Proline-Based Allosteric Inhibitors of Zika and Dengue Virus NS2B/NS3 Proteases

Article abstract of DOI:10.1021/acs.jmedchem.9b01697

The NS2B/NS3 serine proteases of the Zika and Dengue flaviviruses are attractive targets for the development of antiviral drugs. We report the synthesis and evaluation of a new, proline-based compound class that displays allosteric inhibition of both proteases. The structural features relevant for protease binding and inhibition were determined to establish them as new lead compounds for flaviviral inhibitors. Based on our structure-activity relationship studies, the molecules were further optimized, leading to inhibitors with submicromolar IC₅₀ values and improved lipophilic ligand efficiency. The allosteric binding site in the proteases was probed using mutagenesis and covalent modification of the obtained cysteine mutants with maleimides, followed by computational elucidation of the possible binding modes. In infected cells, antiviral activity against Dengue virus serotype 2 using prodrugs of the inhibitors was observed. In summary, a novel inhibitor scaffold targeting an allosteric site shared between flaviviral NS2B/NS3 proteases is presented whose efficacy is demonstrated in vitro and in cellulo.

Full text of DOI:10.1021/acs.jmedchem.9b01697

Subscriber access provided by University of Winnipeg Library
Article
Proline-based allosteric inhibitors of Zika
and Dengue virus NS2B/NS3 proteases
Benedikt Millies, Franziska von Hammerstein, Andrea Gellert, Stefan Hammerschmidt, Fabian Barthels,
Ulrike Göppel, Melissa Immerheiser, Fabian Elgner, Nathalie Jung, Michael Basic, Christian Kersten,
Werner Kiefer, Jochen Bodem, Eberhard Hildt, Maike Windbergs, Ute A. Hellmich, and Tanja Schirmeister
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.9b01697 • Publication Date (Web): 26 Nov 2019
Downloaded from pubs.acs.org on November 27, 2019
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W.,
Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 95
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Proline-based allosteric inhibitors of Zika
and Dengue virus NS2B/NS3 proteases
Benedikt Millies,† Franziska von Hammerstein,† Andrea Gellert,† Stefan Hammerschmidt,†
Fabian Barthels,† Ulrike Göppel,† Melissa Immerheiser, ┴ Fabian Elgner,║ Nathalie Jung┤,
Michael Basic,║ Christian Kersten,† Werner Kiefer,† Jochen Bodem, ┴ Eberhard Hildt,║ Maike
Windbergs┤, Ute A. Hellmich,*,‡,§,# Tanja Schirmeister*,†,#
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
†Institute of Pharmacy and Biochemistry, Johannes Gutenberg University Mainz, Staudinger
Weg 5, 55128 Mainz, Germany
┴Institut für Virology und Immunbiologie, Julius-Maximilians-Universität Würzburg, Versbacher
Str. 7, 97078 Würzburg, Germany
║Department of Virology, Paul-Ehrlich-Institut, 63225 Langen, Germany
‡Institute of Pharmacy and Biochemistry, Johannes Gutenberg University Mainz, Johann-Joachim-
Becherweg 30, 55128 Mainz, Germany
§Center for Biomolecular Magnetic Resonance (BMRZ), Goethe-Universität, 60438 Frankfurt am
Main, Germany
┤Institute of Pharmaceutical Technology and Buchmann Institute for Molecular Life Sciences,
Goethe University, Max-von-Laue-Straße 15, 60438 Frankfurt am Main, Germany
ACS Paragon Plus Environment
1

Journal of Medicinal Chemistry
Page 2 of 95
1
2
3
4
5
6
7
8
KEYWORDS Dengue virus, Zika virus, non-peptidic, allosteric, protease, inhibitor, cysteine
mutants, metabolism
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ABSTRACT The NS2B/NS3 serine proteases of the Zika and Dengue flaviviruses are attractive
targets for the development of antiviral drugs. We report the synthesis and evaluation of a new,
proline-based compound class that displays allosteric inhibition of both proteases. The structural
features relevant for protease binding and inhibition were determined to establish them as new lead
compounds for flaviviral inhibitors. Based on our structure-activity relationship studies, the
molecules were further optimized, leading to inhibitors with submicromolar IC50 values and
improved lipophilic ligand efficiency.The allosteric binding site in the proteases was probed using
mutagenesis and covalent modification of the obtained cysteine mutants with maleimides,
followed by computational elucidation of the possible binding modes. In infected cells, antiviral
activity against Dengue virus serotype 2 using prodrugs of the inhibitors was observed. In
summary, a novel inhibitor scaffold targeting an allosteric site shared between flaviviral
NS2B/NS3 proteases is presented whose efficacy is demonstrated in vitro and in cellulo.
INTRODUCTION In 2015, the Zika virus (ZIKV) became endemic in Brazil and infected
between 400.000 – 1.600.000 people. It is not only transmitted by the mosquito species Aedes
aegypti and Aedes albopticus, typically associated with flaviviral distribution, but also from human
to human.1–4 In 2016, the World Health Organization declared ZIKV a public health emergency
due to its appearance throughout South America and its correlation with a striking increase of
ACS Paragon Plus Environment
2

