Novel Allosteric Activators for Ferroptosis Regulator Glutathione Peroxidase 4

Article abstract of DOI:10.1021/acs.jmedchem.8b00315

Glutathione peroxidase 4 (GPX4) is essential for cell membrane repair, inflammation suppression, and ferroptosis inhibition. GPX4 upregulation provides unique drug discovery opportunities for inflammation and ferroptosis-related diseases. However, rationa

Full text of DOI:10.1021/acs.jmedchem.8b00315

Subscriber access provided by UNIV OF DURHAM
Novel Allosteric Activators for Ferroptosis Regulator Glutathione Peroxidase 4
Cong Li, Xiaobing Deng, Weilin Zhang, Xiaowen Xie, Marcus
Conrad, Ying Liu, José Pedro Friedmann Angeli, and Luhua Lai
J. Med. Chem., Just Accepted Manuscript • Publication Date (Web): 24 Apr 2018
Downloaded from http://pubs.acs.org on April 24, 2018
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W.,
Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Novel Allosteric Activators for Ferroptosis
Regulator Glutathione Peroxidase 4
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Cong Li,†,& Xiaobing Deng,‖,& Weilin Zhang,† Xiaowen Xie,‡ Marcus Conrad,§ Ying Liu,†,‡ José
#
† ‖ ‡
,
Pedro Friedmann Angeli*, , and Luhua Lai*, ,
†BNLMS, State Key Laboratory for Structural Chemistry of Unstable and Stable Species,
College of Chemistry and Molecular Engineering, ‖PekingꢀTsinghua Center for Life Sciences,
and ‡Center for Quantitative Biology, Peking University, Beijing 100871, China
§Institute of Developmental Genetics, Helmholtz Zentrum München, 85764 Neuherberg,
Germany
#Rudolf Virchow Center for Experimental Biomedicine, University of Würzburg, 97080
Würzburg, Germany
&Coꢀfirst authors
ABSTRACT: Glutathione peroxidase 4 (GPX4) is essential for cell membrane repair,
inflammation suppression, and ferroptosis inhibition. GPX4 upregulation provides unique drug
discovery opportunities for inflammation and ferroptosisꢀrelated diseases. However, rational
design of protein activators is challenging. Until now, no compound has been reported to activate
the enzyme activity of GPX4. Here, we identified a potential allosteric site in GPX4, and
ACS Paragon Plus Environment
1

Journal of Medicinal Chemistry
Page 2 of 36
1
2
3
4
5
6
7
8
9
successfully found eight GPX4 activators using a novel computational strategy and experimental
studies. Compound 1 from the virtual screen increased GPX4 activity, suppressed ferroptosis,
reduced proꢀinflammatory lipid mediator production, and inhibited NFꢀκB pathway activation.
Further chemical synthesis and structureꢀactivity relationship studies revealed seven more
activators. The strongest compound 1d4 increased GPX4 activity to 150 % at 20 ꢁM in cellꢀfree
assay and 61 ꢁM in cell extracts. Therefore, we demonstrated that GPX4 can be directly
activated using chemical compounds to suppress ferroptosis and inflammation. Meanwhile, the
discovery of GPX4 activators verified the possibility of rational design of allosteric activators.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
INTRODUCTION
Glutathione peroxidase 4 (GPX4) is the sole antioxidant enzyme that repairs lipid
hydroperoxides, regulates eicosanoid biosynthesis, cytokine signaling,1 and ferroptosis.2
Activating GPX4 is a promising strategy to suppress eicosanoid generation and inflammation,3ꢀ5
whereas GPX4 insufficiency triggers ferroptosis—a nonꢀapoptotic form of cell death
characterized by increased lipid hydroperoxide levels.2,6,7 Ferroptosis has attracted considerable
attention due to its implication in ischemia reperfusion injury and neurodegeneration.8
Ferroptosis suppression requires GPX4 peroxidase activity, which is responsible for prevention
of arachidonoyl and adrenoyl phosphatidylethanolamines peroxidized species formation.9
Several compounds that can inhibit ferroptosis and the eicosanoid storm generated during its
execution have been developed. The ferroptosis suppression strategies that have been used can
be divided into three groups: 1) suppression of lipid peroxidation using radical trapping
antioxidants,10 2) decrease of esterified polyunsaturated fatty acids such as arachidonic and
ACS Paragon Plus Environment
2

