Screening, synthesis, crystal structure, and molecular basis of 6-amino-4-phenyl-1,4-dihydropyrano[2,3-c]pyrazole-5-carbonitriles as novel AKR1C3 inhibitors

Article abstract of DOI:10.1016/j.bmc.2018.10.044

AKR1C3 is a promising therapeutic target for castration-resistant prostate cancer. Herein, an evaluation of in-house library discovered substituted pyranopyrazole as a novel scaffold for AKR1C3 inhibitors. Preliminary SAR exploration identified its derivative 19d as the most promising compound with an IC₅₀ of 0.160 μM among the 23 synthesized molecules. Crystal structure studies revealed that the binding mode of the pyranopyrazole scaffold is different from the current inhibitors. Hydroxyl, methoxy and nitro group at the C4-phenyl substituent together anchor the inhibitor to the oxyanion site, while the core of the scaffold dramatically enlarges but partially occupies the SP pockets with abundant hydrogen bond interactions. Strikingly, the inhibitor undergoes a conformational change to fit AKR1C3 and its homologous protein AKR1C1. Our results suggested that conformational changes of the receptor and the inhibitor should both be considered during the rational design of selective AKR1C3 inhibitors. Detailed binding features obtained from molecular dynamics simulations helped to finally elucidate the molecular basis of 6-amino-4-phenyl-1,4-dihydropyrano[2,3-c]pyrazole-5-carbonitriles as AKR1C3 inhibitors, which would facilitate the future rational inhibitor design and structural optimization.

Full text of DOI:10.1016/j.bmc.2018.10.044

Accepted Manuscript
Screening, Synthesis, Crystal Structure, and Molecular Basis of 6-Amino-4-
phenyl-1,4-dihydropyrano[2,3-c]pyrazole-5-carbonitriles as Novel AKR1C3
Inhibitors
Xuehua Zheng, Zan Jiang, Xiaolin Li, Chen Zhang, Zhe Li, Yinuo Wu, Xinhua
Wang, Chao Zhang, Hai-bin Luo, Jun Xu, Deyan Wu
PII:
S0968-0896(18)31389-0
https://doi.org/10.1016/j.bmc.2018.10.044
BMC 14603
DOI:
Reference:
Bioorganic & Medicinal Chemistry
To appear in:
Received Date:
Revised Date:
Accepted Date:
3 August 2018
15 October 2018
30 October 2018
Please cite this article as: Zheng, X., Jiang, Z., Li, X., Zhang, C., Li, Z., Wu, Y., Wang, X., Zhang, C., Luo, H-b.,
Xu, J., Wu, D., Screening, Synthesis, Crystal Structure, and Molecular Basis of 6-Amino-4-phenyl-1,4-
dihydropyrano[2,3-c]pyrazole-5-carbonitriles as Novel AKR1C3 Inhibitors, Bioorganic & Medicinal Chemistry
(2018), doi: https://doi.org/10.1016/j.bmc.2018.10.044
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Screening, Synthesis, Crystal Structure, and Molecular Basis of
6-Amino-4-phenyl-1,4-dihydropyrano[2,3-c]pyrazole-5-carbonitriles as
Novel AKR1C3 Inhibitors
Xuehua Zhenga, Zan Jianga, Xiaolin Lib, Chen Zhangb, Zhe Lib, Yinuo Wub, Xinhua Wanga, Chao
Zhanga, Hai-bin Luob, Jun Xu b,*, Deyan Wub,†
aKey Laboratory of Molecular Target & Clinical Pharmacology, School of Pharmaceutical Sciences & the Fifth Affiliated
Hospital, Guangzhou Medical University, Guangzhou 511436, China
bSchool of Pharmaceutical Sciences, Sun Yat-sen University, Guangzhou, 510006, China
Abstract: AKR1C3 is a promising therapeutic target for castration-resistant prostate cancer. Herein,
an evaluation of in-house library discovered substituted pyranopyrazole as a novel scaffold for
AKR1C3 inhibitors. Preliminary SAR exploration identified its derivative 19d as the most promising
compound with an IC50 of 0.160 μM among the 23 synthesized molecules. Crystal structure studies
revealed that the binding mode of the pyranopyrazole scaffold is different from the current inhibitors.
Hydroxyl, methoxy and nitro group at the C4-phenyl substituent together anchor the inhibitor to the
oxyanion site, while the core of the scaffold dramatically enlarges but partially occupies the SP
pockets with abundant hydrogen bond interactions. Strikingly, the inhibitor undergoes a
conformational change to fit AKR1C3 and its homologous protein AKR1C1. Our results suggested
that conformational changes of the receptor and the inhibitor should both be considered during the
rational design of selective AKR1C3 inhibitors. Detailed binding features obtained from molecular
dynamics
simulations
helped
to
finally
elucidate
the
molecular
basis
of
6-Amino-4-phenyl-1,4-dihydropyrano[2,3-c]pyrazole-5-carbonitriles as AKR1C3 inhibitors, which
would facilitate the future rational inhibitor design and structural optimization.
Key words: AKR1C3 inhibitors; Synthesis; Crystal structure; Castration-resistant prostate cancer;
1,4-dihydropyrano[2,3-c]pyrazole
*Corresponding author. Jun Xu, Tel.: +086-020-3994-3023, Fax: +86-20-3994-3000, and E-mail: junxu@biochemomes.com.
Deyan Wu, Tel.: +086-020-3994-3062, Fax: +86-20-3994-3000, and E-mail: wudeyan3@mail.sysu.edu.cn.

