflux, then 4 N HCl, 4 h, reflux.
Results and Discussion
(22-30, 41-45, and 52) were obtained by a stepwise reaction
of 2,2-bis-methylthio-1-nitroethene with the corresponding
amines (Scheme 5).33 The synthesis of the 2-thioxopyrimi-
din-4(1H)-ones (31-35, 46-50, and 53) was accomplished
via the corresponding (5-alkyloxycarbonyl)arylthiourea de-
rivatives by the formation of an amide bond leading to a ring
closure at the 1-nitrogen of the thiourea moiety (Scheme 6).
Structure-Activity Relationship. The cyanoguanidines
13-21 (Table 1) were found to be of medium potency.
Compounds containing alkyl substituents like methyl (13)
and cyclopropylmethyl (14) exhibited a potency comparable
to that of phenyl (15), 4-bromophenyl (16), and 3,4-di-
methoxyphenyl (20) derivatives. A 6-fold decrease was
found for the 4-trifluormethylphenyl derived inhibitor
17. In contrast to that, the 4-isopropylphenyl- (18), the
4-methoxyphenyl (19), and the tethering of the 3- and 4-
methoxy substituent as in (21) had a slightly improved
potency.
The influence of the variable substituent was found to be
more pronounced in the case of the nitrovinyldiamine deri-
vatives (Table 1). In this case, the methyl derivative 22 was
found to be inactive. The compounds with aromatic substitu-
ents (24-27, 29, and 30) were of comparable potency with
exceptions found in the plain phenylnitrovinyldiamine (24)
and 4-methoxynitrovinyldiamine (28), which were found to
be less active. The character of the para-substituent at the
phenyl ring was of low influence in this group of compounds,
as the 4-chloronitrovinyldiamine and the 4-trifluormethyl-
phenylnitrovinyldiamine had comparable inhibitory po-
tency. Surprisingly 23, containing cyclohexyl substituents,
was found to be as active as the phenyl derivatives 24-27, 29,
and 30.
Chemistry. The methylaminopropylimidazole precursors
were synthesized starting from 4-methylimidazole via a modi-
fied Gabriel synthesis,26 whereas a regioselective alkylation
route was applied in order to yield the respective methyl
substituted derivatives 6 and 8.
In the case of the 5-methyl derivative 6, 4-methyl-1-
H-imidazol was reacted with trityl chloride and triethyl-
amine according to Scheme 1, yielding 1-trityl-4-methylimi-
dazole (3) exclusively. After quaternation the resulting
1-trityl-3-[3-(1,3-dioxo-1,3-dihydro-2H-isoindol-2-yl)pro-
pyl]-4-methyl-1H-imidazol-3-ium chloride 4 was deprotected
yielding 5. The subsequent hydrazinolysis led to the 3-(5-
methyl-1H-imidazol-1-yl)propyl-1-amine 6.
The synthesis of 4-methyl derivative 8 (Scheme 2) started with
the alkylation of sodium 4-methyl-1H-imidazolate with 2-(3-
bromopropyl)isoindolin-1,3-dione. The resulting mixture of 2-
(3-(5-methyl-1H-imidazol-1-yl)propyl)isoindoline-1,3-dione
and 2-(3-(4-methyl-1H-imidazol-1-yl)propyl)isoindoline-1,3-
dione was reacted with trityl chloride, leading to a quaterna-
tion product only in the case of the 2-(3-(5-methyl-1H-imida-
zol-1-yl)propyl)isoindoline-1,3-dione (4). The quaternized
species could be separated by means of flash chromatography,
yielding 2-(3-(4-methyl-1H-imidazol-1-yl)propyl)isoindoline-
1,3-dione (7), which was then converted into 3-(4-methyl-
1H-imidazol-1-yl)propyl-1-amine (8) by hydrazinolysis.
The preparation of the thiourea (9 and 10) and the
thioamide (11 and 12) derivatives was accomplished accord-
ing to Scheme 3.26 The cyanoguanidine derivatives (13-21,
36-40, and 51) were prepared according to Scheme 4.32
Sodium cyanamide and the respective isothiocyanate were
reacted leading to a thiourea intermediate. The respective
1-alkylimidazol moiety was introduced by means of water-
soluble carbodiimide (WSCD). The nitrovinyldiamines
When the classes of the cyanoguanidines and nitrovinyl-
diamines are compared, major differences are found with the

Article
Journal of Medicinal Chemistry, 2009, Vol. 52, No. 22 7071
Scheme 3a
a Reagents and conditions: (a) 6 or 8, EtOH, 2 h, reflux; (b) 1-bromo-2-chloroethane, TEBA, KOH, 2 d, room temp; (c) KOH, ethylene glycol, 12 h,
reflux; (d) 6 or 8, CAIBE, NMM, THF, 5 min, -5 ꢀC, 10 h, room temp; (e) Laweson0s reagent, 1,4-dioxane, 8 h, reflux.
Scheme 4a
a Reagents and conditions: (a) potassium cyanamide, EtOH, 2 h, reflux; (b) 3-(1H-imidazol-1-yl)propan-1-amine, 6 or 8, WSCD, DMF-EtOH, 2 h,
room temp.
Scheme 5a
a Reagents and conditions: (a) 1,1-bis(methylthio)-2-nitroethene, EtOH, 24 h, reflux; (b) 3-(1H-imidazol-1-yl)propan-1-amine, 6 or 8, EtOH, 24 h,
reflux.
Scheme 6a
The incorporation of the known thiourea motif26 into a
condensed ring system as thioxopyrimidine led to potent
compounds. Only little differences in the potency between
the derivatives were seen, depending on the ring size and the
character of the condensed ring (Table 2). The phenylthio-
phene (31) and the 3-methylthiophene derivative (32) as well
as the 2,3-dihydrobenzo[b][1,4]dioxin containing compound
(33) were of comparable potency. Only a little decrease of the
potency was observed in the case of an extension or reduction
of the attached saturated ring (34, 35).
The introduction of a methyl group in position 4 of the
imidazole did not affect the inhibitory potency compared to
the nonsubstituted equivalents. This finding holds true for
all examples of the molecular scaffolds investigated (Tables 2
and 3). Only a slight improvement by 2-fold was visible in
the case of the thioamide 12 in comparison to the nonmethy-
lated analogue 2. In the case of the cyanoguanidine example
51 and the nitrovinyldiamine 52 the activity dropped by
a factor of 2.5 compared to the unsubstituted derivatives 21
and 30.
a Reagents and conditions: (a) 3-(1H-imidazol-1-yl)propan-1-amine,
6 or 8, EtOH, 2 h, reflux.
alkyl-type derivatives 13, 14, 22, and 23 with a preference for
the methyl substitution in the case of the cyanoguanidines
(13) and the cyclohexyl substitution in the case of the
nitrovinyldiamines (23). The trifluoromethylphenyl substi-
tuent, as in 17 and 26, having a lower electron density in the
aromatic ring, led to more active inhibitors when incorpo-
rated in a nitrovinyldiamine such as in 26. The 4-methoxy-
phenyl substitution was better accepted in cyanoguanidines
(19), whereas the 3,4-dimethoxyphenyl nitrovinyldiamine
inhibitor 29 was found to be more potent as the correspond-
ing cyanoguanidine equivalent 21.