Page 3 of 95
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
neurological disorders, such as microcephaly in neonates.5,6 Although the main symptoms of a
ZIKV infection, fever, rash, and conjunctivitis, are relatively mild and most infections remain
asymptomatic, some of the (adult) patients develop severe neurological conditions such as the
Guillain-Barré syndrome.7–9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ZIKV, the four different serotypes of the Dengue virus (DENV1 – 4), and West-Nile Virus
(WNV) belong to the family of flaviviridae. The viral RNA encodes a polyprotein precursor that
is processed by viral and cellular proteases. Here, the viral NS2B/NS3 serine protease plays an
essential role. It cleaves the polyprotein precursor into the structural proteins that are components
of the virion, and the non-structural (NS) proteins involved in viral replication and maturation.
Additionally, the NS2B/NS3 protease suppresses the human immune response by cleaving the
stimulator of interferon genes (STING).10,11 These functions render the flaviviral NS2B/NS3
protease a promising target for the development of antiflaviviral drugs.12
The protein NS3 consists of a helicase and a serine protease domain. However, NS3 requires the
NS2B protein as a cofactor for full catalytic activity.13,14 In structures of the NS2B/NS3 complex,
two different conformations have been observed, the open and the catalytically active closed
conformation, where NS2B is wrapped around NS3, forming a β-turn with its C-terminus, which
contributes to the P2 and P3 binding sites of the protease.13,15 In the X-ray structures of the open
conformation, only the N-terminus of NS2B is resolved. The closed, active conformation is
obtained upon binding of a substrate or a competitive inhibitor, while an equilibrium between the
two conformations is observed in solution.16–18
Most of the reported competitive inhibitors for the ZIKV NS2B/NS3 protease, or the related
protease from e.g. WNV and DENV proteases19–21, share a peptidic or peptidomimetic structure.
Some, but not all of them additionally contain electrophilic warheads targeting the active site serine
ACS Paragon Plus Environment
3

Journal of Medicinal Chemistry
Page 4 of 95
1
2
3
4
5
6
7
8
9
residue. However, because of the preference of the S1 and S2 binding pockets for cationic residues,
effective competitive inhibitors display low membrane permeabilities.22 As an alternative route to
obtain antiflaviviral compounds, screening campaigns also identified non-competitive
inhibitors.23–25 In case of the DENV protease, non-competitive inhibitors led to a shift of the
protein conformation equilibrium to the open conformation. The putative allosteric binding site
was proposed to be located close to Asn152. This residue was hence termed the “molecular switch”
between the open and closed conformation.15,26
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
In previous studies,27 we reported a new series of non-competitive inhibitors for the DENV
NS2B/NS3 protease (e.g. 1, Figure 1). Since the use of these inhibitors as lead compounds for
further drug development and as biophysical probes for structural and dynamic studies of flaviviral
proteases was found to be hindered by their comparably high hydrophobicity, we aimed to
optimize these inhibitors in terms of affinity, hydrophilicity and lipophilic ligand efficiency (LLE).
In this regard, the resulting new proline-based inhibitors indeed (Figure 1) showed improved
properties.27
H
H
N
O
N
N
N
O
N
O

O
S
HO
HO
S
S
O
HO
HO
R
OH
1
7
15
-
slogP = 5.61
12
(R)- : R = 4-NO2
IC50 (ZIKV) = 1.42  0.16 µM
IC50 (DENV) = 4.2  0.44 µM
LLE = 0.24
slogP = 2.41
IC50 (ZIKV) = 0.32  0.05 µM
IC50 (DENV) = 10.0  0.98 µM
LLE = 4.09
Figure 1. The inhibitor of the DENV NS2B/NS3-protease 127 was modified to improve affinity,
hydrophilicity and lipophilic ligand efficiency (LLE).
ACS Paragon Plus Environment
4

Page 5 of 95
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
The antiflaviviral efficiency of the new inhibitors was demonstrated in cell-based assays. Based
on the metabolism profiles of the compounds and the differences in inhibition potency between
the methoxy-protected and the non-protected compounds, we could determine the relevance of
specific hydroxyl groups for allosteric inhibition capability.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The binding to the proposed allosteric site was verified using single cysteine point mutations
and their reaction with maleimides, which block the allosteric site for the inhibitors and,
intriguingly, also led to inhibition of the enzyme by itself, thus demonstrating cross-talk between
the allosteric and the active site and paving the way for the future development of covalent
inhibitors at this site.
The possible binding modes of the inhibitors were elucidated by molecular docking and
molecular dynamic simulations.
RESULTS AND DISCUSSION
CONSIDERATIONS ON NEW LEAD STRUCTURES
Because of the high global incidence numbers and their severity, new drugs against flaviviral
diseases are urgently needed. The flaviviral NS2B/NS3 protease essential for viral maturation
seems to be a particularly promising inhibitor target. Due to the high sequential (>50% identity)
and structural (1.11 Å RMSD) similarity between DENV and ZIKV NS2B/NS3 protease (DENV2,
PDB: 2FOM28 and ZIKV, PDB: 5GXJ15), it should in principle be possible to use the same or
structurally very similar inhibitors against both targets. Indeed, compound 1, first developed by us
against DENV protease was also active against ZIKV protease and showed an even lower IC50
value (ZIKV: 1.41 ± 0.16 µM; DENV2: 4.2 ± 0.44 µM27). However, this molecule, based on a 5,6-
ACS Paragon Plus Environment
5

Journal of Medicinal Chemistry
Page 6 of 95
1
2
3
4
5
6
7
8
9
dihydroxybenzo[d]thiazole-2-amine scaffold linked to a diarylether moiety via an amide bond,
suffers from poor water solubility and high lipophilicity. These issues significantly hamper not
only the pharmacological use of this compound, but also its application in co-crystallization or as
a probe in e.g. NMR studies. Nonetheless, compound 1 is interesting because it displays a non-
competitive inhibition against both DENV2 and ZIKV protease. It thus presumably interacts with
the same, previously proposed, putative allosteric site. In this study, we therefore set out to
systematically vary compound 1 for improved protease affinity and hydrophilicity. Since we
previously identified the dihydroxy-substituted benzothiazole unit as an important structural
feature, this fragment was kept constant.27
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The ortho-substituted aromatic ring system in the center of 1 was exchanged against a proline
residue to enhance solubility. Additionally, this opened the possibilities for further derivatization
via facile amine coupling reactions to sulfonamides and carboxylic acid amides (Figure 1).
CHEMISTRY The benzothiazole building block 2-amino-5,6-dimethoxybenzo[d]thiazole 2 was
synthesized as reported previously.27 TBTU mediated amide coupling with either (S)- or (R)-boc-
protected proline resulted in the benzothiazole amides 3 which were deprotected at the amine group
with HCl in dioxane yielding the amines 4. These were coupled with benzoyl chloride to obtain
amides 5 which contain two methoxy groups at the benzothiazole fragment. The inhibitors 6 with
two free hydroxyl groups at the benzothiazole fragment were obtained by reaction with BBr3
(Scheme 1).
ACS Paragon Plus Environment
6