1. Introduction
AKR1C3, one member of aldo-keto reductase (AKR) superfamily, has also been designated as
17β-hydroxysteroid dehydrogenase type 5 (17β-HSD5) due to its 17- ketosteroid reductase activity.
The substrates of AKR1C3 include the precursors in all pathways to the potent androgens
testosterone (T) and 5α-dihydrotestosterone (DHT) in the prostate, such as dehydroepiandrosterone,
△4-androstene-3,17-dione, androstan-3,17-dione and androsterone1. Recent studies have established
that the reactivation of the androgen axis contributes to castrate resistant prostate cancer (CRPC),
and led to FDA approval of abiraterone (steroidogenic enzyme CYP17A1 inhibitor) and
enzalutamide (new androgen receptor antagonist) for the treatment of patients with CRPC2-4. The
involvement of AKR1C3 in downstream androgen biosynthesis suggests that AKR1C3 plays a
critical role in CRPC. Recent investigations have indicated that AKR1C3 is overexpressed in cell
lines deprived of androgens, in prostate tumor xenografts in castrate mice, and in CRPC patients.
Knockdown or inhibition of AKR1C3 resulted in suppression of tumor cell growth within a castrate
environment and a decrease in intra-tumoral testosterone production in castrated nude mice induced
by androstenedione5-13. Yepuru et al. have identified that AKR1C3 is also a unique AR-selective
coactivator to promote CRPC growth and AKR1C3-selective competitive inhibitors inhibit this
coactivator function14. Moreover, AKR1C3 is implicated in resistance to abiraterone and
enzalutamide therapies and enzalutamide resistance in vitro and in vivo can be overcome with
AKR1C3 competitive inhibitors7, 8. Collectively, these data suggest that AKR1C3 inhibition may
have distinct advantages over the current therapeutics for the treatment of CRPC.
Considerable efforts have been invested into the discovery of compounds inhibiting AKR1C3
first for research purposes, but later also for therapeutic applications1, 15, 16 17-23. Penning et al. have
extensively studied the inhibitory activities of nonsteroidal anti-inflammatory drugs (NSAIDs) and
discovered N-phenylanthranilate and indomethacin analogs as AKR1C3 inhibitors with high
selectivity versus COX-1,217-23. The high-throughput screening conducted by Jamieson et al.
identified 3-(3,4-dihydroisoquinolin-2(1H)-ylsulfonyl)benzoic acid as a new carboxylate inhibitor of
AKR1C3 24. Several non-carboxylate inhibitors including isoquinolines, morpholylureas, pyrrolidine,
and indole derivative have also been reported14, 25-27. At the same time, there have been significant
efforts to explore the interaction of existing inhibitors with the enzyme by single-crystal techniques.

Studies conducted by us and other groups have reported 43 crystal structures of AKR1C3 in complex
with different inhibitors28-35. The ligand binding pocket of AKR1C3 can be divided into oxyanion
site, steroid channel, and three sub-pockets, SP1, SP2, and SP31. Most of inhibitors are anchored to
the oxyanion site by hydrogen bonding with Tyr55 and His117 and occupy the SP1 pocket, where
usually result in conformational changes of Trp227, Phe306 and Phe31115. Nevertheless, no
AKR1C3 inhibitors are currently in clinical use for the treatment of CRPC, and superior AKR1C3
inhibitors are still in need.
To seek novel AKR1C3 inhibitors, in the current study, we conducted a screening on an
in-house compound library consisting of 298 small molecules by using enzymatic assay. Nine
inhibitors were identified with new scaffolds compared to the current AKR1C3 inhibitors.
Preliminary SAR studies were performed on the best compound and identified 19d as the most
potent AKR1C3 inhibitor with an IC50 of 0.160 μM among the synthesized 23
6-Amino-4-phenyl-1,4-dihydropyrano-[2,3-c]pyrazole-5-carbonitriles. Crystal structure, molecular
docking, together with binding free-energies and per-residue contributions studies revealed a new
insight into the molecular basis for selective inhibition of AKR1C3, which would be helpful in future
design and optimization of new AKR1C3 inhibitors.
2. Materials and methods
2.1. Enzyme preparation and inhibition assays
Human recombinant AKR1C1 and AKR1C3 were expressed in the E. coli BL21 (Condon Plus)
and were purified using the procedures as described before34. These enzymes in vitro reduce
9,10-phenanthrenequinone (PQ) with high catalytic efficiency in the presence of the coenzyme
NADPH. Initial velocities were determined with a Flex Station® 3 Multi-Mode Microplate Reader
(Molecular Devices) by measuring the decrease in NADPH emission of 460 nm with an excitation of
340 nm. The potency of the compounds was determined by their ability to inhibit the reduction
reaction. The volume of the reaction mixture was 100 μL that contained 0.1 M phosphate buffer (pH
= 6.0), 0.15 mM NADPH, 8.0 μM PQ, 0.33 μM AKR1C3 or 0.39 μM AKR1C1, and 1% DMSO with
or without compound. Screening was carried out at 25 μM compounds and IC50 value were measured
for a strong inhibition corresponding to more than 50% AKR1C3 inhibition. Selectivity towards
AKR1C1 was further determined for the most promising compounds with IC50 less than 5.0 μM.

2.2. Chemistry
As shown in Scheme 1-2, we developed an efficient route to obtain the expected
6-amino-4-phenyl-1,4-dihydropyrano[2,3-c]pyrazole-5-carbonitriles 1, 13a-k, and 19a-l. Our initial
efforts focused on synthesis of C4-substituted and N1-substituted derivatives from
6-amino-4-phenyl-3-propyl-1,4-dihydropyrano[2,3-c]pyrazole-5-carbonitrile (Schemes 1). Ethyl
3-oxohexanoate in ethanol or glacial acetic acid was treated with the appropriate hydrazine to afford
37
the N-substituted-3-propyl-1H-pyrazol-5(4H)-ones (12a-g), respectively36,
Then, 6-amino-4-
phenyl-3-propyl-1,4-dihydropyrano[2,3-c]pyrazole-5-carbonitriles 1 and 13a-k were respectively
synthesized by reaction of 12a-g, aldehydes, and malononitrile in ethanol. Compounds 19a-l were
synthesized through the route outlined in Scheme 2. Meldrum's acid was reacted with acyl chloride
in the presence of CH2Cl2 at 0 °C using pyridine as a base to afford the 16a-l, which was then
alcoholysis in ethanol to get the β-ketoester 17a-l38. Finally, use the same procedure as scheme 1 to
obtained the target compounds 6-amino-4-phenyl-1,4-dihydropyrano-[2,3-c]pyrazole- 5-carbonitriles
19a-l, respectively.
Scheme 1. Synthesis of compound 1 and 13a-ka
aReagents and conditions: (a) hydrazine, EtOH, 70 °C, 2 h; or AcOH, AcONa, reflux 12 h; (b)
aldehyde, malononitrile, piperidine, EtOH, 70 °C, 1 h.

