Inhibition of the Cysteine Protease Human Cathepsin L by Triazine Nitriles: Amide???Heteroarene π-Stacking Interactions and Chalcogen Bonding in the S3 Pocket

Article abstract of DOI:10.1002/cmdc.201600563

We report an extensive “heteroarene scan” of triazine nitrile ligands of the cysteine protease human cathepsin L (hCatL) to investigate π-stacking on the peptide amide bond Gly67–Gly68 at the entrance of the S3 pocket. This heteroarene???peptide bond stacking was supported by a co-crystal structure of an imidazopyridine ligand with hCatL. Inhibitory constants (K_i) are strongly influenced by the diverse nature of the heterocycles and specific interactions with the local environment of the S3 pocket. Binding affinities vary by three orders of magnitude. All heteroaromatic ligands feature enhanced binding by comparison with hydrocarbon analogues. Predicted energetic contributions from the orientation of the local dipole moments of heteroarene and peptide bond could not be confirmed. Binding of benzothienyl (K_i=4 nm) and benzothiazolyl (K_i=17 nm) ligands was enhanced by intermolecular C?S???O=C interactions (chalcogen bonding) with the backbone C=O of Asn66 in the S3 pocket. The ligands were also tested for the related enzyme rhodesain.

Full text of DOI:10.1002/cmdc.201600563

Accepted Article
Title: Inhibition of the Cysteine Protease Human Cathepsin L by
Triazine Nitriles: Amide···Heteroarene π-Stacking Interactions
and Chalcogen Bonding in the S3 Pocket
Authors: Maude Giroud, Jakov Ivkovic, Mara Martignoni, Marianne
Fleuti, Nils Trapp, Wolfgang Haap, Andreas Kuglstatter,
Jörg Benz, Bernd Kuhn, Tanja Schirmeister, and Francois
Diederich
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.201600563
Link to VoR: http://dx.doi.org/10.1002/cmdc.201600563
A Journal of
www.chemmedchem.org

10.1002/cmdc.201600563
ChemMedChem
ChemMedChem Full Paper
Inhibition of the Cysteine Protease Human Cathepsin L by
π
Triazine Nitriles: Amide···Heteroarene -Stacking Interactions
and Chalcogen Bonding in the S3 Pocket
Maude Giroud,[a] Jakov Ivkovic,[a] Mara Martignoni,[a] Marianne Fleuti,[a] Nils
Trapp,[a] Wolfgang Haap,[b] Andreas Kuglstatter,[b] Jörg Benz,[b] Bernd Kuhn,[b]
Tanja Schirmeister,[c] François Diederich*[a]
[a] Dr. Maude Giroud, Dr. Jakov Ivkovic, Mara Martignoni, Marianne Fleuti, Dr. Nils Trapp,
Prof. Dr. François Diederich
Laboratorium für Organische Chemie, ETH Zurich,
Vladimir-Prelog-Weg 3,
8093 Zurich (Switzerland)
Fax: (+41)44-623-1109
E-mail: diederich@org.chem.ethz.ch
[b] Dr. Andreas Kuglstatter, Dr. Jörg Benz, Dr. Bernd Kuhn, Dr. Wolfgang Haap
F. Hoffmann-La Roche Ltd.
Pharma Research and Early Development (pRED), Therapeutic Modalities,
Roche Innovation Center Basel,
Grenzacherstrasse 124,
4070 Basel (Switzerland)
[c] Prof. Dr. Tanja Schirmeister
Institut für Pharmazie und Biochemie, Johannes Gutenberg-Universität Mainz,
Staudinger Weg 5,
55128 Mainz (Germany)
Supporting Information for this article is available on the WWW under xxx.
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
We report an extensive "heteroarene scan" of triazine nitrile ligands of the cysteine protease
human cathepsin L (hCatL) to investigate π-stacking on the peptide amide bond Gly67–Gly68
at the entrance of the S3 pocket. This heteroarene···peptide bond stacking was supported by a
co-crystal structure of an imidazopyridine ligand with hCatL. Inhibitory constants Ki are
strongly influenced by the diverse nature of the heterocycles and specific interactions with the
local environment of the S3 pocket. Binding affinities vary by three orders of magnitude. All
heteroaromatic ligands feature enhanced binding by comparison with hydrocarbon analogues.
Predicted energetic contributions from the orientation of the local dipole moments of
heteroarene and peptide bond could not be confirmed. Binding of benzothienyl (Ki = 4 nM)
and benzothiazolyl (Ki = 17 nM) ligands was enhanced by intermolecular C–S···O=C
interactions (chalcogen bonding) with the backbone C=O of Asn66 in the S3 pocket. The
ligands were also tested for the related enzyme rhodesain.
2
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
Introduction
In 2013, we reported a computational study which predicted that the π-stacking of
heteroarenes on peptide amide bonds,[1–4] approximated by N-methylacetamide (NMAC),
becomes increasingly favorable with (a) decreasing π-electron density of the heterocycle and
(b) stronger dipole moment of the heterocycle, when aligned in an antiparallel orientation to
the local dipole moment of the peptide bond.[1b, 5–7] Experimental model host-guest chemistry
studies with Rebek-type imide platform receptors[8,9] recently validated the first prediction
(a).[10]
We subsequently described a fluorine scan (F-scan) of triazine nitrile inhibitors at the S3
pocket of the cysteine protease human cathepsin L (hCatL),[11,12] a versatile enzyme in our
group for biomolecular recognition studies.[13] In this F-scan, we analyzed the interactions of
differently fluorinated phenyl rings, which stack onto the planar peptide amide bond Gly67–
Gly68 at the entrance of the S3 pocket of hCatL. This π-stacking had previously been
confirmed by the co-crystal structure of a triazine nitrile ligand bearing a 4-chlorophenyl
substituent occupying the S3 pocket (PDB ID: 4AXM, 2.8 Å resolution).[14] By varying
degree and pattern of fluorine substitution, binding affinities (inhibitory constants Ki) could be
enhanced by up to a factor 13 (1.8 kcal mol–1) from the weakest to the best binder.[11] We
explained the variation in binding affinity by (i) energetically more favorable π-stacking of
less electron-rich, fluorinated rings on the peptide amide bond, (ii) favorable quadrupolar
interactions of the fluorinated phenyl rings with the π-stacking amide bond, and (iii) specific
interactions with the local environment in the S3-pocket of the enzyme. Contributions of
dipolar interactions between the local dipole moments of the fluorinated rings and the Gly67–
Gly68 peptide bond could not be confirmed, however, which could be due to the close
3
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
proximity of a second flat peptide bond fragment, Gly68–Leu69, with an opposite direction of
its local dipole moment.