Page 3 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
adrenic acid through acylꢀCoA synthetase longꢀchain family member 4 (ACSL4) inhibition,11 or
3) lipoxygenase (LOX) inhibition.12 As previous studies suggest GPX4 activation suppresses
lipid peroxidation generated by lipoxygenase enzymes,3ꢀ5 we rationalized that an alternative way
to suppress cell death and the eicosanoids formed during ferroptosis would be through increasing
GPX4 peroxidase activity.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
However, small molecule activator discovery is a much more challenging task than inhibitor
discovery. Many inhibitors that bind to the substrate pockets of protein targets have been
identified, whereas most activators function by allosteric regulation. However, allosteric site
information is lacking for most enzymes, including GPX4. To date, the most commonly used
method for identifying activators relies on laborious highꢀthroughput experimental screening.13
Therefore, a rational approach for discovering activators is urgently needed.14,15
In this study, we used a rational strategy to discover GPX4 allosteric activators. We first
computationally identified a potential allosteric site using ligandꢀbinding site detection and
motion correlation analysis for allosteric site prediction. We then used this predicted site to
virtually screen a chemical library to identify potential GPX4 activators. Our initial enzyme
assay screen identified one GPX4 activating compound. We demonstrated this GPX4 activator
could inhibit ferroptosis and suppress the generation of proꢀinflammatory lipid mediators in cell
assays. This compound was further optimized using chemical synthesis and structureꢀactivity
relationship (SAR) analysis.
RESULTS AND DISCUSSION
Allosteric Site Prediction and Virtual Screening. To find suitable binding pockets for
designing allosteric activators, we first used the binding site detection program CAVITY16,17 to
ACS Paragon Plus Environment
3

Journal of Medicinal Chemistry
Page 4 of 36
1
2
3
4
5
6
7
8
9
identify surface cavities in the crystal structure of human GPX4 U46C mutant (hGPX4ꢀC, PDB
entry 2OBI18). These surface cavities were analyzed using the CorrSite method based on the
motion correlation between potential ligandꢀ and substrateꢀbinding sites that we developed
recently.19 This procedure predicted one allosteric site localized on the opposite side of the
substrateꢀbinding site (Figure 1a), and the two sites were separated by the central βꢀsheet
consisting of five βꢀstrands. The predicted potential allosteric site has the highest CAVITY score
and a CorrSite score of 1.49 (cavities with CorrSite score > 0.5 were predicted as potential
allosteric sites in the CorrSite method which was developed based on a list known allosteric
proteins). The volume of the oliveꢀshaped allosteric site is 278.4 Å3 with a predicted maximal
pKd of 7.48. We performed molecular dynamics simulations to analyze the dynamic changes of
the predicted allosteric site, and confirmed that this site was persistent in volume with high
motion correlations with the substrateꢀbinding site (Supporting Information, Figure S1). The
allosteric site is surrounded by three acidic residues (D21, D23, and D101), two basic residues
(K31 and K90), and seven apolar residues (I22, A93, A94, V98, F100, M102, and F103).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment
4

Page 5 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 1. GPX4 activator identification. a) The predicted allosteric site (shown in green) using
CAVITY and CorrSite is on the opposite side from the substrateꢀbinding site (shown in red)
across the protein (PDB entry 2OBI18). The connections between D23 and C46, and K90 and
Q81 are shown in magentas and yellow respectively. b) Chemical structure of 1. c) Doseꢀ
response curve for 1 in cellꢀfree assays. Data shown represent the mean ± SEM (n = 6). d) SPR
doseꢀresponse curves of 1. e) Data in Figure 1d at equilibrium were fitted with the Hill model.
The color of the dots are consistent with the color of the response curves in Figure 1d (KD = 63 ±
5 ꢁM). f) Predicted binding mode of 1. The hydrogen bonds between the activator (green) and
GPX4 (pale cyan, PDB entry 2OBI18) are shown as dashed orange lines.
We then used the molecular docking program Glide to virtually screen against the SPECS
compounds library (pan assay interference compounds20 excluded). After performing a twoꢀstep
ACS Paragon Plus Environment
5

Journal of Medicinal Chemistry
Page 6 of 36
1
2
3
4
5
6
7
8
9
molecular docking scheme, we selected 251 compounds for experimental testing according to the
following criteria: compounds should have good interactions with pocket residues with (1) at
least one hydrogen bond, and (2) good hydrophobic interactions (see Experimental Section).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Biological Evaluation. CellꢀFree Enzyme Activity Assay. The activating activities of the 251
selected compounds were tested in the cellꢀfree assay. Compound PKUMDLꢀLCꢀ101 (1) (Figure
1b) increased the hGPX4ꢀC enzymatic activity in a doseꢀdependent manner (Figure 1c) to 263 %
of the original enzyme activity. Compound 1 was shown to directly bind to hGPX4ꢀC using the
surface plasmon resonance (SPR) assay, with 63 ± 5 ꢁM dissociation constant (Figure 1d and
Figure 1e).
Prevention of ErastinꢀInduced Ferroptosis. Activating GPX4 may be beneficial during erastin
(Era) induced ferroptosis. We tested whether 1 can prevent Eraꢀinduced ferroptosis in human
HTꢀ1080 fibrosarcoma cells.2 We incubated the cells with 1 and subjected them to Era (10 ꢁM)
treatment. After 24 hours, cell viability was determined using the MTS assay (Figure 2a).
Compound 1 could inhibit Eraꢀinduced ferroptosis, thus supporting the use of GPX4 activator as
antiferroptotic agent.
ACS Paragon Plus Environment
6

