Coumarin derivatives act as novel inhibitors of human dipeptidyl peptidase III: Combined in vitro and in silico study

Article abstract of DOI:10.3390/ph14060540

Dipeptidyl peptidase III (DPP III), a zinc-dependent exopeptidase, is a member of the metalloproteinase family M49 with distribution detected in almost all forms of life. Although the physiological role of human DPP III (hDPP III) is not yet fully elucida

Full text of DOI:10.3390/ph14060540

pharmaceuticals
Article
Coumarin Derivatives Act as Novel Inhibitors of Human
Dipeptidyl Peptidase III: Combined In Vitro and
In Silico Study
1
,
1
1
1
2
3
3
Dejan Agi c´ * , Maja Karnaš
, Domagoj Šubari c´ , Melita Lon cˇ ari c´
, Sanja Tomi c´ , Zrinka Kara cˇ i c´ ,
Drago Bešlo 1 , Vesna Rastija , Maja Molnar , Boris M. Popovi c´ and Miroslav Lisjak
2
4
1
1
Faculty of Agrobiotechnical Sciences Osijek, Josip Juraj Strossmayer University of Osijek,
31000 Osijek, Croatia; maja.karnas@fazos.hr (M.K.); domagoj.subaric@fazos.hr (D.Š.);
drago.beslo@fazos.hr (D.B.); vesna.rastija@fazos.hr (V.R.); miroslav.lisjak@fazos.hr (M.L.)
Faculty of Food Technology Osijek, Josip Juraj Strossmayer University of Osijek, 31000 Osijek, Croatia;
melita.loncaric@ptfos.hr (M.L.); maja.molnar@ptfos.hr (M.M.)
Division of Organic Chemistry and Biochemistry, Ru d¯ er Boškovi c´ Institute, 10000 Zagreb, Croatia;
sanja.tomic@irb.hr (S.T.); zrinka.karacic@irb.hr (Z.K.)
2
3
4
Department of Field and Vegetable Crops, Faculty of Agriculture, University of Novi Sad,
21000 Novi Sad, Serbia; boris.popovic@polj.uns.ac.rs
*
Correspondence: dejan.agic@fazos.hr
ꢀ
ꢁꢂꢀꢃꢄꢅꢆꢇ
Abstract: Dipeptidyl peptidase III (DPP III), a zinc-dependent exopeptidase, is a member of the
metalloproteinase family M49 with distribution detected in almost all forms of life. Although
the physiological role of human DPP III (hDPP III) is not yet fully elucidated, its involvement in
pathophysiological processes such as mammalian pain modulation, blood pressure regulation, and
cancer processes, underscores the need to find new hDPP III inhibitors. In this research, five series
of structurally different coumarin derivatives were studied to provide a relationship between their
inhibitory profile toward hDPP III combining an in vitro assay with an in silico molecular modeling
study. The experimental results showed that 26 of the 40 tested compounds exhibited hDPP III
ꢀ
ꢁꢂꢃꢄꢅꢆ
Citation: Agi c´ , D.; Karnaš, M.;
Šubari c´ , D.; Lon cˇ ari c´ , M.; Tomi c´ , S.;
Kara cˇ i c´ , Z.; Bešlo, D.; Rastija, V.;
Molnar, M.; Popovi c´ , B.M.; et al.
Coumarin Derivatives Act as Novel
Inhibitors of Human Dipeptidyl
Peptidase III: Combined In Vitro and
In Silico Study. Pharmaceuticals 2021,
inhibitory activity at a concentration of 10
µM. Compound 12 (3-benzoyl-7-hydroxy-2H-chromen-2-
one) proved to be the most potent inhibitor with IC50 value of 1.10
µM. QSAR modeling indicates
1
4, 540. https://doi.org/10.3390/
ph14060540
that the presence of larger substituents with double and triple bonds and aromatic hydroxyl groups
on coumarin derivatives increases their inhibitory activity. Docking predicts that 12 binds to the
region of inter-domain cleft of hDPP III while binding mode analysis obtained by MD simulations
revealed the importance of 7-OH group on the coumarin core as well as enzyme residues Ile315,
Ser317, Glu329, Phe381, Pro387, and Ile390 for the mechanism of the binding pattern and compound
Academic Editor: Salvatore
Santamaria
Received: 30 April 2021
Accepted: 2 June 2021
Published: 5 June 2021
12 stabilization. The present investigation, for the first time, provides an insight into the inhibitory
effect of coumarin derivatives on this human metalloproteinase.
Keywords: dipeptidyl peptidase III; coumarin derivatives; inhibitor; molecular modeling; metallo-
proteinase
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
1
. Introduction
Dipeptidyl peptidase III (DPP III) is a zinc-hydrolase that cleaves dipeptides se-
quentially from the N-terminal of different bioactive peptides [ ]. As a member of the
metalloproteinase family M49, DPP III distribution is detected in almost all forms of life [ ].
Human DPP III is a very well-characterized member of this family in terms of biochemistry,
structural biology and computational chemistry [ 10]. Due to the relative non-specificity
of the peptide substrates as well as the lack of selective inhibitors of the metallopeptidases
of the M49 family, the physiological substrates of DPP III have not been accurately identi-
fied, and its fundamental physiological role has not been precisely determined. However,
1
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
2
3–
4
.0/).
it is assumed that it is involved in post-proteasomal intracellular protein catabolism [3],
Pharmaceuticals 2021, 14, 540. https://doi.org/10.3390/ph14060540
https://www.mdpi.com/journal/pharmaceuticals

Pharmaceuticals 2021, 14, 540
2 of 19
defense against oxidative stress [11
,
12] mammalian pain modulatory system [
4
,13], malig-
nant processes [14 17] and blood pressure regulation [18–20]. Because of its involvement
–
in biological processes, hDPP III has become interesting to investigate as a potential drug
target. To obtain information on the mechanism of action of hDPP III, the influence of
selected mutations on the enzyme activity was tested, and research was conducted to
find potential inhibitors of this mammalian metalloproteinase [21–23]. It has recently
been shown that newly synthesized guanidiniocarbonylpyrrole–fluorophore conjugates
could be used for enzyme sensing and bio-activity inhibiting (theragnostic) studies of DPP
III [24]. In search of new inhibitors from natural sources, we previously reported that
luteolin has the best inhibitory effect against hDPP III (IC50 = 22 µM) of the 17 flavonoids
tested, and that the number and exact distribution of –OH groups on the flavonoid core is
important for their inhibitory properties [25]. The latest study of the biological activity of
cornelian cherry fruit extracts showed inhibitory activity against hDPP III with bioactive
constituent pelargonidin 3-robinobioside with the best binding energy [26]. Coumarin or
2
H-chromen-2-one and its derivatives represent an important group of oxygen-containing
heterocycles with benzopyrone skeleton [27]. They can be isolated from plant material [28
]
or synthesized [29]. Coumarin derivatives possess various beneficial biological activities,
for example, anticoagulant, anticancer, analgesic, anti-inflammatory, bactericidal, antifun-
gal, anticonvulsant, anti-hypertensive, muscle relaxant, antioxidant, etc. [28]. It is known
that coumarins exhibit an inhibitory effect on the enzymes such as acetylcholinesterase,
β-secretase, and monoamine oxidase [30]. Studies have also shown that simple coumarin
derivatives influence the activity of some zinc-dependent metalloproteinases [31–33].
Because of all mentioned above as well as our efforts to find new hDPP III inhibitors,
we report the investigation of 40 structurally different coumarin compounds and provide
a relationship between their inhibitory profile toward hDPP III combining in vitro assay
with Quantitative Structure–Activity Relationship (QSAR) analysis. Additionally, docking
and MD simulations were conducted to explore the mechanism of the most potent inhibitor
binding into the active site of hDPP III.
2. Results and Discussion
2.1. DPP III Inhibitory Activity
In the current study, we evaluated forty various coumarin compounds for their
inhibitory potential towards hDPP III. Results in Table 1 showed that substituted 3-acetyl-
H-chromen-2-ones with a bromo group at the C6 position (compound ) was the most
active with the inhibition rate of 28.5%, while the presence of a hydroxyl group at the same
position on compound reduced (12.8%) the inhibitory potential. Shifting a hydroxyl
group to C7 in resulted in a slightly increased inhibitory potential (16.2%) as compared
to while the presence of diethylamino group at this position in compound was found
to be inactive. Additionally, compounds and which possess a hydroxyl and ethoxy
group at C8 were completely inactive at the concentration of 10 M. Compound bearing
2
1
2
4
2
3
5
6
µ
7
unsubstituted 3-acetyl-2H-chromen-2-one showed a weak (7.8%) inhibitory activity.
The most potent inhibitory potential of substituted 3-benzoyl-2H-chromen-2-ones
was obtained with compound 12 (100.0%) where the hydroxyl group is present at C7.
However, the substitution of C7 with the benzoyl (11) and methoxy (13) group caused a
decrease in hDPP III inhibitory (22.8% and 16.5%, respectively) activity. Moderate (67.5%)
to weak (4.4%) enzyme inhibition was observed with compounds 10 and
a hydroxyl and chloro group at the C6 position, respectively. Compound
group at the C6 and C8 as well as compound 14 with an ethoxy group on C8 did not exhibit
8
which possess
9
with a bromo
inhibition effects on enzyme activities. Unsubstituted 3-benzoyl-2H-chromen-2-one (15
showed only a weak inhibitory activity (9.6%) as compared to 12.
)
Of the seven substituted 2-oxo-2H-chromene-3-carbonitriles tested, only compounds
21 and 18 which only differ with hydroxyl group positions (C8 and C6, respectively)
moderately inhibited enzyme with an inhibition rate of 62.6% and 44.6%, respectively.
In the case where the ethoxy group is at the C8 position (22), no inhibitory activity was

