Insight into the binding mode of HIF-2 agonists through molecular dynamic simulations and biological validation

Article abstract of DOI:10.1016/j.ejmech.2020.112999

Hypoxia-inducible factor-2 (HIF-2), a heterodimeric transcriptional protein consisting of HIF-2α and aryl hydrocarbon receptor nuclear translocator (ARNT) subunits, has a broad transcriptional profile that plays a vital role in human oxygen metabolism. M1001, a HIF-2 agonist identified by high-throughput screening (HTS), is capable of altering the conformation of Tyr281 of the HIF-2α PAS-B domain and enhancing the affinity of HIF-2α and ARNT for transcriptional activation. M1002, an analog of M1001, shows improved efficacy than M1001. However, the cocrystal structure of M1001 and HIF-2 has some defects in revealing the agonist binding mode due to the relatively low resolution, while the binding mode of M1002 remained unexplored. To in-depth understand agonist binding profiles, herein, the molecular dynamic (MD) simulations was applied to construct a stable agonist-protein model, and a possible binding mode was proposed through the analysis of the binding free energy and hydrogen bonding of the simulation results. Nine compounds were then synthesized and evaluated to verify the proposed binding mode. Among them, compound 10 manifested improved agonistic activity and reduced toxicity compared to M1002. This study provides deep insight into the binding mode of such HIF-2 agonists, which would be useful for designing novel agonists for HIF-2.

Full text of DOI:10.1016/j.ejmech.2020.112999

European Journal of Medicinal Chemistry xxx (xxxx) xxx
Contents lists available at ScienceDirect
European Journal of Medicinal Chemistry
journal homepage: http://www.elsevier.com/locate/ejmech
Insight into the binding mode of HIF-2 agonists through molecular
dynamic simulations and biological validation
Yancheng Yu 1, Quanwei Yu 1, Simeng Liu, Chenyang Wu, Xiaojin Zhang*
Jiangsu Key Laboratory of Drug Design and Optimization, and Department of Chemistry, China Pharmaceutical University, Nanjing, 211198, China
a r t i c l e i n f o
a b s t r a c t
Article history:
Hypoxia-inducible factor-2 (HIF-2), a heterodimeric transcriptional protein consisting of HIF-2a and aryl
hydrocarbon receptor nuclear translocator (ARNT) subunits, has a broad transcriptional profile that plays
a vital role in human oxygen metabolism. M1001, a HIF-2 agonist identified by high-throughput
Received 27 September 2020
Received in revised form
30 October 2020
Accepted 4 November 2020
Available online xxx
screening (HTS), is capable of altering the conformation of Tyr281 of the HIF-2
a PAS-B domain and
enhancing the affinity of HIF-2 and ARNT for transcriptional activation. M1002, an analog of M1001,
a
shows improved efficacy than M1001. However, the cocrystal structure of M1001 and HIF-2 has some
defects in revealing the agonist binding mode due to the relatively low resolution, while the binding
mode of M1002 remained unexplored. To in-depth understand agonist binding profiles, herein, the
molecular dynamic (MD) simulations was applied to construct a stable agonist-protein model, and a
possible binding mode was proposed through the analysis of the binding free energy and hydrogen
bonding of the simulation results. Nine compounds were then synthesized and evaluated to verify the
proposed binding mode. Among them, compound 10 manifested improved agonistic activity and reduced
toxicity compared to M1002. This study provides deep insight into the binding mode of such HIF-2
agonists, which would be useful for designing novel agonists for HIF-2.
Keywords:
HIF-2
Agonist
Binding mode
Molecular dynamic simulations
© 2020 Elsevier Masson SAS. All rights reserved.
1. Introduction
as ischemia [7] and tumorigenesis [8]. The use of small-molecule
modulators to regulate HIF-2 levels has been proved as an effec-
The hypoxia response signaling axis acts as a critical role in
promoting adaptation to hypoxia by regulating processes such as
oxygen metabolism, oxygen delivery, angiogenesis, and cell sur-
vival [1e3]. Hypoxia-inducible factor-2 (HIF-2), a member of the
tive strategy for the treatment of HIF-2 related diseases, of which
renal anemia [7] and renal cell carcinoma [9] are most studied.
Currently, there are two main types of HIF-2 small-molecule
modulators, prolyl hydroxylase domain (PHD) inhibitors [10] and
HIF-2 allosteric inhibitors [11]. Under normoxic conditions, the
HIF family proteins, is
composed of an oxygen-dependent HIF-2
a
heterodimeric transcription factor
subunit and a stably
a
proline residues, Pro405 and Pro531, of HIF-2
PHD enzymes and, consequently, the hydroxylated HIF-2
a
are hydroxylated by
is
expressed aryl hydrocarbon receptor nuclear translocator (ARNT)
subunit [4,5]. HIF-2 plays a vital role in the hypoxia response
signaling pathway [6], which has a wide range of target genes
involved in many physiological and pathological processes. The
disorder of HIF-2 contributes to corresponding pathogenesis, such
a
recognized for ubiquitination and proteasome degradation. PHD
inhibitors stabilize HIF by inhibiting the PHD-mediated hydroxyl-
ation process and increase the expression levels of HIF downstream
genes, such as erythropoietin (EPO), cytochrome b (Cytb), and the
ferrous ion membrane transport protein divalent metal transporter
1 (DMT1), thus promoting erythropoiesis and iron metabolism and
thereby alleviating renal anemia [10].
Abbreviations: ARNT, aryl hydrocarbon receptor nuclear translocator; Cytb, cy-
tochrome b; DMT1, divalent metal transporter 1; ESP, electrostatic potential; EPO,
erythropoietin; HIF-2, hypoxia-inducible factor-2; HREs, hypoxia response ele-
ments; MD, molecular dynamics; MM-PBSA, molecular mechanics/Poisson-
Boltzmann surface area; RMSD, root mean square deviation; SAR, structure-activ-
ity relationship; PHD, prolyl hydroxylase domain; PTS, protein thermal shift.
* Corresponding author.
HIF-2
transcription of HIF-2 by changing the conformation of the HIF-2
subunit. In 2009, a buried cavity of approximately 290 Å3 was
recognized inside the HIF-2 -PASB domain (Fig. 1B) [12]. HIF-2
a allosteric inhibitors can inhibit the dimerization and
a
a
a
allosteric inhibitors occupy the hydrophobic cavity and push the
side chain of Met252 (an internal residue) to the binding surface of
E-mail address: zxj@cpu.edu.cn (X. Zhang).
These authors contributed equally.
1
https://doi.org/10.1016/j.ejmech.2020.112999
0223-5234/© 2020 Elsevier Masson SAS. All rights reserved.
Please cite this article as: Y. Yu, Q. Yu, S. Liu et al., Insight into the binding mode of HIF-2 agonists through molecular dynamic simulations and
biological validation, European Journal of Medicinal Chemistry, https://doi.org/10.1016/j.ejmech.2020.112999

Y. Yu, Q. Yu, S. Liu et al.
European Journal of Medicinal Chemistry xxx (xxxx) xxx
Fig. 1. (A) Representative structures of HIF-2 modulators. (B) Schematic diagram of the HIF-2-PASB domain (PDB ID:4ZP4) [17]. The cyan cartoon shows the HIF-2
a-PASB domain,
and the salmon cartoon shows the ARNT-PASB domain. The two oval shapes refer to the internal cavity of the HIF-2 -PASB domain and the binding interface of PASB domains of two
a
subunits, respectively. The enlarged image on the right shows the three key hydrogen bonds (green dotted line representation) at the binding interface. (C) Schematic diagram of the
mechanism of HIF-2 allosteric inhibitor (PDB ID: 6E3S) [15]. The inhibitor is represented by a yellow ball and a stick. The cyan and salmon sticks show the residues in the cocrystal
structure of inhibitor PT2385 with HIF-2 (PDB ID: 6E3S) [15]. The purple sticks show Met252, Asp251, Met250 in apo protein (PDB ID: 4ZP4) [17]. PT2385 pushes the side chain of
Met252 to the binding surface to break the hydrogen bond between Asp251 and Asn448. (D) Schematic diagram of the mechanism of the HIF-2 allosteric agonist (PDB ID: 6E3U)
[15]. The agonist is represented by white ball and sticks. The cyan and salmon sticks show the residues in the cocrystal of agonist M1001 and HIF-2 (PDB ID: 6E3U) [15]. The purple
sticks show Tyr281 in apo protein (PDB ID: 4ZP4) [17]. M1001 pushes the side chain of Tyr281 to the binding surface to form a new hydrogen bond with Tyr456. (For interpretation
of the references to color in this figure legend, the reader is referred to the Web version of this article.)
HIF-2
Met252 disrupts a critical hydrogen bond (Fig. 1C, red line repre-
sentation) between Asp251 (HIF-2 -PASB domain) and Asn448
a
-PAS-B and ARNT-PAS-B. Such conformational change of
cell carcinoma [14].
It was worth noting that in 2019, Wu et al. first identified a HIF-
2a allosteric agonist M1001 (3, Fig. 1A) based on an affinity
a
(ARNT-PASB domain), thereby impairing the dimerization and
transcriptional activity of HIF-2 (Fig.1C) [11]. The fastest-promoting
HIF-2a allosteric inhibitor PT2385 (1, Fig. 1A) [11] and MK-6482 (2,
selectionemass spectrometry (ASeMS) screening [10]. The X-ray
crystal structure (PDB ID: 6E3U) of HIF-2 in complex with M1001
has been resolved; the cocrystal structure shows that M1001 also
formerly known as PT2977, Fig. 1A) [13] has shown promising
therapeutical benefits for the treatment of advanced clear cell renal
binds to the large internal cavity in HIF-2
pushes the side chain of Tyr281 to the junction of the PASB domains
a-PASB [15]. M1001
2

