New aziridine-based inhibitors of cathepsin L-like cysteine proteases with selectivity for the Leishmania cysteine protease LmCPB2.8

Article abstract of DOI:10.1016/j.ejmech.2018.07.012

In the present work a series of aziridine-2,3-dicarboxylate inhibitors of papain-like cysteine proteases was designed, synthesized and tested. The compounds displayed selectivity for the parasitic protozoon Leishmania mexicana cathepsin L-like cysteine protease LmCPB2.8. The computational methods of homology modelling and molecular docking predicted some significant differences in the S2 pocket of LmCPB2.8 and cruzain, a related enzyme from Trypanosoma cruzi. Due to the presence of Tyr209 in LmCPB2.8 rather than Glu208 in cruzain sterically demanding, lipophilic ester groups (inhibitor 7d, 9d, 12d and 14d) are predicted to occupy the S2 pocket of the Leishmania protease, but do not form favorable interactions in cruzain, which is in common with our experimental results. Further, inhibitor 18 bearing a free carboxylic acid attached to the aziridine moiety showed a time-dependent inhibition of LmCPB2.8 (K_i = 0.41 μM; k_2nd = 190,569 M^?1 min^?1). Docking results suggested a strong ionic interaction with the positively charged His163 of the active site. Biological and theoretical data confirm that the novel selective aziridine-based inhibitors are promising candidates for further optimization as LmCPB2.8 inhibitors.

Full text of DOI:10.1016/j.ejmech.2018.07.012

Accepted Manuscript
New aziridine-based inhibitors of cathepsin L-like cysteine proteases with selectivity
for the Leishmania cysteine protease LmCPB2.8
Philipp Fey, Roula Chartomatsidou, Werner Kiefer, Jeremy C. Mottram, Christian
Kersten, Tanja Schirmeister
PII:
S0223-5234(18)30568-3
10.1016/j.ejmech.2018.07.012
EJMECH 10547
DOI:
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 7 March 2018
Revised Date: 3 July 2018
Accepted Date: 6 July 2018
Please cite this article as: P. Fey, R. Chartomatsidou, W. Kiefer, J.C. Mottram, C. Kersten, T.
Schirmeister, New aziridine-based inhibitors of cathepsin L-like cysteine proteases with selectivity for
the Leishmania cysteine protease LmCPB2.8, European Journal of Medicinal Chemistry (2018), doi:
10.1016/j.ejmech.2018.07.012.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT
Graphical abstract

ACCEPTED MANUSCRIPT
New aziridine-based inhibitors of cathepsin L-like
cysteine proteases with selectivity for the
Leishmania cysteine protease LmCPB2.8
§
§
§
#
§
Philipp Fey, Roula Chartomatsidou, Werner Kiefer, Jeremy C. Mottram, Christian Kersten,
§
Tanja Schirmeister *
§
Institut für Pharmazie und Biochemie, Johannes Gutenberg-Universität Mainz, Staudingerweg
, 55268 Mainz, Germany
5
#
Centre for Immunology and Infection, Department of Biology, University of York, Wentworth
Way, Heslington, York, YO10 5DD, UK.
Keywords: cysteine protease, inhibitor, aziridine, rhodesain, cruzain, Leishmania
Abbreviations:
DCC (Dicyclohexylcarbodiimide), DMAP (4-(Dimethylamino)pyridine), DPSI ((S,S)-
Diphenylsulfimine), GB/VI methodology (Generalized Born/Volume Integral methodology), LE
(ligand efficiency), LLE (lipophilic ligand efficiency), MMFF94 (Merck Molecular Force Field),
1

ACCEPTED MANUSCRIPT
MOE (Molecular Operating Environment), ni (no inhibition), nd (not determined), PDB (Protein
Data Bank), PPA (Propylphosphonicanhydride), RMSD (root-mean-square deviation)
Highlights
•
Development of new aziridine-based inhibitors of cathepsin L-like cysteine proteases
with selectivity for the Leishmania mexicana LmCPB2.8
•
•
Analysis of binding sites for LmCPB2.8 and cruzain reveal differences in the S2 pockets
Predicted ionic interaction with the histidinium ion of the active site of LmCPB2.8 leads
to improved affinity and to time-dependent inhibition
Graphical abstract
Abstract
In the present work a series of aziridine-2,3-dicarboxylate inhibitors of papain-like cysteine
proteases was designed, synthesized and tested. The compounds displayed selectivity for the
parasitic protozoon Leishmania mexicana cathepsin L-like cysteine protease LmCPB2.8. The
computational methods of homology modelling and molecular docking predicted some
2