Page 7 of 95
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
O
O
d
O
O
N
N


HO
HO
O
O
S
N
S
N
NH
NH
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
6
5
c
H2
N
Cl
O
O
NH
O

O
O
a
b
S
N
O
S
N
O
N
NH2

O
O
S
O
NH
N
2
3
4
f
e
R2
R1
O
S
O
S
H2
N
O
O
Br
f
O

O
O
N
N
HO
HO


S
HO
HO
O
O
S
S
N
NH
NH
NH
N
N
7
15
16
17 20
-
-
R1 =
R2 =
- 4-C6H4-CH3 71
- C6H5
- 4-C6H4-F
- 4-C6H4-Cl
- 4-C6H4-OCH3 111
- C6H4-NO2
- 4-C6H4-C6H5
- 4-C6H4-NO2
15
- CH3 171
- 4-C10H6-NHAc
- NO2
- OCH
- Cl
81
91
101
121
131
14
181
3 191
201
Scheme 1. Synthesis of proline-based inhibitors 6, 7 – 15 and 17 – 20. Reagents and conditions:
(a) boc-(S)- or -(R)-Pro, TBTU, HOBt, DIPEA; EtOAc, r.t., 97 – 99%; (b) HCl, dioxane, r.t. 91 –
94%; (c) benzoyl chloride, pyridine, r.t., acetonitrile, 90 – 96%; (d) BBr3, DCM, –78 °C – r.t., 13
– 60%. (e) BBr3, DCM, –78 °C – r.t., 86 – 97%; (f) R1SO2Cl, pyridine, acetonitrile, r.t. 39 – 99%.
The sulfonamide derivatives 7 – 15 were obtained by deprotection of either (S)- or (R)-3 with
BBr3 yielding concomitantly the free amino and the free hydroxyl groups (compounds 16), and by
ACS Paragon Plus Environment
7

Journal of Medicinal Chemistry
Page 8 of 95
1
2
3
4
5
6
7
8
9
subsequent reaction with various aromatic sulfonyl chlorides (Scheme 1). The different sulfonyl
moieties were chosen to cover a broad range of compounds with polar and unpolar residues.The
respective sulfonylchlorides were either commercially available or prepared by aromatic
substitution with chlorosulfonic acid.29–31 The dimethoxy-substituted compounds with
sulfonamide structure (17 – 20) were synthesized by deprotection of the boc-protected proline
fragment with HCl in dioxane and reaction with the respective aromatic sulfonyl chlorides
(Scheme 1). In order to check if the amide coupling with boc-proline or one of the deprotection
steps led to racemization, the enantiomeric purity, expressed as enantiomeric ratio (er), was
determined exemplarily for (R/S)-7 and (R/S)-18 by HPLC using a chiral column. For (R)- and (S)-
18, no racemization was detected, while for (R)- and (S)-7, in both cases, a ratio of 93:7 of the
enantiomers was determined (Figure SI1), indicating that not the proline coupling step, but the
subsequent deprotection reaction may lead to partial racemization. The nitro-substituted
derivatives (R)- and (S)-21 and (R)-22 were synthesized starting from the nitro-substituted
benzothiazole 23 by a TBTU mediated reaction with boc-protected proline to 24, subsequent
deprotection with HCl in dioxane yielding 25, and final coupling with benzoyl chloride and tosyl
chloride, respectively, to yield the benzamide and the sulfonamide derivatives (Scheme 2).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment
8

Page 9 of 95
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
O
O
O2N
a
b
S
N
O
N
NH2

O2N
S
N
NH
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
23
24
O
O
N

O2N
S
NH
21
N
c
H2
N
Cl
O
NH

O2N
d
S
N
O
O
S
N
O
O2N
S
N
NH
25
22
Scheme 2. Synthesis of inhibitors with 7-nitrobenzo[d]thiazole unit 21 and 22. Reagents and
conditions: (a) boc-(S)- or -(R)-Pro, TBTU; HOBt, DIPEA, DMF, r.t., 59 – 93%; (b) HCl, dioxane,
r.t., 99%; (c) benzoyl chloride, pyridine, acetonitrile, r.t., 68 – 79%, (d) TosCl, pyridine,
acetonitrile, 88%.
(RS)-26 was synthesized as a racemic mixture starting from nicotinic acid, which was reduced
by H2 with PtO2-catalysis, followed by boc-protection to yield boc-protected nipecotic acid (RS)-
27. After coupling with 2 to yield (RS)-28 and deprotection with BBr3 to yield (RS)-29, (RS)-26
was obtained by conversion with tosyl chloride (Scheme 3).
ACS Paragon Plus Environment
9