We previously identified triazine nitriles as potent inhibitors of rhodesain,[11,14,15] a central
cysteine protease from Trypanosoma brucei rhodesiense and a potential drug target against
human African trypanosomiasis (sleeping sickness).[16,17]
While π-stacking on peptide amide bonds by heteroaromatic rings is ubiquitous in protein-
ligand complexation,[6,18] systematic experimental investigation to identify the best
(hetero)aromatic partner for stacking on a peptide amide bond is scarce. Following our
computational predictions,[1b] we therefore initiated an extensive heteroarene scan of triazine
nitrile inhibitors, addressing the π-stacking of different heterocycles on the peptide amide
bonds in the flat dipeptide backbone fragment Gly67–Gly68–Leu69 at the entrance of the S3
pocket of hCatL. Here, we report the synthesis and biochemical evaluation of a larger series
of 35 new ligands, which contain different or differently attached heteroarenes to fill the S3
pocket, while the S1 and S2 pocket substituents departing from the central triazine scaffold
are kept constant. We compare their complexation to a series of ligands with benzenoid
hydrocarbon substituents or lacking the S3 substituent entirely. We analyze their binding
affinities to decipher effects of nature, size, and type of attachment of the heterocycles,
explore contributions from favorable orientations of local dipole moments in the π-stacking
array and specific interactions of the heteroarenes with the local environment of the S3
pocket. In this comprehensive analysis, we obtained strong evidence for intermolecular S···O
chalcogen bonding interactions with the backbone C=O of Asn66 at the entrance of the
pocket,[19,20] stabilizing the binding of sulfur-containing heteroarenes. A similar analysis of
the complexation of the new ligands to the cysteine protease rhodesain (RD) is presented in
the Supporting Information (SI).
4
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
Results and Discussion
Ligand Design
Triazine nitriles (1–46, Table 1) undergo reversible-covalent binding with the catalytic Cys25
at the active site of hCatL attacking the nitrile group and forming a thioimidate stabilized by
the oxyanion hole, as seen in the previously mentioned co-crystal structure (PDB ID: 4AXM,
2.8 Å resolution).[14] In all ligands, the substituents for the S1 and S2 pockets, departing from
the central triazine scaffold, were kept identical. A morpholinyl substituent reaches into the
S1 sub-pocket, enhancing solubility rather than providing additional binding affinity. The S2
pocket is filled by a cyclopentyl ring, which makes a large contribution to the stability of the
complex with hCatL.[11,14,15] The reference inhibitor 1 contains a phenyl substituent reaching
into the S3 pocket, with an inhibitory constant Ki (hCatL) = 520 nM.[11] This substituent was
systematically changed in the other ligands. Modeling with MOLOC[21] using the MAB force
field suggested that the overall binding mode remained similar for all ligands. During
optimization, the protein coordinates as well as the position of the cyano group in the ligand,
covalently bound to Cys25 in the form of a thioimidate, were kept fixed, except for the
flexible Glu63 side chain in the S3 pocket, as previously reported.[11] The phenyl ring of
reference system 1 was predicted to stack on the Gly67–Gly68 peptide bond at a distance of
3.6 Å (Figure S1 in the SI),[11] which is the optimal distance according to our initial
computational study.[1b] This model system adopts a similar orientation to the previously
reported experimental structure (PDB ID: 4AXM)[14] (Figure S1c).
_________________
Table 1
_________________
5
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
Synthesis
Triazine nitriles 1–46 were prepared according to two known synthetic routes (Scheme
1).[11,14,15] These synthetic approaches were chosen for their robustness and reliability. In the
first approach, by which 24 compounds were prepared, reductive amination[22] of
cyclopentylamine (47) and an aromatic aldehyde (48), in the presence of NaHB(OAc)3 and
molecular sieves (4 Å), generated a secondary amine of general structure 49. Sequential
addition of amine 49 and morpholine to cyanuric chloride, in the presence of Hünig base,
gave triazine chloride 50. Cyanodehalogenation of 50 with KCN in Me2SO or KCN/DABCO
in Me2SO/H2O 9:1 at 80–120 °C provided triazine nitrile 51. In the second approach, by
which 21 compounds were prepared, addition of benzylic bromides or chlorides 52 to triazine
nitrile 10,[11] pretreated with NaH, gave the desired final compounds of general formula 51.
Similarly, ligand 10 was methylated with MeI to afford product 11. A complete description of
synthesis and analytical characterization of the new ligands and their precursors is included in
Section S6 in the SI. Small-molecule X-ray crystallographic structures were obtained for
triazine nitriles 10, 43, and 45 upon slow evaporation of EtOH/CH2Cl2 2:1 solutions,
confirming their assigned structures (Section S5 in the SI).
6
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
Approach 1
O
O
N
N
N
N
N
N
Cl
NH2
O
N
N
N
b)
N
N
c) or d)
a)
N
H
+
H
47
48
49
50
51
= (hetero)cycle
24 examples
Approach 2
O
O
N
N
N
N
N
N
a)
N
N
N
N
X
+
HN
N
52
X = Cl, Br, I
10
51
21 examples
= (hetero)cycle, H
Scheme 1. Overview of the approaches used to prepare ligands 2–46. Approach 1: a) 1.
CH2Cl2, 4 Å MS, 1 h at 23 °C, 2. NaHB(OAc)3, CH2Cl2, MS (4 Å), 4–20 h at 23 °C; b) 1.
Cyanuric chloride, iPr2NEt, CH2Cl2, 2–6 h at 0 °C, 2. Morpholine, iPr2NEt, CH2Cl2, 2–16 h at
23 °C; c) KCN, Me2SO, 18–20 h at 120 °C; d) KCN, DABCO, 9:1 Me2SO/H2O, 18–20 h at
80–120 °C. Approach 2: a) 1. 10, NaH, DMF, 30 min to 1 h at 23 °C, 2. 52, 3–18 h at 23 °C.
DABCO = 1,4-Diazabicyclo[2.2.2]octane, MS = molecular sieves, DMF = N,N-
dimethylformamide.
Biochemical and Physicochemical Studies
The binding affinities of ligands 1–46 were tested against hCatL in a fluorimetric assay[14,23]
in aqueous buffer (50 mM tris pH 6.5, 5 mM EDTA, 200 mM NaCl, and 0.005% Brij35) at
298 K (Table 1). Physicochemical data were also acquired: distribution coefficients logD7.4
were measured in a Carrier Mediated Distribution System (CAMDIS©) assay[24] and the
solubility in a Lyophilisation Solubility Assay (LYSA).[25] In addition, the clogD7.4 values
were predicted for all reported compounds with a Roche in-house developed machine-
learning tool based on a tree-based ensemble method.