ACCEPTED MANUSCRIPT
significant differences in the S2 pocket of LmCPB2.8 and cruzain, a related enzyme from
Trypanosoma cruzi. Due to the presence of Tyr209 in LmCPB2.8 rather than Glu208 in cruzain
sterically demanding, lipophilic ester groups (inhibitor 7d, 9d, 12d and 14d) are predicted to
occupy the S2 pocket of the Leishmania protease, but do not form favorable interactions in
cruzain, which is in common with our experimental results. Further, inhibitor 18 bearing a free
carboxylic acid attached to the aziridine moiety showed a time-dependent inhibition of
-1
-1
i
LmCPB2.8 (K = 0.41 µM; k2nd = 190,569 M min ). Docking results suggested a strong ionic
interaction with the positively charged His163 of the active site. Biological and theoretical data
confirm that the novel selective aziridine-based inhibitors are promising candidates for further
optimization as LmCPB2.8 inhibitors.
1
. Introduction
Cysteine proteases play essential roles in the life cycles of various protozoan organisms making
them attractive targets for new antiparasitic drugs. Especially for the neglected diseases HAT
(Human African Trypanosomiasis) and the various forms of leishmaniasis, but also for malaria
and Chagas disease, new drugs are urgently needed. The parasites causing these diseases express
numerous related cysteine proteases that belong to the papain family; according to the MEROPS
database system [1,2] they are classified as clan CA, family C1 proteases and share structural
similarity to human cathepsins. In detail, these are the cathepsin L-like proteases rhodesain from
Trypanosoma brucei rhodesiense, falcipains 2 and 3 from Plasmodium falciparum, cruzain from
Trypanosoma cruzi, and various cathepsin L-like proteases from Leishmania species, e.g.
LmCPB2.8 from Leishmania mexicana. There are also some cathepsin B-like proteases
expressed by the parasites, e.g. TbCatB from T. b. rhodesiense or the Leishmania CPCs. Some of
3

ACCEPTED MANUSCRIPT
these proteases have already been shown to be suitable targets for new antiparasitic drugs with
the vinyl sulfone K11777 which is in preclinical studies against Chagas disease as one of the
most promising compounds.[3-5]
In previous studies we identified aziridine-based peptides as inhibitors selective for cathepsin L-
like cysteine proteases.[6-9] These consist of a trans-configured dibenzyl aziridine-2,3-
dicarboxylate fragment Azi(OBn) (i.e. (S,S) or (R,R) configuration) and the dipeptide Boc-Leu-
2
Pro attached to the aziridine nitrogen. Among the various stereoisomers of the peptide Boc-Leu-
Pro-Azi(OBn) some displayed selectivity for parasite proteases over human cathepsin L (Table
2
1
). Some compounds concomitantly showed antileishmanial activity against L. major
promastigotes and amastigotes,[8-10] namely compounds I, II, V (Table 1). Among these
derivatives the most promising is compound V which is selective for the leishmanial cathepsin
L-like enzyme over the mammalian cathepsins, and which displays good antileishmanial activity.
Table 1: Inhibition of mammalian cathepsins and parasite cathepsin-like proteases by aziridine-2,3-dicarboxylates,
and antileishmanial activity against L. major promastigotes and amastigotes.[8-10]
No. Configuration at
CL
K
i
CB
LmCPB2.8
RD
CRZ
FP-2
K
i
L. m. (p)/(a)
EC50
atoms 1a, 1b, 2, 3:
K
i
K
i
K
i
K
i
[
µM]
[µM]
[µM]
[µM]
[µM]
[µM]
[µM]
I
S,S,R,S
6.0±0.8
ni
1.7±0.2
1.2±0.2
0.9
43.9±5.2
39.6/2.2±1.5
4

ACCEPTED MANUSCRIPT
II
III
IV
V
S,S,S,R
S,S,S,S
S,S,R,R
R,R,R,S
R,R,S,R
R,R,S,S
4.0±0.2
ni
ni
ni
ni
ni
2.1±0.4
nd
1.1±0.2
2.3±0.3
2.8±0.3
1.7±0.1
0.79
nd
28.3±12.0
35.4/2.7±0.7
ni/nd
ni
ni
ni
ni
ni
ni
nd
nd
ni/nd
3.8±0.1
3.4±0.2
nd
ni
ni
37.4/2.3±0.6
95/nd
VI
VII
15.2±1.9 30.9±8.2
0.5±0.1 nd
ni
0.4±0.2 114±18
0.4±0.1
ni/nd
ni = no inhibition at 20 µM, nd = not determined, CL = cathepsin L, CB = cathepsin B, LmCPB2.8 = Leishmania
mexicana cathepsin L-like protease, RD = rhodesain from Trypanosoma brucei rhodesiense, CRZ = cruzain from
Trypanosoma cruzi, FP-2 = falcipain-2 from Plasmodium falciparum; L. m. (p)/(a) = L. major promastigotes /
amastigotes.
In an ongoing study derivatives with various dipeptides attached to the aziridine nitrogen were
synthesized in which the (S)-Leu residue of compound V was exchanged against another
hydrophobic (S)-configured amino acid (Ala, Val, Ile, Nva, Nle, Chg, Cha, Phg, Phe, hPhe,
Trp).[10] Most of these derivatives showed inhibition of the Leishmania enzyme in the same
range as compound V, but only the inhibitors with Val and Ile also displayed antileishmanial
activity against L. major promastigotes and amastigotes: (EC [µM] amastigotes/promastigotes:
5
0
Val: 34.8/2.2±0.6; Ile: 9.8/2.0±0.6: toxicity against J774.4 macrophages > 250 µM for both).
Compound V, in contrast to its Val and Ile analogs, inhibits both the cathepsin L-like CPB2.8
and the L. major cathepsin B-like enzyme CPC (K = 18.2 µM).[10] Also in L. major lysates,
i
inhibition of both cathepsin L- and cathepsin B-like Leishmania proteases had been observed
with this compound [10] in contrast to compounds I and II. Thus, we decided to extend our
studies towards Leishmania protease inhibitors based on this most promising inhibitor V.
Previous studies had already shown that dibenzyl esters are superior to diethyl esters in terms of
enzyme inhibition and antiparasitic activity.[22] On the other hand, a free acid group at the
5

