Introduction
ROS induce damage to biological systems and were shown to contribute to ageing in
cells. Aberrant ROS levels are linked to a variety of cancers [1 ] and many other diseases, e.g., vascular disease [2 ], [3 ], [4 ], [5 ], [6 ]. Whereas ROS are obligatory byproducts of mitochondrial oxidative metabolism, Nox2,
the founding member of the Nox family [7 ], [8 ], was found responsible for the oxidative burst of phagocytes, where ROS are deliberately
produced by Nox2 to enforce pathogen elimination [7 ], [8 ]. In humans, the Nox family consists of seven enzymes, i.e., five NADPH oxidases
(Nox1 through Nox5; for review, see [7 ], [9 ]) and two Nox homologues (Duox1 and Duox2) [10 ], along with several subunits and regulatory proteins. All Nox proteins are transmembrane
proteins with six (Nox1–Nox5) or seven (Duox1,2) transmembrane domains. In each case,
four conserved histidines bind two heme groups that are essential for the electron
transport from cytosolic NADPH through membranes to generate ROS. An integral part
is also covered by the FAD and NADPH binding domains that are similarly conserved
[2 ], [7 ]. Both Nox2, a multisubunit NADPH oxidase best known for its role in the host defense
against pathogens, and Nox5 require activation by upstream stimuli, such as PMA (Nox2)
or calcium agonists (Nox5) for full activity; in contrast, Nox4, in complex with its
obligatory subunit p22Phox , is constitutively active [7 ]. Nox4 was identified in kidney [11 ] and was later found to be expressed in a variety of different cell types [12 ], [13 ], [14 ], [15 ], [16 ], [17 ], [18 ], [19 ], and regulation occurs primarily at the mRNA level [20 ]. Whereas altered Nox4 activity has been found in a large variety of diseases, including
pulmonary fibrosis, diabetic nephropathy (for review, see [4 ]), and benign prostatic hyperplasia [21 ], experiments with Nox4-/- mice revealed a protective role of Nox4 in the vascular system [22 ], and Nox4 deletion was shown to predispose to diet-induced obesity [23 ].
Given the pleiotropic effects of Nox4 on a variety of physiological and pathophysiological
processes, isotype-specific inhibitors of Nox4 would be potentially very useful [2 ], [3 ], [24 ], [25 ]. Whereas extracts of Piper sarmentosum regulate expression of Nox4 [26 ], inhibition of Nox4 catalytic activity was not described for such extracts. Various
efforts to search for new Nox4 inhibitors have been published in recent years, but
so far there are no Nox4 inhibitors clinically available [21 ], [27 ], although preclinical studies seem to support potential benefits in certain settings
[28 ]. Also most known inhibitors are either unspecific or toxic [29 ], [30 ]. However, there are attempts to chemically synthesise inhibitors from promising
lead compounds, but the claimed effects so far remain uncertified [24 ], [25 ], [31 ], [32 ], [33 ], [34 ]. This study aimed to identify formerly unknown Nox4 inhibitors from edible plants,
in an attempt to satisfy ADMET criteria and avoid toxicity, which is becoming a high
priority in this field [35 ].
Materials and Methods
Chemicals
All reagents were purchased from Sigma-Aldrich, unless stated otherwise.
Test compounds
The majority of the compounds used for the screening were provided as purified compounds
in solid state by AnalytiCon Discovery. For details of sample preparation, see Supplementary
Methods as Supporting Information. As control compounds, we used DPI (Sigma), VAS2870
(Vasopharm), as well as honokiol and magnolol (both kindly provided by Hermann Stuppner,
Innsbruck University). All control compounds were provided at > 95 % purity.