Journal of Medicinal Chemistry
Page 10 of 95
1
2
3
4
5
6
7
8
9
O
OH
Rac
O
N
O
Rac
O
a
b
c
O
OH
O
S
N
N
O
NH
N
O
O
27
28
(RS)-
(RS)-
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
O O
S
N
O
NH HBr
d
HO
HO
O
S
N
HO
HO
Rac
NH
S
N
NH
Rac
29
(RS)-
26
(RS)-
Scheme 3. Synthesis of inhibitor (RS)-26. Reagents and conditions: (a) 1. PtO2, H2, HCl, r.t. 2.
boc2O, NaOH, Dioxan, H2O, r.t., 17%; (b) 2-amino-5,6-dimethoxybenzo[d]thiazole, TBTU,
DIPEA, EtOAc, r.t., 84%; (c) BBr3, DCM, – 78 °C – r.t., 99%; (d) TosCl, pyridine, r.t., 27%.
The syntheses of the inhibitors (R)-30, (S)-30, 31 and 32 were carried out similarly, starting from
the commercially available amino acids or their boc-protected derivatives (Figure 2).
ACS Paragon Plus Environment
10

Page 11 of 95
Journal of Medicinal Chemistry
1
2
3
4
5
6
O
O
7
O S
O S
8
9
O
N
O
N
HO
HO
HO
HO
S
S
N
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
NH
30
NH
N
30
(S)-
(R)-
O
O
O
O
HO
HO
HO
HO
O O
N S
S
N
S
NH HN S
NH
N
31
32
Figure 2. Overview of synthesized inhibitors derived from different amino acids.
Derivative 33 was obtained by coupling of 2 with cyclohexanecarbonyl chloride to 34, followed
by deprotection of the methoxy groups with BBr3 (Scheme 4).
O
O
O
O
a
b
S
N
O
O
HO
HO
S
N
S
N
NH2
NH
NH
2
34
33
Scheme 4. Synthesis of 33. Reagents and conditions: (a) cyclohexanecarbonyl chloride, NEt3,
DCM, r.t., 71%; (b) BBr3, DCM, – 78 °C – r.t., 85%.
Compound 35, which is
a
hybrid molecule consisting of the 2-amino-5,6-
dihydroxybenzo[d]thiazole moiety and the reported competitive inhibitor 36,19 was synthesized as
shown in Scheme 5. First, the amino group of anthranilic acid 37 was boc-protected to yield 38.
ACS Paragon Plus Environment
11

Journal of Medicinal Chemistry
Page 12 of 95
1
2
3
4
5
6
7
8
9
TBTU mediated amide coupling with 2-amino-5,6-dimethoxybenzo[d]thiazole resulted in 39
which was deprotected at the amine and methoxy groups with BBr3 generating amine 40. The
sulfonamide 35 was obtained by reaction with tosyl chloride.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Cl
O
S
NH
N
HN O
S
O
O
36
O
OH
NH2
O
OH
NHBoc
O
a
b
c
O
O
S
N
NH
BocHN
37
38
39
O
HO
HO
O
S
N
d
HO
HO
S
N
NH
NH
H2N
HN O
S
O
40
35
Scheme 5. Synthesis of 35, a hybrid molecule consisting of the 5,6-dihydroxybenzothiazole
moiety and the reported inhibitor 36.19 Reagents and conditions: (a) boc2O, dioxane, water, r.t.,
69%; (b) 2-amino-5,6-dimethoxybenzo[d]thiazole, TBTU, HOBt, DIPEA, DMF, r.t., 97%; (c)
BBr3, DCM, -78 °C – r.t., 86%; (d) TosCl, pyridine, r.t., 41%.
1 and its precursor 41 (Figure 3) were synthesized as reported previously.27
ACS Paragon Plus Environment
12

Page 13 of 95
Journal of Medicinal Chemistry
1
2
3
4
H
N
O
N
5
6
S
O
O
7
8
O
O
9
41
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 3. Structure of 41.
STRUCTURE-ACTIVITY RELATIONSHIP The newly synthesized inhibitor library
consisting of 38 compounds was tested on the purified ZIKV and DENV2 NS2B/NS3 proteases
in enzymatic assays employing the fluorogenic AMC-substituted peptide boc-GRR-7-amino-4-
methylcoumarin (boc-GRR-AMC) (Table 1).
Table 1. Inhibition of ZIKV and DENV2 proteases.a
R1
O
N

R2
R3
S
N
NH
ZIKV
DENV2
Cpd ID
R1
R2
R3 Configuration
IC50 [µM] IC50 [µM] or
or %a
%
a
H
N
O
N
1.41 ± 0.16 4.2 ± 0.4427
S
O
1
HO
HO
OH
(R)-6
(S)-6
OH OH
OH OH
R
S
0.94 ± 0.22
1.55 ± 0.18
33%
28%
ACS Paragon Plus Environment
13

Journal of Medicinal Chemistry
Page 14 of 95
1
2
3
4
5
6
7
8
9
O
C
(R)-21
(S)-21
22
NO2
NO2
NO2
H
H
H
R
S
21.9 ± 0.9
44%
33.9 ± 0.6
41%
R
31%
18%
O
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
S
(R)-7
(S)-7
(R)-8
OH OH
OH OH
OH OH
R
S
1.32 ± 0.10
2.32 ± 0.19
1.86 ± 0.49
19.7 ± 1.5
25.2 ± 3.0
21%
O
O
R
S
(S)-8
OH OH
OH OH
OH OH
S
R
R
1.43 ± 0.29
1.03 ± 0.17
1.40 ± 0.25
27%
17%
34%
O
O
S
F
9
O
O
(R)-10
Cl
O
S
O
(S)-10
(R)-11
OH OH
OH OH
S
0.93 ± 0.06
0.86 ± 0.15
39%
28%
O
S
R
(S)-11
(R)-12
(S)-12
(R)-13
(S)-13
(R)-14
OH OH
OH OH
OH OH
OH OH
OH OH
OH OH
S
R
S
1.32 ± 0.21
36%
O
O
0.32 ± 0.05 10.00 ± 0.98
O2N
S
0.93 ± 0.04
2.47 ± 0.29
3.17 ± 1.58
1.11 ± 0.23
8.58 ± 1.37
10.9 ± 2.3
13.5 ± 1.6
33%
O
O
S
R
S
O
O
R
S
O
(S)-14
(R)-15
(S)-15
OH OH
OH OH
OH OH
S
R
S
1.78 ± 0.52
1.35 ± 0.28
2.95 ± 0.45
25%
18%
29%
O2N
O
S
O
HN
O
aIC50 values are indicated as means ± standard deviation from three independent measurements,
each performed in duplicates. %, Percentaged inhibition was determined at 20 µM inhibitor
concentration.
ACS Paragon Plus Environment
14

