Optimization of Substrate-Analogue Furin Inhibitors

Article abstract of DOI:10.1002/cmdc.201700596

The proprotein convertase furin is a potential target for drug design, especially for the inhibition of furin-dependent virus replication. All effective synthetic furin inhibitors identified thus far are multibasic compounds; the highest potency was found for our previously developed inhibitor 4-(guanidinomethyl)phenylacetyl-Arg-Tle-Arg-4-amidinobenzylamide (MI-1148). An initial study in mice revealed a narrow therapeutic range for this tetrabasic compound, while significantly reduced toxicity was observed for some tribasic analogues. This suggests that the toxicity depends at least to some extent on the overall multibasic character of this inhibitor. Therefore, in a first approach, the C-terminal benzamidine of MI-1148 was replaced by less basic P1 residues. Despite decreased potency, a few compounds still inhibit furin in the low nanomolar range, but display negligible efficacy in cells. In a second approach, the P2 arginine was replaced by lysine; compared to MI-1148, this furin inhibitor has slightly decreased potency, but exhibits similar antiviral activity against West Nile and Dengue virus in cell culture and decreased toxicity in mice. These results provide a promising starting point for the development of efficacious and well-tolerated furin inhibitors.

Full text of DOI:10.1002/cmdc.201700596

Accepted Article
Title: Optimization of substrate-analog furin inhibitors
Authors: Teodora Ivanova, Kornelia Hardes, Stephanie Kallis, Sven
O. Dahms, Manuel E. Than, Sebastian Künzel, Eva Böttcher-
Friebertshäuser, Iris Lindberg, Guan-Sheng Jiao, Ralf
Bartenschlager, and Torsten Steinmetzer
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: ChemMedChem 10.1002/cmdc.201700596
Link to VoR: http://dx.doi.org/10.1002/cmdc.201700596
A Journal of
www.chemmedchem.org

10.1002/cmdc.201700596
ChemMedChem
Optimization of substrate-analog furin inhibitors
Teodora Ivanova[a], Kornelia Hardes[a], Stephanie Kallis[b, c], Sven O. Dahms[d, e], Manuel E.
Than[d], Sebastian Künzel[f], Eva Böttcher-Friebertshäuser[g], Iris Lindberg,[h] Guan-Sheng Jiao,[i, j]
Ralf Bartenschlager,[b, c] Torsten Steinmetzer[a]*
[a]Institute of Pharmaceutical Chemistry, Philipps University, Marbacher Weg 6, D-35032
Marburg, Germany
[b]Department of Infectious Diseases, Molecular Virology, Heidelberg University, Im
Neuenheimer Feld 345, 69120 Heidelberg, Germany
[c]German Center for Infection Research, Heidelberg Partner Site, Im Neuenheimer Feld 345,
69120 Heidelberg, Germany
[d]Protein Crystallography Group, Leibniz Institute on Aging - Fritz Lipmann Institute,
Beutenbergstr. 11, 07745 Jena, Germany
[e]Department of Molecular Biology, University of Salzburg, Billrothstrasse 11, A-5020 Salzburg,
Austria
[f]Faculty of Engineering Sciences, Hochschule Ansbach, Residenzstraße 8, D-91522 Ansbach,
Germany
[g]Institute of Virology, Philipps University, Hans-Meerwein-Str. 2, Marburg, Germany
[h]Department of Anatomy and Neurobiology, University of Maryland Medical School, Baltimore,
Maryland 21201
[i]Department of Chemistry, Hawaii Biotech, Inc., Honolulu, Hawaii, United States
[j]MedChem ShortCut LLC, Pearl City, Hawaii, United States
*Corresponding author
steinmetzer@uni-marburg.de, phone: +49-6421-2825900, fax: +49-6421-2825901
Keywords
furin inhibitors, proprotein convertases, enzyme kinetics, antiviral activity, toxicity
1
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
Abstract
The proprotein convertase furin is a potential target for drug design, especially for the inhibition
of furin-dependent virus replication. All effective synthetic furin inhibitors identified thus far are
multibasic compounds; the highest potency was found for our previously developed inhibitor 4-
(guanidinomethyl)-phenylacetyl-Arg-Tle-Arg-4-amidinobenzylamide (MI-1148). A first study in
mice revealed a narrow therapeutic range for this tetrabasic compound, while significantly reduced
toxicity was observed for some tribasic analogs. This suggests that the toxicity depends at least to
some extent on the overall multibasic character of this inhibitor. Therefore, in a first approach, the
C-terminal benzamidine of MI-1148 was replaced by less basic P1 residues. Despite reduced
potency, a few compounds still inhibit furin in the low nanomolar range, but display negligible
efficacy in cells. In a second approach, the P2 arginine was replaced by lysine; compared to
MI-1148, this furin inhibitor has a slightly reduced potency but exhibits a similar antiviral activity
against West Nile and Dengue virus in cell culture and reduced toxicity in mice. These results
provide a promising starting point for the development of efficacious and well-tolerated furin
inhibitors.
2
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
Introduction
The cellular endoproteinase furin contains a subtilisin-like serine protease domain and belongs to
the family of proprotein convertases (PCs). Together with six additional family members (PC1/3,
PC2, PC4, PACE4, PC5/6, and PC7) it is involved in the maturation of numerous precursor
proteins, which are preferentially cleaved at multibasic sequences.[1-2] Furin is a type I
transmembrane protein, becomes autoactivated, travels through different cell compartments in the
secretory pathway and also reaches the cell membrane. After endocytosis, it can be either retrieved
to the trans-Golgi network (TGN) or recycled back to the cell surface. During this trafficking, furin
activates many precursor proteins and therefore, plays a key role in embryogenesis and cellular
homeostasis. In a few cases, furin cleavage also leads to the inactivation of certain proteins such
as endothelial lipase[3] and PCSK9.[4-5]
Furin performs a few unique functions, such as activating the Ac45 subunit of the vacuolar
H+-ATPase in pancreatic β-cells.[6] However, some of furin’s functions are redundant with those
of other related proprotein convertases. In addition to its role in numerous crucial physiological
processes, furin and other PCs are also involved in various pathological conditions like
Alzheimer’s disease, diabetes, arthritis, and some cancer types.[1-2, 7] Moreover, through its ability
to cleave exogenous substrates like bacterial toxins and viral glycoproteins, furin contributes to
the pathogenicity of many bacteria and viruses.[1-2, 7] For instance, the protective antigen of the
anthrax toxin needs to be cleaved by furin on the cell surface; in contrast, Pseudomonas exotoxin
A, shiga toxin and diphtheria toxin are processed by endosomal furin. Examples for viral
glycoproteins activated by furin are the precursors of H5 and H7 hemagglutinins of highly
pathogenic avian influenza virus (HPAIV) strains, of the membrane protein M of flaviviruses, or
of the E protein of alphaviruses including chikungunya virus. The maturation of these proteins is
required for viral fusion-competence. Thus, an efficient furin inhibition might be a suitable
therapeutic strategy for the treatment of numerous diseases and could enable the development of
broad-spectrum antiviral drugs.