Page 7 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Figure 2. The effects of 1 in cellular model of erastinꢀinduced ferroptosis and human PMN
assay. a) Prevention of erastinꢀinduced ferroptosis by GPX4 activator 1 in HTꢀ1080 cells.
Compound 1 prevented cell death with a lethal concentration of erastin (Era, 10 ꢁM). Data
shown represent the mean ± SEM (n = 5). b) Doseꢀdependent GPX4 activation in PMN cells.
Data shown represent the mean ± SEM (n = 4). The statistical significance was determined using
a twoꢀtailed tꢀtest, ***p < 0.001.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Influence on Eicosanoid Biosynthesis. During the course of ferroptosis, arachidonic acid (AA)
oxidation products are released,7 which may drive a secondary inflammatory damage in
ferroptosisꢀrelated pathologies. In the inflammationꢀrelated AA metabolic network, AA is
oxidized to 5ꢀHpETE, 12ꢀHpETE, and 15ꢀHpETE by 5ꢀLOX, 12ꢀLOX, and 15ꢀLOX,
respectively (Supporting Information, Figure S2).21 GPX4 converts HpETEs to the
corresponding HETEs. And stimulating the release of these antiꢀinflammatory lipid mediators is
an additional beneficial effect of GPX4 activators. Thus, we tested the effects of 1 on eicosanoid
biosynthesis using human polymorphonuclear leucocytes (PMN). We used this assay to assess
the eicosanoids formation without interference from cell death. We stimulated the 5ꢀLOX
pathway using calcium ionophore A2318722,23 and used LCꢀMS/MS to detect the formation of
major eicosanoids derived from AA. Compound 1 induced GPX4 activating effects in a doseꢀ
dependent manner including an increased 5ꢀHETE/leukotriene B4 (LTB4) ratio, 12ꢀHETE level,
and 15ꢀHETE level (Figure 2b and Supporting Information Figure S3). To confirm that 1 does
not affect the other enzymes in the AA network, we tested the effects of 1 on cyclooxygenase 1
(COXꢀ1), cyclooxygenase 2 (COXꢀ2), microsomal prostaglandin E2 synthase 1 (mPGESꢀ1), 5ꢀ
LOX, leukotriene A4 hydrolase (LTA4H), 12ꢀLOX, and 15ꢀLOX. Compound 1 did not show
ACS Paragon Plus Environment
7

Journal of Medicinal Chemistry
Page 8 of 36
1
2
3
4
5
6
activity against COXꢀ1, COXꢀ2, 5ꢀLOX, 12ꢀLOX, or 15ꢀLOX, and only slightly inhibited
mPGESꢀ1 and LTA4H (inhibition = 23 % and 17 % at 500 ꢁM, respectively).
7
8
9
Inhibition of the NFꢀκB Pathway. To further demonstrate that GPX4 activator can modulate
GPX4ꢀmediated events, we tested its effect in cytokine signaling. Direct involvement of
hydrogen peroxide in NFꢀκB activation has been discussed,24 and GPX4 might prevent the TNF
or ILꢀ1 mediated NFꢀκB activation.25,26 We used the dualꢀluciferase reporter assay system to test
the effects of 1 on TNF or ILꢀ1 induced activation of NFꢀκB pathway in HEK293T cells.
Compound 1 inhibited the TNF and ILꢀ1 induced NFꢀκB pathway activation doseꢀdependently
(Figure 3).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 3. Doseꢀdependent inhibition of NFꢀκB pathway activation by GPX4 activator. a)
Inhibition of TNF induced NFꢀκB pathway activation. b) Inhibition of ILꢀ1 induced NFꢀκB
pathway activation. Data shown represent the mean ± SEM (n = 5). The statistical significance
was determined using a twoꢀtailed tꢀtest, **p < 0.01, ***p < 0.001.
ACS Paragon Plus Environment
8

Page 9 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Structure−Activity Relationship. The binding mode of 1 was predicted by molecular docking
studies (Figure 1f). To establish the SAR for the GPX4 activator, several analogues of 1 were
tested. Three compounds (1s1, 1s2, and 1s3) retained the capacity to increase GPX4 activity in a
concentrationꢀdependent manner in the cellꢀfree assay (Table 1 and Supporting Information
Figure S4). As a general feature, compounds with a bulkier R1 group were less potent. To further
explore the SAR of 1, we designed and synthesized 12 analogues (Scheme 1, see Experimental
Section for details) and tested their effects on cellꢀfree GPX4 enzyme activity (Table 1 and
Supporting Information Figure S5).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 1. Activity of 1 analoguesa
Cmpd Name
R1
R2
Max.
Conc.
Activation
activation
required to with
reach 150 % ꢁM
200
of
activation
(ꢁM)
compound
PKUMDLꢀ
LCꢀ101
H
H
H
H
263 %
300 %
200 %
225 %
248
255
285
147
127 %
129 %
115 %
167 %
1
PKUMDLꢀ
LCꢀ101ꢀS01
1s1
1s2
1s3
1d1
PKUMDLꢀ
LCꢀ101ꢀS02
PKUMDLꢀ
LCꢀ101ꢀS03
PKUMDLꢀ
activation with 200 ꢁM
ACS Paragon Plus Environment
9

