Approaches to design non-covalent inhibitors for human granzyme B (hGrB)

Article abstract of DOI:10.1039/c4ob01874e

A structure-based design campaign for non-covalent small molecule inhibitors of human granzyme B was carried out by means of a virtual screening strategy employing three constraints and probe site-mapping with FTMAP to identify ligand "hot spots". In addition, new scaffolds of diverse structures were subsequently explored with ROCS shape-based superposition methods, following by Glide SP docking, induced fit docking and analysis of QikProp molecular properties. Novel classes of moderately active small molecule blockers (≥25 μM IC₅₀ values) from commercially available libraries were identified, and three novel scaffolds have been synthesized by multi-step procedures. Furthermore, we provide an example of a comprehensive structure-based drug discovery approach to non-covalent inhibitors that relies on the X-ray structure of a covalently bound ligand and suggest that the design path may be compromised by alternative and unknown binding poses. This journal is

Full text of DOI:10.1039/c4ob01874e

Organic &
Biomolecular Chemistry
View Article Online
View Journal | View Issue
PAPER
Approaches to design non-covalent inhibitors for
human granzyme B (hGrB)†
Mi-Sun Kim,a Lauriane A. Buisson,b Dean A. Heathcote,c Haipeng Hu,a
D. Christopher Braddock,b Anthony G. M. Barrett,b Philip G. Ashton-Rickardtc and
James P. Snyder*a,d
Cite this: Org. Biomol. Chem., 2014,
12, 8952
A structure-based design campaign for non-covalent small molecule inhibitors of human granzyme B
was carried out by means of a virtual screening strategy employing three constraints and probe site-
mapping with FTMAP to identify ligand “hot spots”. In addition, new scaffolds of diverse structures were
subsequently explored with ROCS shape-based superposition methods, following by Glide SP docking,
induced fit docking and analysis of QikProp molecular properties. Novel classes of moderately active
small molecule blockers (≥25 µM IC50 values) from commercially available libraries were identified, and
three novel scaffolds have been synthesized by multi-step procedures. Furthermore, we provide an
example of a comprehensive structure-based drug discovery approach to non-covalent inhibitors that
relies on the X-ray structure of a covalently bound ligand and suggest that the design path may be com-
promised by alternative and unknown binding poses.
Received 3rd September 2014,
Accepted 22nd September 2014
DOI: 10.1039/c4ob01874e
www.rsc.org/obc
GVHD by killing normal cells and suppresses GVL by destroy-
ing useful tumor-specific CTL.2 Therefore, inhibition of GrB
Introduction
Over the past several decades, allogeneic bone marrow trans- has been considered an attractive mechanism to promote GVL
plantation (BMT) has emerged as an important therapeutic and prevent GVHD. GrB is unique among mammalian serine
option for a number of malignant diseases. Of particular proteases as a result of its strong requirement for Asp in the
importance are acute myeloid leukemia (AML), acute lymphoid substrate P1 position similar to a number of members of the
leukemia (ALL) and chronic myeloid leukemia (CML). The caspase family, cysteine proteases involved in apoptosis. This
success of BMT relies on the graft-versus-leukemia (GVL) effect, preference is complemented by the presence of Arg226 in the
which eradicates remaining malignant cells via immunologic enzyme’s S1 pocket. In addition, IEPD (Ile-Glu-Pro-Asp) was
mechanisms. However, BMT also faces a major obstacle; discovered as the preferred tetrapeptide substrate sequence
namely, in acute graft-versus-host disease (GVHD), the func- based on combinatorial library screening.3
tional immune cells in the transplanted marrow recognize the
With this background, Rotonda et al. (Merck) generated the
corresponding tetrapeptide aldehyde inhibitor 1 (Fig. 1a) and
recipient as “foreign” and lead to immunologic attack.1
Cytotoxic T cells (CTL), which belong to a sub-group of T carried out an X-ray crystallographic characterization, which
lymphocytes (a type of white blood cell), are capable of recog- clearly exhibits covalent interaction between the active site
nizing specific infected target cells and inducing apoptosis, serine and the P1 aldehyde carbon of 1.4 Following this initial
i.e. programmed cell death. Recent work with granzyme B study, a series of inhibitors equipotent to 1, but lacking both
(GrB) deficient mice and humans has demonstrated a critical an electrophilic serine trap and caspase activity, were prepared
role for this enzyme, a lymphocyte serine protease and one of to improve activity against human granzyme B.5
the most abundant granzymes in CTL granules. GrB causes
1,2,3 Triazole substitution at P1 (modification of the serine
trap) and a rigid tricyclic group in the P2–P3 position of tri-
azole 2 furnished a 2-fold improvement (KiGrB = 38 nM) over 1
(KiGrB = 80 nM). By adding hydrophobic phenyl and benzothio-
phene groups at the P5 position of triazole 2 to produce tri-
azoles 3 and 4, respectively, enzyme and cellular activities were
improved 3 to 5-fold and 5 to 10-fold, respectively. According
to docking studies we have performed, this is most likely due
to a strong non-covalent cation–π interaction6 between the
phenyl of 3 and benzothiophene ring of 4 and Arg217 (Fig. 2b).
aDepartment of Chemistry, 1515 Dickey Drive, Emory University, Atlanta, GA 30322,
USA. E-mail: jsnyder@emory.edu
bDepartment of Chemistry, Imperial College London, London, SW7 2AZ, UK
cSection of Immunobiology, Division of Inflammation and Immunology, Department
of Medicine, Faculty of Medicine, Imperial College London, London SW7 2AZ, UK
dEmory Institute for Drug Development, 1515 Dickey Drive, Emory University,
Atlanta, GA 30322, USA
†Electronic supplementary information (ESI) available: 1H and 13C NMR spectra
for the synthesized compounds. See DOI: 10.1039/c4ob01874e
8952 | Org. Biomol. Chem., 2014, 12, 8952–8965
This journal is © The Royal Society of Chemistry 2014