Recently, we have developed highly potent substrate-analog furin inhibitors containing a
C-terminal 4-amidinobenzylamide (4-Amba) as shown in Figure 1.[8-10]
3
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
Figure 1: Structures of the known furin inhibitors 1 - 4, containing an N-terminal phenylacetyl
(Phac) or 4-guanidinomethyl-phenylacetyl (4-GMe-Phac) residue and valine or tert-leucine (Tle)
in P3 position.[10]
The binding mode of several of these inhibitors in complex with furin was characterized
by crystal structure determination.[10-12] The 4-Amba group binds in a deep, well-defined S1 pocket
and is involved in numerous contacts with furin, thereby contributing to the picomolar potency of
these analogues. Thus far, the strongest potency against anthrax or diphtheria toxin cytotoxicity
and the inhibition of virus replication after infection of cells by HPAIV of the type H5N1 and
H7N1, canine distemper virus, chikungunya virus, Semliki-Forest virus, Dengue- or West-Nile
virus was found for compound 4 (MI-1148).[10, 13-14]
The 4-Amba residue has also been used for the design of N-terminally elongated furin and
PACE4 inhibitors by the Day group.[15-17] However, increased toxicity was found for these
derivatives compared with analogues containing an Arg-amide or 2,3-dehydroagmatine as P1
groups. For their best compound acetyl-DLeu-(Leu)3-Arg-Val-Lys-4-Amba, which inhibits furin
and PACE4 with Ki values of 9.8 nM and 4.9 nM, respectively, a maximum tolerated dose of ≈10
mg/kg was observed after intravenous treatment of mice.[17]
Although our best compound 4 did not exhibit any noteworthy cytotoxicity up to a
concentration of 50 µM in cell culture tests, a pronounced toxicity in mice was observed. While
an intraperitoneally (i.p.) administered dose of 2.5 mg/kg was tolerated, three of four mice died
within the first hour after treatment with 5 mg/kg. We speculated that this toxicity is caused by the
strongly multibasic character of the molecule and the presence of the benzamidine in the P1
position, as previously suggested for longer peptidic inhibitors.[16] Therefore, this group was
replaced by residues with reduced basicity in this study. Some of these compounds exhibited low
4
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
nanomolar Ki values. Despite this relatively potent furin inhibition, only a weak inhibitory effect
in cells could be observed for these analogs. In spite of its limitations, the benzamidine group has
been used in the design of the direct thrombin inhibitors melagatran and dabigatran, with both
compounds being utilized as benzamidine-prodrugs (ximelagatran and dabigatran-etexilate).
Therefore, as a second approach, we retained the P1 4-Amba residue and attempted to reduce the
overall basicity by replacement of the P2 arginine with lysine. The inhibitor obtained thereby was
only slightly less active in enzyme kinetic measurements and cell culture assays compared to
inhibitor 4. However, as shown below, its toxicity was reduced.
Results
Synthesis
In a first series, the benzamidine was replaced by less basic P1 groups; some of these were
previously used in the development of inhibitors for trypsin-like serine proteases. The structures
of the newly prepared inhibitors 5 – 27 and the calculated pKa values for the acetylated P1 residues
are provided in Tables 1 and 2. For comparison, two strongly basic benzguanidine analogues have
also been incorporated in compounds 7 and 9, which were accessible from their aniline precursors
6 and 8. Most P1 residues were attached to the P5-P2 segments Phac-Arg-Val-Arg or 4-GMe-
Phac-Arg-Val-Arg and provided inhibitors 6, 8, 10 – 17 (Table 1). In a second series containing
the N-terminal 4-GMe-Phac residue, selected P1 groups were combined with the slightly better
Tle in P3 position, as known from inhibitor 4 (Table 2). The non-basic bicyclic benzo[d]isoxazol-
3-amine was derived from the P1 ligand of the factor Xa inhibitor razaxaban[18] and incorporated
in inhibitor 23.
Most inhibitors were prepared by a combination of solid phase and solution synthesis. For
the inhibitors shown in Scheme 1, the P5-P2 segment was synthesized on 2-chlorotrityl chloride
resin and cleaved from the solid support under mild acidic conditions maintaining the side chain
protection. The P1 residue was subsequently coupled in solution. After side chain deprotection and
preparative HPLC, the final inhibitors were obtained as TFA salts.
5
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
Scheme 1: Preparation of inhibitors 5, 10-13, 15-20, 23, and 24 containing neutral (A) or basic
(B) P5 residues on 2-chlorotrityl chloride resin. Reagents and conditions: (a) 1 % TFA in DCM, 3
ꢀ 30 min; (b) (i) 1.5 equiv of the corresponding amine (benzylamine for inhibitor 5, compound
33 for inhibitor 10, compound 34 for inhibitor 11, and compound 37 for inhibitor 12), 1.65 equiv
of PyBOP, 4.5 equiv of 6-Cl-HOBt, DIPEA (pH ~ 9), DMF, 15 min at 0°C, 4 h at rt, (ii)
TFA/TIS/H2O (95:2.5:2.5 v/v/v), 3 h at rt; (c) (i) 20 % piperidine in DMF, (ii) 3.0 equiv of N,N'-
Bis-Boc-1-guanylpyrazole, 4.0 equiv of DIPEA, DMF, overnight, (iii) 1 % TFA in DCM, 3 ꢀ 30
min; (d) (i) 1.5 equiv of the corresponding amine (benzylamine for inhibitor 18, compound 33 for
inhibitors 15 and 20, compound 34 for inhibitor 16, compound 37 for inhibitor 17, compound 32
for inhibitors 13 and 19, compound 42 for inhibitor 23, and compound 44 for inhibitor 24), 1.65
equiv of PyBOP, 4.5 equiv of 6-Cl-HOBt, DIPEA (pH ~ 9), DMF, 15 min at 0°C, 4 h at rt, (ii)
6
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
TFA/TIS/H2O (95:2.5:2.5 v/v/v), 3 h at rt, (iii) only for inhibitors 13 and 19: 0.5 mL of 1 M NaOH
in 3 mL of dioxane/water (2:1 v/v), 2 h at rt.
*) Prepared as described previously.[8]
**) Prepared in analogy to the corresponding N-terminally Boc protected derivative described
previously.[9]
A few compounds were prepared on trityl chloride resin, which was initially loaded with
the Fmoc protected P1 residues. After mild acidic cleavage from resin the free amino groups were
converted into guanidines by N,N'-Bis-Boc-1-guanylpyrazole, followed by final deprotection
(Scheme 2, inhibitors 7 and 9). A few inhibitors were entirely synthesized on resin; following
strong acidic cleavage, compounds 6, 8, and 14 were obtained.
7
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
Scheme 2: Synthesis of inhibitors 6-9 and 14 on trityl chloride resin. The initial resin loading was
performed with the Fmoc derivatives 45 and 46, respectively. Reagents and conditions: (a)
TFA/TIS/H2O (95:2.5:2.5 v/v/v), 3 h at rt; (b) (i) 1 % TFA in DCM, 3 ꢀ 30 min, (ii) 8.0 equiv of
N,N'-Bis-Boc-1-guanylpyrazole, DIPEA (pH ~ 9), DMF, 72 h at rt, (iii) TFA/TIS/H2O (95:2.5:2.5
v/v/v), 3 h at rt.