7
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
Exploring Halogen and Chalcogen Bonding to the Backbone C=O of Gly61 in the S3
Pocket
In a previous systematic study with a structurally very different class of reversible-covalently
binding nitrile inhibitors,[13] we had observed strong gains in binding potency (up to a factor
74, translating into a gain in free enthalpy of –ΔΔG = 2.6 kcal mol–1) by establishing halogen
bonding (XB)[26] to the conformationally locked backbone C=O of Gly61 in the S3 pocket.[13]
We therefore investigated whether 4-halophenyl substituents of triazine nitrile inhibitors
would also establish such intermolecular interactions. The comparison between ligand 1 and
the halogenated derivatives 3–5 however showed that XB is not very effective in this system
yielding a maximum 7-fold gain in Ki when adding a bromine atom in para position of the
phenyl ring (Table 1). Molecular modeling suggested (Figure S2) that the substituted triazine
nitrile ligand scaffold prevents the halophenyl ring from penetrating sufficiently deep into the
pocket to establish XB with the C=O at an O···X distance (X = Cl, Br, I) significantly below
the sum of the van der Waals radii.[27] Also, in the formerly reported class of hCatL ligands,
the halophenyl ring was much more pre-organized for XB than the corresponding ring in the
triazine nitriles. We therefore abstained from synthesizing ligands with halogenated six-
membered heteroarenes as S3 pocket vectors.
Stimulated by the recent interest in chalcogen bonding,[19,20] we also prepared the 4-
methylsulfanyl derivative 6. With a Ki-value of 13 nM (LipE = 3.4), 6 is among the most
potent of the ligands shown in Table 1. MOLOC proposes a binding geometry in which the
phenyl ring stacks in the middle of the flat dipeptide Gly67–Gly68–Leu69 fragment, orienting
the S-atom of 6 to both chalcogen-bonding distance and geometry with the backbone C=O of
Gly61 (d(S···O=CGly61) = 3.6 Å, α(OGly61···S–C) = 162°) (Figure S3a). The methoxy
derivative 7, which is not capable of undergoing chalcogen bonding, is however a similarly
strong binder (Ki =18 nM, LipE = 3.9) (Figure S3b). We therefore conclude that the origin of
8
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
the enhanced affinity of 6 and 7 is the interaction of the ether/thioether methyl group with the
local lipophilic environment of the S3 pocket: both groups undergo apolar contacts at van der
Waals distance with the phenol ring of Tyr72 and the isobutyl group of Leu69 lining one side
of the S3 pocket (Figure S3). Probably, enthalpically strained water molecules in this
lipophilic environment are also displaced by the S3 pocket vector in an energetically
beneficial way.[26a]
π-Stacking Interactions of 5- and 6-Membered Ring Heteroarenes with the Gly67–Gly68
Peptide Bond at the Entrance of the S3 Pocket of hCatL
In agreement with the earlier co-crystal structure (PDB ID: 4AXM),[14] modeling with
MOLOC consistently proposed π-stacking of the heteroarene substituents in ligands 12–26
with the Gly67–Gly68 peptide amide bond at the entrance of the S3 pocket of hCatL. Based
on new co-crystal structure data, however (see below), we cannot exclude additional, different
orientations of the monocyclic heteroarene rings, binding outside the S3 pocket.
Conformational flexibility of the ligands in the free state is also supported by the small-
molecule X-ray crystal structures in which the S2 and S3 pocket vectors do not always adopt
the same orientation as in the predicted bound state (Section S5 in the SI, for similar findings,
see [11]).
The triazine nitrile ligands with five- (12–16) or six-membered ring (17–26) heteroarenes
as S3 pocket vector bind to hCatL with similar or higher affinity to the reference phenylated
ligand 1. All Ki-values are in or near the triple-digit nanomolar range (pKi values between 6.4
and 7.2). We prepared fluoropyridine derivatives 24–26 to lower the pKa[28] and prevent
possible protonation of pyridines in the negative electrostatic potential area of the flexible
anionic side chain of Glu63 in the S3 pocket. They possess very similar activity to their
fluorophenyl analogues,[11] and are slightly less potent than their pyridine analogues 17–19.
9
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
The larger series of six-membered N-heteroarenes 17–23 was initially prepared to evaluate
the predicted[1b] energetic contributions from different local dipole moment orientations of the
heteroarene stacking on a peptide amide bond. The dipole moments of the five- and six-
membered heterocycles were calculated at the B3LYP/6-31+G* or MP2/3-21G level (see
Section S1.3 in the SI). Red arrows in Table 1 represent these dipole moment vectors
schematically for the six-membered N-heterocyclic ligands. Because inhibitors with opposite
local dipole moments possess similar binding affinities (e.g., 20 vs. 22), we could exclude
significant effects of the dipole moment. From this and the previous work on a fluorine scan
of S3 pocket substituents,[11] we can conclude that the Gly67–Gly68 peptide amide bond in
hCatL is not suitable for experimentally testing the computational predictions of dipolar
alignment contributing to heteroarene···amide bond π-stacking. The main reason is the close
proximity of a second, coplanar Gly68–Leu69 peptide bond with opposite direction of its
local dipole moment (see Figure S1).
We also did not find any correlation of hCatL binding strength with the quadrupole
moments of the heteroarenes[29] in the ligand series 17–23 (Figure S4). We finally examined
contributions of the experimentally determined lipophilicity of the ligands to binding affinity,
but no correlation between pKi and logD7.4 values was observed (Figure S5).
π-Stacking Interactions of Bicyclic Heteroarenes with the Gly67–Gly68 Peptide Bond at
the Entrance of the S3 Pocket of hCatL
A large difference in binding affinity of three orders of magnitude was observed when the
heteroarene was extended to fused bicycles in ligands 27–46. One of the weakest ligand with
a 2-quinolyl substituent (45) only bound in the micromolar range (Ki = 1.45 µM, LipE = 1.6)
whereas 2-benzothienyl ligand 27 is the strongest prepared, with Ki = 4 nM (LipE = 4.1). For
two heteroarene ligands, 3-benzothienyl derivative 40 and 8-quinolyl derivative 43, as well as
10
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
for 1-naphthyl ligand 8, the inhibition constants could even not be measured with confidence
as they are in the higher micromolar range. Modeling suggested that all these ligands fit
sterically into the S3 pocket in the active site of hCatL. As the stacking on the flat peptide
bond fragment Gly67–Gly68–Leu69 was not much affected by the position of the
heteroatoms in the monocyclic ligands, specific interactions with the local S3 pocket
environment should explain the differences in binding affinity, besides proper orientation of
the bicycle in the pocket, as well as conformational effects (see below).
The high affinity of the piperonyl substituent (in 36, Ki = 41 nM, LipE = 3.6) for the S3
pocket of the related cysteine proteases rhodesain and falcipain-2 had already been observed
in our earlier work on triazine nitrile ligands,[14] but remains poorly understood in the absence
of a co-crystal structure. Binding in the double-digit nanomolar range is also exhibited by the
quinoxaline ligand 46 (Ki = 56 nM, LipE = 3.2) and the indazole derivative 34 (Ki = 72 nM,
LipE = 3.4). The latter could posssibly form a H-bond with its N–H to the flexible side chain
of Glu63, as suggested by modeling.
The strongest binding, however, was observed for four ligands with similar topology and
attachment of the S3 pocket vector to the triazine core: 2-benzothienyl (27, Ki = 4 nM, LipE =
4.1), 2-benzofuranyl (28, Ki = 11 nM, LipE = 3.9), 2-benzoxazolyl (29, Ki = 17 nM, LipE =
3.7), and 2-imidazopyridine (30, Ki = 35 nM, LipE = 3.5). Gratifyingly, a co-crystal structure
of the latter with hCatL was solved.