ACCEPTED MANUSCRIPT
aziridine ring has several advantages over the diesters. First, the water solubility is enhanced in
comparison to diesters. Second, a free acid improves affinity to cysteine proteases by forming an
ionic interaction with the histidinium residue of the active site.[7,11] Furthermore, the free acids
show fast covalent inhibition of cysteine proteases since this ionic interaction leads to a
positioning of the electrophilic aziridine ring, which is favorable for the nucleophilic ring
opening by the active site cysteine residue.[18] The new target compounds thus should consist of
the dipeptide sequence Boc-(S)-Leu-(R)-Pro attached to the aziridine nitrogen, and either of two
hydrophobic ester groups or of a free acid and a hydrophobic ester group at the aziridine ring
(
Figure 1).
Figure 1. General structure of the target cysteine protease inhibitors.
2
2
. Results and discussion
.1. Syntheses
The synthesis of such compounds with an ester and also an acid group at the aziridine building
block (Figure 1) requires an unsymmetric diester as intermediate. From this diester the free acid
6

ACCEPTED MANUSCRIPT
is released in a final synthetic step which, due to the instability of the aziridine and aziridide
structure, should not involve strong alkaline or acidic media. In this regard, a benzyl ester is ideal
since it can be removed by mild hydrogenolysis (Figure 1). The enantioselective synthesis of an
unsymmetric aziridine-2,3-dicarboxylate building block starting from a tartrate needs at least
seven steps including protection of the aziridine nitrogen, transesterification of a symmetric
aziridine diester by selective hydrolysis and re-esterification, and finally deprotection of the
aziridine nitrogen.[12] Thus, we decided to search a new, shorter synthetic route with
introduction of the two different esters at an earlier step of the synthesis. Besides the multi-step
synthesis starting form a tartrate, aziridine-2,3-dicarboxylates can also be synthesized by a one-
step Michael-addition of DPSI to fumarates.[11,13–15] Interestingly, also the cis-configured
maleates yield the racemic trans-configured aziridine-2,3-dicarboxylates due to steric hindrance
of the intermediate DPSI adduct T2b (Figure 2). Thus, also the maleates can be used as starting
materials.
7

ACCEPTED MANUSCRIPT
Figure 2. Mechanism of aziridine synthesis by Michael addition of DPSI to either fumarates or maleates. Due to steric hindrance
the adduct T2b rotates to the thermodynamically more stable adduct T2a which yields the trans-configured aziridine B or –after
tautomerization to T1– the enamine A. At higher temperatures (> 90 °C) the aziridine B is converted into enamine A.
The unsymmetric maleates or fumarates were synthesized as follows. Reaction of maleic
anhydride with benzyl alcohol yielded the maleic acid benzyl ester (Figure 3). Steglich
esterification of this half ester with various alcohols yielded the unsymmetric fumarates and / or
maleates (Figure 3). Depending on reaction time, reaction temperature and amount of DMAP as
catalyst either the trans- or the cis-configured unsymmetric diesters could be obtained. With
lower amounts of DMAP (0.1 eq), low temperature (0 °C), and shorter reaction times (1 - 2 d),
the maleates are the main products, while with higher amounts of DMAP (0.8 eq), room
temperature and reaction times of several days the fumarates were formed exclusively due to Z/E
isomerization catalyzed by DMAP [16,17] (Table 2, Figure 3).
3
Figure 3. Synthetic route to unsymmetric, racemic trans-configured aziridine-2,3-dicarboxylates. i. 1.0 eq. BnOH, 1.1 eq. NEt ,
1
h RT; ii. 1.0 eq. ROH, 1.1 eq. DCC, DMAP; iii. 1.2 eq. DPSI, 3 mL toluene/mmol DPSI, 24 h 80 °C
Table 2: Synthesized fumarates (a) and maleates (b) starting from maleic acid benzyl ester (I) or fumaric
acid (II), reaction conditions leading to either the E- or Z-configured olefin.
Cpd.
Synthesis
R
E or Z
Eq.
React.
React.
Yield %
DMAP time 0 °C
time RT
8

ACCEPTED MANUSCRIPT
a
1a
1b
2a
3a
-
Ethyl
Ethyl
E
Z
E
E
0.8
0.1
0.8
0.8
1 h
16 h
1 h
17 h
-
71
42
54
55
I
I
I
tert-Butyl
19 d
14 d
Cyclohexyl
methyl
1 h
3
b
I
Cyclohexyl
methyl
Z
0.1
30 h
-
31
4a
5b
6a
I
I
I
Phenyl
Benzyl
E
Z
E
0.8
0.1
0.8
1 h
48 h
1 h
48 h
-
38
36
25
Naphthalene-
72 h
2
-yl
7
b
I
Naphthalene-
-yl-methyl
Z
0.1
22 h
-
52
2
8a
8b
9a
I
I
I
2-Phenylethyl
2-Phenylethyl
E
Z
E
0.8
0.1
0.8
1 h
48 h
1 h
14 d
-
39
50
52
3-
14 d
Phenylpropyl
1
0a
I
Furan-2-yl-
methyl
E
E
E
E
E
E
0.8
0.1
1 h
72 h
-
35
27
99
99
96
99
b
1
1a
I
Pyridine-4-yl-
methyl
24 h
c
1
2a
II
II
II
II
Cyclohexyl
methyl
-
-
-
-
-
24 h
reflux
c
5
a
Benzyl
-
24 h
reflux
c
1
3a
4a
2-Phenylethyl
-
24 h
reflux
c
1
3-
-
24 h
Phenylpropyl
reflux
a
b
synthesized from commercially available monoethylfumarate; E-product obtained under Z-conditions due to
c
catalysis by 4-pyridinemethanol (in analogy to the Z/E isomerization catalyzed by DMAP); 2.5 eq. alcohol, p-
TosOH as catalyst, 24 h reflux in benzene.
9