*) Conversion of the N-terminal amino group into a di-Boc protected guanidine as described in
Scheme 1.
Inhibitors 25 and 26 were also synthesized on trityl chloride resin, which was initially
loaded with 1,3-diaminopropane or 1,4-diaminobutane as described previously[8], followed by
SPPS and strong acidic cleavage from resin including side chain deprotection. Finally, the N- and
C-terminal amino groups were converted into guanidines by 1H-pyrazole-1-carboxamidine · HCl.
Initial attempts to prepare the hydroxyamidine inhibitor 21 by coupling of the unprotected
4-hydroxyamidinobenzylamine to intermediate 30 (Scheme 1), followed by strong acidic side
chain deprotection, resulted in a minor impurity with the amidine inhibitor 4, which was identified
8
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
by mass spectrometry. These species (4 and 21) could not be separated by our standard reversed
phase HLPC method at pH ~1.9 with solvents containing 0.1 % TFA. An analytical separation of
these derivatives was achieved on a standard C18-RP-column by using a linear gradient of pure
acetonitrile (solvent B) in ammonium acetate buffer (pH 4.0, solvent A), the HPLC chromatograms
with this method are provided in the supporting information. Based on detection at 220 nm, the
amount of the amidine side product 4 was determined to be about 1 %. For this reason, inhibitor
21 was prepared by a modified strategy. The protected P5-P2 segment was synthesized by SPPS,
liberated from resin under mild acidic conditions, coupled with 4-cyanobenzylamine · HCl in
solution, and deprotected under strong acidic conditions. Final conversion of the nitrile with
hydroxylamine provided the hydroxyamidine inhibitor 21 lacking any amidine impurity. This step
yielded a minor impurity with the analogous carboxamide, which was isolated and characterized
as inhibitor 22. Moreover, the P2 Lys analogue of compound 4 was prepared from intermediate 31
(4-(Boc2-GMe)-Phac-Arg(Pbf)-Tle-Lys(Boc)-OH), providing inhibitor 27.
9
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
Table 1: Analytical characterization and potency of inhibitors with Val in P3 position in
combination with neutral or basic P5 residues and modified P1 groups.
MS
HPLC[b]
(min)
No.
P1
pKa[a]
R
Ki (n
M
)
IC50 (nM)
[c]
[c]
found
calcd
636.4
[M+H]+
-
H
H
33.0
637.8
652.4
70200 22900
n.d.
283000 56000
96400 21700
5
6
4.36
23.1
24.5
651.4
693.4
10.40
4.27
H
H
348.0[d]
652.4
890 350
n.d.
3160 180
32.0 5.0
7
8
23.9
25.3
651.4
693.4
10.50
H
347.9[d]
950 80
4250 130
9
6.50
6.73
H
H
23.1
22.8
652.4
652.4
327.4[d]
327.4[d]
540 20
950 70
2010 80
10
11
3240 190
8.65
9.28
H
24.1
18.5
693.4
736.5
348.0[d]
737.5
860 40
3340 200
12.9 0.4
12
13
GMe
3.47 0.15
4.27
6.50
GMe
GMe
18.2
17.6
722.5
723.4
362.5[d]
362.8[d]
140 10
550 30
14
15
3.73 0.25
13.0 1.1
10
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
16
6.73
8.65
GMe
GMe
17.1
18.5
723.4
764.5
362.9[d]
383.5[d]
8.32 0.20
n.d.
31.8 3.7
660
2
17
[a] Calculated for the acetylated P1 residues with the ChemAxon’s pKa plugin
(https://chemicalize.com/welcome), e.g., a pKa of 11.48 was calculated for acetyl-4-Amba.
[b] Start HPLC at 1 % solvent B
[c] Data are the mean SD of at least three independent measurements
[d] [M+2H]2+/2
n.d., not determined
11
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
Table 2: Analytical characterization and potency of inhibitors with Tle in P3 position.
MS
HPLC[b]
(min)
[c]
[c]
No.
P1
P2
pKa[a]
Ki (n
M
)
IC50 (nM)
found
calcd
[M+H]+
Arg
Arg
-
28.3
20.3
721.5
750.5
361.9[d]
751.5
n.d.
580 20
18
19
9.28
2.52 0.32
0.80 0.10
n.d.
9.53 0.65
Arg
Arg
Arg
6.50
5.01
-
19.2
20.2
21.8
737.5
779.5
764.5
738.5
780.5
765.5
2.98 0.14
900 80
20
21
22
n.d.
3970 380
Arg
Arg
0.88
9.46
23.8
19.6
777.5
781.5
390.2[d]
392.0[d]
450 10
1800 230
0.56 0.02
23
24
0.0223 0.0057
H
H
N
N
NH2
NH
Arg
Arg
12.41
12.33
19.1
19.5
730.5
744.5
366.3[d]
373.3[d]
0.28 0.01
0.22 0.02
0.88 0.04
0.76 0.06
25
26
Lys
11.48
19.3
735.5
368.7[d]
0.0085 0.0049
0.63 0.19
27
[a] Calculated for the acetylated P1 residues with the ChemAxon’s pKa plugin
(https://chemicalize.com/welcome).
[b] Start HPLC at 1 % solvent B
[c] Data are the mean SD of at least three independent measurements
[d] [M+2H]2+/2
12
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
Furin inhibition
The inhibitory strength against furin was tested in a fluorometric assay.[9-10] The potency of
compounds with Ki values > 0.1 n
was determined from Dixon plots[19] (Figure 2A) and in
M
parallel by fitting the steady-state rates as a function of the inhibitor and substrate concentrations
using the equation for a classical reversible competitive inhibition (Figure 2B). As expected, both
methods provided identical results. For tight-binding inhibitors (Ki < 0.1 nM) the data were fitted
to the Morrison equation (Figures 3E and 3F; equation provided in the supporting information).[10,
20-21]
A few compounds (6, 8, 17, 18, 21, and 22) showed a deviation from the normal kinetic
behavior, exhibiting non-linear Dixon plots (Figure 2C), and the measured data could not be fitted
to the equation of a classical reversible competitive inhibition (Figure 2D). Therefore, only IC50
values could be determined. For better comparison, IC50 values were also determined for all other
inhibitors (Tables 1 and 2, Figure 3). As expected, similar IC50 values were found for the tight-
binding inhibitors, despite significant differences in their Ki values. IC50 values are not well suited
to discriminate between potencies of tight binding inhibitors, because the data do not fit well to
the standard IC50 equation (see supporting information).[21]
Low nanomolar inhibition constants were determined for the agmatine analogue 26 (Ki =
0.22 n
(Ki = 0.80 n
classical conditions. An enhanced inhibitory effect was found for the tight-binding inhibitor 24 (Ki
= 22.3 p ) possessing an ortho-fluoro-substituted benzamidine in P1 position. This modification
M), the noragmatine derivative 25 (Ki = 0.28 n
M), the para-aminopyridine compound 20
M
), and the benzylamine 19 (Ki = 2.52 n ). Their Ki values could be determined under
M
M
reduces the calculated pKa of the benzamidine group to ~9.5 compared to 11.5 for 4-Amba.