7072 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 22
Buchholz et al.
Table 1
A general improvement of the inhibitory potency was
observed when the methyl substitution was introduced at
position 5 of the imidazole ring (Table 4). An elevation of
10-fold was shown for the thiourea 9 in comparison to 1 and
10, up to 15-fold in the case of the thioamides 11 and 12, or
even 35-fold as for 11 and 2. The cyanoguanidine 39 was
found to be 5-fold more active as the unsubstituted 21 and
13-fold more active as the 4-methylimidazole 51. A com-
parable result was found for the nitrovinyldiamines with a
8.1-fold improvement in the case of 30 compared to 45 and
a 19-fold improvement for 52 compared to 45.
3,4-dimethylphenyl in 36 was exchanged by a dioxine as in 39
or a 3,4-dichlorophenyl as in 37. For the nitrovinyldiamines
the activity range was found to be even smaller with a 2-fold
difference between 45 as the most inactive compound and 43
as the most active inhibitor of this class. Therefore, the
character of the substituent had no major influence on the
inhibitory potency. For example, a compound containing
a cyclohexyl moiety (41) was found to be as potent as a
p-chlorophenyl- (42), a p-trifluoromethylphenyl- (43), an
R-naphtyl- (44), or even a 2,3-dihydrobenzo[b][1,4]dioxine
bearing substituents (45).
A similar result was found for the of 2-thioxopyrimidin-
4(1H)one class (Table 2). The 5-methylation as in 46 led to an
improvement of the activity by a factor of 17 in comparison
to 31 and a factor of 13 in comparison to 53, respectively.
The SAR investigation of 5-methyl-1H-imidazol-1-yl)pro-
pylamino inhibitors (Table 4) of the cyanoguanidine and
nitrovinyldiamine type features most of the compounds with
a QC inhibitory activity in the low nanomolar range, where
the differences in the activity of the compounds are found to
be low. In the case of the cyanoguanidines, the activity
difference between the most active compound 40 and
the most inactive compound 37 is only 4-fold. Therefore,
a 3,4-dimethylphenyl moiety (36) is as well tolerated as
a biphenyl (38) or a 3,4,5-trimethoxyphenyl substituent
(40). A slightly decreased activity was observed when the
The class of the 5-methylimidazolpropyl-2-thioxopyrimi-
din-4(1H)-ones (Table 2) showed an activity range that is
comparable to that of the corresponding cyanoguanidines
(Table 4) with a factor of 5 between 50 and 47. For this class a
2,3-dihydro-2-thioxoquinazolin-4(1H)-one as in 46 was spa-
tially extendable by a 2-thioxo-2,3,5,6,7,8-hexahydro[1]-
benzothieno[2,3-d]pyrimidin-4(1H)-one as in 47, retaining
the inhibitory potency. The extension of the cyclohexyl
moiety of 47 into a cyclooctyl group as in 48 caused a slight
loss of activity. When the thieno part was substituted with a
benzyl moiety at position 6 as in 49, no influence on the
activity is detectable in comparison to the unsubstituted
analogues 46 and 47. However, the substitution at the 5-
position with a phenyl ring in 50 led to a slight loss of activity
by a factor of around 5 compared to 47 and 49.

Article
Journal of Medicinal Chemistry, 2009, Vol. 52, No. 22 7073
Table 2
Biology. The assessment for the efficacy of the different
Aβ3(pE)-40,42 in amounts of 5-20% of the total Aβx-40,42 as
quantified by an ELISA assay applying N-terminal specific
antibodies.35 In this setting, the effective hQC inhibitor 1
(PBD150)26,35 led to a 70% reduction of Aβ3(pE)-40,42 in
relation to the untreated controls, and therefore, 1 was used
as a benchmark for the assessment of the current inhibitors.
A total of 23 compounds was assessed regarding their
hQC inhibitors was performed in a cell culture assay using
transfected HEK293 cells expressing human APP and hQC.
The APP expression construct featured a deletion of the
first two amino acids of the Aβ sequence leading to the
direct liberation of Aβ3-40,42 by β- and γ-secretase activity.34
The cotransfection with hQC caused the production of

7074 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 22
Buchholz et al.
Table 3
the whole loop. The hQC model (Figure 3) reveals the typical
“open sandwich” structure adopted from the aminopeptidase
template. The structural topology with a hydrophobic core
consisting of four parallel and two antiparallel β-sheets and
the surrounding eight R-helices represents the typical R/β-hy-
drolase fold. When the homology model of hQC and the
template aminopeptidase are compared, a high similarity was
found not only for the overall structure but also for zinc binding
site and for most of the active site amino acid residues as
represented by the rmsd values (Figure 3). Structural differ-
ences, resulting in higher rmsd values, were found in the loop
region (S288-D305), which was modified by MD simulation.
Further differences were visible at the position of W329 in the
hQC model that lacks an equivalent in the apAP structure.
Other differences between the template and the model were
found in areas away from the active site at the loops K176-S187
and F204-D211.
The active site is characterized by a deep, tight, and angled
cleft, which is defined by the hydrophobic amino acids
W329, V302, D305 (not shown). This cleft opens in a bigger
˚
˚
˚
cavity of about 8 A ꢀ 8 A ꢀ 10 A defined by the distances
between the zinc ion and the atom OD1 of D248, the atom
CD1 of L249 and atom OD2 of D305, and between the atom
NE2 of H140 and atom OH of Y299, respectively. The zinc
ion, located at the bottom of this cavity, is coordinated by
H330, E202, D159, and the imidazole moiety of the inhibitor
(Figures 3 and 4). A stepwise strategy was applied for the
shaping of the active site of the homology model using the
different known inhibitor structures N-methyl imidazole,9
N-benzylimidazole,9 and compound 1. After each step a MD
simulation was performed, keeping only the amino acids in a
efficacy in the inhibition of Aβ3(pE)-40,42 formation on trans-
fected HEK cells. Taking 1 as the benchmark for the potency
in hQC inhibition, the selection included hQC inhibitors
regarded as potent, such as 2, 9, 11, 46, 47, and 49, medium
potent such as 10, 33, 36, 38, 39, 40, 41, 42, 45, and 53, and
low potent inhibitors such as 51, 52, and 53. As a result, the
potency of the compounds at the isolated target was reflected
by the efficacy inhibiting the formation of pGlu containing
Aβ-peptides in general (Figure 2). The most potent hQC
inhibitors 2, 9, 11, 46, 47, and 49 inhibited the formation of
Aβ3(pE)-40,42 with an efficacy above 90% using the concen-
trations indicated. Compounds with a medium inhibitory
potency such as 10, 33, 36, 38, 39, 40, 41, 42, 45, and 53 were
found to be able to prevent the formation of Aβ3(pE)-40,42 by
30-70%, not generally paralleling their potency found dur-
ing hQC inhibition in vitro. For less potent compounds such
as 51, 52, and 53, a Ki value on hQC above 1 μM was revealed
to be not effective in reducing the Aβ3(pE)-40,42 formation.
The most effective compound 47 led to a total blockage of
the Aβ3(pE)-40,42 formation at 1 μM.