Y. Yu, Q. Yu, S. Liu et al.
European Journal of Medicinal Chemistry xxx (xxxx) xxx
of the two subunits, prompting Tyr281 (HIF-2
a
-PASB domain) and
M1001 and M1002 are currently the only two HIF-2 allosteric
agonists, and previous studies have shown that M1002 is signifi-
cantly more potent than M1001 [15]. Thus, M1002 was chosen as
the ligand for subsequent MD simulations in this study. The HIF-2-
PASB domain (242e360) was selected for MD simulation and
extracted from the X-ray crystal structure of M1001 in complex
with HIF-2 (PDB ID: 6E3U) [15]. Thus, M1002 was docked with a
grid-based molecular docking method using the CHARMm force
Tyr456 (ARNT-PASB domain) to form a hydrogen bond (Fig. 1D).
Further, a structure-activity relationship (SAR) approach based on
M1001 led to a more potent analog, M1002 (4, Fig.1A). M1002, with
two trifluoromethyl substituents, manifests better agonistic activity
toward HIF-2a both in the protein-based and cell-based assays.
Quantitative real-time polymerase chain reaction (RT-PCR) and co-
immunoprecipitation (Co-IP) assays have revealed that M1002 can
promote HIF-2 dimerization and thus exert an agonistic effect [15].
field [19]. The M1002-HIF-2a-PASB complex with the highest
The HIF-2
a
allosteric agonists could probably provide a novel
docking score was selected for MD simulation.
strategy for the treatment of renal anemia. However, due to the
relatively low resolution of the cocrystal structure of M1001 in
2.2. MM-PBSA calculated binding free energy of M1002 with HIF-2
complex with HIF-2a, it has some shortcomings in revealing the
binding mode; for instance, the position of structural waters has
not been observed in the X-ray structure [15]. Besides, the binding
mode of M1002 remains unexplored. Given that structural waters
are significant in the binding pattern of HIF-2 inhibitors [11], it may
also contribute to the HIF-2 agonists for the binding to HIF-2a.
However, detailed atomistic insight into the binding area is needed
for designing HIF-2 agonists.
MD simulations provide a very powerful methodology in
analyzing the binding of small-molecules to proteins in atomic
detail. Long MD simulations can provide a detailed binding mode
description and thermodynamic binding properties, such as the
binding free energy. In the past few years, there had been many
excellent works using MD simulations to analyze binding modes
[11,12]. In 2020, research by Anderson’s group utilized MD simu-
lations to confirm that bound waters play important role in stabi-
lizing the protein structure of the HIF-2-PASB domain [16]. In this
work, MD simulations has been carried out to study the agonist-
The model of the M1002-HIF-2a-PASB complex generated from
docking was emerged in a box of waters and the solvated ligand-
protein complex was used as a starting structure for a 500 ns MD
simulation employing the amber99sb force filed [20]. The root
mean square deviation (RMSD) of M1002 and protein relative to the
initial structure was calculated to evaluate the binding stability,
respectively (Fig. 2A). As is observed in Fig. 2A, the protein complex
reached equilibrium after 120 ns, and the RMSD converged to ca.
2 Å. The RMSD of M1002 converges to 2.5 Å in about 120 ns.
Subsequently, g_mmpbsa, a gromacs tool for MM-PBSA calcula-
tions, was utilized to calculate the binding free energy of M1002
and protein using the molecular mechanics/Poisson-Boltzmann
surface area (MM-PBSA) method [21]. The binding free energy
calculated by the MM-PBSA method includes the following four
items: The van der Waals contribution, and the electrostatic energy
calculated with amber99sb force field, and the non-polar contri-
bution in free energy of solvation calculated by empirical model,
and the electrostatic contribution in the free energy of solvation
calculated by Poisson-Boltzmann Equation. We sampled 500
snapshots every nanosecond to calculate the binding free energy
and plotted the RMSD value and binding free energy of M1002
(Fig. 2B). As the conformation of M1002 gradually converged, the
binding free energy gradually decreased from the initial ꢀ90 kJ/mol
(Fig. 2B, RMSD ¼ 0, blue dot), which demonstrated that a more
stable conformation than the initial one was finally obtained. The
final 300 ns binding free energy data were extracted and averaged
(Table 1). The average MM-PBSA binding free energy of M1002 with
protein is ꢀ130.47 12.09 kJ/mol. The binding of M1002 to protein
is dominated by van der Waals force, which is similar to most ligand
binding (Table 1).
HIF-2a-PASB complex. Our computational studies were constituted
by docking studies followed by long-term all-atom MD simulations
and in-depth structural and energetic analysis of the agonist in
complex with a PASB domain model of HIF-2a. The binding energy
calculation, energy decomposition analysis, and hydrogen bond
analysis of the simulation results provide details of the binding
profile of the agonist in complex with HIF-2a-PASB domain. Sub-
sequently, nine compounds were synthesized and biologically
evaluated for their binding affinity and agonist activity against HIF-
2 using the protein thermal shift (PTS) assay and the luciferase
reporter gene assay, respectively. The combination of MD simula-
tion results and experimental data provides a clear binding pattern
of the HIF-2 agonists in complex with HIF-2
suggest a binding mode that provides a basis for the design of HIF-
allosteric agonists.
a-PASB domain. We
Key protein residues for binding with M1002. To understand
the key residues that M1002 binds to protein binding area, the MM-
PBSA binding free energy was decomposed for each protein
residue-ligand interaction (Fig. 3). The results indicated that
Met289 and Tyr307 are the two residues that contributed most to
2
a
2. Results and discussion
2.1. MD simulation results
binding, with binding energy contributions ꢀ9.33
0.12 kJ/mol
and ꢀ7.92
(ꢀ6.07
0.14 kJ/mol, respectively, followed by His248
HIF-2 has an uncommon large hydrophobic cavity inside the
PASB domain. According to the X-ray structure of HIF-2-PASB
domain (PDB ID: 3F1P), there are eight structural water mole-
0.11 kJ/mol) and then Met252, Phe254, and Cys339
(Fig. 3A and Fig. 3B).
Analysis of binding interactions. To investigate how M1002
binds to these six key residues, we first analyzed the hydrogen
bonds between M1002 and protein. The results (Table 2) show that
the oxygen atom of M1002 forms a hydrogen bond with Tyr307 for
76.47% of the 500 ns MD simulation trajectory. It is worth noting
that a stable bound water molecule was observed that forms a
hydrogen bond with the N1 atom of M1002 with 50.98% frequency.
Furthermore, this bound water forms a hydrogen bond with
interaction frequency 115.69% with His248 and a hydrogen bond
with frequency 23.53% with Cys339. This hydrogen bond network
formed between M1002, bound water, His248, and Cys339 should
stabilize considerably the binding of the ligand (See a video
generated with Pymol v.2.3.0 demonstrating the hydrogen bonds in
cules inside the HIF2a-PASB domain, which form a hydrogen bond
network with the residues inside the cavity, including Tyr281
[12,18]. It has been suggested computationally [18] and also
revealed from the X-ray structure of M1001 with HIF-2 (PDB: 6E3U)
[15] that when a small-molecule binds inside this cavity, the bound
water molecules are driven out of the cavity. Consequently, the side
chain of Tyr281 can move freely again. If the small-molecule can fit
the side chain of Tyr281 to the HIF-2a-PASB and ARNT- PASB
binding junction, this small-molecule exhibits agonistic activity
[15]. To explore the binding interactions of agonists, firstly, we
conducted an MD simulation to study the agonist in complex with
HIF-2a-PASB domain interaction at the atomic level.
3