Cell culture
Peripheral blood mononuclear cells were isolated fresh every day from volunteers and
suspended in RPMI medium. Osteosarcoma cells (U2-OS) were acquired from ATCC, and
human embryonic kidney (HEK 293 FT) cells were bought from Invitrogen. Both were grown
in DMEM (Sigma) and provided with fresh medium every 2 and 3 days, respectively. DMEM
was supplemented with 10 % FCS (Biochrome AG, heat inactivated at 56 °C for 40 min),
4 mM L-glutamine (Gibco), and 1/100 penicillin/streptomycin (Gibco). SF9 cells, a
gift from Alexandra Lusser (Division for Molecular Biology, Medical University Innsbruck),
were incubated in SF9 II medium (Gibco) with 10 % FCS, 1.4 mM glutamine, and pen/strep
at 27 °C.
Nox dependent chemiluminescence
Chemiluminescence was performed in white 96-well plates, tissue culture-treated, optical
flat bottom (Matrix, Nunc, or Perkin Elmer), essentially as described before [36 ]. For details, see Supporting Information.
ROS-scavenging assessment by DHE staining
U2-OS cells were treated with rotenone and stained with dihydroethidium to assess
ROS levels, as described [37 ]. For details, see Supporting Information.
Preparation of cell membranes and Nox4 in vitro assay
Membrane fractionation and Nox4 activity measurement was performed similar to a method
previously described [24 ], [27 ], [34 ], [38 ]; for details, see Supporting Information.
Shape-based alignment
In a retrospective analysis of the hits in the Nox-dependent chemiluminescence assay,
we performed a shape-based analysis applying ROCS (version 3.1.1. OpenEye Scientific
Software. http://www.eyesopen.com), as described [39 ], [40 ]. For details, see Supporting Information.
Results and Discussion
To assess the activity of Nox4 in intact cells, HEK cells expressing high levels of
p22Phox were transfected with an expression vector for Nox4. Stable transfection of HEK cells
led to a significant increase of the Nox4 protein and activity, determined by chemiluminescence,
which was sensitive to inhibition by the general Nox inhibitor DPI (Fig. 1S ). This system was applied to determine the inhibitory activity of plant-derived compounds,
selected from a compound library containing 3557 purified compounds, as outlined in
Fig. 2S and briefly described below. For the first screening, a structural diverse subset
of compounds derived from edible plants was selected, with a focus on compounds inside
Lipinskyʼs Rule of Five. Results obtained with the diversity set were used to select
additional compounds for screening, based on chemical similarity. The third screening
was done with compounds mostly not structurally related to the hit compounds.
Selected experimental data are shown in [Fig. 1 ] for three selected plant compounds, representative of the activity spectrum (weak,
intermediate, strong) obtained for the 14 best hits. DMSO vehicle had no effect on
chemiluminescence activity, which was however significantly reduced by DPI when added
at increasing concentrations ([Fig. 1 A ]). Under these conditions, compound ACD005 strongly suppressed Nox4 activity when
assessed at a concentration as low as 0.5 µM (“strong” inhibitory activity) with a
clear dose dependency observed ([Fig. 1 B ]). Under the same conditions, ACD018 induced a clear-cut inhibition of Nox4 activity
when added at a concentration of 7.5 µM (“weak” inhibitory activity) ([Fig. 1 C ], [Table 1 ]), whereas ACD212 displayed significant inhibitory activity when assayed at a concentration
of 2.5 µM (“intermediate” inhibitory activity), and residual inhibitory activity was
still visible at a concentration of 0.5 µM ([Fig. 1 D ]). The results of all measurements are summarised in [Fig. 1 E ] for 14 selected plant-derived compounds and the pharmacological Nox4 inhibitor VAS2870
[41 ], [42 ], and approximative EC50 values were calculated ([Table 1 ]). Apocynin, gomisin C, magnolol, and honokiol, described as potential Nox4 inhibitors
in the literature [25 ], [43 ], [44 ], were included in our screening program; however, these compounds were at best moderately
active in our assays. Thus, EC50 values were 60 µM for apocynin, 25 µM for gomisin C, 13 µM for honokiol, and 150 µM
for magnolol. Because of the relatively low inhibitory activity in comparison to our
best hits, these compounds were not considered further.