To avoid interference of false positive compounds with our subsequent study, PAINS screening
of the designed compounds was performed using an online program (i.e., “PAINSRemover”,
http://www.cbligand.org/PAINS/), and all the compounds passed the filter39.
Scheme 2. Synthesis of compound 19a-la
aReagents and conditions: (a) pyridine, CH2Cl2, 0 °C, 30 min, r.t., 12 h; (b) EtOH, reflux, 4 h; (c)
hydrazine, EtOH, 70 °C, 2 h; or AcOH, AcONa, reflux 12 h; (d) aldehyde, malononitrile, piperidine,
EtOH, 70 °C, 1 h.
2.3. Crystallization and structure determination
Crystallization of AKR1C3/AKR1C1-NADP+-inhibitor complex was achieved by soaking the
AKR1C1/AKR1C3-NADP+
crystals
in
a
saturated
solution
of
the
inhibitor.
AKR1C3/AKR1C1-NADP+ crystals were obtained using hanging drop vapor diffusion method at
298 K. The drops were prepared by mixing 2.0 μL protein-NADP+ mixture with 2.0 μL reservoir
solution. AKR1C3-NADP+ crystals appeared in drops containing 0.1 M MES (pH 6.0), 15-20% (w/v)
PEG 8K, 0.14 M NaCl, while AKR1C1- NADP+ crystals grew in drops consisting 0.1 M Hepes (pH
7.3), 23-27% PEG 4K, 10 mM CaCl2, 0.4 M NaCl.

X-ray data were recorded on our in-house Oxford Diffraction Xcalibur Nova diffractometer and
processed with the program CrysAlis Pro. The structures were built with Phaser by using structure of
4FAM (AKR1C3) or 3NTY (AKR1C1) as a model for molecular replacement, and refined with the
program Phenix and Coot40, 41. Grade Web Server was used to generate the coordinate and restraint
for the inhibitor. Data collection and refinement statistics are shown in Table S1. The coordinate data
were deposited in the Protein Data Bank with accession numbers of 6A7A, 6A7B and 6IJX. Related
figures were produced using PyMOL.
2.4. Molecular docking, molecular dynamics simulations and binding free energy
calculations
Molecular docking was performed using CDOCKER module embedded in Accelrys Discovery
Studio 2.5.5. The crystal structures presented in the present paper were used as the receptor. All the
water molecules and ligands were removed from the target protein except cofactor NADP+ and the
water molecules were found to play an important role in receptor–ligand interactions (See 3.3). A
spherical region with a radius of 10 Å was constructed as the binding site, based on the location of
the co-crystallized ligand. The receptor and the ligands were typed with CHARMm force field and
Momany-Rome partial charges. Other input parameters were set as their default options. The original
ligand was re-docked into the X-ray protein to verify the dock procedure. An RMSD less than 1.0 Å
between the crystal and docked conformations was considered desirable docking protocol.
The complexes obtained from X-ray or molecular docking was subject to the molecular
dynamics (MD) simulation and subsequent binding free energy calculations using the sander module
of Amber 14. The detailed parameters used in this study were similar to those in our previous
studies42. The system was initially energy-minimized using steepest descent method followed by
conjugated gradient method in the presence of positional restraints on both protein and ligand. After
heated to 300 K with positional restraints for the protein, 8ꢀns production simulations were carried
out after 100 ps equilibration simulations. Crystal water molecules mediating hydrogen bonds
between the ligand and receptor were kept during the MD simulation. RMSD for the backbone atoms
was monitored to validate the stability of the complex system. The binding free energies (ΔGbind
)
were calculated with MM-GBSA approach using the final 1 ns of MD trajectories, which was
well-equilibrated43-47. To obtain a detailed view of interactions between the ligand and the receptor,

the total free energy were then decomposed to each residue with the MM-GBSA method.
3. Results and discussion
3.1. In-house library screening identified 1 as AKR1C3 inhibitor with completely
novel molecular scaffold
To identify AKR1C3 inhibitors with new scaffold, we screened 298 compounds at 25 μM in
triplicate from our in house library. Twenty compounds showed more than 50% inhibition towards
AKR1C3. Out of these compounds, 3 compounds produced high fluorescence, 5 compounds were
poorly soluble in the aqueous buffer, and IC50 values for these compounds were unable to be assayed
in our assay system. Among 12 compounds assayed, there were 3 steroids whose analogues were
already identified as AKR1C3 inhibitors by others (data not shown). The active compounds having
novel scaffolds for AKR1C3 inhibitors are summarized in Table 1, together with their inhibitory
activity and selectivity for those with the IC50 values < 5.0 μM.
Nine structurally diverse compounds were identified in the present screening with the IC50
values covering a broad range from 0.227 μM to 24.2 μM. Among them, two compounds showed an
IC50 values < 5.0 μM towards AKR1C3. Compound 2 with an IC50 value of 1.46 μM is a pyrimidine
derivative. For its relatively high IC50 value and complicated chemical structure, we didn’t pay much
attention to it. The best one (compound 1) showed a very low IC50 value of 0.227 μM with a slight
selectivity towards AKR1C1 (0.411 μM, see the explanations in 3.2). Compared to current AKR1C3
inhibitors, 1,4-dihydropyrano[2,3-c]pyrazole derivative 1 has a completely novel scaffold, which is
much more compact and rigid implying different inhibition mode may exist. Besides,
1,4-dihydropyrano[2,3-c]pyrazoles have been found to have antimicrobial, insecticidal,
anti-inflammatory, molluscicidal, and Chk1 kinase inhibitory activities36. The chemistry on
1,4-dihydropyrano[2,3-c]pyrazoles have also received remarkable interest and some efficient
synthetic routes have been reported using commercially available materials with high diversity 36, 48
Therefore, it can be concluded that wide varieties of substituents on
.
1,4-dihydropyrano[2,3-c]pyrazole core could be easily achieved and therefore generating novel
promising AKR1C3 inhibitors.