View Article Online
Organic & Biomolecular Chemistry
Paper
Fig. 1 Structures of Merck inhibitors and their actions against GrB enzyme and CTL apoptosis, respectively. (a) 1 (tetrapeptide aldehyde IEPD-CHO,
KiGrB = 80 nM), (b) 2 (KiGrB = 38 nM/IC50_apoptosis = 30 µM), (c) 3 (KiGrB = 13 nM/IC50_apoptosis = 6.3 µM), and (d) 4 (KiGrB = 7 nM/IC50_apoptosis = 3.1 µM).
Fig. 2 Cation–π interactions at P5. (a) Alignment of GrB complexes of docked triazole 3 (blue) and triazole 4 (magenta) highlighting the ligands sur-
rounded by the binding pocket surface; (b) Cation–π interaction between Arg217 and P5-phenyl and benzothiophene rings, respectively. The pro-
posed poses were generated by ligand docking studies employing the covalent inhibitor 1/GrB bound complex (PDB ID: 1IAU) with the ligand
removed.4
To our knowledge, there has been no follow-up on 1,2,3-
triazoles 2, 3 and 4 and no alternative small molecule
Computational solvent mapping
approaches since 2002. Thus, we initiated the present work to To identify lead compounds from commercially available
identify a novel class of non-covalent small molecule inhibitors libraries, docking studies are frequently employed for virtual
for human granzyme B (hGrB) by employing virtual screening screening. Even though various algorithms and scoring func-
strategies derived by examining the X-ray structure of human tions have been used in attempts to predict realistic binding
GrB bound covalently to a potent tetrapeptide aldehyde inhibi- poses and rank them,10–12 the following questions are still rele-
tor. Since the key determinants of substrate specificity and vant: which poses are true conformations in the receptor and
binding almost universally reside in the non-prime
region,7–9 these sites were incorporated into the strategy.
P
which residues are important for ligand binding.
According to some studies, individual residues in the
We report the discovery of novel, modestly active, small binding region contribute unequally to the free energy of
molecule inhibitors with selectivity against granzyme A as well binding, while smaller regions of the binding pocket can
as caspase-3 and 8 resulting from virtual screening. Moreover, provide major contributions to it.13 These smaller sectors of
we illustrate that a comprehensive structure-based design cam- high or special activity are regarded as ‘hot spots’. Such mole-
paign for non-covalent blockers that relies solely on a covalently cular zones are more likely to bind drug-like molecules with
bound ligand and the corresponding X-ray structure complex higher affinity than the rest of the binding site.14 As a conse-
may well be compromised by the existence of alternative and quence, the identification of hot spots is an important task
unknown binding poses.
for drug design. One solution to this problem utilizes
This journal is © The Royal Society of Chemistry 2014
Org. Biomol. Chem., 2014, 12, 8952–8965 | 8953

View Article Online
Paper
Organic & Biomolecular Chemistry
Fig. 3 Outline of computational solvent mapping algorithm used by FTMAP.
computational solvent mapping to explore the protein’s (Fig. 4a). Solvent mapping shows that probes can be located
binding properties and to assist understanding of interactions deeper in the S1 subsite (orange spheres) for the bound blocker
between proteins and solvent molecules. The FTMAP algor- as represented by the increased number of probes in the spa-
ithm achieves such a result by placing multiple copies of 16 cious S1 pocket induced by ligand binding (Fig. 4b; orange
different small organic molecular probes on a protein surface circle). This is in good agreement with the X-ray results.
by means of an extremely efficient fast Fourier transform (FFT)
The S2 subsite/P2-Pro. The X-ray structure of the covalently
correlation approach.15 For each of the 16 probes, the bound complex illustrates that a hydrophobic wall is generated
mapping algorithm generates 2000 best-bound positions by by Phe99 and the flipping of His57 (Fig. 5; pink tubes),
rigid body docking followed by energy minimization to refine causing P2-Pro to be directed away from the body of the
the probe positions. The resulting conformations are clus- protein so as to make no specific interactions with it. However,
tered, and all clusters are ranked on the basis of average free binding induces a slight kink in the inhibitor allowing the
energy. For each probe, six clusters with the lowest average free P1-Asp to achieve its observed conformation in S1. Solvent
energy are retained. Overlapping clusters of different probes mapping identified the newly generated pocket caused by dis-
are defined as consensus sites (CSs). The largest CS is gener- placements of Phe99 and His57 as well as Pro96, Try94 and
ally located at the most important subsite of the protein Asp102 (orange circle). Accordingly, additional probes (orange
binding site, and the nearby smaller CSs identify other impor- spheres) are posited to bind deeply in the S2 pocket (Fig. 5).
tant subsites (Fig. 3).15
Summary of GrB mapping results
The two X-ray structures of GrB, the apo form (PDB ID:
1FQ3)16 and the covalent inhibitor (Ac-IEPD-CHO, Fig. 2)
Table 1 lists the eight largest consensus subsites (CS) located
bound complex (PDB ID: 1IAU),4 were examined to compare side
in and nearby the active site for each protein structure. The
chain locations in the respective binding sites and subsequently
mapped to identify hot spots. The assessment, discussed below,
reveals which changes occur after covalent binding takes place.
Prior to applying the mapping procedure, the inhibitor and water
molecules in both GrB representations were removed.
two largest CSs for apo GrB are the non-prime binding sites S1
and S4 with 27 and 14 probe clusters, respectively (Table 1 and
Fig. 6a). The newly generated site in the extended S2 (CS6)
groove is populated by 9 probes (Table 1 and Fig. 6a).
Following covalent inhibitor binding, the conformation of
the protein is altered, especially in the binding site. For both
protein X-ray structures, FTMAP identified all GrB subsites
(S4–S2′) that bind organic probes in the full active site (Table 1
Two key features accompany inhibitor binding in the S1 and
S2 subsites
S1 subsite/P1-Asp. Changes in both size and position of the and Fig. 6b). It is noteworthy that inhibitor binding engenders
S1 subsite from the apo form to the bound form in the X-ray the formation of both the primed subsites (S1′–S2′ (CS5 and
structures, especially due to the movements of Arg226 and CS7) in Fig. 6b) and new pockets deep in S2 (CS4, Fig. 6b).
Asp194, increases the volume of S1 allowing three water
Fig. 7 depicts the distribution of non-bonded interactions
molecules (red spheres) to be bound along with the inhibitor between probes and GrB residues in the active site. Prior to
8954 | Org. Biomol. Chem., 2014, 12, 8952–8965
This journal is © The Royal Society of Chemistry 2014