However, compound 24 is ~4-fold less potent than inhibitor 4 containing the non-substituted P1
benzamidine. A very strong inhibition was found for the new 4-Amba-derived inhibitor 27
containing lysine in P2 position (Ki = 8.5 pM), although the fitted curve showed a minor deviation
from the measured data in the elbow region (Figure 2F).
13
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
B
A
2.5
20
15
10
5
2.0
1.5
1.0
0.5
0
0.0
0
5
10
15
20
25
30
0
5
10
15
20
25
30
[inhibitor 23] (µM)
[inhibitor 23] (µM)
18
C
D
25
20
15
10
5
16
14
12
10
8
6
4
2
0
0
0.00
0.05
0.10
0.15
0.20
0.00
0.05
0.10
0.15
0.20
[inhibitor 8] (µM)
[inhibitor 8] (µM)
E
F
16
14
12
10
8
14
12
10
8
6
6
4
4
2
2
0
0
0.000
0.005
0.010
0.015
0.020
0.025
0.030
0.000
0.005
0.010
0.015
0.020
0.025
0.030
[inhibitor 24] (µM)
[inhibitor 27] (µM)
Figure 2: Inhibition of furin (0.95 nM) by selected inhibitors. Dixon plot (A) and fitting to the
equation of classical competitive inhibition (B) for inhibitor 23 in presence of substrate Phac-Arg-
14
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
Val-Arg-Arg-AMC: 5 (ꢁ), 20 (), and 50 (p) µ
M. The analog plots for inhibitor 8 (C, D) reveal
a strong deviation from a normal competitive reversible binding mode (the data in panel C were
simply connected for better visualization). Fits for the tight-binding inhibitors 24 (E) and 27 (F)
using Morrison’s equation in the presence of a single substrate concentration (ꢂ, 12.5 µM).
100
80
60
40
20
0
B
100
80
60
40
20
0
A
1E-4
1E-3
[inhibitor] (µM)
0.01
1E-4 1E-3 0.01
0.1
1
10
100 1,000
[inhibitor] (µM)
Figure 3: IC50 plots for the furin (0.95 n
M) inhibition by selected inhibitors in the presence of
12.5 µ Phac-Arg-Val-Arg-Arg-AMC. (A) A limited fitting accuracy was obtained for the IC50
M
curves of the tight-binding inhibitors 4 (ꢂ), 24 (p, dashed line), and 27 (). (B) IC50 plots of less
potent inhibitors 2 (ꢁ), 20 (ꢂ), 8 (ꢃ), 14 (p), 21 (), 22 (r), and 5 (ꢄ).
Cell culture assays
The most potent inhibitors (Ki values < 10 nM) were examined for their activity in cell culture. The
tested inhibitors were well accepted; at concentrations ≤ 50 µ
M cell viability was not diminished
(see Figure S4).
In an initial assay, selected inhibitors were tested in HEK293 cells, which produce
endogenous furin. The cells were transfected with a plasmid encoding an influenza A virus
hemagglutinin (A/FPV/Rostock/1934; H7N1) containing the sequence PSKKRKKR↓GLFG,
where ↓ represents the scissile peptide bond. The hemagglutinin precursor HA0 is cleaved by furin
15
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
into HA1 and HA2. Results obtained with this screening assay usually correlate well with the
antiviral activity of inhibitors in cell-based infection studies as shown previously for H7N1 and
H5N1 HPAIV[9, 22], canine distemper virus[10], Semliki Forest virus, chikungunya virus[13], West-
Nile virus, and Dengue-2 virus[14]. In the current study, strong differences in potency were found
in this assay, as shown by the Western blot in Figure 4. A negligible activity barely distinguishable
from the control (w/o) was found for inhibitors 13 and 19 containing the less basic P1-benzylamine
group (pKa ~ 9.3) and compounds 15, 16, and 20 possessing aminopyridines (pKa ~ 6.5). Poor
inhibition of H7 cleavage was also observed for the strongly basic noragmatine and agmatine
derivatives 25 and 26. A notable effect was observed for the fluoro-substituted benzamidine 24
(pKa ~ 9.5), which was even more pronounced for the reference inhibitor 4 and the new P2-Lys
analogue 27. Despite some differences in the Ki values of compounds 4 and 27 for furin, a similar
activity was found in this cell-based screening test.
13 15 16 19 20 24 25 26 45 4
kDa
75
HA0
48
35
HA1
25
HA2
48
Actin
Figure 4: Effect of selected furin inhibitors at 25 µ
M
on H7 cleavage in HEK293 cells transfected
with a pCAGGS-H7 plasmid compared to controls without inhibitor (w/o) and without transfection
(mock). Cell lysates were analysed by SDS-PAGE and Western blot at 48 h post transfection using
HA- and actin-specific antibodies, respectively. Identical results were observed in an additional
independent experiment.
16
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
Furthermore, the most potent derivatives were tested for the inhibition of West Nile (WNV)
and Dengue-2 virus (DENV-2) replication in human liver cells (cell line Huh7). WNV and DENV
are mosquito-borne single stranded RNA viruses that belong to the Flavivirus genus and cause
infections with mild to severe symptoms in humans worldwide.[23] Flaviviruses are enveloped
particles containing the surface proteins E and M. While protein E mediates both host cell binding
and membrane fusion[24], the proform of M designated prM forms a heterodimer with E in the
endoplasmic reticulum (ER), where virus assembly takes place.[25] There prM covers the fusion
domain of E and is subsequently cleaved by furin in the acidic environment of the TGN, with M
remaining anchored in the virion membrane and associated with the pr segment. The assembled
virus particle is subsequently transported to the cell surface where pr is released at neutral pH, and
the virus becomes fusion-competent.[26] After endocytosis of the mature virus particle, the
membrane fusion is accomplished by protein E. It has been shown that the prM cleavage - and thus
the membrane fusion and the infectivity of flaviviruses – are blocked by furin inhibitors and in
furin-deficient cells.[27] Therefore, furin activity is crucial for flavivirus infectivity.
A significant effect was observed for inhibitors 24 and 27 in a preliminary screen. A slight
concentration-dependent inhibition was observed for the aminopyridine analogue 20, which
provided ~10-fold reduced DENV-2 titers at 50 µ
compounds 14 17 19, and 23 (data not shown). Therefore, only the most potent inhibitors 24 and
27 were selected for a further dose-response assay, using the recently investigated compound
and the nucleotide analogue ribavirin as reference inhibitors (Figure 5).[14] Also in this assay, the
strongest antiviral efficacy was found for the reference compound , which is more potent than
ribavirin at comparable concentrations. A slightly reduced potency was determined for the P2 Lys
analogue 27; however, its efficacy is in the same range as found for compound . Treatment with
50 µ of inhibitor 27 reduced WNV and DENV replication more than 1000-fold, as determined
M, whereas a negligible inhibition was found for
,
,
4
4
4
M
by plaque assay. A lower but still significant concentration-dependent inhibition was observed for
the ortho-fluoro substituted benzamidine 24 (Figure 5).