Table 1. Structures and inhibitory activities of AKR1C3 inhibitors with novel scaffolds
Inhibitory activity (IC50, μM)
No.
Structure
AKR1C3
AKR1C1
1
0.227±0.013
0.411±0.013
2
1.46±0.29
9.87±1.77
3
4
5.54±1.66
7.22±1.90
-
-
5
6
7
7.80±0.14
14.8±2.9
15.2±2.4
-
-
-

8
15.5±2.9
24.2±5.1
-
-
9
Meclofenamic acida
0.512±0.080(0.54b) 0.740±0.065(3.16b)
aUsed as a positive control;
bIC50 values taken from reference24.
3.2. Preliminary structure activity relationship investigations
Inspection of the reported crystal structures of AKR1C3-NADP+-inhibitor showed that the
carboxylate/carbonyl group on most of the known inhibitors often provides the basic anchorage of
the inhibitor to the oxyanion site by hydrogen bonding with the catalytic tetrad members Tyr55 and
His117. The anchorage may be not obligated but apparently beneficial for the high binding affinities
of the current AKR1C3 inhibitors. Thus, we proposed a hypothesis that the nitro group (on
C4-phenyl substituent) could serve as the anchor and play a role in strong inhibition of 1 towards
AKR1C3. As seen from Table 2, removal of nitro group (13c) resulted in great decrease in the
inhibitory activity. However, 3´-nitrophenyl at C4-position led to loss of inhibition of 13b. Crystal
structure of 4wdt showed an example of hydrogen bonding to the catalytic residues by a combination
of nitro and hydroxyl groups, which would suggest that both of nitro and hydroxyl group may be
important for retaining high binding affinity towards AKR1C3. 13a with 3´-nitro-4´-hydroxylphenyl
was thus synthesized, which as expected displayed comparable inhibitory activity to compound 1.
Table 2. SAR of phenyl-substituted for inhibition of AKR1C3 and AKR1C1

Inhibitory activity(IC50，μM)
Cpd.
R
AKR1C3
AKR1C1
0.661±0.125
> 25
4´-OH-3´-NO2
3´-NO2
0.278±0.054
> 25
13a
13b
13c
13d
13e
13f
4´-OH- 3´-OMe
3´,4´-diOH
4´-OH
16.5±3.8
3.96±0.42
6.73±0.79
> 25
2.49±0.02
5.09±1.45
3.51±1.19
> 25
4´-F
Analogues sets around 1,4-dihydropyrano[2,3-c]pyrazole core were developed for the purpose
of probing the interaction in the SP pockets, which usually relate with inhibitor selectivity in
AKR1Cs. There are two closely related isoforms of AKR1C3 (AKR1C1 and AKR1C2), which are
involved in DHT inactivation, their inhibition may increase the androgenic signal1. Thus, it will be
necessary to inhibit AKR1C3 selectively for the treatment of CRPC. Since only a single residue is
different at their respective active site, AKR1C1 inhibitors could potentially also inhibit AKR1C2.
Selectivity over AKR1C1 for the compounds was evaluated. As seen in Table 3, substituents at
N1-position appear to be disfavored to the inhibitory activity (13g-13k) towards both of AKR1C3
and AKR1C1. The least active compound of this set was 13j, with a pyridyl substitution at
N1-position, which was less potent than 1 about 22- and 6-fold for AKR1C3 and AKR1C1,
respectively. By contrast, a range of substituents at C3-position seems to be very well tolerated, as
many compounds exhibit comparable or superior inhibitory activities to the parent 1 (Table 4).
However, both the short and long chain substituents would decrease the potency. Overall, the most
potent compound 19d, with pentyl substitution at the C3-position, had an IC50 of 0.160 μM. 19d is
also the most selective compound among all the molecules, which illustrated about 3.3-fold
selectivity between AKR1C3 and AKR1C1.
Table 3. SAR of N1-substitutions for inhibition of AKR1C3 and AKR1C1

Inhibitory activity (IC50，μM)
Cpd.
R
AKR1C3
AKR1C1
-CH3
0.510±0.091
1.83±0.37
0.462±0.082
13g
13h
1.05±0.18
1.09±0.07
2.52±0.45
0.778±0.044
5.05±0.56
13i
13j
0.683±0.157
1.27±0.16
13k
Table 4. SAR of C3-substitutions for inhibition of AKR1C3 and AKR1C1
Inhibitory activity (IC50，μM)
Cpd.
R
AKR1C3
AKR1C1
0.533±0.095
0.255±0.025
0.213±0.020
0.160±0.028
-CH3
0.487±0.087
0.790±0.005
0.810±0.046
0.531±0.095
19a
19b
19c
19d
-CH2CH3
-(CH2)3CH3
-(CH2)4CH3
-(CH2)5CH3
0.168±0.011
0.534±0.060
19e
-(CH2)6CH3
0.684±0.014
0.676±0.158
0.730±0.067
0.480±0.047
1.20±0.03
0.870±0.042
1.21±0.10
19f
19g
19h
19i
-(CH2)2CH(CH3)2
-(CH2)3CH(CH3)2
-CH2C(CH3)3
0.844±0.126
0.237±0.042
0.287±0.033
0.433±0.077
0.443±0.077
19j
19k

0.469±0.084
0.801±0.144
19l
3.3. The unique mode of binding between 1 and AKR1C3
To understand how the potency is achieved, we determined the structure of compound 1 bound
to AKR1C3 by soaking it into the AKR1C3-NADP+ crystals. Electron density for compound 1 was
clearly observed in the active site region. As shown in figure 1A, compound 1 binds to the oxyanion
site, SP2 and SP3 pocket, and the hydroxyl group forms hydrogen bonds to Tyr55 (2.6 Å) and
His117 (2.7 Å). The methoxy and nitro group also form hydrogen bond to Tyr55 or His117. It is
firstly observed that hydroxyl, methoxy and nitro groups together anchor the inhibitor to the
oxyanion site. However, as seen from the SAR results, compared to methoxy group, it appears that
nitro group plays a more important role in the binding. A comparison of their chemical properties
identified an important difference between them, i.e. only nitro group being one of the strongest
electron-withdrawing groups. Because of this property, O−H is expected to be more acidic and be a
better
H-bond
acceptor,
thus
mimicking
the
carboxylate
group.
The
1,4-dihydropyrano[2,3-c]pyrazole core enters into the SP2 and SP3 pocket, which is defined by
Trp86, Leu122, Ser129, and Phe311, and by NADP+, Tyr24, Glu192, Ser221 and Tyr305,
respectively, and in which lots of hydrogen bond interactions were formed via water molecules. All
the atoms on the core that are capable of forming hydrogen interactions are involved in this hydrogen
bond net with NADP+, Tyr24, Gln192, Tyr216, Ser217, Gln222, Asp224, and Tyr319. It is likely that
hydrogen bond interactions contribute largely to the inhibitory potency of 1.
Figure 1. Crystal structures of 1 bound to AKR1C3 (A) and AKR1C1 (B). Carbon atoms for the