View Article Online
Organic & Biomolecular Chemistry
Paper
Fig. 4 Water molecules in the X-ray structure S1 pocket and alignment of consensus sites for the apo and inhibitor bound forms of GrB. (a) Protein
residues and inhibitor are displayed with tube and ball and stick representations, respectively; dark blue (before inhibitor binding) vs. light blue (after
inhibitor binding). The volume increase in the bound form allows three water molecules (red spheres) to occupy S1 simultaneously with the inhibitor.
(b) Alignments of the consensus sites (spheres) resulting from solvent mapping; blue spheres (apo structure before inhibitor binding) vs. orange
spheres (bound form after inhibitor binding). More probes can be located deeper in the S1 subsite (orange circle) for the bound blocker as rep-
resented by the increased number of probes in the spacious S1 pocket induced by ligand binding.
Table 1 Summary of GrB mapping results
1FQ3
1IAU
Apo GrB
Sizea
Ac-IEPD-CHO-GrB
CS
Pocket
Sizea
Pocket
1
2
3
4
5
6
7
8
27
14
14
11
9
9
7
S3–S4
S1–S2
17
16
13
12
11
9
S4
Not active site
S1
S2
S1′
Not active site
S2′
Not active site
Not active site
Not active site
Not active site
Extended S2
Not active site
Not active site
8
8
2
a Size: the number of probe clusters. For each of the 16 probes, 2000
best poses rescored by energy minimization are grouped into clusters
using a simple greedy algorithm. These are further ranked on the basis
of averaged energies, and six clusters for each probe with the lowest
average free energies are finally retained. All clusters were
superimposed to identify the consensus sites (CSs). The largest CS1
with the greatest number of probe clusters generally identifies the
most important subsite, namely a hot spot.
Fig. 5 Alignment of consensus sites resulting from solvent mapping;
blue spheres (apo structure before inhibitor binding) and orange spheres
(bound form after inhibitor binding). Protein residues are displayed as
tubes; dark brown (before inhibitor binding) vs. pink (after inhibitor
binding). Additional probes (orange spheres) are posited to bind deeply
in the newly generated pocket (orange circle) of S2 caused by displace-
ments of nearby residues.
interactions with residues of S2′ and S1′ are increased. On the
other hand interactions with Leu171 of S4 disappear, possibly
increasing the hydrophobicity of S4. Surprisingly, the number
of probe contacts with Arg226 in the S1 pocket decreased dra-
matically after inhibitor binding. By contrast, interactions with
Ser214, Gly193 and Ser195 increased.
For both non-bonded and hydrogen bonding interactions,
different conformations of the protein induced by ligand
binding result in changes for the number of probe contacts
even with the same residues. Since the increased contact
numbers imply a higher affinity between a given residue and
the ligand, such residues were considered to be particularly
important for selecting constraints in the subsequent virtual
screening exercises.
inhibitor binding, major residues participating in non-bonded
interactions are His57 of S2, Leu171, Tyr174 and Tyr215 of S4.
After inhibitor binding, probe interactions with Arg41 of S2′,
Lys40 of S1′, Phe191 of S1 and Phe99 of S2 are increased. On
the other hand, interactions with Leu171 of S4 are decreased.
In both cases, Lys192 extends from S1 to S3 and establishes a
barrier between these sites and the solvent pool.
In the same way, Fig. 8 shows the distribution of hydrogen
bonding interactions between probes and GrB residues in both
active sites. Before inhibitor binding, the major residues parti-
cipating in H-bonding interactions are Lys40 of S1′, Arg226,
Ser214, Gly193, Ser195 of S1, His57 of S2, Gly216 of S3, Leu171
of S4 and Lys192 as part of the wall between S1–S3. After
inhibitor binding, dramatic changes were evident: namely,
This journal is © The Royal Society of Chemistry 2014
Org. Biomol. Chem., 2014, 12, 8952–8965 | 8955

View Article Online
Paper
Organic & Biomolecular Chemistry
Fig. 6 Consensus sites in the binding pockets of (a) apo GrB (PDB code 1FQ3) and (b) Ac-IEPD-CHO-GrB (PDB code 1IAU) highlighting reorganiz-
ation of the binding site residues.
Fig. 7 Intermolecular non-bonded interactions between probes and residues for apo GrB (blue) and Ac-IEPD-CHO-bound GrB (red) without the
ligand.
Fig. 8 Intermolecular hydrogen bond interactions between probes and residues for apo GrB (blue) and Ac-IEPD-CHO-bound GrB (red) without the
ligand.
8956 | Org. Biomol. Chem., 2014, 12, 8952–8965
This journal is © The Royal Society of Chemistry 2014

View Article Online
Organic & Biomolecular Chemistry
Paper
Fig. 9 Constraints selected for docking based on probe mapping: (a) Occupation of the deeper S2 pocket generated after inhibitor binding (circle
in yellow), and (b) hydrophobic interactions in the S4 pocket (cubes in pink).
Three constraints
menting the trio of constraints in Glide’s receptor grid file
during the docking exercise.
Based on the mapping results for the Ac-IEPD-CHO bound
protein, three constraints were selected for docking studies. (1)
A hydrogen bond with Arg226 in S1 to identify molecules with
a P1-Asp specificity motif. Although the number of H-bond
probe contacts with Arg226 decreased after inhibitor binding
(Fig. 8), most likely due to the movement of Arg226 deeper
into the larger S1 subsite, it is known that hGrB strongly
prefers P1-Asp based on positional scanning with a combina-
torial peptide library.3 Moreover, the Merck compounds
possess electrophilic groups promoting activity.5 (2) An occu-
pied new pocket of S2 (Fig. 9a); and (3) A hydrophobic inter-
action within the S4 pocket (Fig. 9b) to accommodate the
newly generated hydrophobic centers in the S2 and S4 pockets
upon ligand binding.
Second, extra-precision (XP) docking,23 which was develo-
ped to weed out false positives and provide a better correlation
between good poses and good scores, was performed with the
715 and 3634 hits from Maybridge SC and ChemDiv, respecti-
vely, to select more reasonable poses and reduce the number
of purchased compounds for the first round of bio-assay.
Finally, a total of 511 available compounds (46 from Maybridge
SC, 203 from ActiTarg-P and 262 from ChemDiv) were ordered
for bio-assay. According to the suppliers’ quality control speci-
fications, all commercial compounds ordered from the three
library suppliers were >90% pure with an average of 95%
purity as established by LC/MS.
Non-prime subsites are the focus of inhibitor design as men-
tioned in the introduction. Therefore, interactions with residues
in the prime subsites were not considered in this study.
Bioassay and selectivity test
The 511 selected hits were tested at 100 µM. A total of 28 com-
pounds (21 compounds from TimTec and 7 compounds from
ChemDiv) showing greater than 40% inhibition of GrB were
subjected to serial dilution for determination of IC50 values.
Compounds blocking below 50 µM were subjected to
selectivity testing against the GrA isoform, which is the most
abundant granzyme and expressed earlier than other granzymes
in cytotoxic lymphocytes along with GrB.24,25 In addition, since
GrB (serine protease) and the caspases (cysteine protease) are
the only proteases involved in cell death, selectivity evaluation
was also performed against caspase-3 (effector caspase) and
caspase-8 (initiator caspase which activates effector caspases).
Even though these enzymes show distinct catalytic mechanisms
relative to other classes of enzymes and are unrelated in
sequence, they share both substrate specificity for P1-Asp and a
biological role in promoting apoptosis.
Virtual screening
Four libraries were selected for virtual screening of the con-
strained cleft described above: the Maybridge Screening Collec-
tion17 which consists of over 56 000 organic compounds, the
ActiTarg-P18 set (1520 structures) provided by TimTec that
affords chemical structures reported in the technical or patent
literature to exhibit protease-inhibiting properties, and
ChemDiv19 subsets of 8300 and 430 compounds from the pro-
tease library and those with the aspartate recognition motif,
respectively.
All compounds from the three libraries (total: over 66 000
structures) were screened by Glide20–22 standard-precision (SP)
docking using the three constraints generated by solvent
mapping. As mentioned above, the latter identifies binding
‘hot spots’ that provide major contributions to the binding
free energy following placement of 16 diverse organic mole-
Discussion
cular probes on the protein surface. As a result, a total Of 28 compounds showing IC50 values less than 100 µM, 18 com-
4552 hits (715 from Maybridge Screening Collection, 203 from pounds delivered inhibition below 50 µM and exhibited promis-
ActiTarg-P and 3634 from ChemDiv) were identified by imple- ing selectivity against GrA as well as caspase-3 and 8 (cf. Table 2).
This journal is © The Royal Society of Chemistry 2014
Org. Biomol. Chem., 2014, 12, 8952–8965 | 8957