Y. Yu, Q. Yu, S. Liu et al.
European Journal of Medicinal Chemistry xxx (xxxx) xxx
Fig. 2. (A) RMSD of M1002 (blue) and HIF-2
a
-PASB (red). (B) From 0 ns to 500 ns, the RMSD of M1002 converges to 2.5 Å after 120 ns, and the binding energy decreases
from ꢀ90 kJ/mol at the beginning of the simulations to an average of ꢀ130 kJ/mol at the end of the simulations. (For interpretation of the references to color in this figure legend,
the reader is referred to the Web version of this article.)
Table 1
MM-PBSA binding energy of M1002 to HIF-2a-PASB.
Complex
D
VDWa
D
Elecb
D
Apolc
D
Pold
152.53 8.18
D
Ee
ꢀ130.47 12.09
M1002-PASB
ꢀ214.12 11.78
ꢀ51.09 8.36
ꢀ17.78 0.67
a
D
D
D
D
D
VDW (kJ/mol): change of van der Waals energy in the gas phase upon complex formation.
Elec (kJ/mol): change of electrostatic interactions upon complex formation.
Apol (kJ/mol): change of nonpolar solvation energy upon complex formation.
Pol (kJ/mol): change of polar solvation energy upon complex formation.
E (kJ/mol): MM-PBSA binding energy.
b
c
d
e
Fig. 3. (A) MM-PBSA binding free energy decomposition for the interaction of each protein residue with the ligand. (B) Histogram showing the contribution of the six residues
contributing most to the binding free energy of the M1002-protein complex.
Table 2
Hydrogen bonds analysis of M1002 in complex with HIF-2a
-PASBa
.
Compound
Donor
Acceptor
Occupancy
Tyr307-OH
M1002-N1
Water-O, His248-CE1
Water-O
M1002-O2
Water eO
His248-ND1, Water-O
Cys339-SG
76.47%
50.98%
115.69%
23.53%
a
Hydrogen bonds analysis angel cutoff (degrees) ¼ 30 and donor-acceptor distance cutoff (Å) ¼ 3.5.
4

Y. Yu, Q. Yu, S. Liu et al.
European Journal of Medicinal Chemistry xxx (xxxx) xxx
Supplementary data).
supposed to be the local preferential structure.
Supplementary video related to this article can be found at
http://doi:10.1016/j.ejmech.2020.112999
2.4. Design and synthesis of analogs of M1002 based on the
The snapshot in Fig. 4 shows M1002 forming hydrogen bonds
with Tyr307 and the bound water, which is consistent with the
previous hydrogen bond analysis results. The side chain of Phe254
suggested binding mode
Design of the structure of the new analogs of M1002. Nine
compounds (Fig. 6) were designed and synthesized based on the
MD simulation results for validation of the suggested binding
mode. Our hypothesis used for the design divides the reference
structure of M1002 into three parts: A-ring, atom linker, and B-ring
(Fig. 6). Regarding the B-ring, we speculated that an electron-
and the benzene ring of M1002 have a
p-p stacking interaction
(Fig. 4) [22]. Besides, the interaction between the aromatic ring and
the sulfur atom also plays a role in stabilizing the protein structure
[23,24]. The benzene ring of M1002 which is substituted by two
trifluoromethyl groups may be partially positively charged, and the
lone pairs on the sulfur atoms of residues Met252 and Met289 may
deficient B-ring may increase n/p* interactions with the
form an n/p* interaction with benzene ring [25]. Overall, we
calculate that six residues participating in binding through
hydrogen bonds and hydrophobic interactions.
electron-rich sulfur atoms of Met289 and M252. Thus, we designed
compounds 5e10 by introducing substituents having different
electronic effects for investigating the interaction between the B-
ring and two methionines. Regarding the atom linker, the MD
simulation results suggest that it forms a hydrogen bond network
through a bound water with His248 and Cys339. We methylated
the NH group of M1002 to obtain 12 for examining the effect of this
possible hydrogen bond to binding. We replaced the NH moiety
with an oxygen atom to 11 to investigate whether the linked moiety
should present as a hydrogen bond donor. To investigate SAR for A
ring, we replaced the sulfonamide with an amide moiety to 13 for
exploring if this change should reduce activity through weakening
hydrogen bond with Tyr307.
Synthesis of the designed analogs of M1002. The synthesis of
the target compounds is shown in Scheme 1. Saccharin (1a) was
selected as the raw material for the synthesis of 4e12. 1a was
chlorinated with phosphorus pentachloride to give 2a. Then 2a was
condensed with the corresponding amines and 3,5-
bis(trifluoromethyl) phenol to give 4e11 [26]. 12 was obtained by
methylation of 4 with iodomethane. 13 was synthesized through a
ring-forming condensation reaction, followed by a substitution
reaction [27,28].
2.3. Comparison of the binding modes of M1001 and M1002
In the previous report, M1002 has been proved to exhibit better
HIF-2 binding affinity and agonist activity than M1001 [15]. We
tried to utilize the predicted binding mode of M1002 to explain why
M1002 showed improved binding affinity than M1001. The coc-
rystal structure of M1001 and HIF-2 (6E3U) was superimposed with
the representative binding pose of M1002 extracted from conver-
gent trajectory (Fig. 5) using the protocol "superimpose proteins" in
Discovery Studio software (version 2020).
From the results of superposition, overall, M1002 has a binding
mode similar to M1001 (Fig. 5, left). In detail, due to the steric
hindrance of Tyr278 at the edge of the cavity, the relative position
of M1002 in the cavity has changed compared to M1001. The tri-
fluoromethyl group of M1002 occupies the position of the benzene
ring of M1001, causing the overall position of M1002 to move up-
ward (Fig. 5, right). As a result, the oxygen atom of the sulfone
moiety of M1002 (3.0 Å), which plays an key role in binding in our
prediction, is closer to Tyr307 than that of M1001 (4.2 Å). The close
distance between the sulfone moiety and Tyr307 makes M1002
more capable of forming hydrogen bond with Tyr307, which may
be the reason why M1002 has better affinity and agonistic activity
than M1001. In summary, the 3,5-disubstituted benzene ring is
2.5. Biological assays
Two methods were employed to evaluate the activity of the
target compounds: 1. the protein thermal shift (PTS) assay [29] was
Fig. 4. Snapshot of M1002-protein complex from 500 ns-MD simulation using amber99sb force field. The green dotted line in the figure represents the hydrogen bond interaction,
the yellow dotted line represents the n/ * interaction, and the purple dotted line represents the interaction. (For interpretation of the references to color in this figure legend,
the reader is referred to the Web version of this article.)
p
p-p
5

Y. Yu, Q. Yu, S. Liu et al.
European Journal of Medicinal Chemistry xxx (xxxx) xxx
Fig. 5. Comparison of the binding modes of M1001 and M1002. M1001 is a cyan ball-and-stick representation, and its nearby residues are displayed as cyan sticks. M1002 and the
corresponding residues are grey. The two red dotted lines indicate the distance between the center of the benzene ring of Tyr278 and the trifluoromethyl group of M1002 (3.6 Å)
and the distance between the center of the benzene ring of Tyr278 and the benzene ring of M1001 (3.5 Å), respectively. The two green dashed lines indicate the distance between
the hydroxyl group of Tyr307 and the oxygen atom in the sulfone moiety of M1002 (3.0 Å) and M1001 (4.2 Å), respectively. (For interpretation of the references to color in this figure
legend, the reader is referred to the Web version of this article.)
Fig. 6. Design of analogs of M1002 as probes to explore its binding profile with protein.
selected to evaluate the binding affinity of the target compounds
with HIF-2; 2. The luciferase reporter gene assay was employed to
evaluate the compounds’ agonistic activity on HIF-2.
increase in thermal shift compared to 8 (
D
Tm ¼ 2.05 ꢁC) and 9
(D
Tm ¼ 2.08 ꢁC).
In an attempt to suggest an explanation for the higher binding
Protein thermal shift assay. The PTS data of the target com-
pounds are shown in Fig. 7. We first analyzed compounds 5e10
with various substituents on the B-ring. Compounds 6 and 7 with
methyl and methoxy substituents, respectively, should have
increased the electron cloud density on the B-ring; their thermal
strength of 10 compared to 8 and 9, we docked 10 into the binding
pocket of HIF-2. The best-scored docking pose was chosen as the
initial pose to conduct a 50 ns MD simulation. The ligand and
protein RMSD (compared with their corresponding starting poses)
were calculated to estimate the binding stability (Fig. 8A). The
RMSD calculations showed that during the MD simulation, the
conformation of the 10-protein complex changed very slightly and
stabilized quickly (Fig. 8A). In an attempt to compare binding
modes of 10 and M1002, the stable pose of the 10-protein complex
was extracted to align with the previous predicted M1002-protein
structure. It was found that 10 and M1002 have similar binding
modes to protein (Fig. 8B). The simulation results also revealed that
the bromine atom of 10 may have halogen bond interaction with
shift (
D
Tm ¼ 1.93 ꢁC for 6, and
D
Tm ¼ 1.89 ꢁC for 7) was slightly
weaker than that of 5 (
D
Tm ¼ 2.0 ꢁC) substituted by hydrogen
atoms. The trifluoromethyl-substituted M1002 and the halogen-
substituted compounds 8e10 with electron-deficient B-ring,
generally, are more potent than 5e7. This result is consistent with
the previous speculation that the sulfur atom on methionine is
more likely to interact with the electron-deficient aromatic ring.
Additionally, compound 10 (
D
Tm ¼ 2.52 ꢁC) has a significant
6