Journal of Medicinal Chemistry
Page 10 of 36
1
2
3
4
LCꢀ101ꢀD01
of compound < 115 %
5
6
7
8
PKUMDLꢀ
LCꢀ101ꢀD02
activation with 200 ꢁM
of compound < 115 %
1d2
1d3
1d4
1d5
1d6
1d7
1d8
1d9
1d10
1d11
1d12
PKUMDLꢀ
LCꢀ101ꢀD03
437 %
279 %
29
318 %
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
PKUMDLꢀ
LCꢀ101ꢀD04
20
219 %
PKUMDLꢀ
LCꢀ101ꢀD05
activation with 200 ꢁM
of compound < 115 %
PKUMDLꢀ
LCꢀ101ꢀD06
activation with 200 ꢁM
of compound < 115 %
PKUMDLꢀ
LCꢀ101ꢀD07
activation with 200 ꢁM
of compound < 115 %
PKUMDLꢀ
LCꢀ101ꢀD08
activation with 200 ꢁM
of compound < 115 %
PKUMDLꢀ
LCꢀ101ꢀD09
activation with 200 ꢁM
of compound < 115 %
PKUMDLꢀ
LCꢀ101ꢀD10
209 %
214 %
121
167 %
PKUMDLꢀ
LCꢀ101ꢀD11
301
121 %
PKUMDLꢀ
LCꢀ101ꢀD12
activation with 200 ꢁM
of compound < 115 %
aAll measurements were performed in hextuplicate. Data were given by reading corresponding
values from the curves fitted using the Hill model.
Scheme 1. Synthesis Procedures for the 1 Analogues
ACS Paragon Plus Environment
10

Page 11 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
H
O
R2
O
N
H
Cl
R2
S
N
O
S
S
1)R2NH2
O
S
O
O
Cl
Cl
O
N
2)NaOH(1N), EtOH, reflux
H
THF, H2O, r.t.
H2N
SCN
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
H
O
S
R2
N
1) R1NH2, EtOH
2) HCl, EtOAc
S
O
R1
N
N
H
H
Since the negatively charged residues D21 and D23 in the allosteric site located near the R2
group, we designed compounds with positively charged substitution on the sulfonamide amino
group to enhance the ligandꢀtarget interactions by an additional salt bridge. Compounds 1d1,
1d2, 1d3, 1d4, 1d5, and 1d6 with 2ꢀaminoethyl R2 group have different sizes of R1 groups,
including 4ꢀchlorobenzyl, ethyl, tertꢀbutyl, cyclopentyl, cyclohexyl, and cycloheptyl. Among
them, tertꢀbutyl (1d3) and cyclopentyl (1d4) R1 groups improved the activity, and the other R1
groups larger than cyclopentyl or smaller than tertꢀbutyl were less potent, confirming that steric
effect is crucial and subtle in the binding to the small allosteric site. Compound 1d4 enhanced
GPX4 activity to 279 %. The 1d4 concentration required to activate the enzyme activity to 150
% was approximately one order of magnitude lower than that of 1 (20 ꢁM vs 248 ꢁM). The
dissociation constant of compound 1d4 determined by SPR assay (Supporting Information,
Figure S6) was also lower than that of 1 (19 ± 1 ꢁM vs 63 ± 5 ꢁM).
We also explored other R2 groups including ethyl and hydroxyethyl (1d7, 1d8, and 1d9), both of
them led to the loss of potency, suggesting that electrostatic interaction has a more dominant role
than hydrogen bond in this situation. Starting from 1d4, we further tested the suitable length of
ACS Paragon Plus Environment
11

Journal of Medicinal Chemistry
Page 12 of 36
1
2
3
4
5
6
7
8
9
R2 group. Compounds 1d10 and 1d11 with Nꢀ(2ꢀ(methylamino)ethyl) and 3ꢀaminopropyl
substitutions on the sulfonamide amino group could activate GPX4, while compound 1d12 with
4ꢀaminobutyl substituent was not active. In general, a 2ꢀaminoethyl R2 group fits the best for the
limited space in the allosteric site. Overall, SAR studies supported the predicted binding position
of 1 and allowed us to identify compounds that are more potent.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Mutagenesis Studies. We further corroborated the binding positions using site directed
mutagenesis. According to the molecular docking structures, the activators interacted with the
side chain carboxyl groups of residues D21 and D23 (Figure 4a). We generated two single
mutants (D21A and D23A) and one double mutant (D21A/D23A). GPX4 activation by
compound 1 and 1d4 was reduced for D21A and D23A and was dramatically deteriorated for
D21A/D23A, supporting the predicted hydrogen bond and electrostatic interactions (Figure 4b).
The dissociation constant of 1d4 with D21A/D23A was further determined using SPR assay (KD
= 374 ± 8 ꢁM, Supporting Information, Figure S7). The reduced binding to the mutant enzyme
also supports the docking positions. Interestingly, the D21A, D23A, and D21A/D23A mutants
showed increased enzymatic activities, implying these residues may be involved in regulating
GPX4 enzymatic activity. D23 forms salt bridge with K90 in the allosteric site. Activator
compound interacting with D23 would weaken the salt bridge, and possibly stabilize the
conformation of the catalytic triad (C46, Q81, and W136) favorable for catalysis through the two
pathways indicated in Figure 1a. The biological relevance of this finding warrants further
investigation.
ACS Paragon Plus Environment
12