ACCEPTED MANUSCRIPT
Finally the unsymmetric trans-configured aziridine-2,3-dicarboxylates which contain one benzyl
ester group were synthesized by Michael addition of DPSI to the fumarates or maleates (Figure
3
). Only with a few fumarates (with phenyl (4a), naphthalene-2-yl (6a), furan-2-ylmethyl (10a),
and pyridine-4-ylmethyl (11a) moiety) the aziridine synthesis failed.
Table 3: Synthesized trans-configured, racemic aziridine-2,3-dicarboxylate
building blocks.
1
2
Cpd.
R
R
Yield %
1
2
3
c
c
c
Ethyl
Benzyl
Benzyl
Benzyl
14
14
56
tert-Butyl
Cyclohexyl
methyl
5
7
c
c
Benzyl
Benzyl
Benzyl
68
a
Napthalene-2-yl-
methyl
10
8
9
c
c
2-Phenylethyl
Benzyl
Benzyl
19
17
18
3-Phenylpropyl
1
2c
Cyclohexyl
methyl
Cyclohexyl methyl
1
1
3c
4c
2-Phenylethyl
2-Phenylethyl
18
9
3-Phenylpropyl
3-Phenylpropyl
a
synthesis starting from the corresponding maleate; in all other cases the synthesis started from the fumarate.
The N-acylated aziridines with two different ester groups (Table 4, Figure 4, compounds 1d, 2d,
d, 7d, 8d, 9d) were obtained as diastereomeric mixtures by acylation with the dipeptide Boc-
S)-Leu-(R)-Pro using PPA as coupling reagent (Table 4, Figure 4). The halfesters were finally
3
(
10

ACCEPTED MANUSCRIPT
synthesized by hydrogenolysis of the benzyl ester part of the unsymmetric diesters (Table 4,
Figure 4, compounds 15-19). Additionally, several symmetric diesters were also synthesized,
starting from fumaric acid, esterification with excess of the respective alcohol, aziridine
synthesis with DPSI and finally N-acylation with the dipeptide to yield the symmetric N-acylated
aziridine-2,3-dicarboxylates (Table 2, synthesis II; Table 3, compounds 5c, 12c, 13c, 14c; Table
4
, cpds. 5d, 12d, 13d, 14d).
Table 4: Synthesized N-acylated aziridine-2,3-dicarboxylates.
1
2
Cpd.
R
R
Reaction
time at rt
Yield
diester/acid
a
[
h]
[%]
15
1
2
3
d
d
d
Ethyl
Benzyl
Benzyl
Benzyl
72
48
24
tert-Butyl
26
Cyclohexyl
methyl
13
5
7
d
d
Benzyl
Benzyl
Benzyl
24
48
58
23
Napthalene-2-
yl-methyl
8
9
d
d
2-Phenylethyl
Benzyl
Benzyl
48
48
43
24
3-
Phenylpropyl
1
2d
Cyclohexyl
methyl
Cyclohexyl
methyl
24
22
1
1
3d
4d
2-Phenylethyl
3-
2-Phenylethyl
3-
48
72
31
35
11

ACCEPTED MANUSCRIPT
Phenylpropyl
Ethyl
Phenylpropyl
1
1
1
5
6
7
H
H
H
85
72
74
tert-Butyl
Cyclohexyl
methyl
1
1
8
9
2-Phenylethyl
H
H
95
98
3-
Phenylpropyl
a
step i. / ii. of the synthesis shown in Figure 4.
Figure 4. Synthesis of N-acylated aziridine-2,3-dicarboxylates. i. 1.2 eq. dipeptide, 3.0 eq. PPA, EA, 1 h 0 °C, then several h RT;
2
ii. R = benzyl, 103 mg/mmol Pd/C, MeOH, 30 min, 5 bar H2.
The previous studies [10] on these compounds (Table 1) showed that the diastereoisomers which
only differ in the configuration of the aziridine ring (R,R or S,S) are similarly active against
LmCPB2.8 (I and V; II and VI; see Table 1), indicating that the activity is mainly dependent on
the configuration of the dipeptide and to a lesser extent on the configuration of the aziridine ring.
Thus, we refrained from the rather difficult separation of the diastereoisomers.
12

ACCEPTED MANUSCRIPT
2
.2. Biological activities
The synthesized N-acylated aziridine-2,3-dicarboxylates (Table 4) were tested against the
cysteine proteases LmCPB2.8 (Table 5), against mammalian cathepsins L (Table 5) and B (not
shown), cruzain, falcipain-2 and rhodesain (Table 5).
Table 5: Inhibition of cysteine proteases by N-acylated aziridine-2,3-dicarboxylates and LE and LLE values for
inhibition of LmCPB2.8
Cpd.
LmCPB2.8
LE
LLE
Cruzain
CL
Falcipain-2
Rhodesain
K [µM]
i
1
2
3
5
7
8
9
d
d
d
d
d
d
d
ni
ni
-
-
ni
ni
ni
ni
ni
ni
ni
ni
-
-
ni
1.8±0.4
0.174
0.169
0.170
0.170
0.171
0.186
0.169
0.169
-
-1.62
-0.86
-1.50
-1.09
-1.32
-2.19
-1.40
-1.90
-
7.6±0.7
ni
ni
a
a
a
a
2.8
1.8
12.0
ni
nd
1.5
0.8
1.92±0.06
1.36±0.05
0.77±0.05
1.6±0.1
0.9±0.1
ni
ni
ni
ni
5.5±0.2
ni
ni
ni
ni
ni
ni
ni
1
1
1
2d
3d
4d
ni
ni
ni
ni
7.7±0.7
ni
ni
ni
ni
ni
ni
ni
ni
1
5
6
ni
ni
ni
ni
1
ni
-
-
ni
ni
ni
b
1
1
1
7
1.79±0.43
0.44±0.022
ni
0.207
0.223
-
0.35
1.53
-
0.71±0.015
7.3±2.4
0.45±0.015
1.26±0.22
0.12±0.002
0.16±0.003
ni
c
8
ni
ni
ni
ni
d
9
28.6±3.7
a
b
see Table 1, cpd. 5d is a mixture of the two diastereomers I and V (see Table 1); ni, inhibition at 20 µM < 50%;
-1
-1
k
2nd [M min ] for inhibition of cruzain = 322,531, LmCPB2.8 = 53,590, rhodesain = 494,038, falcipain-2 = 76,789,
13