˚
range of 4.5 A around the respective inhibitor flexible. In the
first step the N3 of 1-methylimidazole was placed on the free
coordination site adopted by the coordinating water in the
case of the free enzyme. After the MD simulation the same
procedure was repeated with N-benzyl-imidazole and com-
pound 1, respectively. The resulting model was then visually
inspected, and the amino acids that were identified to have
geometry failures according to the Ramachandran plot were
adjusted stepwise followed by energy minimization steps.
The 3,4-dimethoxyphenyl moiety of compound 1 was found
to be localized at a region, which is formed by the hydro-
phobic side chains of the residues Y299, F325, P326 and the
acidic side chain of E327. This area corresponds to the
positions of F244, Y248, and Y251 in apAP, which are
involved in forming a hydrophobic pocket discussed to be
the S1-recognition site of aminopeptidases (Figure 3).40
Docking. The binding pose of the imidazole part of 1 resulting
fromthe MDsimulationwas usedas a substructureconstraint for
the docking41 of the inhibitory molecules 1, 2, and 9-12. The
resulting poses were scored applying the ChemScore function
containing an explicit metal binding term. The visual inspection
of the docking results of 1 and 2 suggested an open space between
the imidazole and the active site in the near of the carbon 5, which
could be filled by additional substituents. In contrast to that, the
spacesurrounding the carbon 4 appeared to be smaller (Figure 4).
Supporting this finding, the solutions of the scoring function for
the thiourea derivatives 1, 9, and 10 showed a preference of the
5-methyl compound 9 over the nonsubstituted derivative 1 and
the 4-methylated 10. The 5-methyl derivative 11 again was found
to have the highest score, followed by the unsubstituted imidazole
derivative 2 and the corresponding 4-methyl compound 12. The
low score for the 4-methyl derivatives was mainly caused by a
high contribution of the clash-term to the lower scoring value
(Supporting Information).41
Computational Chemistry
hQC Homology Model. The homology model of hQC was
developed using the X-ray structures of the aminopeptidase
from Aeromonas proteolytica (apAP, PDB codes 1amp and 1rtq)
and a hQC model (PDB code 1moi)12 lacking the information
regarding the catalytic zinc ion. After the superimposition36 of
the active sites of 1rtq and 1moi the catalytic zinc ion (Zinc 1)
with the attached water molecule was transferred to the homol-
ogy model 1moi. The template aminopeptidase contains two
catalytic zinc ions (Zinc 1 and Zinc 2). It was previously shown
that the hQC protein contains only one active zinc ion,9,37 a fact
that is meanwhile proven by the solution of the hQC protein
structure.13 In the case of the hQC homology model it was
postulated that the Zinc 1 of the aminopeptidase corresponds to
the zinc ion found in hQC, as for this position a greater zinc
affinity over position 2 is described.38,39 The template topology
of 1rtq on the loop region between T222 and S228 led to an
unfavorable superimposition of the side chain of the residue
Q304 of hQC (1moi) to the side chain from C237 of apAP (1rtq).
In this conformation the access to the active site was blocked by
the side chain of Q304, making the binding of the known
inhibitor structures unlikely. An explanation was found in the
presence of a disulfide bridge between the C223 and C227 in the
structure of the aminopeptidase template, which is lacking the
hQC protein. For that reason the side chain of Q304 was turned
manually followed by a molecular dynamic simulation of
Discussion
The bioisosteric replacement of the thiourea moiety by
cyanoguanidines and nitrovinyldiamines starting from 1 led

Article
Journal of Medicinal Chemistry, 2009, Vol. 52, No. 22 7075
Table 4
to compounds with a QC inhibiting potency in the submicro-
molar range. The most potent compounds, 21 for the cyano-
guanidines and 23 for the nitrovinyldiamines, were found to
be 10-fold less potent than the active compound 1, belonging
to the thiourea class. In general, the SAR within the classes of
investigated compounds lacks deep activity clefts. An unex-
pected observation was made within the class of the nitrovi-
nyldiamines, wherein saturated carbocyclic substituents as in
23 were found to be more potent as the corresponding phenyl
derivative 24. This finding stands in contrast to the SAR of the
thioureas, where the cyclohexyl substituted thiourea deriva-
tive was found to be of equal potency as the respective
phenyl derivative.26 Differences in the SAR are visible when
comparing similar molecules belonging to different classes.
The methylcyanoguanidine 13 is 13-fold more active than the
corresponding nitrovinyldiamine 22; in contrast to this the

7076 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 22
Buchholz et al.
Figure 2. Efficacy (% of inhibition of pGlu-Aβ3(pE)-40/43 formation, blue bars) of selected hQC inhibitors (1 μM) in hQC and NLE
overexpressing HEK cells compared with their hQC inhibitory potency (pKi, red dots).
Figure 3. Left: Homology model of hQC. The colors correspond to rmsd values between the mass centers of the amino acids after structural
alignment with the aminopeptidase template 1amp. The binding pose of 1 after MD simulation is shown. Right: Possible binding mode of 1 at a
hydrophobic area formed by the amino acid residues: (1) F325, (2) P326, (3) E327, and (4) Y299 (orange) in the homology model. For
comparison prominent amino acids of the hydrophobic S1 pocket of apAP (1amp) are shown in white: (5) F248, (6) F244, and (7) Y251.
Figure 4. Left: Position of 1 after MD simulation. Shown are the residues of the metal-binding site and the catalytic zinc ion of the hQC
homology model (orange). The metal-binding site of apAP (white) bearing two zinc ions is shown for comparison. Right: Docking of 9 (pink)
and 10 (blue) in the active site. The van der Waals surfaces of the imidazoles are visualized. Amino acids (hQC/apAP) are as follows: (1) E201/
151, (2) E202/152, (3) H140/97, (4) D248/179, (5) H330/266, and (6) D159/117.
corresponding phenyl derivatives 15 and 24 display the same
potency. A possible explanation for this could be found in the
different binding mode of the respective methyl derivatives 13
and 22. In this case, the cyano part of 13 could adopt the same
hydrophobic interaction with the protein as the phenyl part of
15. This assumption is supported by the results of docking

Article
Journal of Medicinal Chemistry, 2009, Vol. 52, No. 22 7077
investigations (not shown) and the bioisosteric replacement of
phenyl groups by cyano moieties, which is established to be
possible.28,29
2-Thioxopyrimidin-4(1H)ones were identified as potent
compounds in general, exhibiting inhibitory activities in the
lower micromolar range. For the investigated compounds
only little differences in the inhibitory potency were recogniz-
able, depending onthe size of the annulated saturated ring(see
compounds 33, 34, and 35).
The presented study led to novel inhibitors of hQC with
improved potency and activity in an in vitro cell system.
For the first time a cell based system assessing the hQC
inhibition potency in connection with the generation of
Aβ3(pE)-40,42 was used for the efficacy profiling. The results
again emphasize the high probability of hQC to be the
responsible enzyme for the cyclization of the N-terminal
glutamate of Aβ3,11(pE)-40,42 yielding Aβ3(pE)-40,42 because the
inhibitory potency and the efficacy in the cellular model were
found to be in good correlation. The characterization of the
most effective hQC inhibitors in transgenic animal models is
currently underway.