Fig. 1 Identification of Nox4 inhibitors. A –D Nox4-dependent chemiluminescence measurements of transfected HEK 293 cells using
controls (A DMSO, DPI) and substances with strong (B ACD005), medium (C ACD212), and weak (D ACD018) inhibitory effect. E The signal of selected inhibitors at different concentrations 10 min after the start
of the experiment. All values were derived from 3 independent experiments.
Table 1 EC50 of selected Nox4 inhibitors. This table summarises the calculated EC50 values for several inhibitors, as obtained in experiments with HEK-Nox4 cells. All
values where derived from 3 independent experiments with 3 data points each, except
for × 2 independent experiments with 3 data points each and ** 1 experiment with 3
data points.
Substance
Mean [µM]
± S
ACD 005
0.54
0.10
ACD 018*
7.42
6.59
ACD 042*
2.06
0.76
ACD 047
0.87
0.42
ACD 053
0.92
0.28
ACD 055**
1.88
0.26
ACD 056
1.62
1.12
ACD 063*
2.55
2.19
ACD 066
0.83
0.44
ACD 067
1.12
0.43
ACD 069
2.37
0.69
ACD 084
3.08
2.77
ACD 176**
1.56
0.16
ACD 212
1.59
0.52
VAS 2870
2.65
1.13
Appropriate controls were included to identify inhibitors specific for Nox4 activity.
Thus, we set up a Nox2 activity screen based on normal PBMC isolated from healthy
donors, in which Nox2 activity was stimulated by addition of PMA ([Fig. 2 A ]), and compounds were applied at a concentration suitable for Nox4 inhibition (see
above, [Table 1 ]). Some of the Nox4 inhibitors, such as ACD042, did not inhibit Nox2 to any visible
degree; in contrast, compound ACD018 displayed Nox2 inhibitory activity close to that
of DPI, whereas VAS2780 displayed intermediate inhibitory activity ([Fig. 2 B ]). The relative Nox2 inhibition by selected compounds is displayed in [Fig. 2 C ].
Fig. 2 Assessment of inhibitory activity towards Nox2. A Nox2-dependent chemiluminescence of PBMCs was measured after the addition of 2 µg/mL
PMA in the presence of compounds at the indicated concentration. DMSO 0.1 % was the
vehicle control, DPI 10 µM was the positive control. B A selection of typical Nox2-CL curves over time. 10 µM DPI and 20 µM ACD018 show
strong inhibition, 1 µM ACD005 and 20 µM VAS2870 show reasonably moderate inhibition,
20 µM ACD042 shows no inhibitory capacity similar to the vehicle control. C The chemiluminescence values of the above experiment at t = 10 min as a bar graph,
comprising 14 Nox4 inhibitors, VAS2870, and standards (DMSO vehicle control, 10 µM
DPI, untreated control). (Color figures available online only.)
To further address Nox specificity we also established a HEK-derived cell line constitutively
expressing Nox5. Nox5 expression was verified, and Nox5 activity was detectable in
these cells after addition of ionomycin (Fig. 3S ), as described before [45 ]. In this system, Nox5 activity was completely inhibited by DPI. Compound ACD063
displayed intermediate inhibitory activity, and VAS2780 was inactive (Fig. 3S ).
An important source of potentially confounding effects is the well-known ability of
many plant-derived compounds to unspecifically quench ROS, in many cases by direct
chemical interaction (for recent review of ROS quenching mechanisms, see [46 ]). Such compounds, often referred to as ROS scavengers, neutralise a large variety
of chemical ROS entities and therefore are expected to negatively interfere with the
activity of most if not all experimental systems described in this communication.
To assess the ability of plant-derived compounds to quench mitochondrial ROS production
[37 ], mitochondrial ROS production was stimulated in U2-OS osteosarcoma cells by the
addition of rotenone, an inhibitor of ETC complex I [47 ]. Subsequently, cells were stained with DHE, a redox-sensitive fluorescent dye recorded
at 590 nm [37 ]. Cells were co-stained with Hoechst 33 342, and fluorescence was recorded at 465 nm.