17
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
7
6
5
4
3
2
1
0
B
A
8
7
6
5
4
3
2
1
0
l
0
0
0
0
5
5
0
5
5
5
2
5
0
5
5
5
3
5
2
0
5
5
.
5
3
5
2
o
.
.
.
r
0
1
5
2
5
2
2
1
5
2
2
5
2
2
.
t
2
1
2
.
2
.
2
2
1
1
.
n
1
6
.
1
6
.
1
6
o
c
3
4 (µM)
ribavirin (µM)
24 (µM)
27 (µM)
4 (µM)
ribavirin (µM)
24 (µM)
27 (µM)
Β
Figure 5: Infection of Huh-7 cells with WNV (A) and DENV-2 ( ) in the presence of inhibitor
concentrations specified on the bottom. Cell culture supernatants were harvested 48 h post
infection and virus titers standard deviation (n = 3) were determined by plaque assay in VeroE6
cells. The effects of the reference compounds
methods used, have been described recently.[14]
4 and ribavirin, as well as details of the assay
Toxicity study
For toxicity studies, groups of 4 adult mice (two male and two female) were intraperitoneally
treated with the inhibitors (converted into their physiologically more acceptable HCl salts[28]
)
starting with the lowest dose on day 1. If all mice survived 24 hours after treatment, the next higher
dose was applied on day 2, followed by a further increase of the dose on day 3, etc. Control
treatments with 0.9 % NaCl under identical conditions were well accepted at repeated injections
by all mice (n=4). Different toxicities were observed for the various inhibitors tested in this
manner. Treatment with 2.5 mg/kg of the tetrabasic lead structure
4 was tolerated, whereas 3 of 4
mice died at the next higher dose of 5 mg/kg (Table 3). Previous work describing the toxicity of
benzamidine derivatives[16-17, 29] and of strongly multibasic peptides[30] suggested a test of
compound
2
lacking the guanidinomethyl group on the P5 phenyl ring and the benzylamide
, all
compound 18 without the amidine substitution at the P1 residue. In the case of compound
2
mice tolerated a 4-fold higher dose of 10 mg/kg, whereas 3 of 4 mice died at 15 mg/kg. An even
18
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
lower toxicity was observed for compound 18; all mice tolerated 10 mg/kg, and only 1 of 4 mice
died at 15 mg/kg. These results suggest that the strongly basic substituents at the P5 and P1
positions, which are both very important for furin inhibition, significantly contribute to in vivo
toxicity. Additionally, the P2 Lys inhibitor 27 was tested because of its strong antiviral potency
and slightly reduced basicity. The mice survived 5 mg/kg, but a dose of 10 mg/kg was lethal for
all animals. This suggests that the toxicity of inhibitor 27 is lower than observed for compound
4,
but higher compared to inhibitors and 18. It should be noted that all cases of death occurred
2
within the first hour after administration without specific symptoms, indicating rather acute than
chronic toxicity. However, the observation period of 24 hours is too short to allow statements about
long-term effects that might occur with mice that were given sub-lethal doses.
Table 3: Toxicity study in mice. For each group four mice (two female and two male) were
intraperitoneally treated with the indicated compounds and concentrations.
Tolerated dose Number of deaths at
No.
(mg/kg)
next higher dose
4 of 4 at 5 mg/kg
3 of 4 at 15 mg/kg
1 of 4 at 15 mg/kg
4 of 4 at 10 mg/kg
2.5
10
10
5
4
2
18
27
Discussion
The toxicity study in mice revealed limited safety of compound 4, the most efficient picomolar
furin inhibitor known so far. Therefore, the question arises, whether an effective furin inhibitor
could ever represent an acceptable therapeutic compound. In vivo toxicity is significantly reduced
but still present after deleting the strongly basic amidine substitution in P1 or the guanidine group
in P5 position, with both inhibitors being tolerated at an i.p. dose of 10 mg/kg. The comparable
19
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
toxicity of both compounds and their tremendous differences in inhibitory potency (Ki = 0.17 n
M
for
2
and IC50 = 580 n
M
for 18) strongly suggest that the observed acute toxic effect is not caused
through a direct furin inhibition and might be rather the result of unidentified off-target effects.
Moreover, in contrast to the embryonic lethality of a furin knock-out mouse model (thought to be
due to the absence of growth factors required during development), no obvious adverse effects
have been observed in a liver-specific transgenic furin knock-out model.[31] This suggests that at
least a partial furin inhibition should be a tolerable therapeutic approach. The results in Table 3
also reveal that the aromatic P1 amidine contributes somewhat more strongly to the overall toxicity
than the P5 guanidine, which confirms previous reports, in which P1 benzamidines and guanidines
have been compared in thrombin[29] and PC inhibitors[16-17]. Moreover, inhibitor
4 contains two
additional guanidines in its P2 and P4 side chains and the reduced toxicity of the P2 Lys inhibitor
27 suggests that their presence also contributes to the toxic effects. Whereas the replacement of P2
Arg by Lys is relatively well accepted, a dramatic ~400-fold drop in inhibitory potency occurred
by the same exchange in P4 position or when arginine was changed to the nonbasic citrulline.[32]
Interactions with more than one target have been reported for numerous basic compounds,
while less off-target effects are normally observed for neutral or zwitterionic agents.[33-35] Studies
with some benzamidines and polycationic peptides revealed toxic side effects in mammals,
especially at high dosages. These unspecific reactions range from circulatory collapse caused by
an acute drop in blood pressure[29] to cardiac arrhythmia[36] and pseudo-allergic reactions[37]
.
Although these reactions have been described in detail and repeatedly verified, the underlying
physiology still remains largely unclear.
Benzamidine-induced hypotension and respiratory disturbance have been observed in
studies with thrombin inhibitors in mice and rats, especially when bis-benzamidines have been
tested.[29, 38] In contrast, despite some side effects, the bisbenzamidine pentamidin is approved for
the parenteral treatment of numerous infectious diseases. Its toxicity could be reduced by
inhalative administration, used for the treatment of Pneumocystis jirovecii[39-40]; this could also
represent a suitable administration route for furin inhibitors against respiratory viruses like
HPAIV. Several mechanisms explaining the hypotensive action have been postulated, i.a., the
affinity of the benzamidinium cation for a number of central and peripheral receptors. For instance,
an agonistic action of p ,
-aminobenzamidine on the imidazoline-1 receptor has been reported[41]
20
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
whereas its activation partly mediates a central hypotensive action.[42] Speculation has been offered
as to possible antagonism of bis-amidines and guanidines on alpha-adrenergic receptors.[43] In
addition, a possible adrenergic neuron blockade by guanidines has been discussed.[44] However,
the exact mechanism underlying the hypotensive action of these compounds remains unknown.
Whether our benzamidine-containing compounds would also exhibit hypotensive effects is not
known.