protein and NADP+, DMF, and 1 are colored gray, slate, and green (in AKR1C3)/orange (in
AKR1C1), respectively. Key hydrogen bonds are represented as dotted lines.
The overall structure of AKR1C3-NADP+-1 is similar to the crystal structure of
AKR1C3-NADP+-indomethacin (PDB entry 1S2A) as assessed by inhibitor binding position and the
side chain conformations of mobile residues Phe306, and Phe31130. Both of them induce a rotation of
Phe306, in concert with Phe311, thereby exposing SP3 pocket. However, a more significant rotation
is seen in the mobile residue Trp227 in structure of AKR1C3-NADP+-1. Trp227 undergoes an
approximately 200° rotation from the indomethacin or acetate (PDB entry 1S1P) containing
structures. The broad flips of Phe306, Trp227 and Phe311 dramatically enlarged the binding pocket
and made the active site more accessible. To our knowledge, only bimatoprost (PDB entry 2F38) and
PGD2 (PDB entry 1RY0) with much long side chains induced similar flips in these mobile residues28,
35. Compared to bimatoprost and PGD2, compound 1 is more compact and “closed”. Thus, while
bimatoprost and PGD2 enter deeply into SP1 and SP2 pocket, compound 1 binds mainly in the SP3
pocket, and allowed SP1 and SP2 pockets to be occupied by solvent DMF. According to the above
analysis, we can conclude that the binding pattern of 1 is fairly distinct from those of previously
reported inhibitors and there is enough room for compound 1 to improve the binding affinity by fully
occupying the SP pockets.
3.4 Induced-fit conformational changes of 1 upon binding to AKR1C1
As proposed by Byrns et al., selectivity in AKR1Cs can be gained by taking advantage of the
structural differences in Phe306 15. Because the corresponding residue in AKR1C1 and AKR1C2
(Leu306) has a more rigid side chain that would clash with inhibitors entering the SP3 pocket, such
as selective AKR1C3 inhibitor indomethacin. However, as a SP3 pocket binding inhibitor, compound
1 displays comparable inhibition toward AKR1C1. Crystal structure of compound 1 bound to
AKR1C1-NADP+ was then determined to uncover the potential mechanism. As seen from figure 1B,
compound 1 binds to AKR1C1 in a similar manner to AKR1C3 by interring with residues from
oxyanion site, SP2 and SP3 pocket. However, to avoid the clash with Leu306, compound 1 moves
significantly closer to SP2 pocket. The closest contacts between the attached propyl group and
Phe311 from SP2 pocket were 3.8 Å and 8.0 Å for AKR1C1 and AKR1C3, respectively (Figure 2A),
and, hydrogen bond net in the SP3 pocket is formed directly to the residues of AKR1C1 but via the

water molecules in AKR1C3. In addition, compound 1 was found to undergo a slight rotation. The
bicyclic core displays about 11° torsion difference from the benzene ring plane in the two structures
(Figure 2B).
Figure 2. Superimposition of AKR1C3 and AKR1C1 by residues (A, C) and inhibitors (B, D)
showed that both compound 1 (A, B) and meclofenamic acid (C, D) undergo conformation changes
to suite binding interface. Surrounding residues are represented by gray and slate sticks for AKR1C3
and AKR1C1, respectively. Binding poses are colored in green (in AKR1C3) or orange (in AKR1C1),
while the preferred conformation of compound 1 is rendered as gray ball and stick model.
It is noted that pH-dependent binding modes were observed in indomethacin-AKR1C3
crystals21, 32. As the presented crystal structures were obtained from different pH. The observed
different binding poses of compound 1 are also pH dependent? The pH-dependent enzymatic
inhibition study indicated that compound 1 is more potent at pH 7.5 (IC50 = 0.0869 ± 0.0274), which
may result from the decreased enzymatic activity. However, indomethacin showed lower inhibitory
activity at pH 7.5 (IC50 = 4.30 ± 1.07) than at pH 6.0 (IC50 = 2.12 ± 0.79). It suggested pH likely
would have a great impact on indomethacin binding, but it does not affect compound 1 binding that

much. In addition, molecular modeling predicted a preferred conformation of compound 1 that
differs from the conformations in the two structures. Thus, it is reasonable to assume that compound
1 reshapes itself to suit different binding interfaces. As shown in Figure 2C and 2D, the crystal
structures obtained from similar pH showed that, to avoid the clash with Leu306, meclofenamic acid
(MCF) also adopts a different binding poses when binds to AKR1C1 and AKR1C3 (PDB entry 3R6I),
which clearly explains the lack of isoform specificity reported for this ligand. These findings
suggested that it might also be necessary to consider the conformational changes of the inhibitor in
the rational design of selective AKR1C3 inhibitors, which would make AKR1C3 inhibitors discovery
more difficult. It may be useful in estimating the specificity of the lead and the designs, if the two
binding interfaces were both taken into account. For example, as shown in figure 3,
N-Phenylanthranilic acid based AKR1C3 inhibitors flufenamic acid (FLF) and BT9 showed high
similarities in their chemical structure, AKR1C3 inhibitory activity and binding pattern, which is not
sufficient to explain their big difference in specificity (selectivity ratio of 6 and 378, respectively)24,
29. However, when these two compounds docked into AKR1C1, many different features were
observed. FLF binds to AKR1C1 in a similar manner to MCF, whereas BT9 moves to an outer area
of the protein where little strong interactions can be formed. Thus, FLF gave much better docking
scores than BT9, which explains their big difference in AKR1C1 binding affinity.
Figure 3. Structures of flufenamic acid (FLF) and BT9 bound to AKR1C3 (A) and AKR1C1 (B)
help to explain their big difference in specificity. Carbon atoms for the protein, FLF, BT9 and
meclofenamic acid are represented by gray, cyan, magenta and orange, respectively.