ACCEPTED MANUSCRIPT
c
-1
-1
cathepsin L = 13,939; k2 [M min ] for inhibition of LmCPB2.8 = 190,569, rhodesain = 178,388, falcipain-2 =
nd
d
-1
-1
1
1,100; k [M min ] for inhibition of falcipain-2 = 3,349.
2nd
None of the new N-acylated aziridine-2,3-dicarboxylates inhibited mammalian cathepsin B (20
µM final concentration, not shown), only one inhibited mammalian cathepsin L (the acid 17, K
.3 µM), only three falcipain-2 (the acids 17-19, K = 0.45, 1.26, 62.0 µM), and only two
i
=
7
i
compounds rhodesain (the acids 17, 18, K = 0.12, 0.16 µM). Cruzain was inhibited by the
i
diesters 3d, 8d, 13d (K = 7.6, 5.5, 7.7 µM) and the acid 17 (K = 0.71 µM).
i
i
These results show that the acids, especially the acid 17 with cyclohexylmethyl ester moiety, are
the less selective compounds. Only these acids showed time-dependent inhibition of the
enzymes, which is in agreement with experimental and theoretical studies on oxirane- and
aziridine-2,3-dicarboxylates.[18] These studies had shown that an acid at the electrophilic three-
membered ring leads to a much faster ring opening reaction since the strong ionic interaction
between the inhibitor’s carboxylate and the histidinium ion of the active center of the protease
pushes the electrophilic ring onto the negatively charged active site’s cysteinate residue.[18] For
these acids the second-order rate constants of inhibition k2nd were determined (Table 5).
For the inhibition of LmCPB2.8 the ligand efficiencies (LE) and the lipophilic ligand efficiencies
(
LLE) were calculated using the following equations (Table 5): [19–21]
ꢃꢄꢅ
ꢀ
ꢁ = ꢂꢆꢇ ꢈ ∗ 1.37
and
ꢀꢀꢁ = ꢉꢊ − ꢌꢍꢎꢏꢐ
ꢋ
with HA as heavy atom count and clogP as calculated logP values.
The suitability of the calculated clogP values was verified by comparing the calculated values to
the retention times obtained in RP-HPLC runs (see supplementary data). The results show a good
correlation (Figure 5A) rendering the clogP values suitable for calculation of LLE values.
14

ACCEPTED MANUSCRIPT
6
6
6
,6
,4
,2
6
34
32
30
28
26
24
22
20
18
16
14
12
10
8
Cpd. 18
5
5
5
,8
,6
,4
Cpd. 17
5,2
6
8
2
4
6
8
10
clogP
clogP
A:
B:
Figure 5. A) Correlation of clogP values and retention times in RP-HPLC runs. B) LLE diagram showing correlation between
clogP and pK values. The regression line shows that the enhanced affinity of the diesters is correlating to their lipophilicities.
i
Both parameters, LE and LLE (Table 5), were found to be better for the acids 17 and 18 in
comparison to the diesters. For the diesters, the affinities correlate with the lipophilicities of the
compounds (Figure 5B). Thus, in terms of suitability as hit-to-lead candidate for further
development of LmCPB2.8 inhibitors the acid 18 turned out to be not an optimal but a
reasonably good candidate. This compound also displays a quite good selectivity against other
cysteine proteases.
2
.3. Molecular Docking
The inhibitors can be divided in three groups based on the groups attached to the aziridine ring:
symmetric diesters (5d, 12d, 13d, 14d), unsymmetric diesters (1d, 2d, 3d, 7d, 8d, 9d) and acids
15

ACCEPTED MANUSCRIPT
(halfesters) (15 - 19). For symmetric and unsymmetric diesters selectivity and also affinity for
LmCPB2.8 improve, when the aromatic ester chain gets longer. All inhibitors with two large
lipophilic ester groups (3d – 14d) display good affinity and selectivity for LmCPB2.8. Only
compounds 3d, 5d (which is a mixture of inhibitors I+V, see Table 1), 8d, and 13d are also
active against cruzain. These results point to differences in the binding modes between cruzain
and LmCPB2.8 on the one hand, but also to differences in the binding modes between diesters
and halfesters.
2
.4. Comparison of the active sites of LmCPB2.8 and cruzain
In order to explain these observed differences in inhibition of cruzain and LmCPB2.8 we
performed docking studies with cruzain (PDB 3I06) [23,24] and LmCPB2.8. Since for the
Leishmania enzyme no X-ray structure is available we built a homology model based on its
protein sequence with the Uniprot identifier P36400 (http://www.uniprot.org/uniprot/P36400)
(see Material and Methods).
For identification of relevant amino acids in the binding pockets of both enzymes and to explore
possible differences, we aligned and superposed the homology model structure of L. mexicana
CPB2.8 with the X-ray structure of cruzain (PDB 3I06, resolution 1.1 Å) (Figure 6). Although
the alignment indicates high similarity between the two proteins with a Cα RMSD value of 0.61
Å, some significant differences are found, especially for the S2 subsite. Table 6 shows the key
amino acids of each subsite. As expected, the catalytic triad and the subsites S1 and S1’ are
highly conserved for both parasite cathepsin L-like cysteine proteases.[25,26] Due to the
exchange of Glu208 (cruzain) into Tyr209 (LmCPB2.8) in the S2 pockets of the enzymes,
16