X-Ray Co-Crystal Structure Analyses with hCatL
A co-crystal structure of imidazopyridine ligand 30 bound to hCatL was solved at 1.0 Å
resolution (PDB ID: 5MAJ, Figure 1). The complex crystallized in the P212121 space group,
with one chain of mature enzyme in the asymmetric unit of the crystal. The observed binding
mode of ligand 30 is in accordance with our design. The ligand is covalently bound as a
11
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
thioimidate, formed by nucleophilic attack of Cys25 on the triazine nitrile. The morpholine
moiety addresses the S1 pocket, and the cyclopentyl ring is located within the S2 pocket. The
imidazopyridine moiety is directed into the S3 pocket and stacks on top of the Gly67–Gly68
peptide amide bond, at a distance of 3.3–3.9 Å (Figure 1b). The calculated local dipole of the
heterocycle and the peptide bond Gly67–Gly68 are oriented in an antiparallel fashion (Figure
1c), but as already stated, the presence of the second amide bond Gly68–Leu69 with opposite
local dipole orientation in close proximity complicates the analysis of dipolar contributions to
the π-stacking interactions.
Figure 1. a) Co-crystal structure of 30 with hCatL (PDB ID: 5MAJ, resolution 1.0 Å). b)
Position of the imidazopyridine moiety on top of the Gly67–Gly68 peptide fragment. c)
Representation of the calculated dipole moments (at the B3LYP/6-31+G* level) for the
imidazopyridine ring and NMAC. The arrows are proportional to the strength of the dipole
moment. Color code: gray Cenzyme, green Cligand, red O, blue N, yellow S. Distances given in
Å. Torsion angle τ (N–CH2–Csp–Csp) = 94°. W = water molecule.
An interaction analysis of the co-crystal structure of 30 in complex with hCatL was
performed using the Scorpion network approach.[30] The Scorpion software tries to identify
different types of favorable receptor–ligand interactions and uses as network the analysis of
12
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
these contacts to estimate atom-based energy contributions to ligand binding (see Section S2
in the SI). This analysis identified carbon atoms C(4) and C(7) of the imidazopyridine ring as
binding hot-spots (Figure S6). Most interactions are indeed emerging from the contact of the
imidazopyridine ring with the S3 pocket of hCatL. Dispersive interactions with Gly67, Gly68,
and Leu69, as well as π-stacking interactions with the backbone C=O and NH of Gly68 and
the backbone C=O of Gly67 are detected as binding contributions. Additionally, a C–H···O
H-bond between H–C(4) and the backbone C=O of Asn66 was identified, as well as a N···H–
O H-bond between N(1) of the imidazopyridine ring and a water molecule (Figures 1 and S6).
The thioimidate moiety is stabilized by a H-bond with the side chain NH2 of Gln19 in the
oxyanion hole, and the thioimidate C=NH bond engages in a dipolar interaction with the
backbone C=O of Gly23 (Figure S6b). The central triazine ring stacks on the backbone C=O
of Asp162. The triazine core should engage in O–H···π interactions with resolved water
molecules. The cyclopentyl ring addressing the S2 pocket engages in dispersive interactions
with the Met70 and Ala135 side chains. In the S1 pocket, no favorable or unfavorable
interactions were identified. Some crystal contacts between the imidazopyridine ring in the S3
pocket and a second adjacent protein unit were observed (see below).
Another co-crystal structure was obtained for the pyrazine ligand 23, which is bound to
hCatL in a reversible-covalent manner (PDB ID: 5MAE, 1.0 Å resolution) (Figure 2). This
complex also crystallized in the P212121 space group. Two morpholine conformations in the
S1 pocket were observed with about equal occupancy. The cyclopentyl moiety again fills the
S2 pocket. The pyrazine ring, however, does not stack on the peptide bond at the entrance of
the S3 pocket but adopts another orientation.
13
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
Figure 2. a) Co-crystal structure of 23 in hCatL (PDB ID: 5MAE, resolution 1.0 Å). b)
Binding of the pyrazine ring in the interstitial void. c) Contacts of the pyrazine ring with the
second protein. Color code: gray Cmonomer_1, cyan Cmonomer_2 green Cligand, red O, blue N,
yellow S, red dots: water molecules. Distances given in Å. Two conformations of the
morpholine were observed with about equal occupancy. d) Overlay of the co-crystal stuctures
of 23 (PDB ID: 5MAE) and 30 (PDB ID: 5MAJ) in hCatL, showing their two different
cavities with a second protein monomer. Color codes: green C5MAE, dark green C5MAE_monomer2
cyan C5MAJ, violet C5MAJ,monomer_2 red O, blue N.
There are driving forces for the preference of the pyrazine ring to bind into this void
between proteins. (i) The torsional angle τ for fixation of the pyrazine to the triazine core
(τ(Csp2–Csp2–Csp3–N) = 26°) is close to the corresponding one seen in the crystal structure for
benzene derivative 1 (τ(Csp2–Csp2–Csp3–N) = 11°). A substantial investment of conformational
energy is required to change this angle to 90° for π-stacking on the peptide bond at the
entrance of the S3 pocket (see below); which does not have to be paid upon binding into the
void in the solid state. (ii) If the pyrazine stacks on the Gly67–Gly68 peptide amide bond at
the entrance of the S3 pocket, which can occur in two conformations by rotation of the
heteroarene by 180°, the pyrazine N-atoms undergo repulsive intermolecular lone-pair
interactions, according to MOLOC modeling (Figure S7). In one conformation, one pyrazine
14
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
N-atom approaches the backbone C=O of Asn66; in the second, this N-atom comes close to
the anionic side chain of Glu63, with additional desolvation of the second N-atom by the
lipophilic side chain of Leu69.
In addition, we analyzed the crystal packing to find a rationale for this unexpected binding
geometry in the solid state and observed that two neighboring proteins form an interstitial
cavity in which the pyrazine ring is located and benefits from attractive interactions (Figure
2b,c). In this interstitial pocket, the pyrazine ring undergoes favorable C–H···O interactions
with the backbone C=O of Asn179 (d(C···O=CAsn179) = 3.3 Å), weak ionic H-bonding with a
side chain NH of Lys181 (d(N···H–NLys181) = 4.1 Å), and C–H···π interactions with the CH2
of Asp162 ring (d(C···H–CAsp162 = 4.0–4.3 Å). The cavity at the interface between the two
proteins has shape-complementarity to the six-membered pyrazine but is too small to
accommodate bicyclic heteroarene substituents. The interstitial binding is reminescent of
clathrate formation,[31] often observed in supramolecular and other chemistries. This pyrazine
conformation would not be observed in the hCatL complex in solution, as the heterocycle
would be exposed to bulk water. For the complex of 30 with hCatL, such cavity does not exist
in the same way. In this case, the Asn179 side chain is moved over to the heteroarene (Figure
2d). The imidazopyridine ring undergoes contacts with both Asp178 and Asn179, which
stabilize the ring in the S3 pocket. In addition, the preferred torsional angle (τ(Csp2–Csp2–Csp3–
N) ~ 110°) for the imidazopyridine ring makes it pre-organized for binding in the S3 pocket.