Fluorescence intensity of Hoechst 33 342-stained cells is proportional to cell numbers
and was used to normalise DHE fluorescence. Rotenone-induced signal was reduced to
background by the addition of ascorbate ([Fig. 3 A ]), consistent with previous findings [37 ]. In this experimental system, the selected compounds had no significant activity
as ROS scavengers, with the exception of compound ACD066. Our finding that ACD066
reduced the rotenone-induced DHE signal by roughly 50 % suggests a more general mechanism
rather than specific inhibition of NADPH oxidase activity for ACD066, which was therefore
considered as a general ROS scavenger. Interestingly, the addition of some compounds,
such as ACD176 and ACD055, even increased DHE fluorescence ([Fig. 3 B ]); this observation was not investigated further as it is not directly related to
Nox inhibition.
Fig. 3 Effects of candidate compounds on mitochondrial ROS. A ROS measurement of U2-OS cells after treatment with 0.2 µM rotenone for 30 min (except
for the untreated control). The rotenone control was otherwise untreated, 0.1 % DMSO
was used as the vehicle control. B DHE signal of 14 selected Nox4 inhibitors, the vehicle control (0.1 % DMSO), and
VAS2870 at the indicated concentrations. The untreated control (Panel A ) has been subtracted as blank for all substances and the control in panel B .
For further characterisation of the Nox4 inhibitors identified in this study, we assessed
their ability to inhibit Nox4 activity in vitro . To this end, membranes were isolated from Nox4-expressing HEK cells (HEK-Nox4) and
control HEK cells (HEK-pcDNA), and Nox4 content of the membrane fractions was verified
by Western blot ([Fig. 4 A ]). NADPH oxidase activity of these membrane fractions was assayed using AR fluorescence
measurements, as described [24 ], [27 ], [34 ]. HEK-Nox4 derived membranes oxidised AR, and this activity was sensitive to DPI
inhibition ([Fig. 4 A ]), as expected. Addition of compounds ACD005, ACD047, ACD053, and ACD212 reduced
Nox4 activity in vitro , and the inhibitory activity of ACD053 was similar to that of DPI, whereas VAS2870
showed rather weak activity in this assay ([Fig. 4 B ]). Some of the compounds also inhibited the residual activity obtained with membranes
from control (HEK-pcDNA) cells ([Fig. 4 C ]), which most likely is not related to Nox4. For normalisation, the values obtained
with control-transfected cells were subtracted from the respective values obtained
with Nox4-expressing HEK cells. After normalisation, significant Nox4 inhibition was
still obtained with compounds ACD053 and ACD047, whereas some compounds failed to
inhibit Nox4 activity under these conditions ([Fig. 4 D ]). Addition of ACD066 also significantly reduced AR fluorescence in this assay ([Fig. 4 D ]); however, due to the proven capability of ACD066 to act as general ROS scavenger
(see above, [Fig. 3 ]), this compound was excluded from further analysis. DPI fully inhibited Nox4 activity
in intact cells (see above, [Fig. 1 ]), but after addition of DPI to Nox4-containing membranes, roughly 50 % of the chemiluminescence
signal persisted, probably reflecting some background signals in the in vitro setting. Conceivably, the high background signal is due to the fact that much higher
(roughly 10 fold) cell numbers are required for one measurement in the broken cell
assay, compared to assays performed in intact cells; however, other reasons for the
high background cannot be excluded.
Fig. 4 Nox4 inhibitor assay with Nox4-containing membrane fractions. A Representative experiment. Inserts: Membrane fractions were prepared from HEK 293 cells transiently transfected with
Nox4 and empty vector control, and expression levels of Nox4 and p22Phox were determined by Western blot. Main panel : Membrane fractions (10 µg protein per well) were used to oxidize AR as described
in Materials and Methods; DPI (100 µM) was added at the beginning of the measurement
as indicated. In each case, the reaction was started by the addition of 160 µM NADPH,
and fluorescence was recorded over 20 min. B, C AR measurements were performed using membrane fractions (10 µg protein per well)
prepared from HEK 293 cells transiently transfected with Nox4 and empty vector control,
as shown in panel A. All inhibitors were present at a concentration of 20 µM except
for DPI control (100 µM). 0.1 % DMSO was the vehicle control. The Nox4-dependent oxidation
of AR was started by the addition of 160 µM NADPH. D Summary of the signals at t = 10 min after the experiment was started. The stars
indicate the significance of the difference between the given substance and the DMSO
control from a two-sided Studentʼs test (* p < 0,05; ** p < 0,01; *** p < 0,001).