ACCEPTED MANUSCRIPT
aromatic and lipophilic residues would be more favorable for the leishmanial protease. In
agreement with experimental data and the docking studies (see below), the deep hydrophobic S2
pocket of LmCPB2.8 can be occupied by sterically demanding, aromatic residues, such as
phenylpropyl or methylnaphthalene ester groups.
Table 6: Key active site and binding pocket amino acids for cruzain and LmCPB2.8 (differences are underlined)
Protease
S2 subsite
S1 subsite
Catalytic triad
S1´ subsite
Cruzain
Asn69, Glu208, Ser210
Leu67, Met68, Ala138, Gly163
Cys25, His162, Asn182
Gln19, Trp184
LmCPB2.8 Leu69, Tyr209, Val211
Leu67, Met68, Ala139, Gly164
Cys25, His163, Asn183
Gln19, Trp185
17

ACCEPTED MANUSCRIPT
Figure 6. Alignment of the active sites of cruzain (PDB 3I06) and the LmCPB2.8 homology model; amino acids, which differ
between two enzymes are highlighted in light blue; catalytic residues are highlighted in orange.
2
.5. Docking studies
Docking of the symmetric diesters (-OBn (5d) vs. -OEtPh (13d) vs. -OPrPh (14d)) into the
binding site of cruzain and the homology model of the leishmanial enzyme show, that the
selective LmCPB2.8 inhibitor 14d (K = 0.9 µM) has good FlexX binding scores for LmCPB2.8
i
(-11,89 kJ/mol for the (R,R)- and -10,22 kJ/mol for the (S,S)-diastereomer), while for cruzain the
binding scores are much higher for both diastereomers (1,26 kJ/mol for the (R,R)- and -2,71
kJ/mol for the (S,S)-diastereomer). One of the phenylpropyl ester groups is proposed to bind
18

ACCEPTED MANUSCRIPT
deeply into the hydrophobic S2 pocket of the leishmanial protease and the second ester group
occupies the hydrophobic S1’ pocket, whereas in cruzain the ester group does not fit into the S2
subsite and is solvent exposed (see Figure 7). Docking of compound 13d, with two phenylethyl
ester groups, shows similar binding modes for both enzymes, with the ester groups occupying the
S2 and S1’ pockets. This indicates that the S2 pocket of the leishmanial protease tolerates longer
aromatic ester chains than cruzain, which is in agreement with inhibition data. Also the distance
between the nucleophilic sulfur atom of the catalytic cysteine residue and the electrophilic
aziridine carbon atom is for the leishmanial enzyme in a range, where a covalent bond formation
is possible (4.65 Å for (S,S)- and 4.92 Å for (R,R)-diastereomer).
A
B
C
D
19

ACCEPTED MANUSCRIPT
Figure 7. Binding mode of inhibitor 14d
A Diastereomer with (S,S)-configuration at the aziridine ring into the leishmanial protease (FlexX-Score -9,53 kJ/mol, distance to
the active site Cys residue: 4.65 Å)
B Diastereomer with (R,R)-configuration at the aziridine ring into the leishmanial protease (FlexX-Score -11,83 kJ/mol, distance
to the active site Cys residue: 4.92 Å)
C Diastereomer with (S,S)-configuration of the aziridine ring) into cruzain (FlexX-Score -2,71 kJ/mol, distance to the active site
Cys residue: 5.29 Å)
D Diastereomer with (R,R)-configuration at the aziridine ring into cruzain (FlexX-Score 1,26 kJ/mol, distance to the active site
Cys residue: 9.54 Å)
For the unsymmetric diesters (-OEtPh (8d) vs. -OPrPh (9d) vs. -OMeNaph (7d)) we obtained
i
similar results. The LmCPB2.8-selective inhibitor 7d (K = 0.8 µM) with the sterically most
demanding ester group showed the best FlexX scores for the leishmanial protease (-13,65 kJ/mol
for the (R,R)- and -15,39 kJ/mol for the (S,S)-diastereomer). Both diastereomers placed the
methylnaphthalene ester group in the hydrophobic S2 pocket (see Figure 8). The configuration of
the aziridine ring determines the position of the second ester group. For the (R,R)-diastereomer
the benzylester is placed in the hydrophobic S1’ pocket and for the (S,S)-diastereomer the proline
residue occupies the S1’ pocket. In contrast, no binding mode could be found for cruzain and the
(R,R)-derivate of 7d, and for the (S,S)-derivate the binding score was much higher for cruzain (-
7
,99 kJ/mol), which is in accordance to experimental data.
20