Page 15 of 95
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
All proline-based inhibitors displayed weaker activity against the DENV protease compared to
the reference compound 1. This holds true for the nitro-substituted inhibitors 21-22, as well as for
the dihydroxy-substituted compounds. In contrast, with the exception of the aforementioned nitro-
substituted compounds, similar or even better inhibition was observed against the ZIKV protease
(IC50 values in the range of 0.32 – 3.17 µM), demonstrating a certain selectivity for this protease.
Also, the introduction of the sulfonamide group ((R)-8, (S)-8) instead of the benzamide moiety
((R)-6, (S)-6) yielded similar inhibition of the ZIKV protease. Even if some of the (R)-configured
enantiomers (e.g. (R)-6 and (R)-11) show lower IC50 values compared to their (S)-enantiomers, the
differences between the values are not significant in light of the obtained standard deviations. The
reduced activities of the nitro-substituted compounds (R)-21, (S)-21 and 22 with a 7-
nitrobenzo[d]thiazole moiety are consistent with our previous results.27 Interestingly, a trend was
observed regarding the hydrophobicity of the inhibitors: While the large, hydrophobic biphenyl or
naphthyl groups of inhibitors (R)-13, (S)-13, (R)-15 and (S)-15 led to IC50 values in the low
micromolar range, the more hydrophilic ones, such as (R/S)-10, (R/S)-11 and (R/S)-12, showed
submicromolar IC50 values.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The lowest IC50 value was found for the 4-nitrophenyl substituted sulfonamide (R)-12. However,
for further development, we did not select this nitro compound, but the tosyl-substituted
derivatives (R)- and (S)-7 as new lead structures due to several reasons: Aromatic nitro groups are
considered problematic in drug design campaigns due to possible mutagenicity and toxicity.32,33
Furthermore, nitro-substituted compounds may lead to artefacts in fluorometric assays due to
quenching effects which have to be accounted for by determination of correction factors.34 The
facile synthetic availability of the starting material in case of the tosyl derivative in comparison to
the similarly active p-chloro or p-methoxy analogues (R)- and (S)-10 and (R)- and (S)-11, and the
ACS Paragon Plus Environment
15

Journal of Medicinal Chemistry
Page 16 of 95
1
2
3
4
5
6
7
8
9
better inhibition of the DENV protease by the tosyl compounds led to the choice to select (R)- and
(S)-7 as new lead structures, from which further compounds were derived (Table 2).
Exemplarily for different compound structures, non-competitive inhibition of the ZIKV protease
was confirmed by determining the IC50 values in the presence of different substrate concentrations
for 1 as the first lead compound, (R)-7 as the new lead compound, (R)-21 as one of the compounds
with a nitro group and the hybrid molecule 35 (Figure SI2). The inhibition was found to be
independent of the substrate concentration, demonstrating a non-competitive inhibition with
respect to the substrate, in all cases.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The KD values of the tosyl derivatives’ ((R)-7 and (S)-7) interaction with the ZIKV protease
were also determined by microscale thermophoresis (MST) to be 0.49 ((R)-7) and 0.24 µM ((S)-
7), respectively. IC50 values are a measure of the the inhibitory activities of the compounds,
determined by a functional assay which measures the residual hydrolysis of a substrate in presence
of an inhibitor. KD values, on the other hand, are derived from the biophysical binding events
between inhibitor and enzyme alone, in absence of a substrate. The slightly lower KD values in
comparison to the IC50 values thus show that the inhibitor binding does not lead to a decrease of
the enzymatic activity to the same degree.
Since, as discussed above, the different substituents on the sulfonamide structure did not have a
substantial impact on the affinity, in the next step, the sulfonamide group was kept constant, while
the proline moiety was exchanged against other amino acids.
Inhibitors (R)-30 and (S)-30, containing a pipecolinic acid instead of the proline moiety, showed
IC50 values of 0.58 and 0.51 µM on the ZIKV NS2B/NS3 protease (Table 2). Again, the
configuration of the amino acid did not influence the inhibition.
ACS Paragon Plus Environment
16

Page 17 of 95
Journal of Medicinal Chemistry
1
2
3
4
Table 2. Inhibition of ZIKV and DENV2 proteases by inhibitors based on the exchange of the
5
6
proline structure.a
7
8
9
ZIKV
DENV2
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Cpd ID
Structure
IC50 [µM] or
IC50 [µM] or
a
a
%
%
O
O S
O
N
(R)
(R)-30
0.58 ± 0.05
29 ± 4
HO
HO
S
N
NH
O
O S
O
N
(S)
(S)-30
0.51 ± 0.44
4.55 ± 0.92
13 ± 5
HO
HO
S
N
NH
O O
S
N
O
(RS)-26
34%
HO
HO
S
N
Rac
NH
O
O
HO
HO
N S
S
32
33
5.95 ± 0.65
36%
41%
20%
NH
O
N
O
HO
HO
S
NH
N
O
HO
HO
O
NH HN S
S
N
31
0.81 ± 036
23± 4
O
ACS Paragon Plus Environment
17