Y. Yu, Q. Yu, S. Liu et al.
European Journal of Medicinal Chemistry xxx (xxxx) xxx
Scheme 1. Synthetic routes for 4e13. Reagents and conditions: (i) PCl5, 160 ꢁC, 4 h, 63%; (ii) corresponding amines, pyridine, reflux, 2e5 h, 43%e85%; (iii) pyridine, reflux, 5 h, 40%
(iv) iodomethane, DMF, K2CO3, rt, overnight, 54% (v) N,N-diethylhydroxylamine, chloroform, reflux, 3 h, 55%; (vi) acetic acid, ethanol, reflux, 5 h, 34%.
the hydroxyl oxygen atom and the benzene ring of Tyr278 (Fig. 8C).
The halogen bond between halogen and the benzene ring of tyro-
sine is the most common form of halogen bond found in PDB
database statistics [30]. Statistics show that the distance of the
halogen bond formed by bromine is the sum of van der Waals radii
of bromine and its partner (the sum of van der Waals radii of the
bromine atom and the oxygen atom is approximately 3.5 Å), and
the CeBr$$$Donor bond angle is approximately 155ꢁ [30,31]. In our
simulation results, the bromine atom of 10 can form a halogen bond
with Tyr278 in most frames of simulation. The representative pose
of the 10-protein complex was extracted from the trajectory, and
the distances and angles between the bromine atom and Tyr278
were measured. In the conformation we extracted, the distances of
bromine from the center of the benzene ring and the oxygen atom
both are 3.6 Å, and the bond angles are 164.4ꢁ and 147.7ꢁ, respec-
tively. Additionally, we calculated the electrostatic potential (ESP)
surfaces [32,33] for the docking pose between 10 and Tyr278. The
results show that the positively charged region on the surface of the
bromine atom can interact with the negatively charged region of
Tyr278 (Fig. 8D). These geometric features and ESP surfaces indi-
surfaces of 8, 9, and 10 are calculated and displayed in the sup-
plementary data, Fig. S1). These results may provide an explanation
why compound 10 in which the benzene ring is substituted with Br
is stronger binder compared to 8 and 9, with Cl or F substituents, in
the PTS assay (Fig. 7).
The PTS data of 12 show that removing the hydrogen of the NH
linker is detrimental for its binding to HIF-2, suggesting that the
hydrogen bond network that formed by the NH linker and bound
water and residues is essential for binding. As in 11, replacing the
nitrogen atom with an oxygen atom also leads to significant loss of
activity, suggesting that the linker between the A and B ring should
be a hydrogen bond donor in the formation of a hydrogen bond
with bound water. As in 13, replacing the sulfonamide structure
with an amide also results in a significant loss of activity; it in-
dicates that the geometry of the sulfonamide structure is indis-
pensable in the formation of hydrogen bonds with Tyr307.
Compared with the tetrahedral sulfonamide, the amide is planar
and the orientation of the hydrogen bond acceptor changes, thus
resulting in the inability for 13 to form hydrogen bond interactions
with Tyr307 with a subsequent loss of activity.
cate that a slightly positive region, known as
be generated on the bromine atom, may indeed be capable of
forming a halogen bond with Tyr278 and enhance the binding af-
s
-hole [34] which can
Luciferase reporter gene assay. 786-O-HRE-Luc single clone
cells were used to verify the results of the PTS assay. We evaluated
the agonistic effect of 5e10, and M1002 was selected as the control
finity of 10. In the case of chlorine, as in 9, the
s
-hole is much less
compound. The 786-O-HRE-Luc cells were treated with 2
mM and
positive. For fluorine, as in 8, no positive region is observed (the ESP
20 M compounds, respectively, for 24 h and then tested. The
m
7

Y. Yu, Q. Yu, S. Liu et al.
European Journal of Medicinal Chemistry xxx (xxxx) xxx
Fig. 7. Activity for M1002 derivatives tested by PTS assay.
Fig. 8. The docking pose of 10 and potential halogen binding interaction. (A) The ligand and protein RMSD. (B) The superimposition of the 10-protein complex and the M1002-
protein complex. The 10-protein complex is green and the M1002-protein complex is white. (C) Potential halogen bond interaction between compound 10 and Tyr278. The white
ball-and-stick formula represents 10, and the cyan one is Tyr278. The distances and angles of the possible halogen bonds between them are shown in orange. (D) The electrostatic
potential (ESP) Surfaces of 10 and Tyr278. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
8

Y. Yu, Q. Yu, S. Liu et al.
European Journal of Medicinal Chemistry xxx (xxxx) xxx
results indicate that 5e10 can activate HIF-2 in a concentration-
dependent manner. Compounds 8e10 with electron-withdrawing
halogen-substituted B-ring exhibit improved efficacy than 5e7,
which is consistent with our previous simulation and PTS assay
results. The only exception is M1002. In the PTS assays, 10 showed
better binding affinity than M1002, but in the luciferase reporter
bromo group may form a halogen bond and contribute significantly
to the binding with HIF-2.
3. Conclusions
Taken together, we performed MD simulations on the HIF-2
agonist M1002 in complex with HIF-2a-PASB domain. Through
gene assays, 2
also worth noting that 20
M M1002, probably due to the cytotoxicity of M1002 (Fig. 9). To
mM M1002 showed better efficacy than 2
mM 10. It is
mM M1002 seems to be less effective than
the calculation of binding free energy and residue energy decom-
position, six important amino acid residues that contribute the
most to binding were identified. Through hydrogen bond analysis,
we suggested a binding mode for these six residues and M1002.
Subsequently, we synthesized nine M1002 derivatives and evalu-
ated their activity using protein and cell assay to verify the sug-
gested binding mode. Indeed, compound 10 with bromo-
substituted B-ring exhibited better agonistic activity and lower
toxicity than M1002. The docking results and the ESP calculation
results indicate that 10 may form a halogen bond between the
bromine atom and Tyr278 of the cavity, thereby enhancing the
binding ability. The suggested binding mode of the tested agonists
with HIF-2 may provide the basis for future research on more
effective HIF-2 agonists.
2
m
explain these two issues, multiple concentrations of 10 and M1002
were employed in the luciferase reporter gene assays and cell
proliferation assays (Fig. 10).
The results of cell proliferation assays showed that 10 is non-
cytotoxic against 786-O-Luc cells at a concentration of 20
while M1002 significantly inhibits the proliferation of 786-O-Luc
cells even at 1 M (Fig. 10B). The luciferase reporter gene assays
indicated that 10 increases the fluorescence of 786-O-HRE-Luc cells
by 1.63 times at 20 M with an EC50 of 1.68 0.62 M. The EC50 of
M1002 is 0.44 0.32 M M1002 at 2.2 M can increase the fluo-
rescence by 1.41 times, while the fluorescence values of the cell
treated with 6.6 M and 20 M M1002 decreased significantly
mM,
m
m
m
m
m
m
m
(Fig.10A). The abnormal results of M1002 at a high concentration in
the luciferase reporter gene assays may be due to its cytotoxicity,
which causes a decrease in cell numbers. M1002 exhibits a lower
EC50 value than 10, which is contrary to the binding data in the PTS
assays, probably because the trifluoromethyl substituents in M1002
is beneficial for the cell membrane permeability [35]. Nevertheless,
due to the cytotoxicity of M1002, 10 still show higher agonistic
efficacy at high concentrations. Thus, the cell-based assays could
also validate our hypothesis on the agonist binding mode.
Based on the MD simulation results and activity data, we pro-
posed a 2D binding mode of the benzisothiazole compound and
HIF-2. As shown in Fig. 11, the oxygen atom on the sulfonamide
structure of the A-ring can form a hydrogen bond with Tyr307. The
replacement of sulfonamides to amides results in loss of activity.
4. Experimental
4.1. Molecular docking
The structure of the HIF-2-PASB domain was obtained from the
PDB database website (PDB ID: 6E3U) and then extracted the
242e360 residues of HIF-2
were prepared by the software Discovery Studio (2020). Residues
around 10 Å of the native ligand in the HIF-2 -PASB domain are
a from it. Both protein and compounds
a
defined as binding sites for docking. The CDOCKER protocol of
discovery studio (2020) was employed for docking. All settings in
the docking were by following per under the default. The confor-
mation with the highest CDOCKER score was used for subsequent
dynamic simulation or analysis.
Furthermore, the A-ring may form a
p-p stack interaction with
Phe254. The NH linker can form a hydrogen bond network by
bound water with nearby residues His248 and Cys339. The
methylation of NH or replacing the nitrogen with oxygen leads to a
significant loss of activity, indicating that a hydrogen bond donor is
required at this position for HIF-2 binding. Regarding the B-ring,
the electron-rich sulfur atoms on Met289 and Met252 can form n
4.2. Molecular dynamics simulation
GROMACS package (version 2019.03) [36] was applied to run
conventional MD simulations to investigate the binding mode of
M1002. The force field amber99sb.ff [20] was employed to
parameterize protein. The TIP3P [37] was used for the waters.
M1002 was parameterized using the AnteChamber Python Parser
Interface (ACPYPE) with amber99sb force field. This M1002-protein
complex was solvated in an octahedral water box, and then the
charge of the system was neutralized by adding 0.150 M chloride
and sodium ions. The system was assembled, containing 11,515
water molecules, 37 sodium ions, and 34 chloride ions. First, the
steepest descent minimization method was used to minimize the
energy of the system by 50,000 steps. In the next step, we restricted
the position of heavy atoms to run both NVT equilibration and NPT
equilibration by 50,000 steps. The system temperature was main-
tained at 300 K, and the system pressure was maintained at 1 bar.
Upon completion of the two equilibration phases, the system is
now well-equilibrated at the desired temperature and pressure. a
500-ns unrestrained simulation was carried out. Every 10 ps, the
energy and coordinate system of the trajectory were saved. The
hydrogen bond analysis plug-in of Visual Molecular Dynamics
(VMD) was applied to analyze the trajectory file [38]. In the simu-
lation trajectory, the bond length changes of the hydrogen bonds
formed by M1002 and the corresponding residues were made into
a video using open-source Pymol (version 2.3.0) [39].
/ p* interactions with the benzene ring of the B-ring. In addition,
an electron-donating substituent on the benzene ring, such as CF3,
is superior for binding compared to an electron-withdrawing one; a
Fig. 9. The agonistic activity of 5e10 evaluated by the HRE luciferase reporter assays.
Error bars, mean s.d.; n ¼ 3 (distinct replicates for cell cultures).
9