ACCEPTED MANUSCRIPT
A
B
C
D
No binding mode
Figure 8. Binding mode of inhibitor 7d
A Diastereomer with (S,S)-configuration at the aziridine ring into the leishmanial protease (FlexX-Score -19,92 kJ/mol, distance
to the active site Cys residue: 4.90 Å)
B Diastereomer with (R,R)-configuration at the aziridine ring into the leishmanial protease (FlexX-Score -14,21 kJ/mol, distance
to the active site Cys residue: 4.60 Å)
C Diastereomer with (S,S)-configuration of the aziridine ring) into cruzain (FlexX-Score -7,99 kJ/mol, distance to the active site
Cys residue: 5.66 Å)
D No binding mode into cruzain was found for the diastereomer with (R,R)-configuration of the aziridine ring
21

ACCEPTED MANUSCRIPT
For the acids (15 - 19), as mentioned above, biological data showed a time-dependent inhibition
of the enzymes which means fast irreversible inhibition. In previous studies, this property of
aziridine-carboxylic acids had been explained by a strong ionic interaction between the free
carboxylate group of the ligand and the positively charged histidinium ion of the active site.[11]
This strong ionic interaction may also be responsible for the improved affinity of the acids
compared to the corresponding diesters (17 vs. 3d, 18 vs. 8d / 18 vs. 13d). Docking of inhibitor
1
8 (see Figure 9) with an affinity of 0.4 µM for the leishmanial protease showed indeed for both
diastereomers an ionic and hydrogen bonding interaction with the catalytic His163. The ester
group is placed in the S2 pocket in case of the (S,S)-diastereomer, whereas in case of the (R,R)-
diastereomer the proline moiety occupies this pocket. In contrast, docking into the active site of
cruzain indicated that there is no ionic interaction, which is in accordance with the inhibition
data.
A
B
22

ACCEPTED MANUSCRIPT
C
D
Figure 9. Binding mode of inhibitor 18
A Diastereomer with (S,S)-configuration at the aziridine ring into the leishmanial protease (FlexX-Score -13,60 kJ/mol, distance
to the active site Cys residue: 5.8 Å)
B Diastereomer with (R,R)-configuration at the aziridine ring into the leishmanial protease (FlexX-Score -11,40 kJ/mol, distance
to the active site Cys residue: 4.81 Å)
C Diastereomer with (S,S)-configuration of the aziridine ring into cruzain (FlexX-Score -12,94 kJ/mol, distance to the active site
Cys residue: 4.6 Å)
D Diastereomer with (R,R)-configuration at the aziridine ring) into cruzain (FlexX-Score -10,79 kJ/mol, distance to the active site
Cys residue: 10.50 Å)
3
. Conclusion
Novel and selective aziridine-based inhibitors for the Leishmania mexicana protease LmCBP2.8
were discovered. Compounds 7d (K = 0.8 µM), 9d (K = 1.36 µM), 12d (K = 0.77 µM), 14d (K
i
i
i
i
=
0.9 µM) and 18 (K = 0.441 µM) showed the highest potency and selectivity over related
i
parasite and human homologues within the investigated series. For an analysis of the predicted
binding modes and binding pockets a homology model for the leishmanial protease was built and
compared to the structure of cruzain. This indicated that the S2 pocket of LmCBP2.8 is a deep,
lipophilic pocket, favorably being occupied by sterically demanding and lipophilic aromatic
23

ACCEPTED MANUSCRIPT
moieties of the ligands, while a tyrosine to glutamate substitution was observed in cruzain. This
significant difference between the two enzymes may also be the selectivity-determining feature.
Attachment of a carboxylate group at the aziridine ring leads to a time-dependent inhibition and
-1
-1
to an improved affinity to LmCPB2.8 (18 (K = 0.441 µM ) k = 190,569 M min ) likely to be
i
2nd
caused by an ionic interaction with His163 as indicated by the docking studies performed. This
strong interaction could be the reason for higher affinity and the fast irreversible inhibition.
Furthermore, LE and LLE values were improved, compared to the corresponding diesters.
24

ACCEPTED MANUSCRIPT
4
4
. Experimental
.1. Homology Modelling of LmCPB2.8
A homology model of the Leishmania mexicana LmCPB2.8 was generated to obtain the three-
dimensional structure of the enzyme in order to perform docking studies and to determine
relevant amino acids of the binding pockets. The homology model was built using the Molecular
Operating Environment (MOE) 2015.1001 software. The protein sequence of LmCPB2.8 used
for modeling has the Uniprot identifier P36400. Based on this target sequence a homology search
was performed using MOE Application PDB-Search. To find the most suitable template for
modeling, the sequence of the first nine results were aligned with the target sequence. PDB-ID
3
KKU,[27] solved at 1.28 Å resolution, the crystal structure of cruzain in complex with a ligand,
was selected as template showing a sequence similarity of 74.0% and an identity of 59.5% with
the target sequence (see supplementary data).
The homology model was built based on the protein sequence P36400 and the template structure
3
KKU using the MOE Homology Model application. The calculation was carried out under
default conditions with the AMBER10 [28] force field used for energy minimization. Ten
independent models were constructed by the Boltzmann-weighted randomized modeling
procedure. The final homology model was selected based on the best-scoring intermediate
model. The scoring method used was the Generalized Born/Volume Integral methodology [29]
(GB/VI methodology). The homology model was subsequently aligned to the template structure,
showing a Cα RMSD value of 0.54 Å (see supplementary data).
25