Journal of Medicinal Chemistry
Page 18 of 95
1
2
3
4
5
6
7
8
9
O
HO
HO
S
N
NH
HN O
S
35
6.43 ± 2.76
15.3 ± 2.3
O
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
aIC50 values are indicated as means ± standard deviation from three independent measurements,
each performed in duplicates. %, Percentaged inhibition was determined at 20 µM inhibitor
concentration.
Changing the position of the N-atom of the cyclic amino acid, and thus the conformation of the
molecule, yielded inhibitors (RS)-26 and 32. These compounds showed 10 fold higher IC50 values
of 4.55 and 5.95 µM, respectively, compared to the pipecolinic acid derivatives (R)-30 and (S)-30.
On the other hand, the exchange of the ring system to the open, more flexible glycine-based
inhibitor 31 led to an IC50 value of 0.81 µM. These results suggest that a defined inhibitor
conformation which is either fixed by the rigid pipecolinic acid or which can at least partially be
adopted by the flexible glycine linker might be necessary for good inhibition potency. The
relevance of the sulfonamide part is apparent since compound 33 which lacks this moiety, is only
weakly active.
In a recent study, Lee and coworkers reported the competitive inhibition of ZIKV protease using
non-peptidic active-site directed inhibitors (36) (Scheme 5).19 To evaluate which structural
features are necessary for competitive or non-competitive inhibition, we synthesized a hybrid
compound (35), consisting of the 2-amino-5,6-dihydroxybenzo[d]thiazole moiety and the reported
competitive inhibitor 36. 35 displayed an IC50 value of 6.43 ± 2.76 µM at a substrate concentration
of 100 µM. To test for competitive versus non-competitive inhibition, the IC50 values were
determined at different substrate concentrations (50, 75, 100, 150, 200 µM) according to Dixon.35
Since they were found to be independent of the substrate concentration (5.36 ± 1.40 µM), this
strongly suggests a non-competitive binding mode with respect to the substrate.
ACS Paragon Plus Environment
18

Page 19 of 95
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
The different inhibition properties of (35) and (36) showing non-competitive and competitive
behavior, respectively, demonstrated that the inhibition mode is mainly influenced by the
substitution pattern of the benzothiazole moiety. This further highlights the importance of the
hydroxy groups for non-competitive binding.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Similarly to previously reported compounds,27 the dimethoxy-substituted derivatives 41, 17 - 20
were shown to be inactive at 20 µM.
In summary, our results with the dimethoxy-substituted compounds, the hybrid compound 35
and the nitro-substituted compounds 21, 22 highlight that the dihydroxy substitution pattern at the
benzothiazole fragment is the primary determinant for accomplishing succesful non-competitive
inhibition.
To analyze the hydrophilicity of the new compounds, their SlogP values were calculated and
compared to their retention times, measured with RP-HPLC which showed a good correlation (see
SI part ).36 Using the SlogP and pIC50 values for inhibition of the ZIKV protease, lipophilic ligand
efficiency (LLE) plots were prepared (Figure 4). These data confirmed the significant
improvement of the new inhibitors compared to 1, with several compounds exhibiting LLE values
> 3 or even > 4 ((R)-12).37
ACS Paragon Plus Environment
19

Journal of Medicinal Chemistry
Page 20 of 95
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 4. Lipophilic ligand efficiency (LLE) plots with the SlogP and pIC50 values for inhibition
of ZIKV protease for all compounds with catechol and proline substructure 1, 6 – 15, 26, 30, 31,
32, 35.
PROBING THE PUTATIVE ALLOSTERIC DRUG BINDING SITE BY SITE-DIRECTED
MUTAGENESIS In previous docking studies with inhibitor 1, the hydroxyl group of the
naphthalene moiety was proposed to interact by hydrogen bonding with Thr120 of a putative
allosteric pocket of the DENV protease. Similarly, the two hydroxyl groups of the benzothiazole
moiety were proposed to interact with Asn152 and Lys73 in the same pocket, respectivley.27,38–40
To assess wether this inhibitor indeed binds into this putative allosteric pocket and to probe which
residues of this binding pocket are relevant for interactions, several amino acids (T118, T120,
T122, A164, A166) were mutated individually to cysteine in the DENV protease to be able to
ACS Paragon Plus Environment
20

Page 21 of 95
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
modify the protease at single, specific sites with maleimides. The DENV protease was chosen
because it does not contain any native cysteines. In contrast, the wildtype ZIKV protease inherently
contains the Cys residues Cys80 and Cys143. A mutation of the ZIKV protease’s cysteines to
serines is principally also possible, but affects the catalytic protease activity and thus possibly the
overall structure.13 Due to the overall high similarity of DENV and ZIKV NS2B/NS3, the DENV
protease was thus deemed a suitable target for mutation and modification studies. Importantly, the
inhibitor used for these studies with the maleimide-modified Cys mutants, namely compound 1,
inhibits both proteases in the low micromolar range.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Thr120 was chosen for the mutation studies because of the interactions with the inhibitor proposed
by docking studies. In some lower ranked poses, interactions with Thr118 were also proposed.
Thr122 was chosen to evaluate if the inhibitor performs dynamic interactions with these closely
located residues. Ala164 and Ala166 are located deeper inside the pocket and their Cys-mutations
and subsequent maleimide-modifications were expected to block the binding pocket nearly
completely. Subsequently, the newly introduced cysteine residues were irreversibly modified with
the maleimide compounds N-benzylmaleimide (BMI) and N-ethylmaleimide (EMI). The enzyme
activity of the modified DENV NS2B/NS3 cysteine mutants was determined by monitoring the
hydrolysis of the fluorogenic substrate boc-GRR-7-amino-4-methylcoumarin (boc-GRR-AMC)
(Figure 5A).
ACS Paragon Plus Environment
21