Other known side effects of certain cationic drugs, including benzamidines, are different
forms of arrhythmia due to blockade of the hERG channel.[36, 45] The presence of a positively
charged nitrogen does not necessarily lead to a hERG block, but increases its likelihood since the
basic functionality can be stabilized by cation-π interaction with residue Tyr652 in the potassium
channel. It is uncertain whether a similar hERG channel blockade could be responsible for the
observed toxic effects of our compounds in mice given that little (if any) hERG-like potassium
current has been found in rodent cardiac myocytes, although newer research argues to the
contrary.[46-48]
Moreover, in a non-IgE-dependent manner, polybasic compounds can provoke a direct
mast-cell degranulation and thus allergy-like symptoms like rash, pruritus, but also respiratory
49]
disturbance, cardiac dysrhythmia, hypotension, anaphylaxis, and even death.[37-38,
One
postulated explanation of this phenomenon is an activation of G-proteins and subsequent signal
cascades, which can lead to the release of neurotransmitters (histamine, serotonin), proteolytic
enzymes (tryptase), prostaglandins, leukotrienes, platelet-activating factor (PAF), and cytokines.
Although these reports raise some concern about the safety of benzamidine-containing
drugs, a few have already been approved for human use. In addition to the antimicrobial drug
pentamidin, the thrombin inhibitors dabigatran etexilate and ximelagatran (meanwhile withdrawn)
and the naphthylamidine derivative nafamostat are used as anticoagulants. Notably, a very short
half-life of 8 min is reported for nafamostat, which is used in rare cases for hemodialysis patients
with a tendency to bleed.[50-51] The orally available drugs dabigatran etexilate and ximelagatran are
both applied as amidine prodrugs, which also limits the plasma concentrations of the free
benzamidines. For instance, after oral treatment of 20 mg ximelagatran to healthy volunteers,
plasma levels of 0.2 µmol/L were reached and were sufficient for anticoagulant use. Thus far, we
can only speculate whether a similar prodrug approach, for example with the hydroxyamidino
21
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
inhibitor 21, could reduce toxicity by avoiding high peak levels through slow conversion into the
active drug. The reduction of hydroxyamidines is performed by an enzyme system found in the
liver and other organs consisting of cytochrome b5, a NADH-dependent cytochrome b5 reductase,
and a P450 enzyme[52], but it is not yet known whether these substrate-analog furin inhibitors
would be accepted as substrates.
Our data reveal that the toxicity of compound
4 depends on its specific strongly basic
overall profile required for an efficient furin inhibition and is not simply caused by the presence
of the benzamidine moiety. Notably, treatment of mice with our previously described monobasic
benzamidine-derived uPA inhibitor benzylsulfonyl-D
Ser-Ser-4-Amba[53] (100 mg/kg twice a day
over 12 days) was well tolerated.[54] Thus far, we have not been able to replace the P1 benzamidine
without losing efficacy in cell culture (Figure 4), although select P1 groups still provide furin
inhibitors with low nanomolar potencies in enzyme kinetic assays. If the P1 4-Amba residue is
maintained, toxicity can be only reduced by gradual modification of other inhibitor residues; a first
step in this process is the conversion of Arg to Lys in the P2-position.
For some inhibitors (
enzyme kinetic analysis. The measured rates could not be fitted to the standard equation for a
classical competitive inhibition (e.g., Figure 2D for compound ) and the obtained Dixon plots
6, 8, 17, 18, 21, and 22) an unexpected behavior was observed during
8
revealed rather a parabolic shape (Figure 2C). Originally we had the impression that this tendency
is caused by the weak potency of these analogues, however, normal competitive inhibition was
found for the weakest inhibitor
5 (Ki = 70.2 µM, IC50 = 283 µM) and for the poorly active compound
23 Ki = 450 n ). Since most of these analogues contain weakly and non-basic or sterically more
(
M
demanding P1 residues, which should preclude the strong charge-assisted anchoring in the S1
pocket, it might be also possible that these compounds inhibit furin by a deviating binding mode.
In previous docking studies it was predicted that such inhibitors may also adopt a reverse
conformation when bound to furin.[55] Parabolic Dixon-plots can be interpreted as a situation in
which for some reason at higher inhibitor concentrations a stronger inhibition occurs compared to
a normal competitive inhibition with linear plots. Based on literature reports, this behavior could
be a sign for the presence of an additional allosteric inhibitor binding site. Parabolic Dixon plots
have been described for the mixed-type inhibition of plasma cholinesterase by phenoxazine
dyes[56] and for the inhibition of dipeptidyl peptidase IV by a nonapeptide derived from the
22
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
N-terminus of the TAT protein[57], as well as for the trypsin inhibition by some bis-benzamidine
compounds[58]. In all these cases, mechanisms have been postulated, in which the inhibitors can
occupy two different binding sites in parallel. So far, we have no structural proof that a second
binding site for these substrate-analog inhibitors exists in furin which could explain a similar
enzyme kinetic behavior. Interestingly, in a recent crystal structure analysis of furin in complex
with a tetrabasic 2,5-dideoxystreptamine-based small molecule inhibitor, the parallel binding of
two molecules was found.[59] One molecule spans between the S4 and S2 sites without occupying
the S1 pocket, but makes numerous polar contacts to furin. The second molecule shows a likely
weaker binding to a furin region close above the active site, where only one guanidine interacts
with the side chains of residues Asp228 and Asn192 and the carbonyl oxygen of Asp191. Only for
reference, we have tested this same molecule in our kinetic assay, but could not detect parabolic
Dixon plots. Instead, a normal reversible competitive inhibition was observed and a Ki-value of 12
n
M
was determined under our conditions, which is comparable to the value described in literature
(
Ki = 46 n
M
for compound 1n[60]).
Rather than interpreting the data in the Dixon plot of Figure 2C as a parabolic relationship,
these data can be also considered as biphasic with two linear correlations, indicating a switch
between two states at a concentration of roughly 50 n in the case of inhibitor . Crystal structure
M
8
determinations have revealed that unliganded furin exists in the so-called “off state”.[12] This furin
conformation is characterized by several changes within its active site compared to furin in
complex with bound inhibitors. The “off state” would appear to be incompatible with efficient
substrate cleavage and would probably also hamper the efficient binding of at least some inhibitors.
Therefore, it is also conceivable that – in a concentration-dependent manner – some inhibitors can
form complexes with two different enzyme forms: at low concentrations with a less active state,
and after a conformational switch induced by higher inhibitor concentrations, with a more active
enzyme. In consequence of all these speculations, only IC50 values were determined for these
inhibitors, since their mode of inhibition is still unclear. The abnormal behavior of these
compounds is also visible in their IC50 plots. Their slopes are steeper (Figure 3B, curves with
empty symbols) compared to the plots of the other inhibitors (Figure 3B, graphs with filled
symbols). This actually represents the same effect – a stronger inhibition than expected at high
inhibitor concentrations and a weaker inhibition than expected at low concentrations.