View Article Online
Paper
Organic & Biomolecular Chemistry
Table 2 IC50 values for 5–12 against GrB, GrA and caspase-3 and 8
IC50 (μM)
Compound
GrB
GrA
Caspase-3
Caspase-8
5
6
7
8
35
44
44
25
28
45
30
30
95
>100
>100
>100
>100
>100
>100
>100
>100
82
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
>100
50
9
10
11
12
68
Interestingly, most of the hits incorporate one of the follow-
ing moieties in the structure: thiotetrazole 5 (ST053449,
Fig. 10a), thiotriazole 6 (ST057416, Fig. 10b), diazolidinedione
Fig. 11 Tetrazole analog 5 pose: The predicted binding mode sustains a
hydrogen bond between Arg226 and a tetrazole nitrogen as indicated by
the yellow dotted line (2.45 Å).
7
(ST058025, Fig. 10c), thiazolidinediones 8 (ST057924,
Fig. 10d) and 9 (ST057833, Fig. 10e), or sulfonamide 10 (L581-
0227, Fig. 10f).
Another interesting series is represented by compounds 11
(G202-0149) and 12 (E627-0664) (Fig. 12), which incorporate a
tricyclic ring system as do triazoles 2, 3 and 4 (Fig. 1b–d). In
the latter series, the rigid tricyclic scaffold led to an increase in
activity.5 The predicted binding poses of tricyclics 11 (IC50_GrB
= 30 µM) and 12 (IC50_GrB = 30 µM), when superimposed with
triazole 4 to compare the relative positions of the tricyclic
rings (blue circles), evidenced good alignment (Fig. 12).
The enzyme activities of the tetrapeptide aldehyde inhibitor
IEPD-CHO and compounds 5–12 and their IC50 values against
GrB, GrA, and Caspase-3 and 8 are presented in Fig. 13 and
Table 2, respectively.
In addition, the agents are similar in structure and can be
categorized into several groups. These observations seemed to
imply that the compounds were the basis for an emerging SAR
as inhibitors for GrB. Fig. 11 depicts the predicted binding
mode for thiotetrazole 5, the class with the lowest molecular
weight. The tetrazole ring is located in the P1-Asp recognition
position, forming a weak H-bond with Arg226 (2.45 Å). In
addition, the model suggests there is space to add additional
functional groups for both hydrogen bonding and hydro-
phobic interactions in the deep S1 pocket as well as the poten-
tial to improve the latter at P2 and P4.
Fig. 10 Structures of hit compounds from virtual screening and their IC50 values against GrB. (a) 5 (IC50 = 35 µM), (b) 6 (IC50 = 44 µM), (c) 7 (IC50
=
44 µM), (d) 8 (IC50 = 25 µM), (e) 9 (IC50 = 28 µM), and (f) 10 (IC50 = 45 µM).
8958 | Org. Biomol. Chem., 2014, 12, 8952–8965
This journal is © The Royal Society of Chemistry 2014

View Article Online
Organic & Biomolecular Chemistry
Paper
Fig. 12 Alignment of the predicted GrB-ligand complexes (a) 11 and (b) 12 with 4 (blue). The protein structure is not shown. The relative positions
of the tricyclic rings are aligned at the same P2 positions in the protein.
Fig. 13 % Enzyme activity of GrB treated with tetrapeptide aldehyde inhibitor IEPD-CHO (KiGrB = 80 nM) and the virtual screening hits 5–12.
At this point, with promising scaffolds and maximal inhi- pounds. Thiotetrazole 5 was chosen as a starting point due to
bition from the thiazolidinedione series 8 (IC50 = 25 µM) and its low molecular weight and simple structure. Accordingly, 5
9 (IC50 = 28 µM), we extended the screening to additional and the corresponding bromo-analog 19 (Scheme 1, SI p4†)
libraries in an effort to improve potency by seeking novel were prepared as reference compounds.
scaffolds with dissimilar structures. Unfortunately, the best
Docking of tetrazole 5 into the binding pocket of granzyme
compound proved to be a disappointingly weak blocker with B showed some unoccupied space in the S1 pocket (Fig. 11).
an IC50 value of 73 μM (SI, p2–3†).
This observation led to the design of 1,2,4-triazole 24a with
the introduction of a pyridine ring to improve occupation of S1
after replacing the tetrazole with a triazole. This new design
led to the possibility of an additional hydrogen bond between
the pyridine nitrogen and Arg226. Adding a second nitrogen
atom within the ring led to pyrimidine analog 24b, which gave
a more favorable docking pose due to the symmetrical charac-
ter of this substituent (Scheme 2, SI p4†).
Synthesis and testing of novel
compounds
In addition to applying a scaffold hopping approach to
improve potency, we also initiated the synthesis of novel com-
This journal is © The Royal Society of Chemistry 2014
Org. Biomol. Chem., 2014, 12, 8952–8965 | 8959