Y. Yu, Q. Yu, S. Liu et al.
European Journal of Medicinal Chemistry xxx (xxxx) xxx
Fig. 10. Biological effect and cytotoxicity of 10 and M1002. (A) Results of the luciferase reporter gene for 10 and M1002. (B) Cell proliferation assay results of 10 and M1002 against
the 786-O-HRE-Luc cells. Error bars, mean s.d.; n ¼ 3 (distinct replicates for cell cultures).
4.3. Free energy calculations and residue decomposition employing
the MM-PBSA method
MM-PBSA method has been widely adopted in the estimation of
binding free energy in drug research. In our work, the MM-PBSA
calculation was performed using the g_mmpbsa, a tool of GRO-
MACS for MM-PBSA calculations. To understand the binding of
protein and ligand at the molecular level, we used g_mmpbsa to
decompose the free energy of binding to the contribution of each
residue to the free energy of binding. We sampled the trajectory
every nanosecond to get 500 snapshots. The energy components
EMM, Gpolar, and Gnonpolar of the protein-M1002 complex was
calculated for 500 snapshots.
4.4. Chemical synthesis
All chemical reagents and solvents are purchased through
commercial sources and used as received. Reactions were moni-
tored using TLC silica gel plates (GF254, 0.25 mm) and visualized
under UV light. Hanon MP420 automatic melting point apparatus
was used to determine the melting points. The proton nuclear
magnetic resonance (1H NMR) and carbon nuclear magnetic reso-
nance (13C NMR) spectra were determined using Bruker AV-300
instruments. EI-mass spectra and high-resolution mass spectra
(HRMS) were recorded on a Shimadzu GCMS-2010 mass spec-
trometer and a Water Q-Tof micro mass spectrometer, respectively.
Fig. 11. Schematic diagram of HIF-2 agonist binding mode.
10