Intermolecular S···O Interactions
The sulfur-containing bicyclic heteroarenes benzothiophene 27 (Ki = 4 nM, LipE = 4.1) and
benzothiazole 29 (Ki = 17 nM, LipE = 3.7) are among the strongest binders. Sulfur possesses
unusual stereoelectronic properties,[32] and has been reported to undergo chalcogen
bonding,[19,20] an interaction similar to halogen bonding.[13] Chalcogen-containing
15
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
heterocycles possess a positive electrostatic potential adjacent to the sulfur atom consistent
with the "σ-hole" concept.[19b] The magnitude and volume of this σ-hole increases from
thiophene to thiazole.[20a] A comprehensive summary of intermolecular O···S–C n → σ*
interactions reported in the literature has been provided by Meanwell and co-workers.[19b]
We modeled benzothiophene ligand 27 into the active site of hCatL by modifying the
imidazopyridine 30 in the co-crystal structure and optimizing the new complex with MOLOC.
The resulting structural proposal clearly revealed a chalcogen bond from the benzothiophene
donor to the backbone C=O of Asn66 as acceptor (Figure 3). The position of this backbone
C=O is highly conserved in co-crystal structures of hCatL, which makes it entropically a
favorable interaction partner.[13,33] The interaction occurs at a distance d(S···O=C) = 3.5 Å
and at an angle α(O···S–C) = 158°, well within the range of chalcogen bonds observed in
other protein-ligand complexes (Figure 3).[19] The 2-benzothiazolyl ligand 29 shows a similar
chalcogen bond to the C=O of Asn66 (Figure S8a), but complexation is weaker (Ki = 17 nM,
LipE = 3.7). A weak chalcogen bonding may also explain the good binding of thiazole 16 (Ki
= 69 nM, LipE = 3.2) in the monocyclic heteroarene ligand series (Table 1). Chalcogen
bonding is generally stronger for thiazoles than for thiophenes.[19] While thiazolyl 16 binds
better than thienyl 12 (3-fold), the trend was inversed between benzothiazolyl 29 and
benzothienyl 27. The predicted binding mode of 29 in hCatL (Figure S8a) reveals that the
C–S···N(4) angle (141°) is not ideal for an intramolecular N···S interaction. In the unbound
state, ligand 29 is in a more planar conformation (Figure S8b), enabling a better
intramolecular N···S interaction (α(C–S···N(4)) = 164°). This might cost energy upon
reorganization for binding of 29, and explain the weaker binding by comparison with ligand
27. The benzofuran ligand 28 (Ki = 11 nM, LipE = 3.9) cannot undergo chalcogen bonding
and most probably forms a weak C–H···O H-bond with the C=O of Asn66 (Figure S9), with
16
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
the O-atom becoming solvated by a water molecule as seen for the N-atom in the
imidazopyridine crystal structure.
Figure 3. Predicted binding mode of 27 with hCatL (protein coordinates taken from PDB ID:
5MAJ, resolution 1.0 Å), displaying the chalcogen bond (d(O···S) = 3.5 Å, α(O···S–C) = 158
°, d(S···N(4)) = 3.4 Å). Color code: gray Cenzyme, green Cligand, red O, blue N, yellow S.
Distances given in Å.
Apart from the analysis of intermolecular interactions in the complex, it is important to
consider potential ligand strain when trying to rationalize the observed SAR. When
comparing the ligand conformations in the protein-bound state with the one found in small
molecule crystal structures, we noticed a significant difference in the torsional angle along the
benzylic CH2–Csp2 bond of the S3 pocket vector, τ (N–CH2–Csp–Csp). In both the complex of
the 4-chlorophenyl ligand with hCatL (PDB ID: 4AXM)[14] and in the structure of compound
30 (Figure 1), the dihedral angles are 94°, while in the small-molecule crystal structure of
triazine nitrile 1 it is τ = 11° (CCDC 1449741).[11] A torsion angle of around 90° is required
for stacking onto the peptide amide backbone at the entrance of the S3 pocket. To better
assess the energetic implications of this angle difference, we performed quantum mechanical
17
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
torsion scans for ligands 1 and 27–30 at the B3LYP/cc-pVDZ level using Gaussian 09[35]
(Figure 4).
Figure 4. Torsional angle scans of triazine nitriles 1, 27–30 calculated at the B3LYP/cc-
pVDZ level using Gaussian 09.[35] The definition of the torsion angle τ is shown in red, and
the 2D ligand depictions show the preferred orientation of the S3 pocket vector relative to the
triazine core based on the torsional energy profiles.
For the phenyl derivative 1 the dihedral angle at the minimum of the energy profile is
close to the value observed in the small molecule structure. In contrast, a relative energy of
2.7 kcal mol–1 is calculated for τ = 90° suggesting substantial ligand strain for this
compound at the torsional angle required for stacking in the S3 pocket. Relative energies for
at least one of the torsional angles τ = −90°, 90° are lower for all four heterocycles in 27–30,
indicating that reduced ligand strain for these compounds contributes to their improved hCatL
activity compared to 1. The torsional profiles in Figure 4 provide further information on the
preferred orientation of the heterocycles in a stacked binding conformation in the S3 pocket.
For ligands 27–30, a significant energy difference > 1 kcal mol–1 exists for one of the two
18
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
orientations at τ = −90°, 90°. In all these cases, the preferred orientation corresponds to an
anti alignment of the O and N lone pairs in the heterobicycle with the N-atom (N(4)) of the
triazine ring thereby avoiding electrostatic repulsion. For the imidazopyridine 30 this is
indeed observed in the co-crystal structure (Figure 1). In the two sulfur-containing
heterocycles benzothiophene 27 and benzothiazole 29, the torsional profiles suggest that the S
atom of the ligand is preferentially syn to the N(4) of the triazine nitrile. This is in line with
the postulated chalcogen-bonding geometries of benzothiophene 27 (Figure 3) and
benzothiazole 29 (Figure S8). In these orientations, the ligand sulfur atom would then interact
intermolecularly with the C=O O-atom of Asn66 and intramolecularly with the π-orbital on
the triazine N-atom using its second σ-hole. The torsional ligand strain to adopt the stacked
binding mode in the S3 pocket of hCatL is predicted to be 1.1 kcal mol–1 for benzothiazole
29, and 1.4 kcal mol–1 for benzothiophene 27. Interestingly, this is significantly higher than
for the benzofuran 28 (0.2 kcal mol–1) and imidazopyridine 30 (0.6 kcal mol–1). It suggests
that there are compensating attractive intermolecular interactions, which we attribute mainly
to the C=OAsn66···S–Csp2 chalcogen bonding. Similar conformational strain analyses are
undoubtedly needed when trying to rationalize the large affinity differences – by three orders
of magnitude – between the entire set of 20 ligands with the various heterobicyclic S3 pocket
vectors included in Table 1.