Pharmaceuticals 2021, 14, 540
3 of 19
observed. Furthermore, the methoxy substituent at the C6 position (17) gave an inhibition
rate of 19.8% while its presence at the C7 position (20) was found to be inactive. Substitution
of bromo group at C6 in 16 and benzoyl group at C7 in 19 showed almost similar inhibitory
potential (7.9% and 7.1%, respectively). Unsubstituted 2-oxo-2H-chromene-3-carbonitrile
(23) was not effective in inhibiting hDPP III.
Table 1. Structures of analysed compounds, values of experimentally determined inhibition of hDPP III (at 10 µM
concentration of compounds) and calculated logarithmic values of the % inhibition of hDPP III.
Log (% hDPP III
Inh.) exp.
Log (% DPP III
Inh.) Calc. *
Compound No.
Substituents
DPP III Inh. (%)
1
2
3
4
5
6
7
8
9
3-acetyl; 6-bromo
3-acetyl; 6-hydroxy
3-acetyl; 7-diethylamino
3-acetyl; 7-hydroxy
3-acetyl; 8-ethoxy
3-acetyl; 8-hydroxy
28.5
12.8
NA
16.2
NA
NA
7.8
4.4
1.45
1.11
0.00
1.21
0.00
0.00
0.89
0.64
0.00
1.83
1.36
2.00
1.22
0.00
0.98
0.90
1.30
1.65
0.85
0.00
1.80
0.00
0.00
0.00
1.30
1.78
1.82
1.47
0.00
0.00
0.00
0.81
1.33
1.37
1.00
2.00
0.00
0.35
0.00
0.33
1.29
Excl.
-
1.70
0.35
-
1.01
1.01
-
1.95
1.18
1.89
0.78
-
0.65
0.98
0.83
1.81
1.04
-
1.67
0.17
-
3-acetyl
3-benzoyl; 6-chloro
3-benzoyl; 6,8-dibromo
3-benzoyl; 6-hydroxy
3-benzoyl; 7-benzoyl
3-benzoyl; 7-hydroxy
3-benzoyl; 7-methoxy
3-benzoyl; 8-ethoxy
3-benzoyl
3-cyano; 6-bromo
3-cyano; 6-methoxy
3-cyano; 6-hydroxy
3- cyano; 7-benzoyl
3-cyano; 7-methoxy
3- cyano; 8-hydroxy
3-cyano; 8-ethoxy
NA
67.5
22.8
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
100.0 (1.10 ± 0.05 µM)
16.5
NA
9.6
7.9
19.8
44.6
7.1
NA
62.6
NA
NA
NA
20.1
59.7
66.0
29.4
NA
3-cyano
3-ethoxycarbonyl; 6-bromo
3-ethoxycarbonyl; 6-chloro
3-ethoxycarbonyl; 6-dihydroxyamino
3-ethoxycarbonyl; 6-hydroxy
3- ethoxycarbonyl; 6,8-dibromo
3-ethoxycarbonyl; 7-methoxy
3-ethoxycarbonyl; 8-ethoxy
3- ethoxycarbonyl
3- methoxycarbonyl; 6-bromo
3-methoxycarbonyl; 6-dihydroxyamino
3-methoxycarbonyl; 6-hydroxy
3-methoxycarbonyl; 6-methoxy
3-methoxycarbonyl; 7-hydroxy
3-methoxycarbonyl; 7-methoxy
3-methoxycarbonyl
-
0.93
1.49
1.76
1.82
0.26
0.37
-
0.78
1.46
1.35
0.64
1.50
0.56
0.59
−0.37
0.49
NA
NA
6.5
21.2
23.5
9.9
100.0 (2.14 ± 0.06 µM)
NA
2.3
coumarin
7-hydroxycoumarin
NA
2.1
NA, no activity; Excl.; excluded as outlier; -, excluded from initial dataset; * Calculated by quantitative structure-activity relationship (QSAR)
equation: log % hDPP III inh. = 4.07 + 1.85 (0.59) EEig05x + 1.60 (0.52) Mor10u + 0.56 (0.39) nArOH; numbers in brackets represent IC50 values.
−

Pharmaceuticals 2021, 14, 540
4 of 19
Four of seven substituted 3-acetyl-2H-chromen-2-ones: 24
–27 possess different groups
but in the same C6 position. Compound 27 with hydroxyl group and 26 with dihy-
droxyamino group were found to be more active (66.0% and 59.7% respectively) than
the compounds 25 (20.1%) and 24 (not active) with the chloro and bromo group, respec-
tively. Interestingly, dibromo substituents at C6 and C8 (28) increased inhibitory potential
(
29.4%) as compared to mono substituted analog (24). Compounds 29 and 30 which pos-
sess methoxy group at C7 and ethoxy group at C8, respectively as well as unsubstituted
3-acetyl-2H-chromen-2-one (31) did not exhibit inhibitory potential.
Among the substituted methyl 2-oxo-2H-chromene-3-carboxylates, only 36 containing
a hydroxyl group at C7 completely inhibited enzymatic activity. Changing the methoxy
group at the same position (37) completely reduced inhibitory potential. The compound 34
with dihydroxyamino and compound 33 with hydroxyl group at C6 position were found to
be more efficient in the inhibitory potential (23.5% and 21.2%, respectively) as compared to
the methoxy (9.9%) and bromo (6.5%) substituted analogs 35 and 32, respectively. A very
weak inhibitory potential (2.3%) was found for unsubstituted methyl 2-oxo-2H-chromene-
3-carboxylate (38). Similarly, 7-hydroxycoumarin (39) and coumarin (40) exhibited a strong
decrease (2.1% and not active, respectively) in the inhibitory potential towards hDPP III.
From the above analysis, it can be concluded that the best inhibitory potential had
substituted 3-benzoyl-2H-chromen-2-one (12) and methyl 7-hydroxy-2-oxo-2H-chromene-
3
-carboxylate (36) containing a hydroxyl group at position C7, where they completely
inhibited the enzyme at the concentration of 10 M with IC50 values of 1.10 0.05 M and
.14 0.06 M, respectively (Table 1 and Figure 1). Additionally, comparing the structures
µ
±
µ
2
±
µ
of derivatives 12 and 36 with compounds 39 and 40 suggests that in addition to the presence
of the hydroxyl group at the C7 position, the exact presence of particular substituents at the
C3 position is important for increasing the inhibitory activity of tested coumarin derivatives.
Furthermore, when the hydroxyl group is at the C6 and C8 positions, the compounds mostly
show moderate inhibitory activity. Coumarin derivatives with a substituted bromo, chloro,
and benzoyl group showed lower inhibitory potential compared to C6 hydroxy analogs.
Finally, compounds with a dihydroxyamino group at the C7 position had moderate enzyme
inhibition while most coumarin derivatives with a substituted methoxy and diethylamino
group showed no inhibitory activity against hDPP III.
Figure 1. IC50 value determination of compounds 12 and 36 against hDPP III. Data points represent
the average values of three determinations.