Journal of Medicinal Chemistry
Page 22 of 95
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 5. (A) Percentaged enzymatic activities of wild-type and single-cysteine mutant DENV
proteases modified with N-benzylmaleimide (BMI) or N-ethylmaleimide (EMI). Measurements
were carried out with maleimide concentrations of 20 µM and their activity values were compared
to the respective DMSO controls without maleimide (100%). (B) IC50-values of 1 on wild-type
and mutant DENV proteases in the absence or presence of N-benzylmaleimide (BMI) or N-
ethylmaleimide (EMI). Exact values are indicated above the respective bar. Activity and IC50
values are indicated as means ± standard deviation from three independent measurements, each
performed in duplicates.
For all mutants modified with maleimides except A164C and A166C, no dramatic changes in
activity compared to the WT protein were observed. Intriguingly, the A164C mutant was almost
completely inhibited by both maleimides even in the absence of additional inhibitory drugs, and
the BMI-modified NS2B/NS3 A166C showed a reduction in activity to ~50% of the WT, while
the EMI-modified showed one of ~70%. To demonstrate that the structural integrity of the DENV
NS2B/NS3 A164C and A166C mutants was not compromised, Circular dichroism (CD) spectra
ACS Paragon Plus Environment
22

Page 23 of 95
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
for both enzymes in the presence and absence of BMI or EMI were recorded. The resulting spectra
(Figure SI5) showed, that the respective protein’s secondary structure is retained after maleimide
binding. Overall, the decrease in activity was higher for EMI. While smaller differences between
EMI- and BMI-modified enzymes were observed for T118C, T120C and T122C, the highest
difference was observed for A166C.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
In case of the wild-type enzyme, the IC50 of 1 is the same in absence or presence of a maleimide
as expected for an enzyme that cannot be modified by malemides due to the absence of native
cysteine residues. Furthermore, the IC50 values of the lead structure 1 in the absence and presence
of maleimide modifications were determined for those mutants which still exhibited an activity
> 80% compared to WT in the presence of the maleimide (Figure 5B). While the IC50 values for
inhibition by compound 1 of all constructs in the absence of the maleimide modification were in a
similar range or even slightly lower compared to the WT enzyme, the inhibition of the maleimide
modified mutants was strongly impaired with four- to seven-fold higher IC50 values. These
increased IC50 values of 1 measured on the Cys constructs modified with maleimides strongly
indicate that the interaction of the inhibitor 1 with the allosteric binding pocket is perturbed thus
providing a rational basis for docking studies and further development of inhibitors targeting this
binding site. In the cases of the T120C and T122C mutants, where the modification experiment
was done with both, EMI and BMI, slightly higher IC50 values were observed with BMI. A possible
explanation is the higher steric hindrance by the benzyl group of BMI compared to the smaller
ethyl group of EMI.
A164 and A166 are in the center of the proposed binding pocket (Figure 6).28 The strong
inhibition of the respective cysteine mutant enzymes A164C and A166C, and especially of the
ACS Paragon Plus Environment
23

Journal of Medicinal Chemistry
Page 24 of 95
1
2
3
4
5
6
A164C mutant after their modification by maleimides show that this part of the binding pocket is
highly susceptible to allosteric inhibition.
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 6. The position of the allosteric binding pocket on the DENV NS2B/NS3 protease with
predicted binding modes of 1 in complex with the open conformation of DENV protease with
residues that were mutated to cysteine highlighted in green. (Figure based on the open
conformation, PDB: 2FOM).28 NS2B is shown in red and NS3 in grey. The atoms of compound 1
are coloured according to the CPK color scheme, the carbon atoms are shown in cyan. The amino
acids of the allosteric site, which were mutated to cysteine residues, are indicated in the same way,
while the carbon atoms are shown in green.
ACS Paragon Plus Environment
24

Page 25 of 95
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
DOCKING STUDIES To date, no structure of a flaviviral NS2B/NS3 protease in complex with
a non-competitive, allosteric inhibitor is available. (For a discussion of the recently published
complex structure 6MO0 see below.41–43) Therefore, molecular docking studies were performed to
elucidate the potential binding modes of the inhibitors presented herein. Binding site identification
using the drugability prediction tool LeadIT DoGSite Scorer proposed the same pocket as indicated
by the above described experiments with the Cys mutants.44,45 The allosteric site is in principle
present in both, the open and closed conformation of the protease.38,40 To identify the most
plausible conformation of the enzyme in the complex with the inhibitors and to identify the most
plausible binding mode of the ligands within this complex, docking studies were performed on
both conformations of the DENV and the ZIKV proteases. For the structurally closely related
inhibitors sharing the same 2-amino-5,6-dihydroxybenzo[d]thiazole core scaffold, the key
interactions are expected to be conserved and their binding modes should be similar. However, for
these compounds, the scoring function of the docking procedure should also be able to discriminate
binders from non-binders. Therefore, a visual binding mode inspection and receiver operator
characteristics (ROC) analysis of the compounds reported previously27 and herein was conducted
for the closed and open conformation of the DENV and ZIKV NS2B/NS3 protease, respectively.
The most stable predicted binding mode of ligands carrying the essential 2-amino-5,6-
dihydroxybenzo[d]thiazole scaffold was found for the open conformation of the DENV protease
(PDB code 2FOM). In contrast, poses within the closed DENV (PDB code 3U1I), as well as the
open and closed ZIKV (PDB codes 5LC0, 5GXJ) conformations were less stable (Figure SI6).
The key-interactions between the protein and the ligands are exemplarily described for (R)-7 and
(S)-7 (Figure 7A).
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment
25