3.5. Energy analysis for the binding features of 1
To comprehensively understand the binding features of 1, the MM-GBSA method was applied
to estimate the binding free energy (ΔGbind) and the contribution of key residues in the binding of 1
with AKR1C3 and AKR1C1. Crystal water molecules discussed above were kept to understand the
role of water molecules in the binding. As shown in Table 5, the ΔGbind values for 1 in the binding of
AKR1C3 and AKR1C1 are -37.02 and -37.08 kcal/mol, respectively. For both AKR1C3-NADP+-1
and AKR1C1-NADP+-1 complexes, electrostatic and van der Waals interactions are the major
favorable contributors for the binding, whereas the polar solvation energies have positive values and
led to an overall positive value of the solvation free energy, indicating it opposes binding.
Furthermore, the electrostatic interactions value is slightly larger than that of the van der Waals
interactions.
Table 5. The binding free energies for AKR1C3 and AKR1C1 complex systems by MM-GBSA method
Energies
ΔEele
ΔEvdW
ΔGpolar
ΔGnonpl
ΔGgas
ΔGsol
ΔGbind
IC50 (μM)
(kcal/mol)
AKR1C3
-45.78±2.75 -38.60±3.90 51.38±2.14 -4.01±0.09 -84.38±4.24 47.36±2.15 -37.02±3.34 0.227±0.013
-47.90±5.02 -44.97±2.69 57.18±3.48 -3.98±0.08 -92.87±5.11 53.19±3.49 -39.68±2.90 0.160±0.028
-45.50±4.72 -47.00±3.03 64.57±2.66 -4.39±0.07 -92.50±4.65 60.17±2.66 -32.33±3.84 0.469±0.084
-16.39±4.48 -51.51±2.67 41.92±3.09 -4.63±0.07 -67.90±5.02 37.29±3.08 -30.61±3.47 0.778±0.044
1
19d
19l
13i
AKR1C1
1
-45.20±4.61 -44.26±3.07 57.17±3.39 -4.05±0.09 -89.47±4.23 52.39±3.02 -37.08±3.88 0.411±0.013
-49.202±4.40 -44.85±2.59 61.22±3.38 -4.12±0.07 -93.88±4.88 56.51±3.01 -36.71±3.04 0.531±0.095
-38.53±3.94 -55.08±3.16 62.10±3.26 -4.60±0.12 -93.61±5.03 57.49±3.23 -36.12±3.16 0.801±0.144
19d
19l
13i
-42.59±6.61 -47.89±3.28 59.99±5.52 -4.46±0.17 -90.47±7.07 55.53±5.52 -34.94±3.40
1.09±0.07
ΔEele/ΔEvdW: Electrostatic/van der Waals contributions;
ΔGpolar/ΔGnonpl: Polar/nonpolar contributions to solvation;
ΔGgas = ΔEele, + ΔEvdW
ΔGsol = ΔGpolar + ΔGnonpolar
ΔGbind = ΔGgas + ΔGsol, the binding free energies without entropic contribution.
;
;
As shown in Figure 4, the residues with the most favorable contributions (lower than −1.0

kcal/mol) are located at the active site. These residues were considered as key residues which
included Tyr24, Tyr55, Gln/His222, Asp/Glu224, Trp227, Phe/Leu306 and NADP+. As expected,
NADP+ in both AKR1C3 and AKR1C1 are suggested to have high affinity with 1 by forming strong
electrostatic interactions. Other residues are mostly those involved in the hydrogen bond interactions
with 1, which imply the importance of the hydrogen bond interactions that facilitate the electrostatic
energy contributions. Notably, the contributions of Trp227 are also relatively higher than for other
residues. The further investigation of the binding patterns shows that Trp227 is involved in π-π
stacking interactions with 1 in both AKR1C3 and AKR1C1. The key residues comparison between
AKR1C3 and AKR1C1 identified some different interacting points. Leu308 is a key residue for the
binding to AKR1C1, whereas water molecules are important for AKR1C3 binding, which agrees
with the results from the crystal structures.
Figure 4. Molecular dynamics simulations results of 1. (A) Cα RMSD plots of AKR1C3-1 (red)
and AKR1C1-1 systems (black). (B) The inhibitor–residue interaction spectrum between 1 and the
important residues for AKR1C3 (red) and AKR1C1 (black).
3.6. Binding features of the reprehensive derivatives 13i, 19d and 19l
To better understand the structural basis for the potency and selectivity of pyranopyrazoles, the
binding modes of three chosen representative derivatives 13i, 19d and 19l were investigated using
molecular docking and molecular dynamics simulations. Coordinates of the target protein were
obtained from the co-crystal structure of AKR1C3/C1-NADP+-1. We firstly re-docked the original
ligand to the crystal structure that generated an RMSD between the docked and crystal poses of
0.2224 and 0.1108 for AKR1C3 and AKR1C1, respectively, which validates the reliability of the

docking procedures.
As shown in Figure 5A, 19d would mimic the binding pattern of compound 1 at the greatest
extent. It thus observed that 19d makes similar interactions with the key residues defined in 3.5
including Tyr24, Tyr55, Gln222, Asp224, Trp227, Phe306 and NADP+. Moreover, interactions
between the pentyl and residues Trp227 and Trp86 could be stronger because of their closer contact,
which would contribute to the higher IC50 value of 19d than that of compound 1. However, 13i and
19l moves away from the crystal position of compound 1 to some extent as shown in Figure S1A,
which would lead to the loss of interactions from some key residues that result in their weaker
potencies. MM-GBSA estimated the ΔGbind values of -30.61, -39.68 and -32.33 kcal/mol for 13i, 19d,
and 19l, respectively, which were consistent with the results from the molecular docking.
Figure 5. Comparisons on the binding patterns of 1 and 19d to AKR1C3 (A) and AKR1C1 (B).
Carbon atoms for the protein, NADP+ and 1 are displayed as gray, while 19d is represented by cyan
(in AKR1C3) or magenta (in AKR1C1) carbon atoms.
19d binding to AKR1C1 were shown in Figure 5B, which illustrated 19d would move slightly
away from the crystal position of compound 1 to avoid the clash with Phe311 from SP2 pocket. This
result is consistent with the slightly decreased potency of 19d towards AKR1C1. It was thus
observed that 19d exhibited higher selectivity than compound 1. Docking results of 13i and 19l
towards AKR1C1 were shown in figure S1B, which showed the binding patterns having fewer
interactions that agree with their weaker potency. Also, the estimated ΔGbind values were also found
to be in agreement with their potencies.