View Article Online
Paper
Organic & Biomolecular Chemistry
Compounds 33, 42 and 49 were assayed against GrB (Fig. 14
and 15). While 33 showed weak activity (IC50 = 47 μM), no sig-
nificant antagonism was observed for 42 and 49.
Covalent vs. non-covalent models
Even though intensive computational searches were conducted
and novel synthetic compounds were prepared to identify
promising non-covalent candidates and improve GrB potency,
the activities of prospective hit molecules in the present case
failed to drop below 25 μM. Most likely, this results from using
an inappropriate conformation presented by the covalently-
bound inhibitor, Ac-IEPD-CHO, as a model for scaffold
hopping, docking-directed virtual screening and synthetic
strategies. Indeed, the literature reveals that certain non-
covalent small molecules adopt distinct binding modes when
compared with covalent inhibitors resulting from movements
of residue side chains in the binding pocket (see below).
To be sure, successful cases of the identification of non-
covalent drug candidates using the complexed X-ray structures
with covalent bound inhibitors have been reported. For
example, Lu et al. performed a structure-based virtual screen
with the Maybridge library (58 855 compounds) to discover
novel non-peptide inhibitors for the main coronavirus protease
responsible for severe acute respiratory syndrome (SARS-
CoV Mpro).26 The study employed the structure of SARS-CoV Mpro
in complex with a chloromethyl ketone (CMK) inhibitor, a co-
valently bound substrate analog, for initial screening. Two
active hits defined by subsequent analog searching were crys-
tallized to compare binding modes with a number of co-
valently bound peptidyl inhibitors. The X-ray structure of the
complex with the most potent compound (IC50 = 0.3 μM)
experienced induced protein conformational changes in the
binding pocket, disrupting the catalytic dyad (His41 and
Cys145).
Fig. 14 Dose dependent GrB activity of tetrapeptide aldehyde inhibitor
IEPD-CHO (KiGrB = 80 nM) and compounds 33 and 42.
In an effort to better increase the binding affinity of the
ligand for the protease, the molecule was further extended in
two directions. First, we sought to probe enhanced binding in
the S2 pocket by replacing the cyclohexyl group with different
substituents. Subsequently, the western part of the molecule
was exchanged with the published peptoid Merck side-chain5
to augment binding via an increase in hydrogen-bonding
within the S3 and S4 pockets. The first of these new analogs to
be synthesized was triazole-pyrimidine 33, with
a para-
(hydroxyethyl)phenyl substituent in P2 (Scheme 3, SI p5–6†).
In order to investigate different substitution patterns in the S2
pocket, a smaller aromatic substituent, para-fluorophenyl, was
introduced to provide analog 42 (Scheme 4, SI p7†).
The final analog to be synthesized, pyrimidine-triazole 49,
contains a cyclopropyl substituent on N-4 of the triazole
(Scheme 5, SI p8†). The electronic nature of this substituent
suggested it might impart novel properties during binding,
but the structure required the design of a different synthetic
route. The cyclodehydration step proceeded in very low yield
with the desired triazole, necessitating development of an
alternative strategy. Thus, the Boc-protected amine was
replaced with an azide in order to overcome the low-yield
deprotection step encountered previously, possibly due to
unfavorable solubility.
In some instances, however, even compounds with very
similar structures and binding affinities adopt quite different
binding poses. For example, Mattos et al. demonstrated that
non-covalently bound trifluoroacetyl (TFA)-Lys-Pro-isopropyl-
Fig. 15 Dose dependent GrB activity of the tetrapeptide aldehyde inhibitor IEPD-CHO (KiGrB = 80 nM) and compound 49.
8960 | Org. Biomol. Chem., 2014, 12, 8952–8965
This journal is © The Royal Society of Chemistry 2014

View Article Online
Organic & Biomolecular Chemistry
Paper
Fig. 16 Structures of non-covalent elastase inhibitors. (a) TFA-Lys-Pro-ISO (50), (b) TFA-Lys-Phe-ISO (51) and (c) TFA-Lys-Leu-ISO (52).
anilide (ISO) (Fig. 16, compound 50) binds to elastase in the bound inhibitor as well as a model for the non-covalently
same fashion as blockers that form covalent acyl enzyme inter- bound inhibitor depicted in Fig. 2; namely, compound 4
mediates.27 However, replacing Pro with Phe or Leu (TFA-Lys- docked into the protein of the covalent complex relieved of the
Phe-ISO (Fig. 16, compound 51) and TFA-Lys-Leu-ISO (Fig. 16, bound peptide inhibitor.
compound 52)) results in quite different non-covalent binding
modes (SI p61†).
Fig. 17 depicts a 10 nanosecond MD trace for the apo-form
of GrB housing a sulfate ion in the binding site. The trajectory
The lesson from the two cases just described appears to be (Fig. 17a) reveals that the entire protein, including the binding
that multiple binding poses are available to potential inhibi- site, is highly mobile at 300 K, implying that a variety of
tors of GrB with similar structures. To test this hypothesis for binding poses are available, in principle, to a flexible small
GrB, we performed a series of molecular dynamics (MD) calcu- molecule partner. Once tethered, however, as is the case for
lations for the X-ray structures of apo-GrB and the covalently the Ac-IEPD-CHO covalently bound blocker (Fig. 18b), the
Fig. 17 Molecular dynamics trajectory (10 ns, 300k) for apo-GrB enzyme (1FQ3). (a) Backbone RMSF in Å plotted vs. residue number. Residues with
higher values of RMSF denotes greater mobility; (b) RMSF data mapped on the GrB binding site backbone. Red sectors reveal the larger RMSF values.
The active site Ser 195 nucleophile is shown as a green stick diagram.
Fig. 18 Molecular dynamics trajectory (10 ns, 300 K) for the covalently bound Ac-IEPD-CHO peptide to the GrB enzyme (1IAU). (a) Backbone RMSF
in Å plotted vs. residue number. Residues with higher values of RMSF denotes greater mobility; (b) RMSF data mapped on the GrB binding site back-
bone. Red sectors reveal the larger RMSF values. The active site Ser 195 and the covalent peptide ligand are shown as a green stick diagram.
This journal is © The Royal Society of Chemistry 2014
Org. Biomol. Chem., 2014, 12, 8952–8965 | 8961