The optimized potency of the 5-methylimidazole thiourea
was found to be consistent within the bioisosteric cyanogua-
nidines, nitrovinyldiamines, and the 2-thioxopyrimidin-4-
(1H)ones. The gain of potency ranges between a factor of 13
and 20, by comparing the 4-methyl derivatives with the
respective 5-methylimidazoles. This corresponds to an aver-
age gain of free binding energy of 1.57 ( 0.15 kcal/mol
comparing all 5-methyl derivatives with their corresponding
4-methyl analogues.42 The differences of the inhibitory po-
tency among the different 5-methylimidazoles were low. This
leads to the assumption that the effective binding of the
5-methylimidazoles as metal binding group overrules the
contribution of the different remaining parts of the inhibitor
molecules. 11 with a Ki value of 2.6 nM is the most potent
inhibitor of hQC described so far. A reason for the generally
elevated efficacy of the 5-methylimidazoles could be found in
an optimized binding due to sterical conditions as well as the
improvement of the metal binding properties of the imidazole
N3-nitrogen. The first hint supporting this hypothesis could
Experimental Section
Chemistry. Starting materials and solvents were purchased
from Aldrich and Maybridge Co. All chemical structures were
confirmed by 1H NMR and high-resolution mass spectrometry.
The purity of the compounds was assessed by HPLC and
confirmed to be g95%. The analytical HPLC system consisted
of a Merck-Hitachi device (model LaChrom) utilizing a Li-
Chrospher 100 RP 18 (5 μm), analytical column (length 125 mm,
diameter 4 mm), and a diode array detector (DAD) with λ = 214
nm as the reporting wavelength. The compounds were analyzed
using a gradient at a flow rate of 1 mL/min, where eluent A was
acetonitrile and eluent B was water, both containing 0.1% (v/v)
trifluoroacetic acid and applying the following gradient: method
A, 0-min f 5% A, 5-17 min f 5-15% A, 15-27 min f
15-95% A, 27-30 min f 95% A; method B, 0-15 min f
5-50% A, 15-20 min f 50-95% A, 20-23 min f 95% A;
method C, 0-20 min f 5-60% A, 20-25 min f 60-95% A,
25-30 min f 95% A. The purities of all reported compounds
were determined by the percentage of the peak area at 214 nm.
Semipreparative HPLC was performed on a Merck-Hitachi
device (model LaChrom) equipped with a SP250/21 Nucleosil
100-7 C18 semipreparative column (Machery-Nagel). The com-
pounds were eluted using the same solvent system as described
above, applying a flow rate of 8 mL/min.
ESI mass spectra were obtained with a SCIEX API 365
spectrometer (Perkin-Elmer) utilizing the positive ioniza-
tion mode. The high resolution positive ion ESI mass spectra
were obtained from a Bruker Apex III 70e Fourier transform
ion cyclotron resonance mass spectrometer (Bruker Daltonics,
Billerica, MA) equipped with an Infinity cell, a 7.0 T super-
conducting magnet (Bruker, Karlsruhe, Germany), an rf-only
hexapole ion guide, and an external electrospray ion source (API
Apollo, voltages: end plate, -3.700 V; capillary, -4.400 V;
capillary exit, 100 V; skimmer, 1.15 V; skimmer, 2.6 V). Nitro-
gen was used as drying gas at 150 ꢀC. The sample solutions were
introduced continuously via a syringe pump with a flow rate of
120 μL h-1. All data were acquired with 256K data points and
zero-filled to 1024K by averaging 32 scans.
43
be the shift of the calculated pKa of the imidazole N3-
nitrogen from 6.79 for 1 and 6.65 for 10 toward 7.33 for 9
pointing toward a better metal binding with a more basic,
electron rich coordination site.
The first insight into the possible binding mode of hQC
inhibitors gained by the development of a homology model
partly reflected the findings from the structure-activity re-
lationship studies of hQC inhibitors. Provided that the imi-
dazole acts as the metal-binding group via its N3-nitrogen, the
length of the spacer between the imidazole and the molecular
scaffoldwas optimalwhen consistingof three methylene units.
A shorter linker, like ethyl, would cause a steric clash with the
tight cleft formed by the amino acids W329, V302, and Y299
(not shown). This result is in accordance with previous
findings, wherein the ethyl derivative of 1 is 300-fold less
active.26 The thiourea motif with its ability to flip the thio
amide bond enables a possible interaction of the 3,4-dimethoxy-
phenyl moiety of the inhibitor with a pocket of the enzyme
formed by the side chains of the residues Y299, F325, P326,
and E327. The second zinc ion lacking in the hQC protein, but
present in the aminopeptidase template, leaves a space that
can be filled by substituents located at position 5 of the metal-
interacting imidazole ring of the inhibitor. As this area of the
active site is highly conserved in the hQC sequence, the
resulting homology model has a high probability to reflect
the situation of the real hQC protein structure. A first proof
confirming this hypothesis can be found in the 10-fold higher
measured potency of the corresponding 5-methylimidazoles
as seen for the inhibitors 9 and 1.
The melting points were detected utilizing a Kofler melting
point device and are not corrected. The 1H NMR spectra were
recorded at a Varian Gemini 2000 (400 MHz) or at a Bruker AC
500 (500 MHz) when 500 MHz is indicated. DMSO-d6 was used
unless otherwise specified. Chemical shifts are expressed as parts
per million (ppm) downfield from tetramethylsilane. Splitting
patterns have been designated as follows: s (singulet), d (doub-
let), dd (doublet of doublet), t (triplet), m (multiplet), and br
(broad signal).
3-(5-Methyl-1H-imidazol-1-yl)propan-1-amine 6. Step A. 4-
Methyl-1-trityl-1H-imidazole 3. 4-Methyl-1H-imidazole (36.53
mmol, 1 equiv) was dissolved in 120 mL of dimethylformamide,
and triethylamine (73.06 mmol, 2 equiv) and chlorotriphenyl
methane (40.1 mmol, 1.1 equiv) were added. The mixture was
stirred for 3.5 h. The precipitate was filtered off and then washed
by means of ice-cooled dimethylformamide (2 ꢀ 50 mL) and
water (2 ꢀ 50 mL). After the solvent was removed, the remaining
The evaluation of the biological efficacy of the presented
hQC inhibitors was directed toward the ability of the suppres-
sion of the formation of Aβ3(pE)-40,42 generated by the coex-
pressed hQC. The inhibitory potency of the compounds at the
isolated target was found to be in good accordance with the
activity of the compounds on the cellular level.

7078 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 22
Buchholz et al.
product was dried over P4O10. Yield: 10.65 g (98.2%). The
product was used without further purification.
at this temperature for 30 min. The formed precipitate was
filtered off. The filtrate was cooled down to 0 ꢀC, and solid
NaOH was added until a final pH of 10-12 was reached. The
aqueous solution was extracted by means of CHCl3 (3 ꢀ 50 mL).