23
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
Nevertheless, differences in IC50 values can be used to describe the quality of the furin
inhibitors with varying P1 residues. To our surprise, a relatively potent IC50 value of 32 n
M was
found for inhibitor , which is approximately 15 times lower compared to its analogue 14
8
containing the basic P5 substituent and exhibiting a linear Dixon plot. Notably, this is the only
example among all of our substrate-analog inhibitors from previous and recent work where a
reduced affinity was determined for a P5 4-GMe-Phac inhibitor compared to the unsubstituted
Phac analogue. This can be only explained by differences in the overall binding mode of at least
one of these two compounds, which must deviate from the experimentally determined crystal
structures found for several 4-Amba inhibitors.[10, 12] Surprisingly, exchanging the P1
aniline structure in inhibitor by the -isomer also leads to a 3000-fold higher IC50 value of
compound , although both residues possess nearly identical pKa values.
m-substituted
8
p
6
The corresponding aminopyridine moieties in inhibitors 10 and 11 represent a category of
amidines with reduced basicity, with one of the nitrogen atoms being involved in the aromatic
ring. The
development of thrombin inhibitors.[61] Compared to the 4-Amba analogue
inhibitors 10 and 11 are nearly 1000-fold higher, and their IC50 values are about 10- to 15-fold
higher than the one of inhibitor . But more importantly, despite relatively potent Ki values, all
aminopyridine compounds 15 16, and 20 are completely inactive in cell culture (Figure 4). The
very poor furin inhibition observed for the -amidrazone-substituted compounds 12 and 17, the
guanidines and , the hydroxyamidine 21, and the bicyclic P1 (amino)benzisoxazole in
p
-aminopyridine group was successfully used as a 4-Amba replacement for the
1, the Ki values of
8
,
p
7
9
compound 23 may be explained by their steric hindrance, because all of these residues are larger
than the 4-Amba group. In contrast, a relatively potent furin inhibition with Ki values between 0.2
- 0.3 nM was determined for the agmatine and noragmatine inhibitors 26 and 25, which confirms
previous results for analogues without basic P5 group.[8] However, also these compounds are
nearly inactive in cell culture. Therefore, both of these P1 groups are not suited as 4-Amba
replacements. Some cell culture activity was found for the fluoro-substituted benzamidine
compound 24. However, the synthesis of this P1 residue is relatively laborious, and despite a
reduced P1 basicity (pKa circa 9.5), it provides no advantage compared to the 4-Amba group. A
cell culture inhibitory activity comparable to that of the reference compound 4 was found only for
the P2 Lys inhibitor 27. This compound is also a tight binding inhibitor of furin with a Ki value in
the low picomolar range, and effectively inhibits the propagation of WNV and DENV-2. More
24
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
importantly, it possesses more favorable toxicological properties in mice than the reference
compound
4
.
Conclusion
The observed toxicity of inhibitor
4
restricts its further development as a parenteral drug, although
this compound could still be useful for specific topical application, e.g., in combination with
antibiotics like ciprofloxacin for the prophylaxis and treatment of Pseudomonas aeruginosa
keratitis.[62-63] Two strategies have been pursued to improve its toxicological properties. In one
approach, the P1 benzamidine derivative was replaced by a number of structurally related, but –
in most cases – less basic residues, which were predicted to be less toxic. However, all compounds
in this series were considerably less potent than inhibitor 4. Some demonstrated an unusual
inhibition mechanism deviating from classical reversible competitive inhibition. The second
approach was based on substitution of the highly basic arginine in the P2 position with a less basic
lysine. In comparison to the reference compound 4, this provided the nearly equipotent furin
inhibitor 27, with similar antiviral activity in cell culture and twofold reduced toxicity in mice.
Experimental section
Reagents
Protected amino acids, coupling reagents, resins, further chemicals, and solvents were obtained
from Orpegen, Bachem, Iris Biotech GmbH, Polypeptides, Fluka, Acros, Alfa Aesar, and Aldrich.
Standard analytical methods and preparative HPLC
The used methods were recently described in detail.[10, 13] In short, standard analytical reversed-
phase HPLC measurements were performed on a Shimadzu LC-10A system (column: Nucleodur
C18, 5 µm, 100 Å, 4.6
TFA in water, (B) 1 % TFA in acetonitrile; gradient: 1 % increase solvent B/min; flow rate:
1 ml/min; detection at 220 nm). The final inhibitors were purified ( 95% based on the detection
at 220 nm) using the same solvents on a preparative Varian-HPLC system (column: C18 column
(Nucleodur, 5 µm, 100 Å, 32 250 mm (Macherey-Nagel, Düren, Germany); gradient: 0.5 %
× 250 mm column (Machery-Nagel, Düren, Germany); solvents: (A) 1 %
≥
×
25
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
increase solvent B/min; flow rate: 20 ml/min). All peptides were finally obtained as TFA-salts
after lyophilization. Thin-layer chromatography was performed on silica gel 60 F254 aluminum
plates (Merck, Darmstadt, Germany), UV-absorbance was used for detection. The used solvents
are given in the synthesis description of the corresponding compounds. The molecular mass of the
synthesized compounds was determined on a QTrap 2000 ESI spectrometer (Applied Biosystems,
now life technologies, Carlsbad, CA). NMR spectra were measured on an ECX-400 or ECA-500
(Jeol Inc., USA). Internal solvent signals were used as a reference. 19F NMR spectra were
referenced to external trifluorotoluene (-63.9 ppm).
Analytical HPLC for separation of analogous amidine/hydroxyamidine pairs
The analytical separation of analogous amidine and hydroxyamidine pairs (e.g., compounds
21) was performed on a Shimadzu LC 20A system (column: EC Nucleodur C18 HTec, 5
110 , 4 125 mm (Macherey-Nagel, Düren, Germany); solvents: (A) 20 mM ammonium acetate
4
and
μ
m,
Ǻ
×
buffer pH 4.0, (B) pure acetonitrile; gradient: 1 % increase solvent B/min); flow rate: 1 mL/min,
detection: 220 nm).
Synthesis of the P1 residues
Most of the used free or protected P1 residues were prepared according to known literature
procedures, their synthesis and analytical data are provided in the supporting information, e.g., N-
[64]
(4-(aminomethyl)benzyl)-2,2,2-trifluoroacetamide · HCl (32
)
used for inhibitors 13 and 19
,
,
tert-butyl (5-(aminomethyl)pyridin-2-yl)carbamate · CH3COOH (33
)
[65] used for inhibitors 10
,
15
[66]
and 20, 4-(aminomethyl)pyridin-2-amine (34
)
used for inhibitors 11 and 16, (9H-Fluoren-9-
yl)methyl 4-aminobenzylcarbamate (45) used for inhibitors
6
and , and (9H-Fluoren-9-yl)methyl
7
3-aminobenzylcarbamate (46) used for inhibitor , and 14.
8
,
9
The synthesis of 4-(aminomethyl)benzohydrazonamide (37) used as a P1 residue in inhibitors 12
and 17 started from
p
-cyanobenzylamine · HCl according to Scheme 3.
26
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
Scheme 3: Synthesis of compound 37. Reagents and conditions: (a) 2.0 equiv of DIPEA, Cbz-
OSu, water/acetonitrile (70/30, v/v), 1 h at 0°C, overnight at rt, 88 % (35); (b) i: 4 M HCl/dioxane,
dioxane/isobutanol (50/50, v/v), 1 h at 0°C, 10 days at 5°C, ii: hydrazine · H2O, acetonitrile/water
(80/20, v/v), 2 h at rt, 37 % over two steps (36); (c) 32 % HBr/acetic acid, 2 h at rt, 99 % (37).