View Article Online
Paper
Organic & Biomolecular Chemistry
Fig. 19 Molecular dynamics trajectory (10 ns, 300 K) for non-covalently bound 4 to the GrB enzyme (1IAU). (a) Backbone RMSF in Å plotted vs.
residue number. Residues with higher values of RMSF denotes greater mobility; (b) RMSF data mapped on the GrB binding site backbone. Red
sectors reveal the larger RMSF values. The active site Ser 195 and the non-covalent ligand 4 are shown as a green stick diagram.
entire protein tightens up in response to forming the complex to GrB are fewer and, therefore, cooperatively weaker. Namely,
(Fig. 18a).
while the nanomolar potent compounds are optimally folded
Examination of the non-covalent complex between GrB and into the binding pocket, the latter would seem to be poorly
4 (Fig. 19b) exhibits an MD profile (Fig. 19a) very similar to adapted to the same template. Apparently, the use of the
that displayed by the covalent complex of Fig. 18. This FTMAP hot spots derived from IEPD-CHO not only margina-
outcome, which includes the observation that tethered Ac- lizes H-bonding for the modeled poses of 5, 11 and 12, but
IEPD-CHO and untethered 4 map nicely onto one another, also results in poorly adapted conformations for alternative
reinforces the judgment that a covalent complex is a reason- poses available to the micromolar-active ligands in the actual
able template for non-covalent molecular design as cited for experimental binding event. Thus, it appears that a successful
SARS-CoV Mpro above. However, it appears that the analogy is campaign to identify novel GrB inhibitors by virtual screening
not with the latter case, but with the elastase example in which requires sampling of multiple binding poses that are not rep-
more than one binding pose is available but not captured by resented by the covalently bound IEPD-CHO inhibitor, ones
the 10 ns trajectory of Fig. 18. A much longer MD treatment or that utilize low energy conformers, retain a high level of hydro-
the exploration of alternative binding poses as MD starting phobic contacts and result in increased numbers of H-bonded
points would most likely identify other equally viable or anchor points.
superior binding modes.
To explore this point from another perspective, we per-
formed
a
ligand–protein contact (LPC) analysis28 for
Conclusions
IEPD-CHO, GrB-docked highly potent non-covalent Merck GrB
blockers 3 and 4 and the best compounds uncovered in the
present study (5, 11 and 12). Table 3 summarizes the results,
which are depicted graphically in the ESI (SI p63†).
It is striking that all the compounds engage in a large
number of ligand–protein hydrophobic interactions spread
across the diverse molecular frameworks. However, the extent
of hydrogen bonding between the first three and last three
entries of Table 3 is significantly different, implying that the
essential H-bonded anchor points for binding of 5, 11 and 12
Novel classes of weak small molecule inhibitors against
human GrB have been developed by docking studies employ-
ing binding site hot spots and three constraints (hydrogen
bonding with Arg226, hydrophobic interactions for S2 and S4
subsites) based on computational solvent mapping using
FTMAP. The most distinctive compounds identified in this
work were thiazolidinediones 8 (IC50 = 25 µM) and 9 (IC50 = 28
µM), triazole 6 (IC50 = 44 µM) and diazolidinedione 7 (IC50
=
44 µM). All show selectivity against GrA as well as caspase-3
and 8. In order for the structures to form hydrogen bonds with
Arg226 at S1, benzodioxin (6), methoxy (7), acetamide (8) and
nitro (9) substituents in the P1-Asp recognition position
appear to be needed. The compounds might also be produc-
tively modified for improved activity by adding more hydro-
phobic groups at the P2 and P4 positions, although Table 3
suggests this is unlikely to be fruitful.
Table 3 Ligand–protein contact (LPC) analysis for covalently bound
IEPD-CHO and five non-covalently docked ligands based on FTMAP
“hot spots” derived from the IEPD-CHO X-ray structure
# of
# of
H-bonds
# of hydrophobic
interactions
Compound
Activity
atoms
The 25 μM IC50 achieved following an extensive screening
campaign based on the only known structure for a GrB-bound
inhibitor is disappointing. Analysis of the situation for
SARS-CoV Mpro and elastase, MD calculations for GrB and LPC
interrogation of the ligands in Table 3 strongly suggest that for
IEPD-CHO
3
4
5
11
12
KiGrB = 80 nM
KiGrB = 13 nM
KiGrB = 7 nM
IC50 = 35 μM
IC50 = 30 μM
IC50 = 30 μM
35
41
44
22
33
37
26
17
19
11
4
46
27
25
33
31
36
4
8962 | Org. Biomol. Chem., 2014, 12, 8952–8965
This journal is © The Royal Society of Chemistry 2014

View Article Online
Organic & Biomolecular Chemistry
Paper
certain enzymes the binding site plasticity leads to the gene- Ligand–protein contact analysis
ration of multiple ligand binding poses. Thus, we regard the
The web-based LPC analysis package28 was applied to the
structures summarized in Table 3. In order to compare the co-
valently bound IEPD-CHO inhibitor with the non-covalently
docked ligands, it was necessary remove the CHO fragment in
the 1IAU X-ray complex and eliminate the covalent connection,
while retaining the remaining geometry of the bound peptide.
GrB system as an exception to the similarity-based approach to
discovery of novel inhibitors; an exception that will most likely
be fully understood in the present case only with the gene-
ration of X-ray structures of Granzyme B and potent non-
covalent blockers such as 3 and 4.
Screening assays
Experimental details; methods
Computational details
Materials. Granzyme B assay: Enzolifesciences (BML-SE238),
Ac-IEPD-pNA Enzolifesciences (BML-P133-9090); Ac-IEPD-CHO
Biovision (1119-1).
Granzyme A assay: Enzolifesciences (ALX-201-118-C010);
Z-GPA-pNA Bachem (L-1560.0100); 3,4-dichloroisocoumarin
Enzolifesciences (BML-PI110-0010).
Caspase 3 assay: Enzolifesciences (BML-SE169), Ac-DEVD-
pNA Enzolifesciences (BML-P142-9090); Ac-DEVD-CHO Enzolife-
sciences (BML-P410-9090).
Caspase 8 assay: Enzolifesciences (BML-SE172); Ac-IETD-
pNA Enzolifesciences (BML-P431-9090); Ac-IETD-CHO Enzolife-
sciences (BML-430-9090).
Protein and ligand preparation. Both the protein and the
ligands from libraries were prepared for further modeling
studies since unreasonable structures affect the quality of
results. For the apo-protein with a single sulfate in the binding
site (PDB ID: 1FQ3) and the covalently-bound complex (PDB
ID: 1IAU), bond orders were assigned, hydrogens were added,
and waters were deleted beyond 5 Å from hetero atoms using
“Protein Preparation Wizard” of Schrödinger.29 Meanwhile,
the compounds in the libraries were put through “LigPrep” of
Schrödinger30 in order to generate high quality 3D structures;
hydrogens were added, charged acidic or basic groups were
neutralized, and tautomers and only the lowest energy confor-
mers were also generated.
Assay Buffer (50 mM HEPES (pH 7.4), 100 mM NaCl, 0.1%
CHAPS,
1
mM EDTA, 10% Glycerol) Enzolifesciences
(BML-KI111).
Ligand docking. An appropriately sized grid, representing
the physical properties of the protein binding site for the co-
Methods
Granzyme A/B assay. Protease assays were performed using
valently bound complex (PDB ID: 1IAU) with the ligand recombinant full-length human granzyme B & granzyme A
deleted, was generated. Glide searches20–22 seeking favorable generated at Enzolifesciences (BML-SE238; ALX-201-118-C010
interactions between conformations automatically generated respectively). Protease assays were performed using an assay
for each ligand and the protein within the grid, were followed buffer consisting of 50 mM HEPES (pH 7.4), 100 mM NaCl,
by energy minimization of binding poses. Final scoring was 0.1% CHAPS, 1 mM EDTA, 10% Glycerol, into which were
carried out with GlideScore multi-ligand scoring function.
added 49.5 or 60 Units of active granzyme B or Granzyme A
Scaffold hopping. Pre-computed conformers for each com- respectively. Inhibitors where dissolved in DMSO at concen-
pound in each library were aligned to each query by ROCS to tration to a 10 mM stock. Inhibitors where dissolved in DMSO
find and quantify the maximal overlap of the volume and to a 10 mM stock. Inhibitors where diluted to 100 μM in the
“color” atoms. The most common scoring function, Tanimoto- enzyme/assay buffer and incubated for 10 minutes at 30 °C.
Combo (shape + color) was used because the query and library Assay controls included enzyme/no enzyme and equal volume
molecules are of similar size. Since a high score (high simi- of DMSO as a vehicle. Following incubation the reaction was
larity) from ROCS searching does not guarantee true activity, commenced by addition of purified substrate Granzyme B (Ac-
the top 500 hits from each search were docked into the active IEPD-pNA) or Granzyme A (Z-GPA-pNA) to a final concentration
site of GrB using Glide and the three constraints derived by of 200 µM. Initial reactions where read at 405 nm in a micro-
computational solvent mapping. In order to reevaluate the plate-reading spectrophotometer at time 0 and at 30-minute
structures rejected by the rigid docking, induced fit docking end point. Time 0 reactions subtracted from 30-minute end-
was employed as a second step for the top 20 compounds from point as to eliminate auto-absorbance. All compound inhi-
each subset of 500 hits. In this way, the enzyme is allowed to bitions where calculated against values for enzyme only (with
alter the conformation of binding site side chains so as to DMSO) to determine inhibition profiles. Following initial 100
more closely complement the shape and binding mode of the µM screening, those compounds that indicated greater than
ligand.
50% enzyme inhibition where then screened at the following
Molecular dynamics. The protein-prepared GrB X-ray struc- concentrations (100, 50, 25 & 12.5 µM) to determine IC50.
tures (PDB IDs: 1FQ3 and 1IAU) and the GrB (1IAU)/4 complex Granzyme B inhibition was monitored using a commercial
minus the covalent ligand were subjected to Desmond v3.4 inhibitor Ac-IEPD-CHO generated at Biovision (Cat: 1119-1)
OPLS-2005 molecular dynamics in Maestro initiated with the which has an IC50 of 80 nM for granzyme B. Granzyme A inhi-
protein relaxation option.31,32 Each of the three complexes bition was monitored using a commercial inhibitor 3,4-
were subjected to MD at 300 K, 1 atm, SPC water solvent, peri- dichloroisocoumarin generated at Enzolifesciences (BML-
odic boundary conditions in a cubic box for 10 ns.
PI110-0010) which has an IC50 of 12.5 µM for granzyme A.
This journal is © The Royal Society of Chemistry 2014
Org. Biomol. Chem., 2014, 12, 8952–8965 | 8963