3.7. Molecular basis for the inhibitory potency and selectivity of pyranopyrazoles
Nitro and hydroxyl group on the C4-phenyl ring are required to boost the potency of the
compounds by anchoring the inhibitor to the oxyanion site, whereas substituents at N1- and
C3-position of the core would affect the potency and selectivity. Increasing the length of the alkyl
substituents on C3-position would favor the enhancement of the potency and selectivity by better
occupying the relatively larger SP2 pocket of AKR1C3 but being pushed by Phe311 of AKR1C1.
However, too long or too bulky C3-substituents couldn’t even be accommodated by the SP2 pocket
of the AKR1C3. Thus, 19d and 19e having pentyl or hexyl group at this position displayed both
highest inhibitory activity and selectivity of the set. Furthermore, our results suggested that SP2
pocket could be utilized to just slightly elevate the potency and selectivity of 1. By contrast,
substituents on N1-postion would extend to the entrance of the binding pocket of AKR1C3 and
AKR1C1 that resulted in loss of some interactions. It explains why no apparent improved potency
and selectivity for AKR1C3 was observed for compounds 13j-13k. Crystal and MD studies
suggested that the hydrogen-bond network formed in the SP3 pocket contribute greatly to the potent
inhibitory activity of compound 1, and thus modifications on the related positions would be
unfavorable. Taken together, SP1 pocket would be suggested to be utilized to further improve the
binding affinity and selectivity of the current AKR1C3 inhibitor. This hypothesis supported by the
orientation of the methoxy group of compound 1 to the SP1 pocket, which could thus be replaced to
fully occupy the SP1 pocket. AKR1C3 and AKR1C1 docking studies have already confirmed our
proposal and selective pyranopyrazole AKR1C3 inhibitors are currently being developed by our
group.
4. Conclusions
AKR1C3 plays a critical role in androgen biosynthesis and in the development of CRPC,
making it a promising drug target. Using an AKR1C3 enzymatic assay to evaluate our in-house
chemical library, we identified 6-amino-4-phenyl-1,4-dihydropyrano[2,3-c]- pyrazole-5-carbonitrile
derivative 1 as a potential AKR1C3 inhibitor, which exhibited both high potency and novel chemical
structure from known AKR1C3 inhibitors. Consequently, twenty-three analogues of 1 were
synthesized to explore the structure activity relationships that discovered sixteen compounds with an
IC50 < 1.0 μM, five compounds with an IC50 < 0.25 μM. One of the most promising compound 19d

strongly inhibits the AKR1C3 with an IC50 of 0.160 μM. Crystal structure studies revealed a new
binding mode of compound 1, in which hydroxyl, methoxy and nitro group at the C4-phenyl
substituent together anchor the inhibitor to the oxyanion site, and the core scaffold enlarges but
partially occupies the SP pockets with abundant hydrogen bond interactions. Binding free-energy and
per-residue contribution studies indicate that hydrogen bonds formed in the oxyanion site and SP
pocket can contribute to the high potency of pyranopyrazole analogues towards AKR1C3. In addition,
to bind to the homologous protein AKR1C1, compound 1 was found to reshape itself to fit different
receptors, which elucidate the molecular basis of the selectivity of pyranopyrazole derivatives and
indicates that conformational changes of the inhibitor may also need to be considered in development
of AKR1C3 inhibitors. These observations suggest that 19d is a promising lead against AKR1C3 and
provides a new insight into the molecular basis for future design and optimization of new AKR1C3
inhibitors.
Acknowledgements
This work was supported by the Natural Science Foundation of China (81602968, and
81602955), the Natural Science Foundation of Guangdong Province (2016A030313589, and
2016A030310144), the Medical Scientific Research Foundation of Guangdong Province (A2016201,
and A2016104), Science and Technology Program of Guangzhou, China (201707010049), and
Guangdong Provincial Key Laboratory of Construction Foundation (2017B030314030).
The in-house compound library was provided by Prof. Jun Xu from Research Center for Drug
Discovery at Sun Yat-sen University. Also, Prof. Jun Xu has co-designed experiments, interpreted the
results, reviewed and edited the manuscript, and approved the final manuscript.
References
1. Adeniji, A. O.; Chen, M.; Penning, T. M. J Steroid Biochem Mol Biol 2013, 137, 136.
2. Omlin, A.; Pezaro, C.; Gillessen, S. S. Ther Adv Urol 2014, 6, 3.
3. Chuu, C. P.; Kokontis, J. M.; Hiipakka, R. A.; Fukuchi, J.; Lin, H. P.; Lin, C. Y.; Huo, C.; Su, L.
C. J Biomed Sci 2011, 18, 63.
4. Scher, H. I.; Sawyers, C. L. J Clin Oncol 2005, 23, 8253.
5. Pelletier, G. Best Pract Res Clin Endocrinol Metab 2008, 22, 223.
6. Hofland, J.; van Weerden, W. M.; Dits, N. F.; Steenbergen, J.; van Leenders, G. J.; Jenster, G.;

Schroder, F. H.; de Jong, F. H. Cancer Res 2010, 70, 1256.
7. Liu, C.; Armstrong, C. M.; Lou, W.; Lombard, A.; Evans, C. P.; Gao, A. C. Mol Cancer Ther
2017, 16, 35.
8. Liu, C.; Lou, W.; Zhu, Y.; Yang, J. C.; Nadiminty, N.; Gaikwad, N. W.; Evans, C. P.; Gao, A. C.
Cancer Res 2015, 75, 1413.
9. Mitsiades, N.; Sung, C. C.; Schultz, N.; Danila, D. C.; He, B.; Eedunuri, V. K.; Fleisher, M.;
Sander, C.; Sawyers, C. L.; Scher, H. I. Cancer Res 2012, 72, 6142.
10. Stanbrough, M.; Bubley, G. J.; Ross, K.; Golub, T. R.; Rubin, M. A.; Penning, T. M.; Febbo, P.
G.; Balk, S. P. Cancer Res 2006, 66, 2815.
11. Hamid, A. R.; Pfeiffer, M. J.; Verhaegh, G. W.; Schaafsma, E.; Brandt, A.; Sweep, F. C.;
Sedelaar, J. P.; Schalken, J. A. Mol Med 2013, 18, 1449.
12. Pfeiffer, M. J.; Smit, F. P.; Sedelaar, J. P.; Schalken, J. A. Mol Med 2011, 17, 657.
13. Downs, T. M.; Burton, D. W.; Araiza, F. L.; Hastings, R. H.; Deftos, L. J. Cancer Lett 2011,
306, 52.
14. Yepuru, M.; Wu, Z.; Kulkarni, A.; Yin, F.; Barrett, C. M.; Kim, J.; Steiner, M. S.; Miller, D. D.;
Dalton, J. T.; Narayanan, R. Clin Cancer Res 2013, 19, 5613.
15. Byrns, M. C.; Jin, Y.; Penning, T. M. J Steroid Biochem Mol Biol 2011, 125, 95.
16. Penning, T. M. Expert Opin Ther Pat 2017, 27, 1329.
17. Gobec, S.; Brozic, P.; Rizner, T. L. Bioorg Med Chem Lett 2005, 15, 5170.
18. Adeniji, A. O.; Twenter, B. M.; Byrns, M. C.; Jin, Y.; Winkler, J. D.; Penning, T. M. Bioorg
Med Chem Lett 2011, 21, 1464.
19. Adeniji, A. O.; Twenter, B. M.; Byrns, M. C.; Jin, Y.; Chen, M.; Winkler, J. D.; Penning, T. M.
J Med Chem 2012, 55, 2311.
20. Adeniji, A.; Uddin, M. J.; Zang, T.; Tamae, D.; Wangtrakuldee, P.; Marnett, L. J.; Penning, T.
M. J Med Chem 2016, 59, 7431.
21. Liedtke, A. J.; Adeniji, A. O.; Chen, M.; Byrns, M. C.; Jin, Y.; Christianson, D. W.; Marnett, L.
J.; Penning, T. M. J Med Chem 2013, 56, 2429.
22. Bauman, D. R.; Rudnick, S. I.; Szewczuk, L. M.; Jin, Y.; Gopishetty, S.; Penning, T. M. Mol
Pharmacol 2005, 67, 60.
23. Byrns, M. C.; Steckelbroeck, S.; Penning, T. M. Biochem Pharmacol 2008, 75, 484.