The combined organic layers were dried over Na2SO4 and
filtered, and the solvent was removed. The product was purified
by means of flash chromatography using silica gel and a CHCl3/
MeOH gradient containing aqueous ammonia (2% v/v). Yield:
0.31 g (65.1%). 1H NMR (500 MHz, CDCl3): δ 1.82-1.87 (m,
2H), 2.19 (s, 3H), 2.67-2.71 (m, 2H), 3.91-3.95 (m, 2H), 6.60 (s,
H), 7.24 (s, solv.), 7.33 (s, H). ESI-MS m/z: 140.3 (M þ H)þ,
279.4 (2M þ H)þ. HPLC (214 nm, method A): tR = 0.2 min
(100%).
Step B. 1-Trityl-3-[3-(1,3-dioxo-1,3-dihydro-2H-isoindol-2-
yl)propyl]-1,4-dimethyl-1H-imidazol-3-ium Bromide 4. 4-Meth-
yl-1-trityl-1H-imidazole (3) (32.85 mmol, 1 equiv) was sus-
pended in acetonitrile (10 mL), and 2-(3-bromopropyl)iso-
indoline-1,3-dione (32.85 mmol, 1 equiv) was added. The mix-
ture was kept under reflux overnight. The organic solvent was
removed. Purification was achieved by flash chromatography
using silica gel and a CHCl3/MeOH gradient. Yield: 10.65 g
(63.44%).
Step C. 2-(3-(5-Methyl-1H-imidazol-1-yl)propyl)isoindoline-
1,3-dione 5. 1-Trityl-3-[3-(1,3-dioxo-1,3-dihydro-2H-isoindol-2-yl)-
propyl]-1,4-dimethyl-1H-imidazol-3-ium bromide (4) (7.86 mmol)
was dissolved in a stirred solution containing methanol (20 mL)
and trifluoroacetic acid (4 mL). The mixture was kept under
reflux overnight. The solvent was removed, and the remaining
oil was purified by flash chromatography using silica gel and a
CHCl3/MeOH gradient. Yield: 2.05 g (97.0%).
Synthesis of the Compounds 9-12. The synthesis was accom-
plished as previously described.26
General Procedure for the Synthesis of the Cyanoguanidines
13-21, 36-40, 51. All reactions were performed at a Buchi
Synchor parallel synthesizer. A typical reaction mixture con-
sisted of sodium cyanamide (0.26 g, 4.0 mmol, 1 equiv), and the
corresponding isothiocyanate (4.0 mmol, 1.0 equiv) was dis-
solved in 10 mL of ethanol. The mixture was shaken for 3 h
under reflux and then cooled to room temperature. Either
3-(1H-imidazol-1-yl)propyl-1-amine, 3-(5-methyl-1H-imidazol-
1-yl)propyl-1-amine (6), or 3-(4-methyl-1H-imidazol-1-yl)pro-
pyl-1-amine (8) (each, 4.8 mmol, 1.2 equiv) and N1-
((ethylimino)methylen)-N3,N3-dimethylpropan-1,3-diamin-
hydrochloride (WSCD, 0.92 g, 4.8 mmol, 1 equiv) and di-
methylformamide (5 mL) were added. The mixture was shaken
for 3 h at room temperature, and after that the organic solvent
was removed. The remaining oil was dissolved in trichloro-
methane (30 mL), and the organic layer was washed once with
water (20 mL), dried over Na2SO4, and filtered off, and the
solvent was removed. Purification was performed by applying
flash chromatography on basic aluminum oxide using a
CHCl3/MeOH gradient.
Step D. 3-(5-Methyl-1H-imidazol-1-yl)propan-1-amine 6.
2-(3-(5-Methyl-1H-imidazol-1-yl)propyl)isoindoline-1,3-dione
(5) (8.92 mmol, 1 equiv) and hydrazine monohydrate (17.84
mmol, 2 equiv) were dissolved in dry ethanol (50 mL). The
mixture was kept under reflux overnight. Then this mixture was
concentrated down to a volume of 25 mL. After that hydro-
chloric acid (concentrated, 55 mL) was added and the mixture
was heated to 50 ꢀC and kept at this temperature for 30 min. The
formed precipitate was filtered off. The filtrate was cooled down
to 0 ꢀC, and solid NaOH was added until a final pH of 10-12
was reached. The aqueous solution was extracted by means of
trichloromethane (3 ꢀ 50 mL). The combined organic layers
were dried over Na2SO4 and filtered, and the solvent was
removed. The product was purified by means of flash chromato-
graphy using silica gel and a CHCl3/MeOH gradient. Yield:
1
0.74 g (60%). H NMR (500 MHz, CDCl3): δ 1.79-1.85 (m,
2H), 2.18 (s, 3H), 2.69-2.72 (m, 2H), 3.89-3.92 (m, 2H), 6.73 (s,
H), 7.24 (s, solv), 7.38 (s, H). ESI-MS m/z: 140.3 (M þ H)þ,
279.4 (2M þ H)þ. HPLC (214 nm, method A): tR dead time
(100%).
General Procedure for the Synthesis of the Nitrovinyldiamines
22-30, 41-45, 52. A typical reaction batch consisted of 1,1-
bis(methylthio)-2-nitroethene (0.70 g, 4.2 mmol, 1 equiv) and
the corresponding primary amine (4.0 mmol, 1 equiv) dissolved
in ethanol (20 mL). The mixture was stirred for 24 h under
reflux. 3-(1H-Imidazol-1-yl)propyl-1-amine or 3-(5-methyl-1H-
imidazol-1-yl)propyl-1-amine (6) or 3-(4-methyl-1H-imidazol-
1-yl)propyl-1-amine (8) (each, 4.2 mmol, 1 equiv) was added,
and the mixture was stirred for 24 h under reflux. The solvent
was removed, and the resulting oils were purified by means of
flash chromatography using silica gel as the solid phase and a
CHCl3/MeOH gradient.
General Procedure for the Synthesis of 2-Thioxopyrimidine-
4(1H)-ones 31-35, 46-50, 53. A typical reaction batch consisted
of 3-(1H-imidazol-1-yl)propan-1-amine (4.0 mmol, 1 equiv) or 6 or
8 (each, 1.43 mmol, 1 equiv), and the corresponding isothiocyanate
(1 equiv) was dissolved in dry ethanol (20 mL) and heated under
reflux for 3 h. After the solvent was removed the products were
purified by means of flash chromatography using silica gel and a
CHCl3/MeOH gradient.
Inhibitor Testing. QC activity was evaluated fluorometrically
in a coupled assay according to the method previously de-
scribed.44 Gln-AMC served as substrate and pyroglutamyl
peptidase as auxiliary enzyme. After conversion of Gln-AMC
into pGlu-AMC by QC the pGlu-AMC was hydrolyzed by
pyroglutamyl peptidase (pGAP). The generated AMC was
detected with excitation/emission wavelengths of 380/460 nm.
The described assay was modified as follows for the use in a
higher throughput:
3-(4-Methyl-1H-imidazol-1-yl)propan-1-amine 8. Step A.
2-(3-(4/5-Methyl-1H-imidazol-1-yl)propyl)isoindoline-1,3-dione.