View Article Online
Paper
Organic & Biomolecular Chemistry
Caspase 3 and 8. Protease assays were performed using
recombinant full length human caspase 3 & granzyme 8 gene-
rated at Enzolifesciences (BML-SE169; BML-SE172 respect-
Acknowledgements
ively). Protease assays were performed using an assay buffer We are grateful to Open Eye Scientific Software for generous
consisting of 50 mM HEPES (pH 7.4), 100 mM NaCl, 0.1% provision of computer code enabling molecular comparison
CHAPS, 1 mM EDTA, 10% glycerol, into which were added 50 (ROCS) and many other molecular design tasks. We are also
Units of either active caspase 3 or Caspase 8 respectively. thankful to Professor Dennis Liotta (Emory University) for
Inhibitors where dissolved in DMSO at concentration to a partial support of this work and Dr Pieter Burger for assistance
10 mM stock. Inhibitors where dissolved in DMSO to a 10 mM in setting up the Desmond 3.4 MD calculations.
stock. Inhibitors where diluted to 100 µM in the enzyme/assay
buffer and incubated for 10 minutes at 30 °C. Assay controls
included enzyme/no enzyme and equal volume of DMSO as a
vehicle. Following incubation the reaction was commenced by
addition of purified substrate Caspase 3 (Ac-DEVD-pNA) or
References
Caspase 8 (Ac-IETD-pNA) to a final concentration of 200 µM.
Initial reactions where read at 405 nm in a microplate-reading
spectrophotometer at time 0 and at 30-minute end point. Time
0 reactions subtracted from 30-minute endpoint as to elimin-
ate auto-absorbance. All compound inhibitions where calcu-
lated against values for enzyme only (with DMSO) to
determine inhibition profiles. Following initial 100 µM
screening, those compounds that indicated greater than 50%
enzyme inhibition where then screened at the following con-
centrations (100, 50, 25, & 12.5 μM) to determine IC50. Caspase
3 inhibition was monitored using a commercial inhibitor
Ac-DEVD-CHO generated at B Enzolifesciences (BML-P410-
9090), which has an IC50 of 1.3 nM. Caspase 8 inhibition was -
monitored using a commercial inhibitor Ac-IETD-CHO gene-
rated at Enzolifesciences (BML-430-9090), which has an IC50 of
127 nM.
1 K. R. Cooke, A. Gerbitz, J. M. Crawford, T. Teshima,
G. R. Hill, A. Tesolin, D. P. Rossignol and J. L. M. Ferrara,
LPS antagonism reduces graft-versus-host disease and pre-
serves graft-versus-leukemia activity after experimental
bone marrow transplantation, J. Clin. Invest., 2001, 107,
1581–1589.
2 X. Cao, S. F. Cai, T. A. Fehniger, J. Song, L. I. Collins,
D. R. Piwnica-Worms and T. J. Ley, Granzyme B and
perforin are important for regulatory T cell-mediated
suppression of tumor clearance, Immunity, 2007, 27,
635–646.
3 N. A. Thornberry, T. A. Rano, E. P. Peterson, D. M. Rasper,
T. Timkey, M. Garcia-Calvo, V. M. Houtzager,
P. A. Nordstrom, S. Roy, J. P. Vaillancourt, K. T. Chapman
and D. W. Nicholson, A combinatorial approach defines
specificities of members of the caspase family and gran-
zyme B, J. Biol. Chem., 1997, 272, 17907–17911.
4 J. Rotonda, M. Garcia-Calvo, H. G. Bull, W. M. Geissler,
B. M. McKeever, C. A. Willoughby, N. A. Thornberry and
J. W. Becker, The three-dimensional structure of human
granzyme B compared to caspase-3, key mediators of cell
death with cleavage specificity for aspartic acid in P1,
Chem. Biol., 2001, 8, 357–368.
5 C. A. Willoughby, H. G. Bull, M. Garcia-Calvo, J. Jiang,
K. T. Chapman and N. A. Thornberry, Discovery of potent,
selective human granzyme B inhibitors that inhibit CTL
mediated apoptosis, Bioorg. Med. Chem. Lett., 2002, 12,
2197–2200.
6 D. A. Dougherty, Cation-π interactions involving aromatic
amino acids, J. Nutr., 2007, 137, 1504S–1508S.
7 J. J. Perona and C. S. Craik, Structural basis of substrate
specificity in the serine proteases, Protein Sci., 1995, 4, 337–
360.
Abbreviations
GrB
Granzyme B
BMT
AML
ALL
CML
GVL
GVHD
CTL
IEPD
FFT
Bone marrow transplantation
Acute myeloid leukemia
Acute lymphoid leukemia
Chronic myeloid leukemia
Graft-versus-leukemia
Graft-versus-host disease
Cytotoxic T cells
Ile-Glu-Pro-Asp
Fast Fourier transform
Consensus sites
CSs
SP
Standard-precision
XP
Extra-precision
ROCS
IFD
Rapid overlay of chemical structures
Induced fit docking
8 P. Ingallinella, S. Altamura, E. Bianchi, M. Taliani,
R. Ingenito, R. Cortese, R. D. Francesco, C. Steinkühler and
A. Pessi, Potent peptide inhibitors of human hepatitis C
virus NS3 protease are obtained by optimizing the cleavage
products, Biochemistry, 1998, 37, 8906–8914.
9 R. Zhang, J. Durkin, W. T. Windsor, C. McNemar,
L. Ramanathan and H. V. Le, Probing the substrate speci-
ficity of hepatitis C virus NS3 serine protease by using syn-
thetic peptides, J. Virol., 1997, 71, 6208–6213.
TFA-dipeptide- Trifluoroacetyl-dipeptide-isopropylanilide
ISO
MSACK
Methoxysuccinyl-Ala-Ala-Pro-Ala chloromethyl
ketone
Virtual high-throughput screening
1-Ethyl-3-(3-dimethylaminopropyl)-
carbodiimide
VHTS
EDCI
8964 | Org. Biomol. Chem., 2014, 12, 8952–8965
This journal is © The Royal Society of Chemistry 2014