Pharmaceuticals 2021, 14, 540
5 of 19
2.2. Results of the QSAR Analysis
The best QSAR model obtained for hDPP III inhibition is:
log % hDPP III inh. = −4.03 + 1.82 (0.58) EEig05x + 1.46 (0.49) Mor10u + 0.49 (0.36) nArOH
(1)
where EEig05x is an edge-adjacency index descriptor weighted by edge degrees, Mor10u
is a 3D-MoRSE descriptor (unweighted) and nArOH is the number of aromatic hydroxyl
groups.
The model satisfied the threshold for the fitting and internal validation criteria [34],
but Williams plot revealed one outlier, compound
2 (MolID in QSARINS: 24) as shown in
Figure 2. After the exclusion of this compound from the dataset, the subsequent analysis
produced the improved QSAR model:
log % hDPP III inh. = −4.07 + 1.85 (0.59) EEig05x + 1.60 (0.52) Mor10u + 0.56 (0.39) nArOH
(2)
Figure 2. Williams plot (plot of standardized residuals vs. leverages (h) for each compound) of
applicability domain of the QSAR model for hDPP III inhibition calculated by model 1. The warning
0
0
leverage (h* = 0.444) is defined as 3p /n (n is the number of training compounds, and p the number
of model adjustable parameters).
The variables in Equations (1) and (2) are listed in order of relative importance by
their standardized regression coefficient (β, written in brackets). The statistical parameters
for both models are given in Table 2. The values of the descriptors included in the models
are given in the Supplementary Materials (Table S1). The values of log % hDPP III inh.;

Pharmaceuticals 2021, 14, 540
6 of 19
both experimentally obtained and calculated by Equation (2) are presented in Table 1 and
Table S1.
Table 2. The statistical parameters for QSAR models.
Statistical Parameters
Model 1
27
Model 2
26
Ntr
Nex
R2
5
5
0.746
0.713
0.352
22.565
0.210
0.201
0.325
0.276
0.855
0.650
0.381
0.326
3.923
0.807
0.115
−0.236
0.292
0.255
0.795
0.868
0.783
0.780
0.794
0.614
0.186
0.796
0.768
0.323
28.572
0.190
0.215
0.297
0.253
0.886
0.710
0.354
0.302
3.258
0.844
0.121
−0.244
0.295
0.275
0.785
0.873
0.778
0.776
0.798
0.653
0.174
R2
adj
s
F
Kxx
∆
K
RMSEtr
MAEtr
CCCtr
Q2
LOO
RMSEcv
MAEcv
PRESScv
CCCcv
R2
Yscr
Q2
Yscr
RMSEext
MAEext
2
R ext
CCCext
Q2
Q2
Q2
F1
F2
F3
r2m average
r2m difference
Applicability domain
N outliers
1 (2)
-
-
-
N out of app. domain
N
2
(
coefficient of determination); R
F (Fisher ratio); Kxx (global correlation among descriptors);
∆K (global correlation among descriptors); RMSEtr
(
root-mean-square error of the training set); MAEtr (mean absolute error of the training set); CCCtr (concordance
2
correlation coefficient of the training set); Q LOO (cross-validated explained variance); RMSEcv (root-mean-square
error of the training set determined through the cross validated method); MAEcv (mean absolute error of the
internal validation set); PRESScv (cross-validated predictive residual sum of squares); CCCcv (concordance
2
2
correlation coefficient test set using cross validation); R Yscr (Y-scramble correlation coefficients); Q Yscr (Y-
scramble cross-validation coefficients); RMSEext (root-mean-square error of the external validation set); MAEext
2
(
mean absolute error of the external validation set); R ext (coefficient of determination of external validation
2
2
2
set); CCCext (concordance correlation coefficient of the test set); Q F1, Q F2, Q F3 (predictive squared correlation
average (average value of squared correlation coefficients between the observed and leave-one-
out predicted values of the compounds with and without intercept); r
2
coefficients); r
m
2
m
difference (absolute difference between
the observed and leave-one-out predicted values of the compounds with and without intercept).

Pharmaceuticals 2021, 14, 540
7 of 19
The collinearity of the descriptors in the model was evaluated with a correlation
matrix (Table 3) to exclude the possibility that the improved model is overfitted (correlation
coefficient R
≤
0.7). Furthermore, low collinearity was verified with the low value of Kxx
2
and
∆
K being
≥
0.05. The model satisfied fitting and internal validation criteria: R and
2
2
R
≥ 0.60; CCCtr ≥ 0.85; RMSE and MAE close to zero; RMSEtr < RMSEcv
;
Q
≥ 0.50
adj
LOO
2
2
(with R
−
Q being low); high value of F (Table 2). The value of the cross-validated
2
correlation coefficient (Q
= 0.710) shows that model 2 has a good internal prediction
LOO
2
2
power. The robustness of the improved model was confirmed with both R
and Q
Yscr
Yscr
2
2
values < 0.2, and R
> Q
[34]. Model 2 also satisfied the following external validation
Yscr
Yscr
criteria: R2ext ≥ 0.60; low differences between RMSEtr and RMSEext as well as between
MAEtr and MAEext and between CCCtr and CCCext; Q F1, Q F2, and Q2
2
2
≥
0.60; r
m
2
F3
average
external prediction
≥
0.60 and r2
m
difference
≤
0.20 indicating that this model could be used for
Table 3. Correlation matrix (with correlation coefficient values R) for descriptors used in Equation (2).
Descriptor
EEig05x
Mor10u
EEig05x
1.000
Mor10u
nArOH
−0.264
−0.129
1.000
0.489
nArOH
1.000
The Williams plot for model (2) showed no compounds outside the applicability
domain of the model.
The descriptors from the best model were more closely observed to gain insight into
the factors that contribute to the inhibitory activity of tested compounds. The first variable
in Equation (2) with a high positive contribution is descriptor EEig05x, 5th eigenvalue from
edge adjacency matrix weighted by edge degrees (the bond order of the various edges). It
belongs to the edge-adjacency topological indices derived from the edge adjacency matrix,
which encodes the connectivity between graph edges, and is derived from an H-depleted
molecular graph of molecules. These descriptors are sensitive to the size, shape, branching,
and cyclicity of molecules [35,36]. It is shown that compounds with relatively higher
values of this descriptor tend to exhibit higher inhibition of hDPP III. This indicates that
compounds with larger, aromatic, and substituents with a higher number of double or
triple bonds may exhibit enhanced inhibition. Similar conclusions were also drawn in
previous work [37].
The second variable, Mor10u, belongs to the 3D-MoRSE (Molecule Representation of
Structures based on Electron diffraction) group of descriptors. It has a scattering parameter
−
1
s = 9 Å and since it is unweighted, treats all atoms equally [38]. The positive coefficient
of Mor10u in Equation (2) indicates the importance of the three-dimensional arrangement
of all atoms in a molecule and their pairwise distances. Compounds with higher inhibitory
activity tend to have more positive values of this descriptor. Since larger molecules, with
larger interatomic distances, have higher MoRSE descriptor values, this confirms the above
conclusion about the EEig05x descriptor that larger molecules are more active.
The third variable in the equation is nArOH, a descriptor from the functional group
counts that represents the number of aromatic hydroxyls [38]. The positive coefficient in
Equation (2) indicates that the presence of aromatic hydroxyl groups contributes to the
inhibition of hDPP III. This is in accordance with an earlier study, where the presence of
hydrophilic regions (i.e., hydroxyl groups) in flavonoids increased their inhibitory activity
against hDPP III [25]. Williams plot revealed compound
2 as an outlier since it had a
high predicted residual (predicted value in model (1) was significantly higher than the
experimentally obtained). The presence of the hydroxyl group at position 6 might be the
reason for the increase in the estimated value according to Equation (1), as well as the
high value of Mor10u (Table S1). However, this model equation does not consider the
presence of substituents at position 3, such as the -COCH group in this case, that may have
3

Pharmaceuticals 2021, 14, 540
8 of 19
a negative effect on inhibition since compounds with this substituent exhibited relatively
low inhibition values (Table 1).
Based on the conclusions given in the QSAR analysis, structures of two modified
compounds (41
, 42) with possible improved activity are proposed, log (% inh. hDPP III)
3
.08 and 3.01, respectively (Figure 3). Values of their calculated descriptors, as well as
predicted inhibitory activities of the proposed compounds calculated using Equation (2) are
given in the Supplementary Materials (Table S1). Since their calculated values exceed 100%
of inhibition, these compounds could be potent inhibitors at concentrations lower than
10 µM. Both compounds possess a benzoyl group at the position C-3, and two hydroxyl
groups at the position C-5 and C-7 (41), and at the position C-6 and C-8 (42). Improved
calculated inhibition can be attributed to the introduction of an aromatic substituent and
additional hydroxyl groups, as indicated by QSAR analysis.
Figure 3. Structures of proposed compounds with possible enhanced inhibition of hDPP III.
2
.3. Docking
In order to obtain further information on the possible interactions of the most active
compound (12) with a semi-closed form of hDPP III, we combined docking with MD
−
1
simulations. The best results regarding AutoDock Vina binding energy (
of enzyme–ligand complex predict that compound 12 binds to the inter-domain cleft, near
the lower sheet (residues 389–393) (Figure 4A) and enzyme active site, with the minimum
distance between the catalytic Zn cation and 12 (oxygen atom at C2 of coumarin core) being
7 Å. In this complex, the position of compound 12 closely resembles the substrate position
in the hDPP III active site [ ] which is accommodated similarly to the opioid peptides
in the enzyme binding pocket [
−
8.6 kcal mol
)
β
~
9
5
]. Namely, binding of 12 into the inter-domain cleft is
0
accompanied by its interactions mostly with amino acid residues of the hDPP III S1, S1 , S2,
S2 and S3 substrate binding subsite (Figures S1 and 4B).
0
0