24. Jamieson, S. M.; Brooke, D. G.; Heinrich, D.; Atwell, G. J.; Silva, S.; Hamilton, E. J.; Turnbull,
A. P.; Rigoreau, L. J.; Trivier, E.; Soudy, C.; Samlal, S. S.; Owen, P. J.; Schroeder, E.;
Raynham, T.; Flanagan, J. U.; Denny, W. A. J Med Chem 2012, 55, 7746.
25. Heinrich, D. M.; Flanagan, J. U.; Jamieson, S. M.; Silva, S.; Rigoreau, L. J.; Trivier, E.;
Raynham, T.; Turnbull, A. P.; Denny, W. A. Eur J Med Chem 2013, 62, 738.
26. Flanagan, J. U.; Atwell, G. J.; Heinrich, D. M.; Brooke, D. G.; Silva, S.; Rigoreau, L. J.; Trivier,
E.; Turnbull, A. P.; Raynham, T.; Jamieson, S. M.; Denny, W. A. Bioorg Med Chem 2014, 22,
967.
27. Watanabe, K.; Kakefuda, A.; Yasuda, M.; Enjo, K.; Kikuchi, A.; Furutani, T.; Naritomi, Y.;
Otsuka, Y.; Okada, M.; Ohta, M. Bioorg Med Chem 2013, 21, 5261.
28. Komoto, J.; Yamada, T.; Watanabe, K.; Takusagawa, F. Biochemistry-US 2004, 43, 2188.
29. Chen, M.; Adeniji, A. O.; Twenter, B. M.; Winkler, J. D.; Christianson, D. W.; Penning, T. M.
Bioorg Med Chem Lett 2012, 22, 3492.
30. Lovering, A. L.; Ride, J. P.; Bunce, C. M.; Desmond, J. C.; Cummings, S. M.; White, S. A.
Cancer Res 2004, 64, 1802.
31. Qiu, W.; Zhou, M.; Labrie, F.; Lin, S. X. Mol Endocrinol 2004, 18, 1798.
32. Flanagan, J. U.; Yosaatmadja, Y.; Teague, R. M.; Chai, M. Z.; Turnbull, A. P.; Squire, C. J.
Plos One 2012, 7, e43965.
33. Amano, Y.; Yamaguchi, T.; Niimi, T.; Sakashita, H. Acta Crystallogr D Biol Crystallogr 2015,
71, 918.
34. Zhao, Y.; Zheng, X.; Zhang, H.; Zhai, J.; Zhang, L.; Li, C.; Zeng, K.; Chen, Y.; Li, Q.; Hu, X.
Chem Biol Interact 2015, 240, 310.
35. Komoto, J.; Yamada, T.; Watanabe, K.; Woodward, D. F.; Takusagawa, F. Biochemistry-US
2006, 45, 1987.
36. Lehmann, F.; Holm, M.; Laufer, S. J Comb Chem 2008, 10, 364.
37. Sheng, X.; Hua, K.; Yang, C.; Wang, X.; Ji, H.; Xu, J.; Huang, Z.; Zhang, Y. Bioorg Med Chem
Lett 2015, 25, 3535.
38. Davis, J. M.; Truong, A.; Hamilton, A. D. Org Lett 2005, 7, 5405.
39. Baell, J. B.; Holloway, G. A. J Med Chem 2010, 53, 2719.
40. Adams, P. D.; Afonine, P. V.; Bunkoczi, G.; Chen, V. B.; Davis, I. W.; Echols, N.; Headd, J. J.;

Hung, L. W.; Kapral, G. J.; Grosse-Kunstleve, R. W.; McCoy, A. J.; Moriarty, N. W.; Oeffner,
R.; Read, R. J.; Richardson, D. C.; Richardson, J. S.; Terwilliger, T. C.; Zwart, P. H. Acta
Crystallogr D Biol Crystallogr 2010, 66, 213.
41. Emsley, P.; Cowtan, K. Acta Crystallogr D Biol Crystallogr 2004, 60, 2126.
42. Huang, Y.; Liu, X.; Wu, D.; Tang, G.; Lai, Z.; Zheng, X.; Yin, S.; Luo, H. B. Biochemistry
Pharmacol 2017, 130, 51.
43. Chen, F.; Liu, H.; Sun, H.; Pan, P.; Li, Y.; Li, D.; Hou, T. Phys Chem Chem Phys 2016, 18,
22129.
44. Sun, H.; Li, Y.; Shen, M.; Tian, S.; Xu, L.; Pan, P.; Guan, Y.; Hou, T. Phys Chem Chem Phys
2014, 16, 22035.
45. Sun, H.; Li, Y.; Tian, S.; Xu, L.; Hou, T. Phys Chem Chem Phys 2014, 16, 16719.
46. Xu, L.; Sun, H.; Li, Y.; Wang, J.; Hou, T. J Phys Chem B 2013, 117, 8408.
47. Hou, T.; Wang, J.; Li, Y.; Wang, W. J Chem Inf Model 2011, 51, 69.
48. Li, X.; Guo, H. Chin. J. Org. Chem. 2012, 32, 127.

 Substituted pyranopyrazoles as novel AKR1C3 inhibitors.
 Crystal structure revealed a unique binding pattern.
 Pyranopyrazoles change conformations to suit AKR1C3 and AKR1C1.
 Meclofenamic acid reshapes itself to fit AKR1C3 and AKR1C1.
 Conformational changes of ligand should be considered in design of AKR1C3 inhibitor.