(Color figures available online only.)
Of note, the inhibitory potential of VAS2870 was strongly limited in these experiments,
and ACD005 showed only an indistinct trend, relative to experiments in intact cells
([Fig. 4 D ]); possible reasons for this are discussed below with the chemoinformatic analysis.
The bioavailability and a compoundʼs ability to penetrate into the cell is a big issue,
and not all compounds that can inhibit an enzyme in vitro are expected to reach their protein target in intact cells. This is of particular
importance for Nox4, since it is known that, unlike other members of the Nox family,
a substantial proportion of Nox4 is found in intracellular membranes [7 ], [48 ], [49 ], [50 ].
We also assessed the activity of selected compounds towards a recombinant fragment
of Nox4, which retains the dehydrogenase reaction associated with Nox4 enzymatic activity.
The protein fragment Nox4304–578 [51 ] was cloned in a baculoviral vector, expressed in SF9 insect cells as an MBP fusion
protein and purified by affinity chromatography [51 ]. Upon incubation with NADPH, FAD, and NBT, significant dehydrogenase activity was
detectable, which was abolished by boiling the recombinant protein prior to the assay
(data not shown). The reaction could be triggered by addition of NADPH (Fig. 4S ) or, alternatively, by the addition of FAD in the presence of NADPH (data not shown),
confirming the dependency on both cofactors. Surprisingly, addition of DPI did not
lead to sustained inhibition of the diaphorase activity associated with Nox4DH domain;
instead, DPI addition in this setting even increased signal intensity (data not shown).
The reasons why DPI fails to act as an inhibitor in this setting remain to be clarified.
Whereas compound ACD084 significantly inhibited the dehydrogenase reaction catalysed
by Nox4304–578 , compounds ACD005 and ACD063 displayed weak but significant inhibitory activity (Fig. 4S ); in contrast, compounds ACD069 and VAS2870 were inactive in this assay (Fig. 4S ). This may indicate that the mode of inhibition displayed by these inhibitors differs
from the other compounds described in this communication. As was pointed out by Jaquet
et al. [25 ], inhibition of Nox enzymes including Nox4 can occur at different stages, including
depletion of substrate, interference with transcription or complex assembly, masking
of the reaction product, etc. For compounds that inhibit Nox4 in intact cells but
not in the Nox4DH assay, inhibition apparently involves Nox4 domains not present in
the fragment. Accordingly, the mechanism of inhibition may differ between compounds
described here and remains to be identified in all cases.
Plant-derived compounds described here can be assigned to several activity profiles:
several compounds qualify as highly active Nox4 inhibitors, defined here by an EC50 below 1 µM. This includes ACD005, ACD047, and ACD053, all of which also inhibited
Nox2 activity. Two additional compounds identified here, ACD042 and ACD084, are of
potential interest, because they did not inhibit the activity of Nox2 nor Nox5. They
may be considered as specific Nox4 inhibitors, although their EC50 is slightly higher than for the best Nox4 inhibitors described above. In quantitative
terms, their activity is similar to the recently described Nox4 inhibitor VAS2870
[52 ], which in our hands displayed reasonable activity towards Nox4 and Nox2.
During screening as well as retrospectively, chemoinformatics tools were applied to
guide selection of candidates and to characterize our best hits, respectively. In
a shape-based alignment, we identified ACD018 and ACD067 as suitable query molecules
to enrich hits within inactive compounds during the evaluation showing the highest
area under the receiver operator curve value (Fig. 5S ). A retrospective chemical characterisation was done by calculation of selected chemical
descriptors allowing an estimation of the drug-likeness for oral uptake according
to Lipinski [53 ] (Table 1S ). ACD005 was the only compound showing clear violation of Lipinskiʼs Rule of Five
with more than 10 acceptors, more than 5 donors, and a molecular weight above 500 Da.