Pharmaceuticals 2021, 14, 540
9 of 19
Figure 4.
representation, the lower
of compound 12 with amino acid residues of hDPP III as presented in the 2D scheme (Figure S1). Substrate binding subsites
(
A
) Best docking pose for compound 12 in the inter-domain cleft of hDPP III. Compound 12 is shown in stick
β
sheet is colored yellow, and zinc cation is represented as a green sphere. ( ) Potential interactions
B
0
0
0
S1, S1 , S2, S2 and S3 are indicated.
2.4. MD Simulations
To prove the reliability of the best docking result, the binding mode of compound
2 in complex with hDPP III was investigated by productive MD simulations using the
1
AMBER16 software package. Simulations of complex were performed in three replicates,
each 300 ns long, and used for comparison. Dynamic behavior, protein, and ligand stability
during simulations were analyzed by root mean square deviation (RMSD), while the
analysis of the intermolecular interactions during MD simulations included hydrogen
bonding (H-bond), native contacts, and Gibbs free energy. Representative structures of the
complex were used to describe intermolecular interactions in more detail.
2.4.1. RMSD Profile
The RMSD profiles (Figure 5) calculated during the simulations for the protein back-
bone atoms show similar protein stability in all three runs, with only slightly higher protein
stability in run 1 (average RMSD SD of 1.48 0.18 Å, 1.76 0.23 Å, and 1.96 0.34 Å
±
±
±
±
for run 1, 2, and 3, respectively). According to the RMSD values for the heavy atoms of
compound 12 between replicates, it can be seen (Figure 6) that the stability of 12 is better in
run 1 compared to the other two replicates. The average RMSD
±
SD were 0.37
±
0.15 Å,
0.46 ± 0.12 Å and 0.54 ± 0.15 Å, for run 1, run 2 and run 3, respectively.

Pharmaceuticals 2021, 14, 540
10 of 19
Figure 5. RMSD profile of the protein backbone atoms obtained during 300 ns of MD simulations.
Figure 6. RMSD profile for the heavy atoms (hydrogen atoms were not considered) of compound 12
obtained from MD simulations of complex.
2.4.2. Hydrogen Bond Analysis
Hydrogen bond analysis was undertaken to investigate the stability and occupancy of
hydrogen bonds between compound 12 and the key residues of the binding site of hDPP
III. The results of trajectory H-bonds analysis for all three replicates are listed in Table S2.
In run 1, there were two H-bonds formed during the MD process with the occupation
time >10% (Figure 7). The first H-bond is formed by the OE2 atom of Glu329 and the H-O4

Pharmaceuticals 2021, 14, 540
11 of 19
of 12 with an occupation time of 99%, and the second one is formed by the H-NE2 of Gln566
and the O3 of 12 with an occupation time of 23%. Additionally, atom O2 of compound 12
forms H-bond with H-NE2 of Gln566 and H-OH of Tyr318 with the occupation time of 5%
and 4%, respectively.
Figure 7. Overlay of the hDPP III with compound 12 in its preferable binding mode after 300 ns of
MD simulation for run 1 (cyan), run 2 (green) and run 3 (yellow). Hydrogen bonds are depicted as
green dashed lines.
Three hydrogen bonds were formed during the simulations in run 2 with the occupa-
tion time >10% (Figure 7). The first one is formed by the OE1 atom of Glu329 and the H-O4
of 12 with an occupation time of 85%. The second and third H-bond was formed between
the H-N and H-NH2 of Asn391 and the O2 and O3 of 12 with the occupation time of 17%
and 13%, respectively. One H-bond with an occupation time of 6% is formed by the H-ND2
of Asn391 and the O2 of compound 12.
Only one hydrogen bond (with the occupation time > 10%) was formed during the
simulations in run 3 by the OE2 atom of Glu329 and the H–O4 of 12 with an occupation
time of 99% (Figure 7). Moreover, atom O3 of 12 forms H-bond with H–OH of Tyr417
and H-NE2 of His568, both with the occupation time of 4%, while atom O2 of 12 forms
H-bond with H-NE2 of His568 with the occupation time of 3%. It is worthwhile to note that
0
0
Glu329 and Gln568 are found to be constituents of hDPP III S1, S1 and S2 substrate-binding
subsites [ ]. Besides this, His568 and Asn491 are highly conserved residues among known
DPP IIIs [39].
5
2.4.3. Native Contacts
To further investigate the interactions of hDPP III and compound 12, we calculated
the relative occupancy of native contacts during the MD simulations. Relative occupancy
of native contacts is defined as the sum of fractions of native contacts during the trajectory
for each residue pair relative to the total number of native contacts involved with that pair.
Native contacts were defined as a distance between the atoms of enzyme residues and

Pharmaceuticals 2021, 14, 540
12 of 19
atoms of compound 12 within a distance cutoff of 5 Å. The fractions of native contacts were
calculated using the nativecontacts command in the CPPTRAJ module.
Figures 8 and 9 depict the protein residues that were involved in forming native
contacts with a relative occupancy of more than 30% in the run 1, 2, and 3 of the complex
throughout the simulation time. In all three replicates, the protein forms native contacts
with 12 through residues Ile315, Glu316, Glu329, Phe381, and Pro387, which indicates that
these residues could be quite important for the stabilization of the complex. In runs 1 and
3, protein additionally forms native contacts with compound 12 through Phe 109, Ser317,
and Ile386. The remaining contacts are formed in runs 2 and 3 through residues Gly389
and Ile390, and Gly385 and Tyr417, respectively. Some of the listed amino acid residues
such as Glu316, Ser317, Glu329, Ile386, Pro387 and Tyr417 were found to contribute to the
binding of synthetic inhibitors into the active site of hDPP III [22,24,39].
Figure 8. hDPP III residues involved in native contacts and H-bond formation with compound 12 for run 1 (left), run 2
(
middle) and run 3 (right). H-bonds are depicted as green dashed lines, and compound 12 as light blue sticks.
Figure 9. Native contacts between hDPP III residues and compound 12 with relative occupancy of
more than 30% during 300 ns MD simulations.
2.4.4. Types of Intermolecular Interactions
A detailed analysis of different types of intermolecular interactions between replicates
was performed using the Discovery Studio Visualizer. For this purpose, the extracted