Y. Yu, Q. Yu, S. Liu et al.
European Journal of Medicinal Chemistry xxx (xxxx) xxx
The purity (ꢂ95%) of the target compounds used for biological
testing was evaluated by high-performance liquid chromatography
(HPLC) analysis, performed on an Amethyst C18e P (4.6 ꢃ 150 mm,
4.4.4. 3-((3,5-Dimethoxyphenyl)amino)benzo[d]isothiazole 1,1-
dioxide (7)
White solid. Yield 81.8%. mp 219.0e219.6 ꢁC. 1H NMR (400 MHz,
5
m
M, Merck) column eluting with methanol/water (80:20 v:v) at a
DMSO‑d6)
7.96e7.85 (m, 2H), 7.12 (d, J ¼ 2.2 Hz, 2H), 6.44 (t, J ¼ 2.2 Hz, 1H),
3.79 (s, 6H). 13C NMR (75 MHz, DMSO‑d6)
161.03, 157.14, 141.02,
d
10.72 (s, 1H), 8.48 (d, J ¼ 7.6 Hz, 1H), 8.12e8.06 (m, 1H),
flow rate of 0.5 mL/min; the peaks were detected at 254 nm under
UV. The data collection of a single crystal of 12 was performed on a
Bruker Smart Aapex2 CCD and Rigaku Saturn CCD diffractometer
d
139.67, 134.20, 133.80, 128.81, 124.03, 122.00, 100.91, 97.76, 55.84,
40.77, 40.51, 40.24, 39.96, 39.68, 39.40, 39.12. HRMS (ESI): calcd. for
equipped with graphite monochromated Mo-K
a
radiation
(
l
¼ 0.7107 Å) at room temperature. The structure was solved with
C
15H14N2O4S [M þ H]þ ¼ 319.0747, found 319.0751. tR ¼ 3.378 min,
direct methods and refined using OLEX2 and SHELXL-97 programs.
The non-hydrogen atoms were located in the successive difference
Fourier synthesis. The final refinement was performed by full-
matrix least-squares methods with anisotropic thermal parame-
ters for no-hydrogen atoms on F2. The hydrogen atoms were added
theoretically and riding on the concerning atoms. The parameters
used intensity collection and refinements are summarized in
Table S1; the selected bond lengths and angles of 12 are summa-
rized in Table S2 and Table S3, respectively.
HPLC purity: 98.56%.
4.4.5. 3-((3,5-Difluorophenyl)amino)benzo[d]isothiazole 1,1-
dioxide (8)
White solid. Yield 52.6%. mp > 300 ꢁC. 1H NMR (300 MHz,
DMSO‑d6)
d 11.05 (s, 1H), 8.51e8.37 (m, 1H), 8.15e8.06 (m, 1H),
7.99e7.86 (m, 2H), 7.71e7.56 (m, 2H), 7.26e7.11 (m, 1H). HRMS
(ESI): calcd. for C13H8F2N2O2S [M þ H]þ ¼ 295.0347, found
295.0349. tR ¼ 3.720 min, HPLC purity: 99.63%.
General Procedure for the Preparation of Compounds M1002(4)
ꢀ11. Saccharin (1a, 15 g, 0.082 mol) was added to a round bottom
flask, and then phosphorus pentachloride (50 g, 0.163 mol) was
added. The reaction was heated to 160 ꢁC under a nitrogen atmo-
sphere and stirred for 4 h. The reaction mixture was cooled to room
temperature and then slowly added into ice water (500 mL). A
yellow solid precipitated, and the suspension was filtered. The filter
cake was dried to give the crude product 2a as a yellow solid, which
was directly used in the next reaction without further purification.
Compound 2a was added to a round bottom flask, and the corre-
sponding aniline or phenol (the equivalent ratio of 2a to aniline or
phenol: 1 to 1) was added. Dry pyridine (5 mL) was then added to
the mixture. The reaction mixture was heated to reflux and stirred
for 2e5 h. After the reaction was complete, the solution was cooled
to room temperature and poured into water (25 mL). A precipitate
precipitated, and the suspension was filtered. The resultant filter
cake was purified by chromatography column (hexane: ethyl
acetate ¼ 10: 1) to afford the target compounds 4e11.
4.4.6. 3-((3,5-Dichlorophenyl)amino)benzo[d]isothiazole 1,1-
dioxide (9)
White solid. Yield 43.1%. mp > 300 ꢁC. 1H NMR (300 MHz,
DMSO‑d6)
1H), 8.02e7.86 (m, 4H), 7.54 (t, J ¼ 1.8 Hz,1H). HRMS (ESI): calcd. for
13H8Cl2N2O2S [M 326.9756, found 326.9757.
H]þ
tR ¼ 5.143 min, HPLC purity: 97.80%.
d
11.03 (s, 1H), 8.49e8.40 (m, 1H), 8.12 (dd, J ¼ 6.3, 1.9 Hz,
C
þ
¼
4.4.7. 3-((3,5-Dibromophenyl)amino)benzo[d]isothiazole 1,1-
dioxide (10)
White solid. Yield 60.3%. mp > 300 ꢁC. 1H NMR (300 MHz,
DMSO‑d6)
d
10.97 (s, 1H), 8.42 (d, J ¼ 7.2 Hz, 1H), 8.13 (d, J ¼ 13.7 Hz,
3H), 7.95 (q, J ¼ 6.6 Hz, 2H), 7.75 (s, 1H). 13C NMR (75 MHz,
DMSO‑d6)
d 157.54, 140.78, 140.66, 134.54, 134.09, 130.51, 128.32,
124.16, 123.70, 122.91, 122.19, 40.81, 40.53, 40.25, 39.98, 39.70,
39.42, 39.14. HRMS (ESI): calcd. for
C
13H8Br2N2O2S [M
þ
H]þ ¼ 414.8746, found 416.8725. tR ¼ 5.804 min, HPLC purity:
99.52%.
4.4.1. 3-((3,5-Bis(trifluoromethyl)phenyl)amino)benzo[d]
isothiazole 1,1-dioxide (M1002, 4)
4.4.8. 3-(3,5-Bis(trifluoromethyl)phenoxy)benzo[d]isothiazole 1,1-
dioxide (11)
White solid. Yield 83.2%. mp 279.2e279.4 ꢁC. 1H NMR (300 MHz,
White solid. Yield 40.2%. mp 92.2e92.4 ꢁC. 1H NMR (300 MHz,
DMSO‑d6)
d
11.26 (s, 1H), 8.58 (s, 2H), 8.42 (d, J ¼ 7.5 Hz, 1H),
DMSO‑d6)
8.17e8.13 (m, 1H), 8.08 (td, J ¼ 7.6, 1.5 Hz, 1H), 8.04e7.98 (m, 1H),
7.92e7.89 (m, 2H). 13C NMR (75 MHz, DMSO‑d6)
149.23, 137.17,
d
8.33e8.29 (m, 1H), 8.24 (dq, J ¼ 1.6, 0.8 Hz, 1H),
8.21e8.10 (m, 1H), 7.95 (ddd, J ¼ 15.1, 9.8, 6.0 Hz, 3H). HRMS (ESI):
calcd. for C15H8F6N2O2S [M þ Na]þ ¼ 417.0102, found 417.0098.
tR ¼ 4.993 min, HPLC purity: 98.71%.
d
136.52, 134.95, 134.87, 133.30, 132.85, 132.40, 131.95, 131.78, 124.66,
122.65, 120.95, 115.43, 111.06, 40.79, 40.51, 40.23, 39.95, 39.68,
39.40, 39.12. HRMS (ESI): calcd. for
C
15H7F6N1O3S [M
þ
4.4.2. 3-(Phenylamino)benzo[d]isothiazole 1,1-dioxide (5)
Na]þ ¼ 417.9943, found 417.9947. tR ¼ 6.092 min, HPLC purity:
White solid. Yield 81.8%. mp > 300 ꢁC. 1H NMR (300 MHz,
100%.
DMSO‑d6)
2.0 Hz, 1H), 7.96e7.80 (m, 4H), 7.54e7.43 (m, 2H), 7.27 (t, J ¼ 7.4 Hz,
1H). 13C NMR (75 MHz, DMSO‑d6)
157.23, 141.23, 137.98, 134.22,
d
10.86 (s, 1H), 8.49 (d, J ¼ 7.1 Hz, 1H), 8.08 (dd, J ¼ 6.6,
4.4.9. 3-((3,5-Bis(trifluoromethyl)phenyl)(methyl)amino)benzo[d]
isothiazole 1,1-dioxide (12)
d
133.82, 129.53, 128.78, 126.17, 124.07, 122.58, 121.97, 40.83, 40.52,
40.24, 39.99, 39.69, 39.41, 39.13. HRMS (ESI): calcd. for C13H10N2O2S
[M þ H]þ ¼ 259.0536, found 259.0537. tR ¼ 3.220 min, HPLC purity:
99.49%.
To a solution of M1002 (200 mg, 0.5072 mmol, 1.0 eq) in DMF
(5 mL) was added potassium carbonate (105.1 mg, 0.7608 mmol,
1.5eq) and then slowly added iodomethane (107.9 mg,
0.7608 mmol, 1.5 eq). After the addition, the reaction was stirred
overnight at room temperature. The next day, the reaction was
detected by TLC and the reaction was completed. The resulting
suspension was quenched by adding water (100 mL). The mixture
was extracted with ethyl acetate (20 mL ꢃ 3) and washed with
brine (20 ml). The organic layer was combined and purified with
chromatography column (hexane: ethyl acetate ¼ 3: 1) to afford 12.
White solid. Yield 50.2%. mp 223.9e224.3 ꢁC. 1H NMR (300 MHz,
4.4.3. 3-((3,5-Dimethylphenyl)amino)benzo[d]isothiazole 1,1-
dioxide (6)
White solid. Yield 75.2%. mp > 300 ꢁC. 1H NMR (300 MHz,
DMSO‑d6)
d
10.77 (s,1H), 8.55e8.51 (m,1H), 8.06 (dd, J ¼ 6.3, 2.3 Hz,
1H), 7.90 (ddd, J ¼ 6.5, 4.5,1.4 Hz, 2H), 7.51 (d, J ¼ 1.6 Hz, 2H), 6.92 (s,
1H), 2.33 (s, 6H). HRMS (ESI): calcd. for C15H14N2O2S [M þ
H]þ ¼ 287.0849, found 287.0849. tR ¼ 3.856 min, HPLC purity:
95.49%.
DMSO‑d6)
d
8.62e8.55 (m, 2H), 8.41e8.34 (m, 1H), 8.04 (dt, J ¼ 7.5,
0.9 Hz, 1H), 7.74 (td, J ¼ 7.5, 0.8 Hz, 1H), 7.48 (t, J ¼ 7.8 Hz, 1H), 6.11
11