This candidate Nox4 inhibitor bears a sugar moiety and therefore is very hydrophilic
which is reflected in a low logP (o/w). In contrast, high logP (o/w) values are estimated
for the compounds ACD018 and ACD212. Although these logP (o/w) values are higher than
the criteria of logP ≤ 5, those compounds are classified as drug-like according to
Lipinski as one violation is accepted.
The retrospective analysis of the identified hits relies on molecular descriptors
calculated in silico . Chemical characteristics within the hits presented here give insight in the chemical
requirement for Nox4 inhibitors ([Table 2 ]). So, the occurrence of 6 diarylheptanoids (ACD053, ACD055, ACD063, ACD066, ACD067,
and ACD084) and the presence of phenol substructures within 12 out of 14 presented
compounds indicate that this substructure is beneficial for Nox4 inhibition. For ROCS,
the high AUC values for ACD018 and ACD057 are caused by the presence of 4-substituted
phenol substructures. This can be seen as a common characteristic within almost all
actives except ACD042 and ACD005. In fact, ACD055, which is the only diarylheptanoid
without hydroxylation at the aromatic rings, is less active (EC50 = 1.88 µM) than the chemically related compound ACD067 with a phenol.
Table 2 Selected compounds. Given in the table are the compound numbers as used in the paper
in the first column. The second column gives the trivial name as known so far, the
third column provides the order numbers from AnalytiCon Discovery. Column four shows
the 2D chemical structure, and the last line provides the SMILE (Simplified Molecule
Input Line Entry) string for each compound.
Compound Nr.
Name
Order Nr.
Structure
SMILES
ACD 005
–
NP-000520
([Fig. T1 ])
O[C@@H]([C@H](OCC1C(C2=CC=C(O)C(OC)=C2)C3=CC(O)=C(OC)C=C3CC1CO)O4)[C@H]([C@@H]([C@H]4CO)O)O
ACD 018
Bakuchiol
NP-007515
([Fig. T2 ])
OC1=CC=C(/C=C/[C@](CC/C=C(C)/C)(C)C=C)C=C1
ACD 042
Grindelic acid
NP-002752
([Fig. T3 ])
CC1(CCC[C@]2(C)[C@]3(O[C@@](C)(CC(O)=O)CC3)C(C)=CCC12)C
ACD 047
–
NP-000261
([Fig. T4 ])
CC(C(OC1C2=CC(OC)=C(O)C=C2)C3=CC=C(O)C(OC)=C3)C1C
ACD 053
–
NP-002766
([Fig. T5 ])
OC(C(OC)=C1)=CC=C1/C=C/C(CC(/C=C/C2=CC(O)=C(O)C=C2)=O)=O
ACD 055
Dihydroyashabushiketol
NP-003489
([Fig. T6 ])
O=C(C[C@@H](O)CCC1=CC=CC=C1)CCC2=CC=CC=C2
ACD 056
Gomisin J
NP-015225
([Fig. T7 ])
CC1C(C)CC2=C(C(OC)=C(OC)C(O)=C2)C3=C(C=C(O)C(OC)=C3OC)C1
ACD 063
–
NP-012941
([Fig. T8 ])
OC(C=CC(CCCCC(CCC1=CC=C(O)C=C1)OC(C)=O)=C2)=C2O
ACD 066
–
NP-003620
([Fig. T9 ])
O=C(CCCCC1=CC(O)=C(C=C1)O)CCC2=CC=C(O)C(O)=C2
ACD 067
–
NP-003332
([Fig. T10 ])
O=C(CC(O)CCC1=CC=C(O)C(OC)=C1)CCC2=CC=CC=C2
ACD 069
–
NP-012926
([Fig. T11 ])
CC(C(C(O1)C2=CC=C(O)C=C2)C)C1C3=CC(OC)=C(C=C3)OC
ACD 084
–
NP-003624
([Fig. T12 ])
OC(C=CC(CCCCC(CCC1=CC=C(O)C(O)=C1)OC(C)=O)=C2)=C2O
ACD 176
–
NP-003501
([Fig. T13 ])
C[C@](C1CC2=O)(C3=C2C=C(O)C(O)=C3)CCCC1(C)C
ACD 212
–
NP-014837
([Fig. T14 ])
CC(CC1=CC(OC(C2=CC=C(O)C=C2)=O)=C(O)C=C1)C(C)CC3=CC(O)=C(O)C=C3
VAS 2870
–
–
([Fig. T15 ])
C1(N=C(SC2=NC=NC3=C2N=NN3CC4=CC=CC=C4)O5)=C5C=CC=C1
Fig. T1 ACD 005
Fig. T2 ACD 018
Fig. T3 ACD 042
Fig. T4 ACD 047
Fig. T5 ACD 053
Fig. T6 ACD 055
Fig. T7 ACD 056
Fig. T8 ACD 063