Page 13 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 4. Binding mode of 1d4. a) Predicted binding mode of 1d4 to GPX4. The hydrogen
bonds between the activator (green) and GPX4 (pale cyan) are shown as orange dash lines. b)
Enzyme activities and responses to 1 and 1d4 of GPX4 mutants. aData were given by reading
corresponding values from the curves fitted by the Hill model. Data shown represent the mean ±
SEM (n = 3).
Selectivity Study. Compound 1 and 1d4 were tested in a GPX4ꢀspecific activity assay using
mouse embryonic fibroblasts (MEF) cell extracts, and both of them activated GPX4 doseꢀ
dependently (Figure 5a). The 1d4 concentration required to activate the GPX4 activity to 150 %
was 61 ꢁM. To further evaluate the selectivity and specificity of the compounds in a cellular
context, we measured GPX4ꢀspecific activity in GPX4ꢀknockout mouse embryonic fibroblasts
reꢀconstituted with a mouse wildꢀtype GPX4 and a mock transduced control (Figure 5b). In
addition, when using cholesterol hydroperoxide (ChOOH, a specific GPX4 substrate) as a
stressor, compound 1 and 1d4 could activate GPX4 in intact cells and protect them from ChOOH
toxic effects (Figure 5c).
ACS Paragon Plus Environment
13

Journal of Medicinal Chemistry
Page 14 of 36
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 5. Assessment of the GPX4ꢀspecific activity. a) Doseꢀdependent GPX4 activity assay of
1 and 1d4. GPX4 specific activity is expressed as NADPH (nmoles) consumed per mg of protein
per min. Data shown represent the mean ± s.d. (n = 3). b) Cell extracts were prepared from
GPX4 conditional KO MEFs expressing a GPX4 expressing vector (WT) and a mock empty
control (KO) to assess the GPX4 specific activity induced by 1 and 1d4 (100 µM). c) Cholesterol
hydroperoxide (ChOOH) induced cell death assay. 1 and 1d4 (200 µM) prevented cell death
induced by the GPX4 specific substrate ChOOH (20 µM). Data shown represent the mean ± s.d.
(n = 4). The statistical significance was determined using a twoꢀtailed tꢀtest, *p < 0.05, **p <
0.01, ***p < 0.001.
CONCLUSION
Despite significant advantages of protein activators, few known small molecule activators exist.
Using a structure and computationalꢀbased approach, we identified an allosteric site in GPX4 and
discovered eight GPX4 allosteric activators. To the best of our knowledge, this is the first report
of GPX4 enzyme activators. We also verified that these compounds were not reducing agents or
iron chelators (Supporting Information, Figure S8 and Figure S9). These small molecules can
ACS Paragon Plus Environment
14

Page 15 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
suppress ferroptosis with a mechanism distinct from all other inhibitors described so far.
Ferroptosis inhibitors that function by activating the GPX4 enzyme activity are expected to be
more specific than the lipid ROS scavenger Ferrostatin/Liproxstatin and iron chelators like
deferoxamine and ciclopirox olamine. GPX4 activators do not interfere with cellular iron
concentrations or other ironꢀrelated physiological processes.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
For the eight GPX4 activators, we searched SciFinder for substances with similar structures in
related studies. Seven compounds (1, 1s1, 1s3, 1d3, 1d4, 1d10, and 1d11) have unique structures
not published before. For compound 1s2, it was reported by Vullo et al. as inhibitor of the zincꢀ
containing cancerꢀassociated carbonic anhydrases (CA) isozyme IX, with inhibition constant of
37 nM.27 To rule out possible interference of zincꢀrelated processes, we confirmed that these
GPX4 activators were not zinc chelators (Supporting Information, Figure S10).
In conclusion, we found novel allosteric GPX4 activators using combined computational and
experimental approaches. The activators can increase GPX4 activity and produce beneficial
outcomes in biological processes involving GPX4, supporting further development of these
molecules as promising cytoꢀprotective and antiꢀinflammatory agents. These compounds are also
the first description of small molecules able to suppress ferroptosis through GPX4 activation.
Moreover, our study provides a general strategy for the future development of activators for
other protein targets.
EXPERIMENTAL SECTION
Compound Synthesis and Characterization. The compounds selected from virtual screen were
purchased from SPECS with purity of more than 90 % and for most compounds greater than 95
% (confirmed by SPECS with NMR, LCꢀMS, or both; data is available on the website
ACS Paragon Plus Environment
15