4-Methyl-1H-imidazole (36.53 mmol, 1 equiv) and sodium
hydride (60% in mineral oil, 36.53 mmol, 1.0 equiv) were
dissolved in 80 mL of dimethylformamide. The mixture was
stirred at room temperature for 2 h until the formation of
hydrogen gas deceased. 2-(3-Bromopropyl)isoindoline-1,3-
dione (34.70 mmol, 0.95 equiv) was added, and the mixture
was stirred at 90 ꢀC overnight. The solvent was removed, and the
remaining residue was purified by means of flash chromatogra-
phy using silica gel and a CHCl3/MeOH gradient. Yield: 6.1 g
(62.0%) of a mixture of 2-(3-(4-methyl-1H-imidazol-1-yl)pro-
pyl)isoindoline-1,3-dione and 2-(3-(5-methyl-1H-imidazol-
1-yl)propyl)isoindoline-1,3-dione.
Step B. 2-(3-(4-Methyl-1H-imidazol-1-yl)propyl)isoindoline-
1,3-dione 7. A mixture consisting of 2-(3-(4-methyl-1H-imidazol-
1-yl)propyl)isoindoline-1,3-dione and 2-(3-(5-methyl-1H-imidazol-
1-yl)propyl)isoindoline-1,3-dione (22.65 mmol, 1 equiv) and trityl
chloride (13.6 mmol, 0.6 equiv) was dissolved in 40 mL of dichloro-
methane and kept at a temperature of 0 ꢀC for 10 min and 1.5 h at
room temperature. The solvent was removed, and the remaining
solid was purified by means of flash chromatography using silica gel
and a CHCl3/MeOH gradient. Yield: 0.92 g (15.1%).
Step C. 3-(4-Methyl-1H-imidazol-1-yl)propan-1-amine 8.
2-(3-(4-Methyl-1H-imidazol-1-yl)propyl)isoindoline-1,3-dione
(7) (3.42 mmol,
1
equiv) and hydrazine monohydrate
The used buffer consisted of 50 mM Tris-HCl (ROTH,
Karlsruhe, Germany), with pH 8.0, adjusted with HCl. The
substrate H-Gln-AMC hydrobromide (AMC: 7-amido-4-
methylcoumarin; BACHEM, Bubendorf, Switzerland) was
used in a concentration of 0.625 mM. The auxiliary enzyme
pGAP from Bacillus amyloliquefaciens (10 units, 25 units/mL;
(6.84 mmol, 2 equiv) were dissolved in 20 mL of ethanol, and
the mixture was stirred for 12 h under reflux. The mixture was
kept under reflux overnight and then concentrated to reach
a volume of 25 mL. Hydrochloric acid (concentrated, 55 mL)
was added, and the mixture was heated up to 50 ꢀC and kept

Article
Journal of Medicinal Chemistry, 2009, Vol. 52, No. 22 7079
Qiagen, Hilden, Germany) was prediluted 1:10 in Tris-HCl.
Human glutaminyl cyclase (QC, EC 2.3.2.5) was recombinantly
expressed in Pichia pastoris and purified as previously de-
scribed.45 The protein concentration was adjusted to be 0.6
mg/mL, and the hQC solution was prediluted to 1:1250 in Tris-
HCl. A typical reaction mixture consisted of 100 μL of substrate,
100 μL of the inhibitor (stock solution in DMSO), resulting in a
final DMSO concentration of 5%, and 25 μL of pGAP. After
incubation in NUNC 96 flat bottom transparent microwell
plates (Nunc, Roskilde, Denmark) for 10 min at 30 ꢀC the
reaction was started by adding 25 μL of the hQC solution.
All enzyme activity measurements were performed in triplicate.
A GENios Pro multifunctional microwell plate photometer
working on Magellan software, version 4.0 (top and bottom
reading, TECAN, Switzerland), was used to monitor the pro-
gress curves for 20 min at 30 ꢀC. Pipetting steps were accom-
plished using a Tecan Genesis Freedom 200 workstation
(Gemini software, version 4.0; TECAN, Switzerland). The
kinetic data were evaluated using GraFit (version 5.0.4, Eritha-
cus, Horley, U.K.).
Cell Based Assay. The inhibition of Aβ3(pE)-40,42 formation
was evaluated in HEK293 cells overexpressing mutated human
amyloid precursor protein APP(NLE) and human glutaminyl
cyclase (hQC)35 according to the following protocol: Human
embryonic kidney cells HEK293 were cultured in DMEM (10%
FBS) (Invitrogen) in a humidified atmosphere of 5% CO2 at
37 ꢀC. Then 250 000 cells/500 μL were seeded per well in a
24-well plate. After 24 h the cells were transfected with 0.8 μg of
APP-NLE vector, 0.8 μg of hQC vector, and 4 μL of Lipofecta-
mine 2000 (Invitrogen) in 100 μL of OPTIMEM for 5 h. The
cells were further cultivated in the DMEM medium. The next
day, cells were incubated in the assay medium (DMEM, without
phenol red, without FBS) containing an appropriate amount of
QC inhibitor or control. The inhibitor stock solution (10 mM)
was therefore diluted stepwise with DMSO (Sigma-Aldrich,
Taufkirchen, Germany) and medium. The final DMSO concen-
tration was 0.01%. The inhibitor concentration in the medium
was 1 μM using five replicates. After 24 h of the inhibitor appli-
cation the supernatant (250 μL) was collected and readily mixed
with 25 μL of complete mini protease inhibitor cocktail (Roche,
Basel, Switzerland) supplemented in addition with 10 mM
AEBSF (Roth, Karlsruhe, Germany). After centrifugation
(2000g for 5 min) 2 aliquots (120 μL) of supernatant were
collected and the total AβX-40,42 and Aβ3(pE)-40,42 concentrations
were determined using specific sandwich ELISAs (IBL,
Hamburg, Germany) according to the manufacturers advices.
Computational Chemistry. All calculations were performed
using the software packages MOE, version 2007.08 (CCG,
Montreal, Canada); GOLD, version 4.0 (CCDC Software
Ltd., Cambridge, U.K.); SYBYL, verssion 8.0 (Tripos CL, St.
Louis, MO); ProSa2003, version 4.0 (CAME, Salzburg,
Austria);46,47 PROCHECK, version 3.5.4 (European Bioinfor-
matics Institute, Cambridge, U.K.)48 on a 9 CPU computer
cluster with CentOS 4.6 operation system installed.
Halle, Germany, for recording the high resolution mass
spectra. The authors are also grateful to M. Scharfe and H.
Mosdzen for technical support. This work was supported by
The Federal Ministry of Education and Research, Germany,
Grant Nos. 0313185 and 0315089 (to H.-U.D.).
Supporting Information Available: Detailed analytical data of
the compounds (NMR, MS, high resolution MS, HPLC), yields,
description of the homology model development including the
model characteristic data, and fitness and scoring results of the
docking experiments. This material is available free of charge via
the Internet at http://pubs.acs.org.
References
(1) Awade, A. C.; Cleuziat, P.; Gonzales, T.; Robert-Baudouy, J.
Pyrrolidone carboxyl peptidase (Pcp): an enzyme that removes
pyroglutamic acid (pGlu) from pGlu-peptides and pGlu-proteins.
Proteins 1994, 20, 34–51.