Conclusions
We prepared and investigated a large series of triazine nitrile ligands featuring mono- and
bicyclic heteroarene substituents for binding to the S3 pocket of the cysteine protease human
cathepsin L (hCatL). A similar analysis for the binding of these ligands to the related enzyme
rhodesain (RD) is also reported in the Supporting Information (SI). The ligands undergo
reversible-covalent binding, with the nitrile reacting with the catalytic cysteine under
19
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
formation of a thioimidate moiety. Their substituents for the S1 and S2 pockets of the enzyme
were kept constant. Molecular modeling proposed for all ligands that the heteroarene
substituents could undergo π-stacking with the Gly67–Gly68, and partially also with the
neighboring Gly68–Leu69 peptide amide bond at the entrance of the S3 pocket. Ligands
featuring monocyclic five- or six-membered heteroarenes bind with inhibitory constants Ki
within or near the triple-digit nanomolar range. They show higher potency than the reference
ligand with a phenyl substituent, providing another example for the preference of peptide
amide bonds to undergo π-stacking with heteroaromatic rings. The predicted contributions
from the orientation of the local dipole moments of the heteroarenes and the peptide amide
bond could not be confirmed in this study, which could be due to the close proximity of two
flat peptide bonds, with oppositely oriented local dipole vectors, at the entrance of the S3
pocket.[6a]
An unexpectedly large difference of three orders of magnitudes (Ki values between 4 nM
and higher micromolar) was measured for the binding affinity of ligands featuring bicyclic
heteroaromatic rings. The highest binding affinity was measured for four ligands with similar
topology and attachment of the S3 pocket vector, 2-benzothienyl (27, Ki = 4 nM, LipE = 4.1),
2-benzofuranyl (28, Ki = 11 nM, LipE = 3.9), 2-benzothiazolyl (29, Ki = 17 nM, LipE = 3.7),
and 2-imidazopyridinyl (30, Ki = 35 nM, LipE = 3.5). The X-ray co-crystal structure of the
imidazopyridine ligand in complex with hCatL was determined and confirmed the predicted
binding mode. The binding of these ligands was subjected to a more thorough analysis,
including an interaction network analysis of the co-crystal structure and a detailed
conformational analysis, analyzing the torsional ligand strain when adopting the binding
conformation required for π-stacking with the flat dipeptide fragment at the entrance of the S3
pocket. Molecular modeling based on the crystal structure revealed that the complexes of the
benzothiophenyl and the benzothiazolyl ligands are additionally stabilized by chalcogen
20
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
bonding to the backbone C=O of Asn66. The interaction features suitable geometric
parameters for chalcogen bonding, with the distance d(S···O=CAsn66) = 3.5 Å and the angle
α(OAsn66···S–C) = 158°. The conformational analysis further supports this interaction, which
compensates for the higher torsional strain in the S-containing ligands as compared to the
benzofuranyl and imidazopyridinyl ligands. This study shows that the analysis of
intermolecular interactions needs to be accompanied by conformational strain analysis, when
trying to rationalize the large affinity differences between the entire set of ligands featuring
various heterobicyclic S3 pocket vectors. It confirms that optimal design of the S3 pocket
vector can strongly enhance affinity to the cysteine proteases hCatL and RD.
21
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
Experimental Section
Since the synthetic protocols had been previously developed and applied,[11,14,15] all synthesis
and characterization of the new ligands is found in Section S6 of the SI.
Enzymatic Assays
The determination of the activity of the inhibitors against hCatL and RD was performed in
fluorescence-based assays. In the following, the procedure for the hCatL inhibition assay is
described in detail. The assays with rhodesain were carried out similarly.
The biological activities against hCatL were determined using Cbz-Phe-Arg-AMC as
substrate, which releases AMC (7-amino-4-methylcoumarin) after amide bond cleavage by
the enzyme.[14,23]
The proteolytic activity of the enzyme can be monitored
spectrophotometrically by the increase of fluorescence intensity by release of AMC (emission
at 460 nm) upon hydrolysis.
An initial screen at an inhibitor concentration of 20 µM was performed to identify
ligands with an inhibition of hCatL and RD higher than 80%. For active compounds,
continuous assays at different inhibitor concentrations were carried out. The residual enzyme
activities were plotted against the inhibitor concentrations to obtain the experimentally
determined IC50 values. The inhibitory constants (Ki) were calculated according to the Cheng-
Prusoff equation,[34] as triazine nitriles have previously been determined to be competitive
inhibitors and do not show any time-dependent inhibition:[15,36]
ꢂꢃꢄꢅ
ꢀ =
ꢁ
ꢆ
1 +
ꢀꢇ
with S as substrate concentration and Km as Michaelis-Menten constant[34] (Km = 6.5 µM for
Cbz-Phe-Arg-AMC).
22
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
Determination of hCatL Inhibition. The protein was purchased at Calbiochem. The assay
buffer consisted of 50 mM Tris pH 6.5, 5 mM EDTA, 200 mM NaCl, and 0.005% Brij35, and
the enzyme buffer of 50 mM Tris pH 6.5, 5 mM EDTA, 200 mM NaCl, and 2 mM DTT. The
substrate Cbz-Phe-Arg-AMC (Km = 6.5 µM, Bachem, purchased as HCl salt) was diluted from
a 1 mM stock solution to a final concentration of 25 µM. The assay mixtures had a total
volume of 200 µL and consisted of 180 µL assay buffer, 5 µL enzyme, 10 µL Me2SO (as
negative control) or inhibitor solution (prepared from a stock solution with c = 2 mM), and
5 µL of the substrate solution. An initial screening was performed at an inhibitor
concentration of 20 µM (final inhibitor concentration) over 5 min in two independent
measurements to identify compounds with an inhibition > 80%. Enzyme was added to the
substrate dilutions, and the reaction was initiated upon addition of the substrate. For active
inhibitors, dilution series starting from a 20 mM stock solution in Me2SO at seven different
concentrations were prepared in duplicate for each compound at 25 °C. Fluorescence
emission was measured over 30 min on a Cary Eclipse Fluorimeter (Varian, Darmstadt,
Germany, excitation 380 nm, emission 460 nm) equipped with a heating device in 96-well
microtiter plates. Kinetic constants are the mean values of at least two independent assays,
each performed in duplicate. The GraFit® software (version 5.0.13, 2006, Erithacus Sofware
Ltd., UK) was used for nonlinear regression, and the kinetic constants Ki were calculated
based on the IC50 and the Km values using the Cheng-Prusoff equation.
Determination of RD Inhibition. Assay buffer: 50 mm sodium acetate pH 5.5, 5 mm
EDTA, 200 mM NaCl, and 0.005% Brij35. Enzyme: expressed and purified in the group of
Prof. Caroline Kisker (Rudolf-Virchow-Zentrum, University of Würzburg, Germany).