Y. Yu, Q. Yu, S. Liu et al.
European Journal of Medicinal Chemistry xxx (xxxx) xxx
(s, 1H), 3.65 (s, 3H). 13C NMR (75 MHz, DMSO‑d6)
d
160.28, 144.74,
4.7. Luciferase reporter assay
144.01, 133.64, 133.31, 132.81, 132.36, 130.10, 127.55, 125.80, 125.03,
123.69, 122.54, 121.40, 43.67, 40.79, 40.52, 40.24, 39.96, 39.68,
The 786-O-HRE-Luc single clone cells were obtained by infect-
ing 786-O cells, which were purchased from Cell Bank of Shanghai,
Institute of Biochemistry and Cell Biology, Chinese Academy of
Sciences with commercial lentivirus (CLS-007L, Qiagen) that de-
livers a luciferase gene driven by multiple hypoxia response ele-
ments (HREs) according to the literature methods [11]. 786-O-HRE-
Luc cells were cultured using Roswell Park Memorial Institute
(RPMI)-1640 medium (Gibco) and supplemented with 10% FBS
(Gibco). For the luciferase reporter gene experiment, the cells were
seeded on the 96-well opaque white plate (FCP968, Beyotime) on
the first day. After the cells were attached, the corresponding
compound DMSO solution was added to each well. Three experi-
ments were performed in parallel for each compound. After 24 h of
treatment, One-Lumi™ Firefly Luciferase Assay Kit (RG055, Beyo-
time) was added to each well according to the manufacturer’s in-
structions. The luciferase activity was immediately detected by the
multi-function microplate reader (SpectraMax, Molecular Devices).
EC50 was calculated by using GraphPad Prism 7 software. The
graphs are representative results and this experiment was inde-
pendently repeated three times with similar results.
39.41, 39.13. HRMS (ESI): calcd. for
C
16H10F6N2O2S [M
þ
H]þ ¼ 409.0440, found 409.0442. tR ¼ 3.980 min, HPLC purity:
99.03%. To better understand the structures of these target com-
pounds, the crystal structure of 12 was also determined by X-ray
diffraction. As shown in Scheme 1, the crystal structure of 12 was
well consistent with the characterization data of NMR and HRMS.
The detailed crystal parameters are recorded in the Suppele-
mentary data.
4.4.10. 3-((3,5-Bis(trifluoromethyl)phenyl)amino)-1H-isoindol-1-
one (13)
To a solution of 3a (1.28 g, 10 mmol, 1.0 eq) in chloroform
(20 mL) was added N,N-diethylhydroxylamine (1.8 g, 20 mmol, 2
eq) and then heated to reflux for 5 h. Following that, the mixture
was quenched with water (50 mL). The resultant mixture was
extracted with ethyl acetate (25 mL ꢃ 3) and washed with brine
(25 mL). The combined organic layer was dried over anhydrous
sodium sulfate, filtrated, and concentrated to afford 4a (0.7 g, 55%)
as a light yellow oil, which was used for the next step without
further purification. To a solution of 4a (0.29 g, 2 mmol, 1.0 eq) in
ethanol (5 mL) was added 3,5-bis(trifluoromethyl)aniline (0.5 g,
2.2 mmol,1.1 eq) and acetic acid (1 mL). The reaction was allowed to
be heated to reflux for 5 h. The resultant suspension was quenched
with water (20 mL). The mixture was extracted with ethyl acetate
(20 mL ꢃ 3) and washed with brine (20 mL). The organic layer was
combined and purified with a chromatography column (hexane:
ethyl acetate ¼ 20: 1) to afford 13. White solid. Yield, 34.0%.
4.8. Cell proliferation assay
For cell proliferation assay, the 786-O-HRE-Luc cells were
seeded on the 96-well plate on the first day. After the cells were
attached, the corresponding compound DMSO solution was added
to each well to ensure the final concentration (mM) is: 20, 6.67, 2.22,
0.74, 0.25, 0.08, 0.027, 0.009, and 0.003. Three experiments were
performed in parallel for each compound. After 24 h of treatment,
Cell Counting Kit-8 (C0037, Beyotime) was added to each cell ac-
cording to the manufacturer’s instructions. The 96-well plate was
then incubated at 37 ꢁC for 2 h. The absorbance of the samples was
detected at 450 nm. The graphs are representative results and this
experiment was independently repeated three times with similar
results.
mp > 300 ꢁC. 1H NMR (300 MHz, DMSO‑d6)
d
11.10 (s, 1H), 7.83 (t,
173.20, 169.78,
J ¼ 47.1 Hz, 8H). 13C NMR (75 MHz, DMSO‑d6)
d
150.74, 136.09, 134.22, 133.75, 133.65, 133.51, 133.45, 131.78, 131.35,
123.44, 122.89, 122.60, 117.85, 117.78, 117.73, 117.68, 117.65, 40.78,
40.51, 40.23, 39.95, 39.68, 39.40, 39.12, 25.96. HRMS (ESI): calcd. for
C
16H8F6N2O [M þ H]þ ¼ 359.0614, found 359.0611. tR ¼ 9.687 min,
HPLC purity: 99.49%.
4.9. Electrostatic potential calculation
4.5. Protein expression and purification
The wavefunctions of compounds 8, 9, 10, and Tyr278 were
obtained by Gaussian 09 program [40]. The electrostatic potential
(ESP) surfaces were calculated by the "Quantitative analysis of
molecular surface" method in the Multiwfn 3.8 program [41]. The
color mapped isosurface graphs of ESP and reduced density
gradient (RDG) were rendered by VMD 1.9 program [38].
HIF-2aeARNT complex proteins were purified from bacterial
lysates by a combination of pre-packed His∙Bind resin (Novagen),
SP Sepharose (GE Healthcare), and HiLoad 16/60 Superdex 200 pg
gel-filtration column (GE Healthcare). The specific protein expres-
sion and purification methods were described in detail in refs. 15
and 17 [15,17].
Declaration of competing interest
The authors declare that they have no known competing
financial interests or personal relationships that could have
appeared to influence the work reported in this paper.
4.6. Protein thermal-shift binding assay
Protein thermal-shift (PTS) assay was conducted in 96-well PCR
plates (Thermo Fisher Scientific) using the Protein Thermal Shift
Dye Kit (4461146, Thermo Fisher Scientific) on a StepOnePlus qPCR
machine (Thermo Fisher Scientific). The experimental method of
the instrument was set according to the manufacturer’s manual.
Each well in the experiment contained 1.5M HIF-2-ARNT protein
Acknowledgments
This work was supported by grants from the National Natural
Science Foundation of China (Grant 81773571), and Jiangsu Prov-
ince Funds for Excellent Young Scientists (Grant BK20170088). The
HIF-2 expression plasmid was generously provided by Professor
Dalei Wu (Shangdong University, China).
complex and 10
mM test compound dissolved with DMSO. The
buffer used in the experiment contained 20 mM Tris (PH ¼ 8.0) and
400 mM NaCl. The final DMSO concentration in each well was 0.1%
[15]. Each compound was tested four times in parallel. The Tm was
obtained by calculating the Derivative curve with PTS software 1.3
Appendix A. Supplementary data
(Thermo Fisher Scientific). The final
was obtained by subtracting the blank Tm value.
D
Tm value of each compound
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.ejmech.2020.112999.
12