Journal of Medicinal Chemistry
Page 16 of 36
1
2
3
4
5
6
7
8
9
http://www.specs.net/; for our hits, PKUMDLꢀLCꢀ101, PKUMDLꢀLCꢀ101ꢀS01, PKUMDLꢀLCꢀ
101ꢀS02, and PKUMDLꢀLCꢀ101ꢀS03, NMR and LCꢀMS data are provided in the Supporting
Information). Catalog numbers are provided in the Supporting Information Table S1. For the
chemical synthesis of the analogues of 1 for SAR study, synthesis reagents and solvents were
obtained from J&K Chemical (Beijing, China) and used without further purification.
Characterization of the analogues of 1 are provided in the Supporting Information. 1H and 13C
NMR spectra were recorded on Bruker (400 MHz) instruments, using CDCl3 or DMSOꢀd6 as
deuterated solvents and with the residual solvent as the internal reference. For all NMR
experiments, the deuterated solvent signal was used as the internal lock. Coupling constants (J
values) are given in hertz (Hz). Multiplicities of 1H NMR signals are reported as follows: s,
singlet; d, doublet; dd, doublet of doublets; dt, doublet of triplets; t, triplet; q, quartet; m,
multiplet; br, broad signal. NMR data was analyzed by using MestReNova Software version
11.0. Highꢀresolution mass spectra (HRMS) were acquired on a Brukerꢀ400 M spectrometer
using TMS as internal standard. High resolution mass spectra were recorded on a Bruker Apex
IV FTMS mass spectrometer using ESI (electrospray ionization). Analytical thin layer
chromatography (TLC) was performed on silica gel 60 F254 precoated plates (0.25 mm) from
EMD Chemical Inc., and components were visualized by ultraviolet light (254 nm).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
General Procedures for Preparation of Substituted 4ꢀThioureidobenzenesulfonamide.
Amine (15 mmol) R2NH2 in 250 mL dichloromethane was stirred at 0 °C, 2 mL triethylamine
was added, and 2.33 g (10 mmol) 4ꢀacetamidobenzenesulfonyl chloride was added in portions.
Remove the reaction to r.t. for another 1 h, evaporated under reduced pressure, the crude was
diluted with 50 mL water and 50 mL EtOH, then added 1.60 g (40 mmol) NaOH, the mixture
was stirred at reflux for 3 h; Then it was diluted with water adjusting the pH to 7, extracted with
ACS Paragon Plus Environment
16

Page 17 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
AcOEt. The organic extracts were washed with brine, dried over Na2SO4, filtered, and
evaporated under reduced pressure. The crude was purified by flash chromatography on silica
gel. Afforded 4ꢀaminoꢀNꢀsubstitutedbenzenesulfonamide. 4ꢀaminoꢀNꢀ
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
substitutedbenzenesulfonamide (5 mmol) was dissolved in 20 mL THF, added 5.0 mL H2O, the
mixture was stirred at r.t. for 30 min, then pour into 200 mL iceꢀwater, filtered, the white solid
that separated was collected and dried in vacuo afforded 4ꢀisothiocyanatoꢀNꢀ
substitutedbenzenesulfonamide. 1 mmol 4ꢀisothiocyanatoꢀNꢀsubstitutedbenzenesulfonamide was
dissolved in 5 mL EtOH, 1 mmol R1NH2 was added, the mixture was stirred at r.t., the reaction
was monitored by TLC until the starting materials disappeared. The product was separated by
filtration, dried in vacuo afforded substituted 4ꢀthioureidobenzenesulfonamide. For the Nꢀboc
protecting group, the final products were obtained by hydrochloric acid. The carbamate was
dissolved in 5 mL EtOH, and added 0.5 mL EtOAcꢀHCl, the mixture was stirred at r.t. for 3 h,
filtered, the white solid that separated was collected and dried in vacuo afforded corresponding
hydrochloride products.
Nꢀ(2ꢀAminoethyl)ꢀ4ꢀ(3ꢀ(4ꢀchlorobenzyl)thioureido)benzenesulfonamide Hydrochloride
(1d1). 1H NMR (400 MHz, DMSOꢀd6): δ 10.75 (s, 1H), 9.00 (t, J = 5.8 Hz, 1H), 7.98 (s, 3H),
7.93 – 7.84 (m, 3H), 7.78 – 7.70 (m, 2H), 7.39 (q, J = 8.6 Hz, 4H), 4.76 (d, J = 5.8 Hz, 2H), 2.96
(q, J = 6.3 Hz, 2H), 2.85 (m, 2H). 13C NMR (100 MHz, DMSOꢀd6): δ 181.24, 144.30, 138.18,
133.74, 131.91, 129.72, 128.70, 127.80, 121.68, 46.50, 40.51, 38.98. HRMS (ESI): calcd for
+
C16H20ClN4O2S2 , [(M+H)+], 399.0711, found 399.0722.
Nꢀ(2ꢀAminoethyl)ꢀ4ꢀ(3ꢀethylthioureido)benzenesulfonamide Hydrochloride (1d2). 1H NMR
(400 MHz, DMSOꢀd6): δ 10.60 (s, 1H), 8.57 (s, 1H), 8.25 – 7.59 (m, 8H), 3.54 – 3.40 (m, 2H),
3.02 – 2.79 (m, 4H), 1.13 (t, J = 7.2 Hz, 3H). 13C NMR (100 MHz, DMSOꢀd6): δ 180.44, 144.51,
ACS Paragon Plus Environment
17