Fig. T9 ACD 066
Fig. T10 ACD 067
Fig. T11 ACD 069
Fig. T12 ACD 084
Fig. T13 ACD 176
Fig. T14 ACD 212
Fig. T15 VAS 2870
For ACD005, the low AUC value is caused by the sugar moiety unique within the test
set. ACD005 was characterised by high activity towards Nox4 in the chemiluminescence
assay in the assay with intact cells ([Table 1 ]) but a surprisingly low effect on the isolated membranes ([Fig. 4 D ]). We assume ACD005 to enter the intact cells although it violates the Lipinskiʼs
Rule of Five due to its sugar moiety. The glycosylation is a typical feature of natural
products, and Lipinskiʼs Rule of Five might not appropriately reflect the consequences
on cellular uptake, as this rule of thumb was developed from a statistical analysis
of non-natural drug molecules [53 ]. Additionally, the glycosylic bond of ACD005 might be subject of enzymatic hydrolysis
depending on the environment within the cells, and the aglycon might be the active
form. This hypothesis could explain the reduced activity of ACD005 in the assays performed
with isolated membranes and the isolated Nox4 dehydrogenase domain (Nox4304–578 ) fragment. Besides ACD005, no Lipinskiʼs Rule violations are observed. The two compounds,
ACD042 and ACD084, discussed as specific Nox4 inhibitors, showed promising ADMET characteristics.
Nox4 inhibitors have been described by others before. Thus, it was shown that extracts
of Piper sarmentosum inhibit expression of Nox4 at the transcriptional level [26 ]. In an approach similar to the one chosen here [32 ], cell-based assays were used to characterise a set of synthetic compounds derived
by medicinal chemistry for their ability to inhibit Nox4. Several hits with an IC50 (roughly equivalent to the EC50 described here) in the low micromolar range were reported, and Nox4 activity was
entirely inhibited at a concentration of 10 µM for each of their best hits. The compounds
described here have comparable activity in cell-based assays, but since they are derived
from edible plants, we expect better compatibility with ADMET criteria. Based on a
screening campaign of substance libraries with Nox4-containing membranes, Ki values in the two-digit nanomolar range were calculated for the best hits, including
GKT137831 [25 ], [33 ], which is currently being considered for clinical trials [28 ]. Ki values in general describe the kinetic interaction between enzyme and substrate,
and their calculation is rather difficult in such an approach, given the fact that
the molar concentration of Nox4 in isolated membranes is unknown. The comparison to
the inhibitory activity of compounds described here is further complicated by the
fact that EC50 values for these compounds in cell-based assays are currently unknown. In a recent
study with a cellular model of Nox4-dependent liver fibrosis, it was shown that inhibition
of Nox4 activity in hepatocytes required GKT137831 at a concentration of 20 µM [28 ], suggesting that the most active compounds (ACD005, ACD047, ACD053, ACD66, ACD067;
see [Table 1 ]) identified here are at least as active as GKT137831, which is considered the best
available pharmacological Nox4 inhibitor to date [31 ].