Pharmaceuticals 2021, 14, 540
13 of 19
complex structures obtained from trajectory after 300 ns of MD simulations were optimized
and used as representative. According to Figures S2–S4 representing 2D schemes of
the intermolecular interaction types, compound 12 forms almost the same number of
interactions with amino acid residues in all three replicates. However, in run 1, the largest
number of interactions is formed between the coumarin core of 12 and the hDPP III, which
is not the case in the other two replicates, where the benzoyl group of compound 12
also interacts with the amino acid residues. The above differences in the distributions of
intermolecular interactions of 12 and amino acid residues relate mainly to the van der
Waals interactions as shown in Figures S2–S4.
Comparing the other types of intermolecular interactions between replicates, the
2
D schemes show that Glu329 and Pro387 form H-bonds and
coumarin core in all three replicates, while additional H-bond and
Asn391 and Ile390 in run 2, respectively. Additionally, in run 1, coumarin core forms
stacked interaction with Phe109 and amide stacked interaction with Ile386. The benzoyl
group of compound 12 forms one -donor H-bond with Ile386, one -alkyl with Ala567,
and one π–π shaped interaction in run 1, run 2 and run 3, respectively.
π
-alkyl interactions with
π-alkyl are formed with
π–π
π
π
π
2
.4.5. MM-GBSA Free Energy Calculations
MM-GBSA calculations were used to obtain quantitative estimates of the free binding
energies of compound 12 in the complex with the hDPP III for all three replicates. According
to the results given in Table 4, the electrostatic contribution ( Eele) is the most important
to the Gbind for complex in run 1 and run 2. This is in accordance with the results of
MD simulations because in run 1 and run 2, compound 12 formed two and three H-bonds
with the occupation time > 10%), respectively, relative to run 3 where only one H-bond
is formed. Another important contribution to the Gbind of the complexes is the van der
∆
∆
(
∆
Waals interactions (Evdw) with values similar in all three replicates. These results are in
line with the observed similarities of the native contacts formed during the simulations,
especially with residues Ile315, Phe381, and Pro387 in all three replicates. The unfavorable
polar solvation contribution (EGB) was slightly higher for complex in run 2 compared to
runs 1 and 3, while the favorable nonpolar contribution (ESA) had similar values for all
three replicates.
Table 4. Binding free energy (kcal mol 1) of the complexes obtained during the last 5 ns of MD
−
simulations for all three replicates.
Energy Component
Run 1
Run 2
Run 3
Evdw
Eele
EGB
ESA
−28.33
−30.55
35.94
−3.44
−58.88
32.50
−27.44
−30.76
41.26
−4.06
−58.20
37.20
−26.16
−23.34
34.69
−3.41
−49.50
31.28
∆
Ggas
∆
Gsolv
∆
Gbind
−26.36
−20.98
−18.22
Evdw—van der Waals potential energy; Eele—electrostatic energy; EGB—polar solvation energy; ESA—nonpolar
solvation energy; ∆Ggas—gas phase free energy; ∆Gsolv—solvation free energy; ∆Gbind—binding free energy.
From the estimated values of ∆Gbind between replicates, it can be concluded that the
−
1
reason for the highest free binding energy in run 1 (
−
26.36 kcal mol ) is in the more
favored EGB and Eele contribution compared to those in run 2 and run 3, respectively.
3. Material and Methods
3.1. Synthesis of Coumarin Derivatives
Synthesis and characterization of the coumarin derivatives were performed as de-
scribed previously [40]. Briefly, series of coumarin derivatives were synthesized via Kno-
evenagel condensation starting from various substituted salicylaldehydes and ethyl ace-
toacetate (series 1; compounds 1–7), ethyl benzoylacetate (series 2; compounds 8–15), ethyl

Pharmaceuticals 2021, 14, 540
14 of 19
cyanoacetate (series 3; compounds 16
and dimethyl malonate (series 5, compounds 32
compounds 39 and 40 respectively) were purchased from Sigma Chemical Co. (St. Louis,
MO, USA). The structures of the tested compounds are presented in Table 1.
–
23), diethyl malonate (series 4; compounds 24
–31),
–
38). 7-hydroxycoumarin and coumarin
(
3.2. Heterologous Expression and Purification of Human DPP III
C-terminally truncated human DPP III was expressed and purified as described by
Kumar et al. [7]. Briefly, C-terminally truncated hDPP III gene on a pET28-MHL plasmid,
with an N-terminal His-tag and a TEV protease cleavage site fusion, was expressed in
BL21-CodonPlus (DE3) RIL E. coli strain using 0.25 mM IPTG for induction of expression
◦
◦
at 18 C and 130 rpm shaking. After 20 h, cells were centrifuged and frozen at
−
20 C until
purification. Bacterial cells were lysed by a combination of lysozyme lysis and sonication,
and the lysate was centrifuged to precipitate the cell debris. A brief DNase I treatment was
performed before centrifugation to reduce the viscosity caused by DNA released by lysis.
The lysate was purified on a Ni-NTA column (5 mL prepacked His-trap FF, GE Healthcare)
using a buffer system with 50 mM Tris HCl, pH = 8.0, 300 mM NaCl, and increasing
imidazole concentrations: 10 mM for lysis, 20 mM for wash and 300 mM for elution buffer.
Fractions with hDPP III were pooled and incubated with TEV protease to remove the His-
tag. hDPP III was recovered using flow-through affinity chromatography (TEV protease is
His-tagged), and additionally purified on a 16/60 Superdex-200 gel-filtration column (GE
Healthcare). Main fractions were pooled and desalted. Aliquots of protein in 20 mM Tris
◦
HCl buffer pH = 7.4 were stored at
was presented in Figure S5.
−
80 C until use. SDS PAGE of the purified enzyme
3.3. Assay of Human DPP III Activity
Purified hDPP III (1.5 nM) was preincubated with coumarin derivatives (10
for 5 min at 25 C and then for 10 min at 37 C in 50 mM Tris-HCl buffer, pH 7.4. The
µ
M) first
◦
◦
enzymatic reaction was started with Arg -2NA (40
µ
M) as a substrate, and after the 15 min
2
◦
incubation at 37 C the reaction was stopped and the absorbance was measured using the
spectrophotometric method described before [41]. Percentage enzyme inhibition (% inh.)
was calculated by comparing the enzymatic activity without (control activity), and with
inhibitor (inhibited activity) using the following formula:
%
inh. = [(control activity−inhibited activity)/(control activity)] × 100%
The IC50 values of selected compounds (12 and 36) were determined by the linear
regression of the percentage of enzyme inhibition against the increasing concentrations
(
0.5–3.5 µM) of coumarin derivatives. The IC50 value is defined as the concentration of an
inhibitor that caused a 50% reduction in the enzyme activity under assay conditions. The
stock solutions (8 mM) of coumarin derivatives were freshly prepared in dimethyl sulfoxide
and diluted with 50 mM Tris-HCl buffer, pH 7.4 buffer before assay of enzymatic activity.
3.4. Molecular Modeling
3.4.1. QSAR Analysis
Randomly ordered structures of 38 coumarin derivatives, coumarin, and 7-hydroxyco-
umarin (40 compounds in total) were drawn and optimized using the MM+ molecular
mechanics force field [42]. Afterward, the structures were also subjected to geometry
optimization using the PM3 semi-empirical method [43], using the Polak–Ribiere algorithm,
until the root-mean-square gradient (RMS) was 0.1 kcal/(Åmol). Drawing and optimization
of structures were performed in Avogadro 1.2.0. (University of Pittsburgh, Pittsburgh, PA,
USA) [44].
Descriptor calculation for the resulted minimum energy conformations of compounds
was performed with Parameter Client (Virtual Computational Chemistry Laboratory, an
electronic remote version of the Dragon program) [45]. Logarithmic values of experi-
mentally obtained hDPP III inhibition percentages were taken as response values. The

Pharmaceuticals 2021, 14, 540
15 of 19
generation and validation of QSAR models were performed using QSARINS 2.2.4 (Univer-
sity of Insubria, Varese, Italy) [46].
In order to reduce a large number of calculated descriptors, constant and semi-constant
descriptors, i.e., those with a constant value for more than 85% of compounds, and descrip-
tors that were too intercorrelated (>95%) were rejected by QSARINS. The final number
of remaining descriptors was 514. Due to the high number of inactive compounds (14), 8
of them were randomly chosen and excluded from the dataset. A genetic algorithm (GA)
was used to generate the best model. The number of descriptors in the multiple linear
regression equation was limited to three. The splitting of compounds into the training set
(
n = 27 molecules) and test set (n = 5 molecules) was performed by activity sampling [47].
Compounds were ranked by their activities (from the most active to the least active com-
pound) and then divided into five groups of the approximately same size. One compound
was selected randomly from each group and assigned to the test set. The models were
validated by the internal cross-validation performed using the “leave-one-out” (LOO)
and Yscrambling method [46]. The following evaluation criteria were included: coeffi-
2
2
cient of determination (R ), adjusted coefficient of determination (R adj), cross-validated
correlation coefficient (Q2LOO), inter-correlation among descriptors (Kxx), the difference
of the correlation among the descriptors and the descriptors plus the responses (∆K),
the standard deviation of regression (s), Fisher ratio (F), root-mean-square error (RMSE);
LOO cross-validated root-mean-square error (RMSEcv), concordance correlation coefficient
(CCC), LOO cross-validation concordance correlation coefficient (CCCcv), mean absolute
error of the training set (MAE), mean absolute error of the internal validation set (MAEcv),
and LOO cross-validated predictive residual sum of squares (PRESScv). QSAR model
2
2
robustness was tested using the Y-randomization test, giving R Yscr and Q
values [34].
Yscr
External validation parameters included the coefficient of determination of the test set
2
(
R
ext), external validation set root-mean-square error (RMSEext), external validation set
concordance correlation coefficient (CCCext), external validation set mean absolute error
2
2
2
(
MAEext), predictive squared correlation coefficients (Q F1, Q F2, Q F3) and the average
value of squared correlation coefficients between the observed and LOO predicted values
2
of the compounds with and without intercept (r m) [48].
To identify the possible outliers and compounds out of the warning leverage (h*) in
a model, a leverage plot (plot of standardized residuals vs. leverages (h); the Williams
0
plot) was used. The warning leverage is generally defined as 3p /n (n being the number of
0
training compounds, and p the number of model adjustable parameters [49]. Outliers in
the Williams plot are compounds that have values of standardized residuals higher than
two standard deviation units.
3.4.2. Preparation of the Complex Structure
The complex between the enzyme and compound 12 was built using the semi-closed
conformation of hDPP III obtained earlier [8] by MD simulations of the structure avail-
able in the Protein Data Bank (PDB code: 3FVY), since it has been proved that this is
the most preferable enzyme form in water solution [50]. Before the docking procedure,
the protonation of histidines was checked according to their ability to form hydrogen
bonds with neighboring amino acid residues. All Glu and Asp residues are negatively
charged ( 1) and all Arg and Lys residues are positively charged (+1), as expected at
−
physiological conditions. AutoDock Vina 1.1.2 [51] was used to search for the best pose of
the ligand to the enzyme active site. The docking site was defined as a cubical grid box
3
2+
with dimensions 75
×
75
×
75 Å and the center placed on the Zn . Docking simulation
was done with the standard 0.375 Å resolution and 20 conformations were generated. The
complex with the best AutoDock Vina docking score was chosen for the productive MD
simulations. Parameterization of the complex structure was performed by the AMBER-
Tools16 modules antechamber and tleap using General Amber Force Field (GAFF) [52] and
ff14SB [53] force fields to parameterize the ligand and the protein, respectively. For the zinc
2+
cation, Zn , new hybrid bonded-nonbonded parameters were used from our previous