Benzyl 4-cyanobenzylcarbamate (35
)
4-(aminomethyl)benzonitrile · HCl (2.01 g, 11.92 mmol, 1.0 equiv) was dissolved in 30 mL of a
mixture of water and acetonitrile (70/30, v/v) and cooled to 0°C. After addition of DIPEA (4.15
mL, 23.84 mmol, 2.0 equiv), Cbz-OSu (2.97 mg, 11.92 mmol, 1.0 equiv) was dissolved in 10 mL
of acetonitrile and added dropwise to the reaction mixture within 30 min. The solution was stirred
on an ice bath for 1 h and at room temperature overnight. The solvent was removed in vacuo and
the remaining residue was dissolved in a mixture of EtOAc and 5 % aq. KHSO4. The organic phase
was washed thrice with 5 % aq. KHSO4 and thrice with brine, and dried over Na2SO4. The solvent
was evaporated. Yield: 2.79 g (88 %) of a white powder. HPLC: 44.1 min, start at 10 % B (purity:
99.0 %). 1H NMR (400 MHz, DMSO-d6):
7.45 (d, = 8.1 Hz, 2H), 7.39 – 7.28 (m, 5H), 5.05 (s, 2H), 4.29 (d,
(101 MHz, DMSO-d6):
δ
= 7.92 (t,
J
= 5.8 Hz, 1H), 7.79 (d,
J
= 8.1 Hz, 2H),
13
J
J
= 6.2 Hz, 2H). C NMR
δ
= 156.41, 145.60, 136.99, 132.23 (2C), 128.33 (2C), 127.81 (2C), 127.78
(2C), 127.73, 118.83, 109.58, 65.54, 43.55. MS (ESI, positive): calcd m/z: 266.11; found m/z
:
284.16 [M+NH4]+.
[67-68]
Benzyl (4-carbamohydrazonoylbenzyl)carbamate (36
)
Compound 35 (1.00 g, 3.80 mmol, 1.0 eq) was dissolved in 70 mL of a mixture of dioxane and
isobutanol (50/50, v/v) and cooled to 0°C. After addition of 4 M HCl in dioxane (80 mL, 320
mmol), the reaction mixture was kept for 1 h on an ice bath and then in the refrigerator for 10 days
at 5°C. The solvent was removed in vacuo and the intermediate crude Pinner salt (HPLC: 14.8
27
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
min, start at 30 % B; MS (ESI, positive): calcd m/z: 340.18; found m/z: 341.29 [M+H]+) was
dissolved in 25 mL of a mixture of acetonitrile/water (80/20, v/v) and hydrazine H2O (0.5 mL).
×
The reaction mixture was stirred at room temperature for 2 h. The solvent was evaporated and the
new product was purified by preparative RP-HPLC and lyophilized. Yield: 0.56 g (37 % over two
steps) of a white solid (TFA salt). HPLC: 12.7 min, start at 20 % B (purity: 92.9 %). 1H NMR (400
MHz, DMSO-d6):
δ
= 7.94 (t,
J
= 6.1 Hz, 1H), 7.66 (d,
J = 8.1 Hz, 2H), 7.45 (d, J = 8.2 Hz, 2H),
7.38 – 7.32 (m, 5H), 5.05 (s, 2H), 4.29 (d,
J
= 6.2 Hz, 2H). 13C NMR (101 MHz, DMSO-d6):
δ =
163.15, 156.46, 145.62, 137.01, 128.44 (2C), 128.01 (2C), 127.93, 127.84 (3C), 127.36 (2C),
65.57, 43.48. MS (ESI, positive): calcd m/z: 298.14; found m/z: 299.22 [M+H]+.
4-(aminomethyl)benzohydrazonamide · 2 HBr (37
)
Compound 36 (0.40 g, 0.97 mmol) was treated with 32 % HBr in acetic acid (10 mL) at room
temperature. After 2 h, the white suspension was poured into cooled diethyl ether (30 mL) and
after centrifugation, the precipitate was washed twice with diethyl ether and dried in vacuo. Yield:
0.32 g (99 %) of a white powder (2
1H NMR (400 MHz, DMSO-d6):
2H), 4.17 (q,
= 5.8 Hz, 2H). 13C NMR (101 MHz, DMSO-d6):
ꢀ
HBr salt). HPLC: 4.5 min, start at 1 % B (purity: 80.0 %).
= 8.37 (s, 3H), 7.77 (d, = 8.5 Hz, 2H), 7.69 (d, = 8.5 Hz,
= 163.23, 139.85, 129.77 (2C),
δ
J
J
J
δ
128.70 (2C), 127.31, 42.17. MS (ESI, positive): calcd m/z: 164.11; found m/z: 165.17 [M+H]+.
28
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201700596
ChemMedChem
The multistep synthesis of 6-(Aminomethyl)benzo[
d
]isoxazol-3-amine · TFA (42) used for
inhibitor 23 and of 4-(Aminomethyl)-2-fluorobenzimidamide ·2 HCl (44) incorporated in
compound 24 was performed as shown in Scheme 4.
Boc
O
O
H2N
N
N
N
f
Boc
NH2
NH2
42
41
e
O
O
F
Boc
F
F
F
F
b
c
d
O
N
HO
a
HO
Br
Boc
CN
CN
CN
CN
CN
38
40
39
g
Boc
F
Boc
F
F
Boc
F
N
H
N
H2N
N
j
h
i
NH2
NH
NH2
NH
Boc
NH2
OH
Boc
NH2
N
N
O
O
44
43
Scheme 4: Synthesis of compounds 42 and 44. Reagents and conditions: (a) 1.8 equiv of K2CO3,
1.5 equiv of dimethyl sulfate, acetone, 2 h reflux, 90 % (38), (b) 6.0 equiv of NaBH4, methanol, 3
days at rt, 96 % (39), (c) 2.0 equiv of KBr, 3.0 equiv of P2O5, acetonitrile, 1 h at rt, (d) 2.0 equiv
of di-tert-butyl iminodicarboxylate, 2.0 equiv of NaH, dry THF, 1 h at 0°C, overnight at rt, 98 %
over two steps (40), (e) 3.0 equiv of potassium tert-butoxide, 3.0 equiv of acetohydroxamic acid,
DMF, 48 h at rt, 27 % (41), (f) TFA/TIS/H2O (95:2.5:2.5 v/v/v), 1.5 h at rt, 98.6 % (42), (g) 1.5
equiv of hydroxylamine · HCl, 1.5 equiv of DIPEA, ethanol, 6 h reflux, (h) 3.0 equiv of acetic
anhydride, acetic acid, 1 h at rt, (i) 10 % Pd/C, 90 % acetic acid, H2, 3 days at rt, 22 % over three
steps (43), (j) 1 M HCl/acetic acid, 2 h at rt, 99.8 % (44).
29
This article is protected by copyright. All rights reserved.