Journal of Medicinal Chemistry
Page 18 of 36
1
2
3
4
+
133.35, 127.76, 121.22, 40.50, 38.96, 38.88, 14.38. HRMS (ESI): calcd for C11H19N4O2S2 ,
5
6
[(M+H)+], 303.0944, found 303.0943.
7
8
9
Nꢀ(2ꢀAminoethyl)ꢀ4ꢀ(3ꢀ(tertꢀbutyl)thioureido)benzenesulfonamide Hydrochloride (1d3). 1H
NMR (400 MHz, DMSOꢀd6): δ 10.56 (s, 1H), 8.36 (s, 1H), 8.00 (s, 3H), 7.87 (t, J = 8.5 Hz, 3H),
7.69 (d, J = 8.5 Hz, 2H), 3.09 – 2.74 (m, 4H), 1.49 (s, 9H). 13C NMR (100 MHz, DMSOꢀd6): δ
179.53, 143.95, 134.65, 127.53, 121.80, 53.49, 44.13, 40.53, 28.82. HRMS (ESI): calcd for
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
+
C13H23N4O2S2 , [(M+H)+], 331.1257, found 331.1257.
Nꢀ(2ꢀAminoethyl)ꢀ4ꢀ(3ꢀcyclopentylthioureido)benzenesulfonamide Hydrochloride (1d4). 1H
NMR (400 MHz, DMSOꢀd6): δ 10.26 (s, 1H), 8.51 (s, 1H), 7.86 (d, J = 8.5 Hz, 4H), 7.80 (t, J =
5.7 Hz, 1H), 7.70 (d, J = 8.8 Hz, 2H), 4.60 – 4.43 (m, 1H), 2.94 (q, J = 6.3 Hz, 2H), 2.85 (q, J =
6.1 Hz, 2H), 1.93 (dt, J = 13.2, 6.5 Hz, 2H), 1.74 – 1.62 (m, 2H), 1.61 – 1.43 (m, 4H). 13C NMR
(100 MHz, DMSOꢀd6): δ 179.42, 144.21, 139.62, 127.19, 120.41, 54.88, 40.00, 38.48, 31.98,
23.33. HRMS (ESI): calcd for C14H23N4O2S2, [(M+H)+], 343.1257, found 343.1262.
Nꢀ(2ꢀAminoethyl)ꢀ4ꢀ(3ꢀcyclohexylthioureido)benzenesulfonamide Hydrochloride (1d5). 1H
NMR (400 MHz, DMSOꢀd6): δ 9.96 (s, 1H), 8.17 (s, 1H), 7.84 – 7.73 (m, 4H), 7.68 (d, J = 8.8
Hz, 2H), 7.50 (s, 1H), 4.15 – 4.05 (m, 1H), 3.15 (t, J = 7.0 Hz, 2H), 2.92 (s, 2H), 1.97 – 1.87 (m,
2H), 1.71 (s, 2H), 1.60 – 1.55 (m, 2H), 1.32 – 1.19 (m, 4H). HRMS (ESI): calcd for
C15H25N4O2S2, [(M+H)+], 357.1415, found 357.1415.
Nꢀ(2ꢀAminoethyl)ꢀ4ꢀ(3ꢀcycloheptylthioureido)benzenesulfonamide Hydrochloride (1d6). 1H
NMR (400 MHz, DMSOꢀd6): δ 10.51 (s, 1H), 8.58 (d, J = 8.0 Hz, 1H), 7.95 (s, 3H), 7.90 (d, J =
8.8 Hz, 2H), 7.83 (t, J = 5.9 Hz, 1H), 7.70 (d, J = 8.8 Hz, 2H), 4.28 (m, 1H), 2.95 (q, J = 6.3 Hz,
2H), 2.84 (q, J = 6.3 Hz, 2H), 1.94 – 1.89 (m, 2H), 1.64 – 1.43 (m, 10H). 13C NMR (100 MHz,
ACS Paragon Plus Environment
18