Pharmaceuticals 2021, 14, 540
16 of 19
work [54]. The complex was dipped into the truncated octahedral box filled with TIP3P
water molecules with a margin distance of 11 Å. Besides water molecules, 24 sodium ions
were added to neutralize the system and placed in the vicinity of charged amino acids at
the protein surface.
3.4.3. Molecular Dynamics Simulations
Before the productive MD simulations, the complex was energy-minimized in three
cycles to eliminate or reduce the energy constraints. Firstly, 1500 steps of minimization
were performed, where the first 450 steps were of the steepest descent method, and the rest
was the conjugate gradient. Both the protein atoms and the metal were constrained using
2
a harmonic potential of force constant 32 kcal/(mol Å ), to equilibrate water molecules.
Secondly, 2500 steps were performed and only the first 470 steps of steepest descent were
2
used. The metal and protein backbone were constrained with 32 kcal/(mol Å ). Finally, in
the third cycle, the same number of minimization steps was as in the first cycle, and both
2
protein backbone and metal were constrained with 10 and 32 kcal/(mol Å ), respectively.
Next, the minimized system was heated from 0 to 300 K during 30 ps using a canonical
ensemble (NVT), and then equilibrated 80 ps during which the initial constraints on the
protein and the metal ion were used. This was followed by another equilibration stage of
100 ps, during which the initial constraints on the protein and the metal ion were removed
and the water density was adjusted. The time step during the periods of heating and the
water density adjustment was 1 fs. The equilibrated system was then subjected to 300 ns of
the productive MD simulations (in three replicates) at constant temperature and pressure
(
300 K and 1 atm) using the NPT ensemble, without any constraints. The temperature
−
1
was held constant using Langevin dynamics with a collision frequency of 1 ps . Bonds
involving hydrogen atoms were constrained using the SHAKE algorithm [55]. Simulations
of the complex were performed within the AMBER16 software package [56]. The time
step used for the productive MD simulations was set to 2 fs and the trajectory files were
collected every 10 ps for the subsequent analysis. Trajectory analysis and the binding free
energies (∆Gbind) evaluation was performed by the CPPTRAJ module and MMPBSA.py
script, respectively, from the AmberTools16 program package and examined visually using
VMD 1.9.3 [57] and Discovery Studio Visualizer, version 20.1.0.19295 (BIOVIA, San Diego,
CA, USA) software [58].
4. Conclusions
In summary, the potential hDPP III inhibitory activity of a series of coumarin deriva-
tives was investigated for the first time by combining in vitro and in silico approaches.
Compound 12 (3-benzoyl-7-hydroxy-2H-chromen-2-one) was found to be the most potent
inhibitory molecule with IC50 value of 1.10 µM. The productive MD simulations indicate
that H-bonds between the 7-OH group of compound 12 and the carboxyl group of Glu329
as well as van der Waals interactions with Ile315, Ser317, Phe381, Pro387, and Ile390 are
important for the mechanism of binding. According to the results of QSAR and binding
mode analyses, two new compounds with possible improved activity were proposed.
The discovery of coumarin derivatives as hDPP III inhibitors may provide new clues to
the relationship between the chemical structure and biological activity of these naturally
occurring compounds and their derivatives, and provide guidelines for the development
of novel coumarin scaffolds as potent inhibitors of this mammalian metalloproteinase.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10
.3390/ph14060540/s1. Table S1: The values of the descriptors included in QSAR model equation (2):
log % hDPP III inh. = 4.07 + 1.85 (0.59) EEig05x + 1.60 (0.52) Mor10u + 0.56 (0.39) nArOH; Table S2:
−
Detailed analysis of hydrogen bonds between compound 12 (LIG) and hDPP III residues during
MD simulations for run 1, 2 and 3 obtained by hbond command in CPPTRAJ module; Figure S1:
2
D diagram of compound 12 interactions with the hDPP III residues for the best docking pose;
Figure S2: 2D diagram of compound 12 interactions with the hDPP III residues for run 1; Figure S3:
D diagram of compound 12 interactions with the hDPP III residues for run 2; Figure S4: 2D diagram
2