(2) Abraham, G. N.; Podell, D. N. Pyroglutamic acid. Non-metabolic
formation, function in proteins and peptides, and characteristics of
the enzymes effecting its removal. Mol. Cell. Biochem. 1981, 38,
181–190.
(3) Van Coillie, E.; Proost, P.; Van, A., I; Struyf, S.; Polfliet, M.; De,
M., I; Harvey, D. J.; Van Damme, J.; Opdenakker, G. Functional
comparison of two human monocyte chemotactic protein-2 iso-
forms, role of the amino-terminal pyroglutamic acid and proces-
sing by CD26/dipeptidyl peptidase IV. Biochemistry 1998, 37,
12672–12680.
(4) Busby, W. H.; Humm, J.; Kizer, J. S.; Youngblood, W. W.;
Quackenbush, G. An enzyme(s) that converts glutaminyl-peptides
into pyroglutamyl-peptides. Presence in pituitary, brain, adrenal
medulla, and lymphocytes. J. Biol. Chem. 1987, 262, 8532–8536.
(5) Messer, M.; Ottesen, M. Isolation and properties of glutamine
cyclotransferase of dried papaya latex. Biochim. Biophys. Acta
1964, 92, 409–411.
(6) Fischer, W. H.; Spiess, J. Identification of a mammalian glutaminyl
cyclase converting glutaminyl into pyroglutamyl peptides. Proc.
Natl. Acad. Sci. U.S.A. 1987, 84, 3628–3632.
(7) Pohl, T.; Zimmer, M.; Mugele, K.; Spiess, J. Primary structure and
functional expression of a glutaminyl cyclase. Proc. Natl. Acad. Sci.
U.S.A. 1991, 88, 10059–10063.
(8) Fraser, L. R. Fertilization promoting peptide: an important regu-
lator of sperm function in vivo? Rev. Reprod. 1998, 3, 151–154.
(9) Schilling, S.; Niestroj, A. J.; Rahfeld, J. U.; Hoffmann, T.; Wermann,
M.; Zunkel, K.; Wasternack, C.; Demuth, H.-U. Identification
of human glutaminyl cyclase as a metalloenzyme. Potent inhibi-
tion by imidazole derivatives and heterocyclic chelators. J. Biol.
Chem. 2003, 278, 49773–49779.
(10) Bateman, R. C., Jr.; Temple, J. S.; Misquitta, S. A.; Booth, R. E.
Evidence for essential histidines in human pituitary glutaminyl
cyclase. Biochemistry 2001, 40, 11246–11250.
(11) Schilling, S.; Manhart, S.; Hoffmann, T.; Ludwig, H.-H.; Wasternack,
C.; Demuth, H.-U. Substrate specifity of glutaminyl cyclases from
plants and animals. Biol. Chem. 2003, 384, 1583–1592.
(12) Booth, R. E.; Lovell, S. C.; Misquitta, S. A.; Bateman, R. C., Jr.
Human glutaminyl cyclase and bacterial zinc aminopeptidase share
a common fold and active site. BMC Biol. 2004, 2, 2.
(13) Huang, K. F.; Liu, Y. L.; Cheng, W. J.; Ko, T. P.; Wang, A. H.
Crystal structures of human glutaminyl cyclase, an enzyme respon-
sible for protein N-terminal pyroglutamate formation. Proc. Natl.
Acad. Sci. U.S.A. 2005, 102, 13117–13122.
(14) Huang, K. F.; Liu, Y. L.; Wang, A. H. Cloning, expression,
characterization, and crystallization of a glutaminyl cyclase from
human bone marrow: a single zinc metalloenzyme. Protein Expres-
sion Purif. 2005, 43, 65–72.
(15) Song, I.; Temple, J. S.; Burns, K. H.; Bateman, R. C., Jr. Absence of
an essential thiol in human glutaminyl cyclase: implications for
mechanism. Korean J. Biol. Sci. 1998, 2, 243–248.
hQC Homology Model. The Charmm22 force field (MOE,
version 2007.08) was set as default for all modeling steps, if not
mentioned otherwise. The implicit general Born solvation49 was
turned on in all MD simulations (Supporting Information).
Docking. The imidazole residue of the inhibitor 1 obtained
from the homology model was used as a scaffold constraint. All
docking runs were performed 20 times for each molecule not
allowing early termination using the ChemScore as the scoring
function implemented in the software GOLD. The best-ranking
poses were used for further analysis of the docking solutions
(Supporting Information).
(16) Cynis, H.; Rahfeld, J. U.; Stephan, A.; Kehlen, A.; Koch, B.;
Wermann, M.; Demuth, H.-U.; Schilling, S. Isolation of an iso-
enzyme of human glutaminyl cyclase: retention in the Golgi com-
plex suggests involvement in the protein maturation machinery.
J. Mol. Biol. 2008, 379, 966–980.
(17) Bockers, T. M.; Kreutz, M. R.; Pohl, T. Glutaminyl-cyclase
expression in the bovine/porcine hypothalamus and pituitary.
J. Neuroendocrinol. 1995, 7, 445–453.
Acknowledgment. The authors thank Dr. F. Rosche and
Dr. R. Wolf for providing the ESI-MS spectra and Dr. J.
Schmidt from the Leibniz Institute of Plant Biochemistry,
(18) Saido, T. C.; Iwatsubo, T.; Mann, D. M.; Shimada, H.; Ihara, Y.;
Kawashima, S. Dominant and differential deposition of distinct

7080 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 22
Buchholz et al.
beta-amyloid peptide species, A beta N3(pE), in senile plaques.
Neuron 1995, 14, 457–466.
(34) Shirotani, K.; Tsubuki, S.; Lee, H. J.; Maruyama, K.; Saido,
T. C. Generation of amyloid beta peptide with pyroglutamate at
position 3 in primary cortical neurons. Neurosci. Lett. 2002, 327,
25–28.
(35) Cynis, H.; Scheel, E.; Saido, T. C.; Schilling, S.; Demuth, H.-U.
Amyloidogenic processing of amyloid precursor protein: evidence
of a pivotal role of glutaminyl cyclase in generation of pyrogluta-
mate-modified amyloid-beta. Biochemistry 2008, 47, 7405–7413.
(36) Henikoff, S.; Henikoff, J. G. Amino acid substitution matrices
from protein blocks. Proc. Natl. Acad. Sci. U.S.A. 1992, 89, 10915–
10919.
(37) Schilling, S.; Cynis, H.; von Bohlen, A.; Hoffmann, T.; Wermann,
M.; Heiser, U.; Buchholz, M.; Zunkel, K.; Demuth, H.-U. Isola-
tion, catalytic properties, and competitive inhibitors of the zinc-
dependent murine glutaminyl cyclase. Biochemistry 2005, 44,
13415–13424.
(38) Prescott, J. M.; Wagner, F. W.; Holmquist, B.; Vallee, B. L.
Spectral and kinetic studies of metal-substituted Aeromonas ami-
nopeptidase: nonidentical, interacting metal-binding sites. Bio-
chemistry 1985, 24, 5350–5356.