View Article Online
Organic & Biomolecular Chemistry
Paper
10 T. Cheng, X. Li, Y. Li, Z. Liu and R. Wang, Comparative
assessment of scoring functions on a diverse test set,
J. Chem. Inf. Model., 2009, 49, 1079–1093.
11 J. B. Cross, D. C. Thompson, B. K. Rai, J. C. Baber,
K. Y. Fan, Y. Hu and C. Humblet, Comparison of several
molecular docking programs: pose prediction and virtual
screening accuracy, J. Chem. Inf. Model., 2009, 49, 1455–
1474.
scoring. 2. Enrichment factors in database screening,
J. Med. Chem., 2004, 47, 1750–1759.
23 R. A. Friesner, R. B. Murphy, M. P. Repasky, L. L. Frye,
J. R. Greenwood, T. A. Halgren, P. C. Sanschagrin and
D. T. Mainz, Extra Precision Glide: docking and scoring
incorporating
a model of hydrophobic enclosure for
protein-ligand complexes, J. Med. Chem., 2006, 49, 6177–
6196.
12 O. Korb, T. S. G. Olsson, S. J. Bowden, R. J. Hall, 24 J. A. Prendergast, C. D. Helgason and R. C. Bleackley, Quan-
M. L. Verdonk, J. W. Liebeschuetz and J. C. Cole, Potential
and limitations of ensemble docking, J. Chem. Inf. Model.,
2012, 52, 1262–1274.
titative polymerase chain reaction analysis of cytotoxic cell
proteinase gene transcripts in T cells, J. Biol. Chem., 1992,
267, 5090–5095.
13 P. J. Hajduk, J. R. Huth and S. W. Fesik, Druggability 25 C. Kam, D. Hudig and J. C. Powers, Granzymes (lymphocyte
indices for protein targets derived from NMR-based screen-
ing data, J. Med. Chem., 2005, 48, 2518–2525.
14 W. L. DeLano, Unraveling hot spots in binding interfaces:
serine proteases): characterization with natural and syn-
thetic substrates and inhibitors, Biochim. Biophys. Acta,
2000, 1477, 307–323.
progress and challenges, Curr. Opin. Struct. Biol., 2002, 12, 26 I. Lu, N. Mahindroo, P. Liang, Y. Peng, C. Kuo, K. Tsai,
14–20.
H. Hsieh, Y. Chao and S. Wu, Structure-based drug design
and structural biology study of novel nonpeptide inhibitors
of severe acute respiratory syndrome coronavirus main pro-
tease, J. Med. Chem., 2006, 49, 5154–5161.
15 R. Brenke, D. Kozakov, G. Chuang, D. Beglov, D. Hall,
M. R. Landon, C. Mattos and S. Vajda, Fragment-based
identification of druggable ‘hot spots’ of proteins using
Fourier domain correlation techniques, Bioinformatics, 27 C. Mattos, B. Rasmussen, X. Ding, G. A. Petsko and
2009, 25, 621–627.
D. Ringe, Analogous inhibitors of elastase do not always
16 E. Estébanez-Perpiñá, P. Fuentes-Prior, D. Belorgey,
bind analogously, Nat. Struct. Biol., 1994, 1, 55–58.
M. Braun, R. Kiefersauer, K. Maskos, R. Huber, H. Rubin 28 (a) V. Sobolev, A. Sorokine, J. Prilusky, E. E. Abola and
and W. Bode, Crystal structure of the caspase activator
human Granzyme B, a proteinase highly specific for an
Asp-P1 residue, Biol. Chem., 2000, 381, 1203–1214.
17 http://www.maybridge.com/portal/alias__Rainbow/lang__
en/tabID__146/DesktopDefault.aspx.
18 http://www.actimol.com/protease-inhibitors-actitarg-p-library.
html.
19 http://us.chemdiv.com/.
M. Edelman, Automated analysis of interatomic contacts in
proteins, Bioinformatics, 1999, 15, 327–332; (b) http://ligin.
weizmann.ac.il/cgi-bin/lpccsu/LpcCsu.cgi; accessed 8/11/
2014.
29 Schrödinger Suite 2010 Protein Preparation Wizard; Epik
version 2.1, Schrödinger, LLC, New York, NY, 2010; Impact
version 5.6, Schrödinger, LLC, New York, NY, 2010; Prime
version 2.2, Schrödinger, LLC, New York, NY, 2010.
20 Glide, version 5.6, Schrödinger, LLC, New York, NY, 30 LigPrep, version 2.4, Schrödinger, LLC, New York, NY, 2010.
2010.
31 Desmond Molecular Dynamics System, version 3.4, D. E. Shaw
Research, New York, NY, 2014; Maestro-Desmond Interoper-
ability Tools, version 3.4, Schrödinger, New York, NY, 2014.
21 R. A. Friesner, J. L. Banks, R. B. Murphy, T. A. Halgren,
J. J. Klicic, D. T. Mainz, M. P. Repasky, E. H. Knoll,
D. E. Shaw, M. Shelley, J. K. Perry, P. Francis and 32 K. Bowers, E. Chow, H. Xu, R. Dror, M. Eastwood,
P. S. Shenkin, Glide: a new approach for rapid, accurate
docking and scoring. 1. Method and assessment of
docking accuracy, J. Med. Chem., 2004, 47, 1739–1749.
B. Gregersen, J. Klepeis, I. Kolossvary, M. Moraes,
F. Sacerdoti, J. Salmon, Y. Shan and D. Shaw, Scalable
Algorithms for Molecular Dynamics Simulations on Com-
modity Clusters. Proceedings of the ACM/IEEE Conference
on Supercomputing (SC06), Tampa, Florida, November
11–17, 2006.
22 T. A. Halgren, R. B. Murphy, R. A. Friesner, H. S. Beard,
L. L. Frye, W. T. Pollard and J. L. Banks, Glide: a new
approach
for
rapid,
accurate
docking
and
This journal is © The Royal Society of Chemistry 2014
Org. Biomol. Chem., 2014, 12, 8952–8965 | 8965