Pharmaceuticals 2021, 14, 540
17 of 19
of compound 12 interactions with the hDPP III residues for run 3; Figure S5: SDS-PAGE demostrating
purity of hDPP III sample on a 10% gel: M. PageRuler Prestained protein marker. with 72 kDa band
in red; lane 1. hDPP III sample after affinity chromatography; lanes 2–5. fractions of the main hDPP
III peak after gel-filtration.
Author Contributions: Conceptualization, D.A.; Methodology, D.A., V.R., M.M.; Software, S.T., D.A.,
and M.K.; Validation, D.A. and V.R.; Formal Analysis, D.A. and M.K.; Investigation, D.A., M.K., D.Š.;
Resources, M.L. (Melita Lon cˇ ari c´ ), Z.K., D.B., B.M.P. and M.L. (Miroslav Lisjak); Data Curation, D.A.
and M.K.; Writing–Original Draft Preparation, D.A., M.K., and S.T.; Visualization, D.A. and M.K.;
Supervision, D.A.; Project Administration, S.T. and M.M.; Funding Acquisition, S.T., M.M., D.A.,V.R.
and D.B.; Writing-review and editing, D.A., M.K., S.T., V.R., Z.K. All authors have read and agreed to
the published version of the manuscript.
Funding: This work has been supported by the Croatian Science Foundation under the projects
IP-2018-01-2936 (S.T.) and UIP-2017-05-6593 (M.M.).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author.
Acknowledgments: The authors acknowledge the Isabella cluster (http://www.srce.unizg.hr/en/
isabella/, accessed on 9 September 2020) for computer time.
Conflicts of Interest: The authors declare no conflict of interest.
References
1
.
Chen, J.M.; Barrett, A.J. Dipeptidyl-peptidase III. In Handbook of Proteolytic Enzymes, 2nd ed.; Barrett, A.J., Rawlings, N.D.,
Woessner, J.F., Eds.; Elsevier Academic Press: Amsterdam, The Netherlands, 2004; Volume 1, pp. 809–812.
Rawlings, N.D.; Barrett, A.J.; Thomas, P.D.; Huang, X.; Bateman, A.; Finn, R.D. The MEROPS database of proteolytic enzymes,
their substrates and inhibitors in 2017 and a comparison with peptidases in the PANTHER database. Nucleic Acids Res. 2018, 46,
D624–D632. [CrossRef]
2
.
3.
4.
5.
Abrami c´ , M.; Zubanovi c´ , M.; Vitale, L. Dipeptidyl Peptidase III from Human Erythrocytes. Biol. Chem. Hoppe Seyler 1988, 369,
2
9–38. [CrossRef]
Baršun, M.; Jaj cˇ anin, N.; Vukeli c´ , B.; Špoljari c´ , J.; Abrami c´ , M. Human dipeptidyl peptidase III acts as a post-proline-cleaving
enzyme on endomorphins. Biol. Chem. 2007, 388, 343–348. [CrossRef]
Bezerra, G.A.; Dobrovetsky, E.; Viertlmayr, R.; Dong, A.; Binter, A.; Abramic, M.; Macheroux, P.; Dhe-Paganon, S.; Gruber, K.
Entropy-driven binding of opioid peptides induces a large domain motion in human dipeptidyl peptidase III. Proc. Natl. Acad.
Sci. USA 2012, 109, 6525–6530. [CrossRef] [PubMed]
6
7
8
9
1
1
1
.
.
.
.
Kara cˇ i c´ , Z.; Špoljari c´ , J.; Rozman, M.; Abrami c´ , M. Molecular determinants of human dipeptidyl peptidase III sensitivity to thiol
modifying reagents. Biol. Chem. 2012, 393, 1523–1532. [CrossRef]
Kumar, P.; Reithofer, V.; Reisinger, M.; Wallner, S.; Pavkov-Keller, T.; Macheroux, P.; Gruber, K. Substrate complexes of human
dipeptidyl peptidase III reveal the mechanism of enzyme inhibition. Sci. Rep. 2016, 6, 23787. [CrossRef] [PubMed]
Tomi c´ , A.; González, M.; Tomi c´ , S. The Large Scale Conformational Change of the Human DPP III–Substrate Prefers the “Closed”
Form. J. Chem. Inf. Model. 2012, 52, 1583–1594. [CrossRef] [PubMed]
Tomi c´ , A.; Tomi c´ , S. Hunting the human DPP III active conformation: Combined thermodynamic and QM/MM calculations.
Dalton Trans. 2014, 43, 15503–15514. [CrossRef]
0. Tomi c´ , A.; Kova cˇ evi c´ , B.; Tomi c´ , S. Concerted nitrogen inversion and hydrogen bonding to Glu451 are responsible for protein-
controlled suppression of the reverse reaction in human DPP III. Phys. Chem. Chem. Phys. 2016, 18, 27245–27256. [CrossRef]
1. Liu, Y.; Kern, J.T.; Walker, J.R.; Johnson, J.A.; Schultz, P.G.; Luesch, H. A genomic screen for activators of the antioxidant response
element. Proc. Natl. Acad. Sci. USA 2007, 104, 5205–5210. [CrossRef]
2. Mati c´ , S.; Kekez, I.; Tomin, M.; Bogár, F.; Šupljika, F.; Kazazi c´ , S.; Hani c´ , M.; Jha, S.; Brki c´ , H.; Bourgeois, B.; et al. Binding of
dipeptidyl peptidase III to the oxidative stress cell sensor Kelch-like ECH-associated protein 1 is a two-step process. J. Biomol.
Struct. Dyn. 2020, 1–12. [CrossRef]
1
1
1
3. Sato, H.; Kimura, K.; Yamamoto, Y.; Hazato, T. Activity of DPP III in human cerebrospinal fluid derived from patients with pain.
Masui. Jpn. J. Anesthesiol. 2003, 52, 257–263.
4. Šimaga, Š.; Babi c´ , D.; Osmak, M.; Ili c´ -Forko, J.; Vitale, L.; Mili cˇ i c´ , D.; Abrami c´ , M. Dipeptidyl peptidase III in malignant and
non-malignant gynaecological tissue. Eur. J. Cancer 1998, 34, 399–405. [CrossRef]
5. Šimaga, Š.; Babi c´ , D.; Osmak, M.; Šprem, M.; Abrami c´ , M. Tumor cytosol dipeptidyl peptidase III activity is increased with
histological aggressiveness of ovarian primary carcinomas. Gynecol. Oncol. 2003, 91, 194–200. [CrossRef]

Pharmaceuticals 2021, 14, 540
18 of 19
1
6. Šafranko, Ž.M.; Sobo cˇ anec, S.; Šari c´ , A.; Jaj cˇ anin-Jozi c´ , N.; Krsnik, Ž.; Aralica, G.; Balog, T.; Abrami c´ , M. The effect of 17
β-estradiol
on the expression of dipeptidyl peptidase III and heme oxygenase 1 in liver of CBA/H mice. J. Endocrinol. Investig. 2015, 38,
71–479. [CrossRef]
4
1
1
1
7. Prajapati, S.C.; Chauhan, S.S. Human dipeptidyl peptidase III mRNA variant I and II are expressed concurrently in multiple
tumor derived cell lines and translated at comparable efficiency in vitro. Mol. Biol. Rep. 2016, 43, 457–462. [CrossRef]
8. Allard, M.; Simonnet, G.; Dupouy, B.; Vincent, J.D. Angiotensin II Inactivation Process in Cultured Mouse Spinal Cord Cells. J.
Neurochem. 1987, 48, 1553–1559. [CrossRef] [PubMed]
9. Pang, X.; Shimizu, A.; Kurita, S.; Zankov, D.P.; Takeuchi, K.; Yasuda-Yamahara, M.; Kume, S.; Ishida, T.; Ogita, H. Novel
Therapeutic Role for Dipeptidyl Peptidase III in the Treatment of Hypertension. Hypertension 2016, 68, 630–641. [CrossRef]
[PubMed]
2
0. Jha, S.; Taschler, U.; Domenig, O.; Poglitsch, M.; Bourgeois, B.; Pollheimer, M.; Pusch, L.M.; Malovan, G.; Frank, S.; Madl, T.;
et al. Dipeptidyl peptidase 3 modulates the renin–angiotensin system in mice. J. Biol. Chem. 2020, 295, 13711–13723. [CrossRef]
[PubMed]
2
2
1. Matovina, M.; Agi c´ , D.; Abrami c´ , M.; Mati c´ , S.; Kara cˇ i c´ , Z.; Tomi c´ , S. New findings about human dipeptidyl peptidase III based
on mutations found in cancer. RSC Adv. 2017, 7, 36326–36334. [CrossRef]
2. Mati c´ , J.; Šupljika, F.; Tir, N.; Piotrowski, P.; Schmuck, C.; Abrami c´ , M.; Piantanida, I.; Tomi c´ , S. Guanidiniocarbonyl-pyrrole-aryl
conjugates as inhibitors of human dipeptidyl peptidase III: Combined experimental and computational study. RSC Adv. 2016, 6,
8
3044–83052. [CrossRef]
3. Šmidlehner, T.; Kara cˇ i c´ , Z.; Tomi c´ , S.; Schmuck, C.; Piantanida, I. Fluorescent cyanine-guanidiniocarbonyl-pyrrole conjugate with
pH-dependent DNA/RNA recognition and DPP III fluorescent labelling and inhibition properties. Mon. Chem. Chem. Mon. 2018
2
2
2
2
,
1
49, 1307–1313. [CrossRef]
´
4. Cehi c´ , M.; Sajko, J.S.; Kara cˇ i c´ , Z.; Piotrowski, P.; Šmidlehner, T.; Jeri c´ , I.; Schmuck, C.; Piantanida, I.; Tomi c´ , S. The
guanidiniocarbonylpyrrole–fluorophore conjugates as theragnostic tools for dipeptidyl peptidase III monitoring and inhibition. J.
Biomol. Struct. Dyn. 2019, 38, 3790–3800. [CrossRef] [PubMed]
5. Agi c´ , D.; Brki c´ , H.; Tomi c´ , S.; Kara cˇ i c´ , Z.; Špoljarevi c´ , M.; Lisjak, M.; Bešlo, D.; Abrami c´ , M. Validation of flavonoids as potential
dipeptidyl peptidase III inhibitors: Experimental and computational approach. Chem. Biol. Drug Des. 2017, 89, 619–627. [CrossRef]
[PubMed]
6. Blagojevi c´ , B.; Agi c´ , D.; Serra, A.T.; Mati c´ , S.; Matovina, M.; Bijeli c´ , S.; Popovi c´ , B.M. An in vitro and in silico evaluation of
bioactive potential of cornelian cherry (Cornus mas L.) extracts rich in polyphenols and iridoids. Food Chem. 2021, 335, 127619.
[CrossRef]
2
2
2
3
3
3
3
7. Detsi, A.; Kontogiorgis, C.; Hadjipavlou-Litina, D. Coumarin derivatives: An updated patent review (2015–2016). Expert Opin.
Ther. Pat. 2017, 27, 1201–1226. [CrossRef]
8. Borges, F.; Roleira, F.; Milhazes, N.; Santana, L.; Uriarte, E. Simple Coumarins and Analogues in Medicinal Chemistry: Occurrence,
Synthesis and Biological Activity. Curr. Med. Chem. 2005, 12, 887–916. [CrossRef] [PubMed]
9. Lon cˇ ari c´ , M.; Gašo-Soka cˇ , D.; Joki c´ , S.; Molnar, M. Recent Advances in the Synthesis of Coumarin Derivatives from Different
Starting Materials. Biomolecules 2020, 10, 151. [CrossRef]
0. Anand, P.; Singh, B.; Singh, N. A review on coumarins as acetylcholinesterase inhibitors for Alzheimer’s disease. Bioorg. Med.
Chem. 2012, 20, 1175–1180. [CrossRef]
1. Parellada, J.; Suárez, G.; Guinea, M. Inhibition of Zinc Metallopeptidases by Flavonoids and Related Phenolic Compounds:
Structure-Activity Relationships. J. Enzym. Inhib. 1998, 13, 347–359. [CrossRef]
2. Ali, Y.; Seong, S.H.; Jung, H.A.; Choi, J.S. Angiotensin-I-Converting Enzyme Inhibitory Activity of Coumarins from Angelica
decursiva. Molecules 2019, 24, 3937. [CrossRef]
3. Xu, X.-M.; Zhang, Y.; Qu, D.; Feng, X.-W.; Chen, Y.; Zhao, L. Osthole suppresses migration and invasion of A549 human lung
cancer cells through inhibition of matrix metalloproteinase-2 and matrix metallopeptidase-9 in vitro. Mol. Med. Rep. 2012, 6,
1
018–1022. [CrossRef]
34. Gramatica, P. Principles of QSAR models validation: Internal and external. QSAR Comb. Sci. 2007, 26, 694–701. [CrossRef]
35. Todeschini, R.; Consonni, V. Handbook of Molecular Descriptors, 1st ed.; Wiley VCH: Weinheim, Germany, 2000; pp. 91–92, 131–132.
36. Estrada, E.; Ramírez, A. Edge Adjacency Relationships and Molecular Topographic Descriptors. Definition and QSAR Applica-
tions. J. Chem. Inf. Comput. Sci. 1996, 36, 837–843. [CrossRef]
3
3
7. Rastija, V.; Agi c´ , D.; Tomi c´ , S.; Nikolic, S.; Hranjec, M.; Karminski-Zamola, G.; Abrami c´ , M. Synthesis, QSAR, and Molecular
Dynamics Simulation of Amidino-substituted Benzimidazoles as Dipeptidyl Peptidase III Inhibitors. Acta Chim. Slov. 2015, 62,
8
67–878. [CrossRef] [PubMed]
8. Schuur, J.H.; Selzer, P.; Gasteiger, J. The Coding of the Three-Dimensional Structure of Molecules by Molecular Transforms and
Its Application to Structure-Spectra Correlations and Studies of Biological Activity. J. Chem. Inf. Comput. Sci. 1996, 36, 334–344.
[CrossRef]
3
4
9. Abrami c´ , M.; Špoljari c´ , J.; Šimaga, Š. Prokaryotic homologs help to define consensus sequences in peptidase family M49. Period.
Biol. 2004, 106, 161–168.
0. Lon cˇ ari c´ , M.; Sušjenka, M.; Molnar, M. An Extensive Study of Coumarin Synthesis via Knoevenagel Condensation in Choline
Chloride Based Deep Eutectic Solvents. Curr. Org. Synth. 2020, 17, 98–108. [CrossRef]