Enzyme buffer: 50 mM sodium acetate pH 5.5, 5 mM EDTA, 200 mM NaCl, and 2 mM DTT.
Substrate: Cbz-Phe-Arg-AMC (Km = 825 nM, Bachem, purchased as HCl salt).
23
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
Co-Crystallization of hCatL in Complex with 23 and 30
hCatL (produced as previously described)[13a] at a concentration of 2.5 µM in 100 mM sodium
acetate pH 5.5, 5 mM DTT, 5mM EDTA, 0.02% NaN3 was incubated with ligands 23 and 30
in a 12 to 15-fold molar excess overnight at 4 °C under argon. Prior to crystallization
experiments, the protein-ligand mixtures were concentrated to 27–32 mg mL–1 and
centrifuged at 20,000 g. The crystallization droplets were set up at 22 °C by mixing 0.15 µL
of protein solution with 0.3 µL reservoir solution and 0.02 µL seed solution in sitting drop
vapour diffusion experiments. For compound 23 crystals were obtained within two days out
of 25% PEG 3350, 0.1 M Bis-Tris pH 6.5. The complex with compound 30 crystallized within
one day out of 25% PEG3350, 0.2 M ammonium acetate, 0.1 M HEPES pH 7.5. Details on
data collection, processing, and refinement statistics for derivatives 23 and 30 are listed in
Table 2.
Table 2. Statistics for X-ray data processing and model refinement for co-crystal structures of
triazine nitriles 23 and 30 with hCatL.
30
23
PDB accession number
Data Processing[a]
Space Group
5MAJ
5MAE
P212121
P212121
Unit cell axes [Å]
Resolution limits [Å]
Completeness [%]
I/σ(I)
45.7/57.1/75.3
45.51–1.00 (1.09–1.00)
96.1 (90.3)
45.5/56.6/73.9
45.00–1.00 (1.09–1.00)
99.6 (98.6)
13.8 (2.0)
11.2 (2.1)
Multiplicity
6.1 (5.4)
6.1 (5.4)
Refinement
R/Rfree [%]
Rmsd bond length [Å]
Rmsd bond angles [°]
[a] Number in parenthesis are values for the highest of ten resolution shells.
13.6/17.0
0.028
2.4
14.7/18.2
0.030
2.5
logD7.4 Determinations[24]
This protocol was taken from reference [24]. Distribution coefficients are determined using
the CAMDIS© (CArrier Mediated DIstribution System, EP2005102211A) method, which is
derived from the conventional ‘shake flask’ method. CAMDIS© is carried out in 96-well
24
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
microtiterplates in combination with DIFI©-tubes (Weidmann Plastics Technology AG,
Rapperswil, Switzerland), which provide a hydrophobic layer for the 1-octanol phase. The
hydrophobic layer (0.45 µm PVDF membranes) fixed on the bottom of each DIFI©-tube is
coated (Microfluidic Dispenser BioRAPTR, Bechman Coulter) with 1.0 µL of 1-octanol.
Next, the filter membranes are dipped into a 96-well plate, which has been prefilled with 150
µL of aqueous buffer solution (25 mM Phosphate, pH 7.4) containing the compound of
interest at a starting concentration of 100 µM. The plate is sealed and shaken for 24 h at room
temperature (23 °C) to ensure that the partition equilibrium is reached. The next day, the
DIFI©-tubes are removed from the 96-well plate and an aliquote of the aqueous solution is
analyzed by LC/MS. The distribution coefficient is calculated from a control experiment
without 1-octanol and the remaining compound concentration in the aqueous phase, which
was in equilibrium with 1-octanol. Sample preparation is carried out using a TECAN robotic
system (RSP 100, 8 channels).
Lyophilisation Solubility Assay (LYSA)[25]
This procedure was taken from reference [25]. Samples were prepared in duplicate from 10
mM Me2SO stock solutions. After evaporation (1 h) of Me2SO with a centrifugal vacuum
evaporator (Genevac Technologies), the compounds were dissolved in 0.05 M phosphate
buffer (pH 6.5), stirred for one hour, and shaken two hours. After 18 h, the solutions were
filtered using a microtiter filter plate (Millipore MSDV N65) and the filtrate and its 1/10
dilution were analyzed by direct UV measurement or by HPLC-UV. In addition, a four-points
calibration curve is prepared from the 10 mM stock solutions and used for the solubility
determination of the compounds. The results are expressed in µg mL–1.
25
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
Acknowledgements
This research was supported by F. Hoffmann-La Roche Ltd., Basel, and the ETH Research
Council. We thank Ulrike Nowe, Sabine Maehrlein, and Nicole Heindl from the Johannes-
Gutenberg Universität Mainz for support with the biological assays, and Mattia Valentini for
the preparation of two final compounds. Martine Stihle is acknowledged for support with the
co-crystallization experiments. We are grateful to Dr. Bruno Bernet for correcting the
Experimental Section in the Supporting Information. J. I. acknowledges the doctoral program
W901-B05 DK Molecular Enzymology, funded by the Austrian Science Fund (FWF).
26
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
_______________________
[1]
a) S. Cotesta, M. Stahl, J. Mol. Model. 2006, 12, 436–444; b) M. Harder, B. Kuhn, F.
Diederich, ChemMedChem 2013, 8, 397–404; c) Y. Wang, J. Wang, L. Yao, J. Phys.
Chem. A 2015, 119, 3471–3478.
[2]
a) W. H. James, III, C. W. Müller, E. G. Buchanan, M. G. D. Nix, L. Guo, L. Roskop,
M. S. Gordon, L. V. Slipchenko, S. H. Gellman, T. S. Zwier, J. Am. Chem. Soc. 2009,
131, 14243–14245; b) W. H. James, III, E. G. Buchanan, L. Guo, S. H. Gellman, T. S.
Zwier, J. Phys. Chem. A 2011, 115, 11960–11970.
[3]
a) G. Duan, V. H. Smith, Jr., D. F. Weaver, J. Phys. Chem. A 2000, 104, 4521–4532;
b) L. Bendová, P. Jurečka, P. Hobza, J. Vondrášek, J. Phys. Chem. B 2007, 111,
9975–9979; c) Y. N. Imai, Y. Inoue, I. Nakanishi, K. Kitaura, J. Comput. Chem. 2009,
30, 2267–2276.
[4]
[5]
C. He, J. Chen, L. An, Y. Wang, Z. Shu, L. Yao, J. Chem. Inf. Model. 2015, 55, 104–
113.
L. M. Salonen, D. D. Medina, E. Carbó-Argibay, M. G. Goesten, L. Mafra, N.
Guldris, J. M. Rotter, D. G. Stroppa, C. Rodriguez-Abreu, Chem Commun. 2016, 52,
7986–7989.