(19) Schilling, S.; Zeitschel, U.; Hoffmann, T.; Heiser, U.; Francke, M.;
Kehlen, A.; Holzer, M.; Hutter-Paier, B.; Prokesch, M.; Windisch,
M.; Jagla, W.; Schlenzig, D.; Lindner, C.; Rudolph, T.; Reuter,
G.; Cynis, H.; Montag, D.; Demuth, H.-U.; Rossner, S.
Glutaminyl cyclase inhibition attenuates pyroglutamate Abeta
and Alzheimer0s disease-like pathology. Nat. Med. 2008, 14,
1106–1111.
(20) Miravalle, L.; Calero, M.; Takao, M.; Roher, A. E.; Ghetti, B.;
Vidal, R. Amino-terminally truncated Abeta peptide species are the
main component of cotton wool plaques. Biochemistry 2005, 44,
10810–10821.
(21) Schilling, S.; Hoffmann, T.; Manhart, S.; Hoffmann, M.; Demuth,
H.-U. Glutaminyl cyclases unfold glutamyl cyclase activity under
mild acid conditions. FEBS Lett. 2004, 563, 191–196.
(22) Schilling, S.; Lauber, T.; Schaupp, M.; Manhart, S.; Scheel, E.;
Bohm, G.; Demuth, H.-U. On the seeding and oligomerization of
pGlu-amyloid peptides (in vitro). Biochemistry 2006, 45, 12393–
12399.
(23) Youssef, I.; Florent-Bechard, S.; Malaplate-Armand, C.; Koziel,
V.; Bihain, B.; Olivier, J. L.; Leininger-Muller, B.; Kriem, B.; Oster,
T.; Pillot, T. N-Truncated amyloid-beta oligomers induce learning
impairment and neuronal apoptosis. Neurobiol. Aging 2007, 29,
1319–1333.
(24) Russo, C.; Violani, E.; Salis, S.; Venezia, V.; Dolcini, V.; Damonte,
G.; Benatti, U.; D’Arrigo, C.; Patrone, E.; Carlo, P.; Schettini, G.
Pyroglutamate-modified amyloid beta-peptides—AbetaN3(pE)—
strongly affect cultured neuron and astrocyte survival. J. Neuro-
chem. 2002, 82, 1480–1489.
(25) Schilling, S.; Appl, T.; Hoffmann, T.; Cynis, H.; Schulz, K.; Jagla,
W.; Friedrich, D.; Wermann, M.; Buchholz, M.; Heiser, U.; von
Horsten, S.; Demuth, H.-U. Inhibition of glutaminyl cyclase pre-
vents pGlu-Abeta formation after intracortical/hippocampal mi-
croinjection in vivo/in situ. J. Neurochem. 2008, 106, 1225–1236.
(26) Buchholz, M.; Heiser, U.; Schilling, S.; Niestroj, A. J.; Zunkel, K.;
Demuth, H.-U. The first potent inhibitors for human glutaminyl
cyclase: synthesis and structure-activity relationship. J. Med.
Chem. 2006, 49, 664–677.
(39) Bayliss, M. E.; Prescott, J. M. Modified activity of Aeromonas
aminopeptidase: metal ion substitutions and role of substrates.
Biochemistry 1986, 25, 8113–8117.
(40) Chevrier, B.; Schalk, C.; D’Orchymont, H.; Rondeau, J. M.;
Moras, D.; Tarnus, C. Crystal structure of Aeromonas proteolytica
aminopeptidase: a prototypical member of the co-catalytic zinc
enzyme family. Structure 1994, 2, 283–291.
(41) Cole, J. C.; Nissink, W. M.; Taylor, R. Protein-Ligand Docking
and Virtual Screening with GOLD. In Virtual Screening in Drug
Discovery; Alvarez, J., Shoichet, S., Eds.; Taylor & Francis: Boca
Raton, FL, 2005; pp 379-415.
(42) Ajay; Murcko, M. A. Computational methods to predict binding
free energy in ligand-receptor complexes. J. Med. Chem. 1995, 38,
4953–4967.
(43) Szegezdi, J.; Csizmadia, F. A Method for Calculating the pKa
Values of Small and Large Molecules. http://www.chemaxon.com/
conf/Calculating_pKa_values_of_small_and_large_molecules.pdf
(2009); ChemAxon.
(44) Schilling, S.; Hoffmann, T.; Wermann, M.; Heiser, U.; Waster-
nack, C.; Demuth, H.-U. Continuous spectrometric assays for
glutaminyl cyclase activity. Anal. Biochem. 2002, 303, 49–56.
(45) Schilling, S.; Hoffmann, T.; Rosche, F.; Manhart, S.; Wasternack,
C.; Demuth, H.-U. Heterologous expression and characteriza-
tion of human glutaminyl cyclase: evidence for a disulfide bond
with importance for catalytic activity. Biochemistry 2002, 41,
10849–10857.
(27) Thornber, C. W. Isosterism and molecular modification in drug
design. Chem. Soc. Rev. 1979, 8, 563–580.
(28) Patani, G. A.; LaVoie, E. J. Bioisosterism: a rational approach in
drug design. Chem. Rev. 1996, 96, 3147–3176.
(29) Lima, L. M.; Barreiro, E. J. Bioisosterism: a useful strategy for
molecular modification and drug design. Curr. Med. Chem. 2005,
12, 23–49.
(30) Durant, G. J.; Emmett, J. C.; Ganellin, C. R.; Miles, P. D.; Parsons,
M. E.; Prain, H. D.; White, G. R. Cyanoguanidine-thiourea
equivalence in the development of the histamine H2-receptor
antagonist, cimetidine. J. Med. Chem. 1977, 20, 901–906.
(31) Duncan, W. A.; Parsons, M. E. Reminiscences of the development
of cimetidine. Gastroenterology 1980, 78, 620–625.
(32) Atwal, K. S.; Ahmed, S. Z.; O’Reilly, B. C. A facile synthesis of
cyanoguanidines from thioureas. Tetrahedron Lett. 1989, 30, 7313–
7316.
(33) Stankovic, S.; Skobaljic, N.; Stojanovic, N.; Stojicic, S. Proce-
dure for Synthesis of N-[2[[[5-[(Dialkylamino)methyl]-2-furanyl]-
methyl]thyo]ethyl]-N0-alkyl-2-nitro-1,1-alkenediamines and Their
Hydrochlorides. PCT/YU99/00011, 2000; pp 1-11.
(46) Sippl, M. J. Recognition of errors in three-dimensional structures
of proteins. Proteins 1993, 17, 355–362.
(47) Hendlich, M.; Lackner, P.; Weitckus, S.; Floeckner, H.;
Froschauer, R.; Gottsbacher, K.; Casari, G.; Sippl, M. J. Identi-
fication of native protein folds amongst a large number of incorrect
models. The calculation of low energy conformations from poten-
tials of mean force. J. Mol. Biol. 1990, 216, 167–180.
(48) Laskowski, R. A.; MacArthur, M. W.; Moss, D. S.; Thornton,
J. M. PROCHECK: a program to check the stereochemical quality
of protein structures. J. Appl. Crystallogr. 1993, 26, 283–291.
(49) Pellegrini, E.; Field, M. J. A generalized-Born solvation model for
macromolecular hybrid-potential calculations. J. Phys. Chem. A
2002, 106, 1316–1326.