Journal of Medicinal Chemistry
Page 26 of 95
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 7. Predicted binding modes of compounds (R)-7 (green carbon atoms) and (S)-7 (cyan
carbon atoms) (A) or (R)-12 (B) in complex with the open conformation of DENV NS2B/NS3
protease (PDB code 2FOM, grey carbon atoms) based on molecular docking. Electrostatic
interactions illustrated as yellow dashes for (R)-7.
Here, the catechol moiety of either compound (R)-7 or (S)-7 forms H-bonds to the backbone
oxygen of Ala164 and Lys74. The hydroxyl group at the 5-position also acts as an H-bond acceptor
for the Asn152 side-chain. For compounds carrying a 5-nitro moiety instead of the catechol
substructure, the nitro-oxygen atoms act as H-bond acceptors for Asn152 side-chain and Leu149
backbone (Figure SI7A). Residues Leu76, Gly148 and Ile165 surround the benzothiazole scaffold
resulting in a hydrophobic contact area forming van-der-Waals interactions. The amide moiety
acts as an H-bond donor to form indirect interactions mediated through a water molecule (Water-
247). This water molecule’s presence is supported by electron density (found in the 2Fo-Fc map,
ACS Paragon Plus Environment
26

Page 27 of 95
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
PDB code 2FOM) in the crystal structure and it is likely to act as an H-bond donor to backbone
oxygen atoms of Gly87, Val146 or Val147 and as an acceptor for a second water molecule (Water-
238) which is likely to be displaced by the inhibitors (Figure SI7B). Depending on the respective
proline configuration, the tosyl moiety is either oriented towards Trp83 ((R)-7) or Ala166 and
Glu88 ((S)-7) to form hydrophobic interactions. Notably, the sulfonamide moiety does not form
extensive H-bonds, only for a few ligands interactions with Lys74, Trp84 or Asn167 side-chains
were predicted (Figure 7). These results indicate that the type of linker (amide or sulfonamide),
as well as the configuration of the proline sub-structure, is of minor importance as long as
hydrophobic moieties can address one of these mainly hydrophobic sub-pockets. This also explains
the overall similar binding affinity of inhibitors containing a hydrophobic moiety attached to their
proline-linker (Table 1). For (R)-12 an additional H-bond to Lys74 side chain was found, which
might be the reason for the slightly increased affinity of this compound. Further, the predicted
binding modes agree with the observations for the mutated proteases, where the introduced
cysteine residues covalently react with EMI or BMI. C164-EMI/BMI and C166-BMI seem to
occupy the allosteric site reducing the enzymatic activity, while C120/122-BMI block or at least
alter the binding site for the inhibitors reducing their IC50-values (Figure 5, Figure 6).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Besides a stable binding mode of the core scaffold, also the discrimination between binders and
non-binders was most evident for the docking against the DENV NS2B/NS3 open conformation
(Figure 8, Table SI3).
ACS Paragon Plus Environment
27

Journal of Medicinal Chemistry
Page 28 of 95
1
2
3
4
5
6
7
8
9
1,0
0,8
0,6
0,4
0,2
0,0
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2FOM
3U1I
5GXJ
5LC0
random
0,0
0,2
0,4
0,6
0,8
1,0
fraction of decoys selected
Figure 8. Receiver operator characteristics (ROC) curves for docking against DENV NS2B/NS3
open (PDB code 2FOM, blue line, ROC area under curve (AUC) 0.93), closed (PDB code 3U1I,
orange line, ROC AUC 0.87), ZIKV NS2B/NS3 open (PDB code 5GXJ, yellow line, ROC AUC
0.81) and closed conformation (PDB code 5LC0, green line, ROC AUC 0.88). The dataset
consisted of 47 binders and 16 non-binders (defined as <10 % inhibition at 20 µM concentration
in enzyme inhibition assay). Docking was performed with LeadIT-2.3.2.46,47
In ROC studies, high values up to a maximum of 1 mean a strong discrimination between binders
and non-binders and low numbers indicate a reduced discrimination. Using this protein as receptor,
a ROC AUC of 0.93 was obtained whereas it was 0.87, 0.81 and 0.88 for the DENV protease
closed, ZIKV protease open and closed conformation, respectively. Notably, the most negative
docking scores were calculated for the binders in complex with the DENV protease open
ACS Paragon Plus Environment
28

Page 29 of 95
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
conformation, as well (Table SI3). Considering the docking scores, the ROC values and the
stability of the binding modes, it seems very likely that the inhibitors prefer to bind to the open
conformation of the protein.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
However, there seem to be slight differences between the proteins that may explain the slightly
higher inhibitor affinities for ZIKV protease and the lower ROC AUC (0.81) for ZIKV NS2B/NS3
open in the docking studies. The crystallographically determined open conformations of the
proteases differ to some extent (Cα RMSD = 0.78 Å, Figure SI11A, PDB-IDs 3U1I for DENV
and 5GXJ for ZIKV), while the closed conformations are highly similar (Cα RMSD = 0.61 Å,
Figure SI11B, PDB-IDs 3U1I for DENV and 5LC0 for ZIKV). This especially affects the
proposed allosteric binding site (Figure SI11C). The C-terminal residues subsequent to the
Asn152, which have been described as a switch between the active and inactive state15, of DENV
NS2B/NS3 form a loop towards the active site, while in ZIKV these residues subsequent to the
corresponding Asn1152 are oriented directly towards residues 1088 – 1092 (Figure SI11B). As
the crystal structure of the ZIKV NS2B/NS3 consists of a non-physiological homo-dimer in the
asymmetric unit, the second monomer may have affected the open conformation in this structure
due to crystal contacts. Indeed, a structure analogous to the one found in DENV NS2B/NS3 is
prevented by ZIKV residues 1029 – 1032 from one monomer by clashing with the residues
subsequent to Asn1152 of the second monomer (Figure SI11C). We therefore hypothesize that
for ZIKV NS2B/NS3 proteases in solution, a conformation highly similar to that observed for
DENV protease should be present and that this is the conformation targeted by our inhibitors.
Assuming that a DENV-analogue open conformation of NS2B/NS3 also exists for the ZIKV
protease, we compared the ZIKV protease allosteric site sequence with the open DENV protease
ACS Paragon Plus Environment
29