[6]
a) A. M. Hartman, M. Mondal, N. Radeva, G. Klebe, A. K. H. Hirsch, Int. J. Mol. Sci.
2015, 16, 19184–19194; b) B. S. Lauber, L. A. Hardegger, A. K. Asraful, B. A. Lund,
O. Dumele, M. Harder, B. Kuhn, R. A. Engh, F. Diederich, Chem. Eur. J. 2016, 22,
211–221; c) L. Fang, X. Wang, M. Xi, T. Liu, D. Yin, Mol. BioSyst. 2016, 12, 1174–
1182; d) Z. Qiu, B. Kuhn, J. Aebi, X. Lin, H. Ding, Z. Zhou, Z. Xu, D. Xu, L. Han, C.
Liu, H. Qiu, Y. Zhang, W. Haap, C. Riemer, M. Stahl, N. Qin, H. C. Shen, G. Tang,
ACS Med. Chem. Lett. 2016, 7, 802–806; e) M. Volgraf, B. D. Sellers, Y. Jiang, G.
Wu, C. Q. Ly, E. Villemure, R. M. Pastor, P.-w. Yuen, A. Lu, X. Luo, M. Liu, S.
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
Zhang, L. Sun, Y. Fu, P. J. Lupardus, H. J. A. Wallweber, B. M. Liederer, G.
Deshmukh, E. Plise, S. Tay, P. Reynen, J. Herrington, A. Gustafson, Y. Liu, A.
Dirksen, M. G. A. Dietz, Y. Liu, T.-M. Wang, J. E. Hanson, D. Hackos, K. Scearce-
Levie, J. B. Schwarz, J. Med. Chem. 2016, 59, 2760–2779; f) M. Mondal, N. Radeva,
H. Fanlo-Virgós, S. Otto, G. Klebe, A. K. H. Hirsch, Angew. Chem. Int. Ed. 2016, 55,
9422–9426; Angew. Chem. 2016, 128, 9569–9574.
[7]
[8]
O. Korb, B. Kuhn, J. Hert, N. Taylor, J. Cole, C. Groom, M. Stahl, J. Med. Chem.
2016, 59, 4257–4266.
a) R. K. Castellano, V. Gramlich, F. Diederich, Chem. Eur. J. 2002, 8, 118–129; b) R.
Faraoni, M. Blanzat, S. Kubicek, C. Braun, W. B. Schweizer, V. Gramlich, F.
Diederich, Org. Biomol. Chem. 2004, 2, 1962–1964; c) R. Faraoni, R. K. Castellano,
V. Gramlich, F. Diederich, Chem. Commun. 2004, 4, 370–371.
[9]
a) J. Rebek, Jr., B. Askew, P. Ballester, C. Buhr, S. Jones, D. Nemeth, K. Williams, J.
Am. Chem. Soc. 1987, 109, 5033–5035; b) B. Askew, P. Ballester, C. Buhr, K. S.
Jeong, S. Jones, K. Parris, K. Williams, J. Rebek, Jr., J. Am. Chem. Soc. 1989, 111,
1082–1090; c) K. Williams, B. Askew, P. Ballester, C. Buhr, K. S. Jeong, S. Jones, J.
Rebek, Jr., J. Am. Chem. Soc. 1989, 111, 1090–1094; d) J. Rebek, Jr., Acc. Chem. Res.
1990, 23, 399–404; e) J. Rebek, Jr., Angew. Chem. Int. Ed. Engl. 1990, 29, 245–255;
Angew. Chem. 1990, 102, 261–272.
[10] M. Harder, M. A. Carnero Corrales, N. Trapp, B. Kuhn, F. Diederich, Chem. Eur. J.
2015, 21, 8455–8463.
[11] M. Giroud, M. Harder, B. Kuhn, W. Haap, N. Trapp, W. B. Schweizer, T.
Schirmeister, F. Diederich, ChemMedChem 2016, 11, 1042–1047.
28
This article is protected by copyright. All rights reserved.

10.1002/cmdc.201600563
ChemMedChem
[12] a) D. R. Sudhan, M. B. Rabaglino, C. E. Wood, D. W. Siemann, Clin. Exp.
Metastasis, 2016, 33, 461–473; b) D. R. Sudhan, C. Pampo, L. Rice, D. W. Siemann,
Int. J. Cancer 2016, 138, 2665–2677.
[13] a) L. A. Hardegger, B. Kuhn, B. Spinnler, L. Anselm, R. Ecabert, M. Stihle, B. Gsell,
R. Thoma, J. Diez, J. Benz, J.-M. Plancher, G. Hartmann, D. W. Banner, W. Haap, F.
Diederich, Angew. Chem. Int. Ed. 2011, 50, 314–318; Angew. Chem. 2011, 123, 329–
334; b) L. A. Hardegger, B. Kuhn, B. Spinnler, L. Anselm, R. Ecabert, M. Stihle, B.
Gsell, R. Thoma, J. Diez, J. Benz, J.-M. Plancher, G. Hartmann, Y. Isshiki, K.
Morikami, N. Shimma, W. Haap, D. W. Banner, F. Diederich, ChemMedChem 2011,
6, 2048–2054.
[14] V. Ehmke, E. Winkler, D. W. Banner, W. Haap, W. B. Schweizer, M. Rottmann, M.
Kaiser, C. Freymond, T. Schirmeister, F. Diederich, ChemMedChem 2013, 8, 967–
975.
[15] a) V. Ehmke, C. Heindl, M. Rottmann, C. Freymond, W. B. Schweizer, R. Brun, A.
Stich, T. Schirmeister, F. Diederich, ChemMedChem 2011, 6, 273–278; b) V. Ehmke,
J. E. Q. Quinsaat, P. Rivera-Fuentes, C. Heindl, C. Freymond, M. Rottmann, R. Brun,
T. Schirmeister, F. Diederich, Org. Biomol. Chem. 2012, 10, 5764–5768.
[16] a) I. D. Kerr, J. H. Lee, C. J. Farady, R. Marion, M. Rickert, M. Sajid, K. C. Pandey,
C. R. Caffrey, J. Legac, E. Hansell, J. H. McKerrow, C. S. Craik, P. J. Rosenthal, L. S.
Brinen, J. Biol. Chem. 2009, 284, 25697–25703; b) I. D. Kerr, P. Wu, R. Marion-
Tsukamaki, Z. B. Mackey, L. S. Brinen, PLoS Negl. Trop. Dis. 2010, 4, e701.
[17] a) R. Ettari, L. Tamborini, I. C. Angelo, N. Micale, A. Pinto, C. De Micheli, P. Conti,
J. Med. Chem. 2013, 56, 5637–5658; b) R. Ettari, A. Pinto, S. Previti, L. Tamborini, I.
C. Angelo, V. La Pietra, L. Marinelli, E. Novellino, T. Schirmeister, M. Zappalà, S.
Grasso, C. De Micheli, P. Conti, Bioorg. Med. Chem. 2015, 23, 7053–7060; c) R.
29
This article is protected by copyright. All rights reserved.