ACCEPTED MANUSCRIPT
4
.2. Molecular Docking
A successful covalent inhibitor of cysteine proteases in its pre-reactive form must first be able to
fit non-covalently into the binding site of the enzyme in order to bring the electrophilic center in
close proximity to the cysteine sulfur atom. Non-covalent binding interactions, such as hydrogen
bonding, hydrophobic, ionic, van der Waals interactions and (de-)solvation processes, are
responsible and important for a correct ligand binding. Molecular modeling routinely addresses
these non-covalent contributions to ligand binding without taking into consideration the
contribution of covalent reaction. In analogy to previous studies,[6] in the present docking study
the potential for a possible covalent bond formation was determined for each binding pose by
measuring the distance between the nucleophilic sulfur atom of the active site cysteine residue
and the electrophilic atoms of the ligand, namely the aziridine ring carbon atoms. Predicted
binding modes with a distance less than 5 Å were interpreted as possible start geometry for a
covalent bond formation.
The docking program FlexX (LeadIT, version 2.1.6) in the default mode was used to dock the
inhibitors flexibly into the active site of cruzain (PDB 3I06) and the leishmanial protease
LmCPB2.8 homology model. In the structure of cruzain, solved with a resolution of 1.1 Å, a
purine nitrile inhibitor is covalently bound to the enzyme. The binding site was defined as 6.5 Å
shell around this reference ligand. The covalent bond and the inhibitor were manually deleted
from the structure. Water molecules were removed and hydrogens were added to the protein
molecule to simulate physiological conditions at pH 6-7. The catalytic histidine residue was
protonated at both imidazole nitrogen atoms. As already successfully employed in previous
studies, the active site cysteine residue was mutated to alanine to avoid clashes.[6] After the
docking session it is mutated back to the original amino acid for illustration of the pre-covalent
26

ACCEPTED MANUSCRIPT
binding mode and measurement of the distance between the cysteine residue and aziridine
ring.[6] The aziridine-based inhibitors were each docked as both diastereomers (with (S,S)- and
(R,R)- configured aziridine ring). Energy minimization of the inhibitors prior to docking, using
the MMFF94 force field,[30] and visualization were performed with Molecular Operating
Environment (MOE 2015.1001).
For validation of the FlexX algorithm for current docking studies, the nitrile inhibitor, extracted
from the structure 3I06, was redocked into the enzyme. The highest-scoring docking pose
showed good agreement with the native binding mode (RMSD = 0.98 Å, see supplementary
data).
4
4
.4. Syntheses
.4.1. General
Reagents and solvents were obtained from commercial sources (Acros Organics, Alfa Aesar,
AllessaChemie, Bachem, Fluka, Iris Biotech, Merck, Roth or Sigma-Aldrich). Solvents were
1
13
distilled before use, while the other chemicals were used as received. H and C NMR spectra
were determined on a Bruker Fourier 300 or 400 MHz using CDCl or MeOD as solvent.
3
Chemical shifts δ are reported in parts per million (ppm) using residual proton peaks of the
1
13
1
13
solvent as internal standard (CDCl
9.00 ppm)). The coupling constants (J) are given in Hz, and the splitting patterns are designated
as follows: s (singulet); bs (broad singulet); d (doublet); dd (double doublet); t (triplet); m
multiplet). TLC was performed on Merck silica gel (60F254 or RP-C18) precoated plates (0.20
3
( H: 7.26 ppm, C: 77.16 ppm), MeOD ( H: 4.87 ppm, C:
4
(
27

ACCEPTED MANUSCRIPT
mm). Column chromatography was performed on Merck silica gel 60 (0.040-0.063 mm or 0.063-
0
.2 mm). Melting points were determined on a SMP3 Stuart and are uncorrected. Optical
rotation was measured on a P3000 A. Krüss Optronic polarimeter. ESI mass spectra were
performed on an Ultima 3 Micromass Flash Waters QTOF mass spectrometer. HPLC analyses
were run on a Varian ProStar with an analytical/preparative linear upsacle HPLC system (0.05-
5
0 mL/min at 275 bar back pressure with ScaleUp-mast) and a 2-channel UV detector (detection
at 220 nm and 254 nm). For analytical runs a Phenomenex Hyperclone 5u 0DS (C18) 120A
(250mm x 4.6mm) 5 Micron and for semipreparative runs a VARIAN DYNAMAX 250x21.4 MM
(
LxID) Microsorb 60-8 C18 column were used. The eluent system was millipore V3 water (A)
and acetonitrile HPLC grade with 0.1% formic acid (B) (for HPLC method details see
supplementary data).
4
.4.2. General procedure for the syntheses of fumarates 2a-4a, 6a and 8a-10a of Table 2 (2a
as an example)
A solution of maleic acid benzyl ester (20.0 g, 97.0 mmol), tert-butanol (7.19 g, 97.0 mmol) and
DMAP (9.48 g, 77.6 mmol) in 250 mL dichlormethane was stirred at 0 °C. Subsequently, DCC
(22.0 g, 106 mmol) was added. The reaction mixture was stirred 1 h at 0 °C and several hours at
room temperature (see Table 2). After filtration, the solution was washed (10% hydrochloric acid
solution, saturated NaHCO solution and water), dried over anhydrous Na SO and concentrated
3
2
4
under reduced pressure. The residue was purified by column chromatography on silica gel
petroleum ether/EtOAc) to yield the desired product 2a as a colorless oil (yield: 54%).
(
-1
1
-Benzyl-4-tert-butyl-(2E)-but-2-endioate (2a, Table 2): colorless liquid (54 %); FT-IR: ν [cm ]
1
=
2974, 1711, 1638, 1499, 1454, 1368, 1290, 1249, 1135, 972, 841, 751; H NMR (300 MHz,
28