Pharmaceuticals 2021, 14, 540
19 of 19
4
1. Špoljari c´ , J.; Salopek-Sondi, B.; Makarevi c´ , J.; Vukeli c´ , B.; Agi c´ , D.; Šimaga, Š.; Jaj cˇ anin-Jozi c´ , N.; Abrami c´ , M. Absolutely
conserved tryptophan in M49 family of peptidases contributes to catalysis and binding of competitive inhibitors. Bioorg. Chem.
009, 37, 70–76. [CrossRef]
2
42. Hocquet, A.; Langgård, M. An Evaluation of the MM+ Force Field. J. Mol. Model. 1998, 4, 94–112. [CrossRef]
43. Stewart, J.J.P. Optimization of parameters for semiempirical methods I. Method. J. Comput. Chem. 1989, 10, 209–220. [CrossRef]
44. Hanwell, M.D.; E Curtis, D.; Lonie, D.C.; Vandermeersch, T.; Zurek, E.; Hutchison, G.R. Avogadro: An advanced semantic
chemical editor, visualization, and analysis platform. J. Chemin. 2012, 4, 17. [CrossRef] [PubMed]
4
5. Tetko, I.V.; Gasteiger, J.; Todeschini, R.; Mauri, A.; Livingstone, D.; Ertl, P.; Palyulin, V.; Radchenko, E.; Zefirov, N.S.; Makarenko,
A.; et al. Virtual Computational Chemistry Laboratory–Design and Description. J. Comput. Mol. Des. 2005, 19, 453–463. [CrossRef]
[PubMed]
4
4
4
4
6. Gramatica, P.; Chirico, N.; Papa, E.; Cassani, S.; Kovarich, S. QSARINS: A new software for the development, analysis, and
validation of QSAR MLR models. J. Comput. Chem. 2013, 34, 2121–2132. [CrossRef]
7. Golbraikh, A.; Shen, M.; Xiao, Z.; Xiao, Y.-D.; Lee, K.-H.; Tropsha, A. Rational selection of training and test sets for the development
of validated QSAR models. J. Comput. Mol. Des. 2003, 17, 241–253. [CrossRef] [PubMed]
8. Chirico, N.; Gramatica, P. Real External Predictivity of QSAR Models. Part 2. New Intercomparable Thresholds for Different
Validation Criteria and the Need for Scatter Plot Inspection. J. Chem. Inf. Model. 2012, 52, 2044–2058. [CrossRef]
9. Eriksson, L.; Jaworska, J.; Worth, A.P.; Cronin, M.; McDowell, R.M.; Gramatica, P. Methods for reliability and uncertainty
assessment and for applicability evaluations of classification- and regression-based QSARs. Environ. Health Perspect. 2003, 111,
1
361–1375. [CrossRef]
5
5
5
5
5
5
5
0. Tomi c´ , A.; Berynskyy, M.; Wade, R.C.; Tomi c´ , S. Molecular simulations reveal that the long range fluctuations of human DPP III
change upon ligand binding. Mol. BioSyst. 2015, 11, 3068–3080. [CrossRef] [PubMed]
1. Trott, O.; Olson, A.J. AutoDock Vina: Improving the speed and accuracy of docking with a new scoring function, efficient
optimization, and multithreading. J. Comput. Chem. 2010, 31, 455–461. [CrossRef] [PubMed]
2. Wang, J.; Wolf, R.M.; Caldwell, J.W.; Kollman, P.A.; Case, D.A. Development and testing of a general amber force field. J. Comput.
Chem. 2004, 25, 1157–1174. [CrossRef]
3. Maier, J.A.; Martinez, C.; Kasavajhala, K.; Wickstrom, L.; Hauser, K.E.; Simmerling, C. ff14SB: Improving the Accuracy of Protein
Side Chain and Backbone Parameters from ff99SB. J. Chem. Theory Comput. 2015, 11, 3696–3713. [CrossRef] [PubMed]
4. Tomi c´ , A.; Horvat, G.; Ramek, M.; Agi c´ , D.; Brki c´ , H.; Tomi c´ , S. New Zinc Ion Parameters Suitable for Classical MD Simulations
of Zinc Metallopeptidases. J. Chem. Inf. Model. 2019, 59, 3437–3453. [CrossRef] [PubMed]
5. Ryckaert, J.-P.; Ciccotti, G.; Berendsen, H.J.C. Numerical integration of the cartesian equations of motion of a system with
constraints: Molecular dynamics of n-alkanes. J. Comput. Phys. 1977, 23, 327–341. [CrossRef]
6. Case, D.; Betz, R.; Cerutti, D.; Cheatham, T.E., III; Darden, T.; Duke, R.E.; Giese, T.; Gohlke, H.; Goetz, A.W.; Homeyer, N.; et al.
AMBER 2016; University of California: San Francisco, CA, USA, 2016.
5
5
7. Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual molecular dynamics. J. Mol. Graph. 1996, 14, 33–38. [CrossRef]
8. Dassault Systèmes BIOVIA. Discovery Studio Visualizer; Release 2019; Dassault Systèmes: San Diego, CA, USA, 2019.