Y. Yu, Q. Yu, S. Liu et al.
European Journal of Medicinal Chemistry xxx (xxxx) xxx
backbone parameters, Proteins 65 (2006) 712e725.
[21] R. Kumari, R. Kumar, C. Open, Source Drug Discovery, A. Lynn, g_mmpbsa–a
GROMACS tool for high-throughput MM-PBSA calculations, J. Chem. Inf.
Model. 54 (7) (2014) 1951e1962.
References
[1] F.J. Gonzalez, C. Xie, C. Jiang, The role of hypoxia-inducible factors in metabolic
diseases, Nat. Rev. Endocrinol. 15 (2018) 21e32.
[2] T. Hashimoto, F. Shibasaki, Hypoxia-inducible factor as an angiogenic master
switch, Front. Pediatr. 3 (2015) 33e47.
[3] F. Melendez-Rodriguez, O. Roche, R. Sanchez-Prieto, J. Aragones, Hypoxia-
inducible factor 2-dependent pathways driving von hippel-lindau-deficient
renal cancer, Front. Oncol. 8 (2018) 214e221.
[4] H. Tian, S.L. McKnight, D.W.J.G. Russell, development, Endothelial PAS domain
protein 1 (EPAS1), a transcription factor selectively expressed in endothelial
cells, Genes Dev. 11 (1997) 72e82.
[5] P.J. Erbel, P.B. Card, O. Karakuzu, R.K. Bruick, K.H. Gardner, Structural basis for
PAS domain heterodimerization in the basic helix–loop–helix-PAS transcrip-
tion factor hypoxia-inducible factor, Proc. Natl. Acad. Sci. U. S. A. 100 (2003)
15504e15509.
[6] L. Schito, G.L. Semenza, Hypoxia-inducible factors: master regulators of cancer
progression, Trends Cancer 2 (2016) 758e770.
[7] M.J. Koury, V.H. Haase, Anaemia in kidney disease: harnessing hypoxia re-
sponses for therapy, Nat. Rev. Neurol. 11 (2015) 394e410.
[8] Y. Yu, Q. Yu, X. Zhang, Allosteric inhibition of HIF-2a as a novel therapy for
clear cell renal cell carcinoma, Drug Discov. Today 24 (2019) 2332e2340.
[9] E.M. Wallace, J.P. Rizzi, G. Han, P.M. Wehn, Z. Cao, X. Du, T. Cheng,
R.M. Czerwinski, D.D. Dixon, B.S. Goggin, J.A. Grina, M.M. Halfmann,
M.A. Maddie, S.R. Olive, S.T. Schlachter, H. Tan, B. Wang, K. Wang, S. Xie, R. Xu,
H. Yang, J.A. Josey, A small-molecule antagonist of hif2alpha is efficacious in
preclinical models of renal cell carcinoma, Canc. Res. 76 (2016) 5491e5500.
[10] N. Gupta, J.B. Wish, Hypoxia-inducible factor prolyl hydroxylase inhibitors: a
potential new treatment for anemia in patients with CKD, Am. J. Kidney Dis.
69 (2017) 815e826.
[11] P.M. Wehn, J.P. Rizzi, D.D. Dixon, J. Grina, S.T. Schlachter, B. Wang, R. Xu,
H. Yang, X. Du, G. Han, K. Wang, Z. Cao, T. Cheng, R.M. Czerwinski, B.S. Goggin,
H. Huang, M.M. Halfmann, M.A. Maddie, E.L. Morton, S.R. Olive, H. Tan, S. Xie,
T. Wong, J.A. Josey, E.M. Wallace, Design and activity of specific hypoxia-
[22] S. Grimme, Do special noncovalent
pep stacking interactions really exist?
Angew. Chem., Int. Ed. Engl. 47 (2008) 3430e3434.
[23] T.P. Tauer, M.E. Derrick, C.D. Sherrill, Estimates of the ab initio limit for sulfur-
p
interactions: the H2S-benzene dimer, J. Phys. Chem. 109 (2005) 191e196.
[24] C.Q. Wan, J. Han, T.C.W. Mak, Intermolecular S/ interactions in crystalline
sulfanyl-triazine derivatives, New J. Chem. 33 (2009) 707e712.
[25] S.K. Singh, A. Das, The n / * interaction: a rapidly emerging non-covalent
p
p
interaction, Phys. Chem. Chem. Phys. 17 (2015) 9596e9612.
[26] M.S. Elsayed, M.E. El-Araby, R.T. Serya, K.A. Abouzid, Virtual screening and
synthesis of new chemical scaffolds as VEGFR-2 kinase inhibitors, Arznei-
mittelforschung 62 (2012) 554e560.
[27] K.V. Luzyanin, V.Y. Kukushkin, M.N. Kopylovich, A.A. Nazarov, M. Galanski,
A.J.L. Pombeiro, Novel and mild route to phthalocyanines and 3-
iminoisoindolin-1-onesvia n,n-diethylhydroxylamine-promoted conversion
of phthalonitriles and a dramatic solvent-dependence of the reaction, Adv.
Synth. Catal. 350 (2008) 135e142.
[28] Z.Q. Zhang, S. Uth, D.J. Sandman, B.M. Foxman, Structure, polymorphism and
thermal properties of phenyliminoisoindolines, J. Phys. Org. Chem. 17 (2004)
769e776.
[29] K. Huynh, C.L. Partch, Analysis of protein stability and ligand interactions by
thermal shift assay, Curr. Protein Pept. Sci. 79 (2015), 28.9.1ꢀ28.9.14.
[30] N.K. Shinada, A.G. de Brevern, P. Schmidtke, Halogens in protein-ligand
binding mechanism:
a structural perspective, J. Med. Chem. 62 (2019)
9341e9356.
[31] E. Margiotta, S.C.C. van der Lubbe, L. de Azevedo Santos, G. Paragi, S. Moro,
F.M. Bickelhaupt, C. Fonseca Guerra, Halogen bonds in ligand-protein systems:
molecular orbital theory for drug design, J. Chem. Inf. Model. 60 (2020)
1317e1328.
[32] J.S. Murray, P. Politzer, The electrostatic potential: an overview, Wiley Inter-
discip. Rev. Comput Mol Sci 1 (2011) 153e163.
[33] Y.L. Lin, A. Aleksandrov, T. Simonson, B. Roux, An overview of electrostatic free
energy computations for solutions and proteins, J. Chem. Theor. Comput. 10
(2014) 2690e2709.
inducible factor 2
cell carcinoma: discovery of clinical candidate
hydroxy-7-(methylsulfonyl)-2,3-dihydro-1h-inden-4-yl)oxy)-5
robenzonitrile (PT2385), J. Med. Chem. 61 (2018) 9691e9721.
a
(HIF-2
a
) inhibitors for the treatment of clear cell renal
s)-3-((2,2-difluoro-1-
(
-fluo-
[34] H. Wang, W. Wang, W.J. Jin, s-hole bond vs p-hole bond: a comparison based
on halogen bond, Chem. Rev. 116 (2016) 5072e5104.
[35] M. Bassetto, S. Ferla, F. Pertusati, Polyfluorinated groups in medicinal chem-
istry, Future Med. Chem. 7 (2015) 527e546.
[36] M.J. Abraham, T. Murtola, R. Schulz, S. Pall, J.C. Smith, B. Hess, E.J.S. Lindahl,
GROMACS: high performance molecular simulations through multi-level
parallelism from laptops to supercomputers, SoftwareX 1 (2015) 19e25.
[37] W.L. Jorgensen, J. Chandrasekhar, J.D. Madura, R.W. Impey, M.L. Klein, Com-
parison of simple potential functions for simulating liquid water, J. Chem.
Phys. 79 (1983) 926e935.
[38] W. Humphrey, A. Dalke, K. Schulten, VMD: visual molecular dynamics, J. Mol.
Graph. 14 (1996) 33e38.
[39] W.L. DeLano, PyMOL: an open-source molecular graphics tool, CCP4 News-
letter on Protein Crystallography 40 (2002) 82e92.
[40] Gaussian 09, Revision A.02 M.J. Frisch, G.W. Trucks, H.B. Schlegel,
G.E. Scuseria, M.A. Robb, J.R. Cheeseman, G. Scalmani, V. Barone,
G.A. Petersson, H. Nakatsuji, X. Li, M. Caricato, A. Marenich, J. Bloino,
B.G. Janesko, R. Gomperts, B. Mennucci, H.P. Hratchian, J.V. Ortiz,
A.F. Izmaylov, J.L. Sonnenberg, D. Williams-Young, F. Ding, F. Lipparini,
F. Egidi, J. Goings, B. Peng, A. Petrone, T. Henderson, D. Ranasinghe,
V.G. Zakrzewski, J. Gao, N. Rega, G. Zheng, W. Liang, M. Hada, M. Ehara,
K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao,
H. Nakai, T. Vreven, K. Throssell, J.A. Montgomery Jr., J.E. Peralta, F. Ogliaro,
M. Bearpark, J.J. Heyd, E. Brothers, K.N. Kudin, V.N. Staroverov, T. Keith,
R. Kobayashi, J. Normand, K. Raghavachari, A. Rendell, J.C. Burant, S.S. Iyengar,
J. Tomasi, M. Cossi, J.M. Millam, M. Klene, C. Adamo, R. Cammi, J.W. Ochterski,
R.L. Martin, K. Morokuma, O. Farkas, J.B. Foresman, D.J. Fox, Gaussian, Inc.,
Wallingford CT (2016).
[12] T.H. Scheuermann, D.R. Tomchick, M. Machius, Y. Guo, R.K. Bruick,
K.H. Gardner, Artificial ligand binding within the HIF2 PAS-B domain of the
HIF2 transcription factor, Proc. Natl. Acad. Sci. U. S. A. 106 (2009) 450e455.
[13] R. Xu, K. Wang, J.P. Rizzi, H. Huang, J.A. Grina, S.T. Schlachter, B. Wang,
P.M. Wehn, H. Yang, D.D. Dixon, R.M. Czerwinski, X. Du, E.L. Ged, G. Han,
H. Tan, T. Wong, S. Xie, J.A. Josey, E.M. Wallace, 3-[(1S,2S,3R)-2,3-Difluoro-1-
hydroxy-7-methylsulfonyl-indan-4-yl]oxy-5-fluoro-benzonitrile (PT2977),
hypoxia-inducible factor 2 (HIF-2 ) inhibitor for the treatment of clear cell
renal cell carcinoma, J. Med. Chem. 62 (2019) 6876e6893.
a
a
a
[14] E. Jonasch, E.K. Park, S. Thamake, M. Hirmand, W.M. Linehan, R. Srinivasan, An
open-label phase II study to evaluate PT2977 for the treatment of von Hippel-
Lindau disease-associated renal cell carcinoma, J. Clin. Oncol. 37 (2019).
TPS680ꢀTPS680.
[15] D. Wu, X. Su, J. Lu, S. Li, B.L. Hood, S. Vasile, N. Potluri, X. Diao, Y. Kim,
S. Khorasanizadeh, F. Rastinejad, Bidirectional modulation of HIF-2 activity
through chemical ligands, Nat. Chem. Biol. 15 (2019) 367e376.
[16] A.S. Chong, P.C. Anderson, Molecular dynamics simulations of the hypoxia-
inducible factor pas-b domain confirm that internally bound water mole-
cules function to stabilize the protein core for ligand binding, Biochemistry 59
(2020) 450e459.
[17] D. Wu, N. Potluri, J. Lu, Y. Kim, F. Rastinejad, Structural integration in hypoxia-
inducible factors, Nature 524 (2015) 303e308.
[18] F. Correa, J. Key, B. Kuhlman, K.H. Gardner, Computational repacking of HIF-
2alpha cavity replaces water-based stabilized core, Structure 24 (2016)
1918e1927.
[19] G. Wu, D.H. Robertson, C.L. Brooks Iii, M. Vieth, Detailed analysis of grid-based
molecular docking: a case study of CDOCKERda CHARMm-based MD docking
algorithm, J. Comput. Chem. 24 (2003) 1549e1562.
[20] V. Hornak, R. Abel, A. Okur, B. Strockbine, A. Roitberg, C. Simmerling, Com-
parison of multiple Amber force fields and development of improved protein
[41] T. Lu, F. Chen, Multiwfn: a multifunctional wavefunction analyzer, J. Comput.
Chem. 33 (2012) 580e592.
13