Page 19 of 36
Journal of Medicinal Chemistry
1
2
3
4
DMSOꢀd6): δ 179.14, 144.71, 127.70, 120.99, 54.39, 38.98, 34.13, 28.21, 24.11. HRMS (ESI):
5
6
+
calcd for C16H27N4O2S2 , [(M+H)+], 371.1570, found 371.1560.
7
8
9
4ꢀ(3ꢀCyclopentylthioureido)ꢀNꢀethylbenzenesulfonamide (1d7). 1H NMR (400 MHz, DMSOꢀ
d6): δ 9.66 (s, 1H), 8.10 (s, 1H), 7.77 – 7.63 (m, 4H), 7.43 (t, J = 5.7 Hz, 1H), 4.50 (s, 1H), 2.76
(qd, J = 7.2, 5.6 Hz, 2H), 1.95 (dq, J = 12.9, 7.2 Hz, 2H), 1.73 – 1.43 (m, 6H), 0.97 (t, J = 7.2
Hz, 3H). 13C NMR (100 MHz, DMSOꢀd6): δ 179.97, 143.96, 127.58, 121.45, 55.68, 37.98,
32.40, 23.86, 15.19. HRMS (ESI): calcd for C14H21N3O2S2, [(M+H)+], 328.1153, found
328.1146.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
4ꢀ(3ꢀCyclopentylthioureido)ꢀNꢀ(2ꢀhydroxyethyl)benzenesulfonamide (1d8). 1H NMR (400
MHz, DMSOꢀd6): δ 9.66 (s, 1H), 8.11 (s, 1H), 7.80 – 7.64 (m, 4H), 7.47 (t, J = 6.0 Hz, 1H), 4.68
(t, J = 5.6 Hz, 1H), 4.53 – 4.47 (m, 1H), 3.37 (q, J = 6.2 Hz, 2H), 2.77 (q, J = 6.2 Hz, 2H), 1.99 –
1.91 (m, 2H), 1.72 – 1.41 (m, 8H). 13C NMR (100 MHz, DMSOꢀd6): δ 179.97, 143.97, 127.59,
60.38, 55.68, 45.53, 32.39, 23.86. HRMS (ESI): calcd for C14H21N3O3S2, [(M+H)+], 344.1097,
found 344.1096.
4ꢀ(3ꢀ(4ꢀChlorobenzyl)thioureido)ꢀNꢀ(2ꢀhydroxyethyl)benzenesulfonamide (1d9). 1H NMR
(400 MHz, DMSOꢀd6): δ 10.02 (s, 1H), 8.53 (s, 1H), 7.71 (d, J = 3.5 Hz, 4H), 7.51 (t, J = 5.2 Hz,
1H), 7.39 (q, J = 8.7 Hz, 4H), 4.75 (d, J = 5.2 Hz, 2H), 4.69 (t, J = 5.5 Hz, 1H), 3.44 – 3.31 (m,
2H), 2.87 – 2.74 (m, 2H). 13C NMR (100 MHz, DMSOꢀd6): δ 181.16, 143.52, 138.20, 135.27,
131.94, 129.79, 128.70, 127.71, 122.27, 60.39, 46.88, 45.55, 40.15, 39.94, 39.73. HRMS (ESI):
calcd for C16H18ClN3O3S2, [(M+H)+], 400.0551, found 400.0551.
4ꢀ(3ꢀCyclopentylthioureido)ꢀNꢀ(2ꢀ(methylamino)ethyl)benzenesulfonamide Hydrochloride
(1d10). 1H NMR (400 MHz, DMSOꢀd6): δ 10.18 (s, 1H), 8.60 (s, 2H), 8.44 (d, J = 7.7 Hz, 1H),
ACS Paragon Plus Environment
19

Journal of Medicinal Chemistry
Page 20 of 36
1
2
3
4
5
6
7
8
7.86 – 7.82 (m, 3H), 7.76 – 7.65 (m, 2H), 4.55 – 4.45 (m, 1H), 3.00 – 2.95 (m, 5H), 2.54 (t, J =
5.2 Hz, 2H), 1.93 (dt, J = 13.3, 6.5 Hz, 2H), 1.74 – 1.35 (m, 6H). 13C NMR (100 MHz, DMSOꢀ
d6): δ 180.95, 144.68, 134.80, 126.49, 121.96, 51.17, 48.40, 40.29, 33.39, 30.65, 23.35. HRMS
9
+
(ESI): calcd for C15H26N4O2S2 , [(M+H)+], 357.1413, found 357.1415.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Nꢀ(3ꢀAminopropyl)ꢀ4ꢀ(3ꢀcyclopentylthioureido)benzenesulfonamide Hydrochloride (1d11).
1H NMR (400 MHz, DMSOꢀd6): δ 10.55 (s, 1H), 8.71 (s, 1H), 8.07 – 7.56 (m, 8H), 4.52ꢀ4.47
(m, 1H), 2.82ꢀ2.73 (m, 4H), 1.96 – 1.85 (m, 2H), 1.73ꢀ1.63 (m, 4H), 1.58ꢀ1.43 (m, 4H). 13C
NMR (100 MHz, DMSOꢀd6): δ 179.90, 144.42, 127.49, 120.86, 55.37, 36.98, 32.49, 27.89,
+
23.83. HRMS (ESI): calcd for C15H25N4O2S2 , [(M+H)+], 357.1413, found 357.1406.
Nꢀ(4ꢀAminobutyl)ꢀ4ꢀ(3ꢀcyclopentylthioureido)benzenesulfonamide Hydrochloride (1d12).
1H NMR (400 MHz, DMSOꢀd6): δ 10.73 (s, 1H), 8.87 (d, J = 23.5 Hz, 1H), 8.15ꢀ8.00 (m, 4H),
7.90 (d, J = 8.6, 2H), 7.67 (d, J = 8.6 Hz, 2H), 4.53 ꢀ 4.46 (m, 1H), 3.45 (q, J = 6.6 Hz, 2H), 2.70
(t, J = 7.0 Hz, 2H), 1.94 – 1.83 (m, 4H), 1.74ꢀ1.64 (m, 4H), 1.60 – 1.49 (m, 4H). 13C NMR (100
MHz, DMSOꢀd6): δ 179.90, 144.41, 133.97, 127.44, 120.68, 55.25, 51.62, 42.40, 38.74, 31.07,
26.49, 24.71, 23.89. HRMS (ESI): calcd for C16H27N4O2S2 , [(M+H)+], 371.1570, found
+
371.1564.
Reagents and Cell Culture. Reagents were purchased from SigmaꢀAldrich unless otherwise
noted. The pQEꢀ30 bacterial expression plasmid of the U46C mutant of human cytosolic GPX4
was a generous gift from Professor Hartmut Kuhn (University Medicine BerlinꢀCharité,
Germany). IPTG, PMSF, DTT, EDTA, and glutathione were from Amresco. Standard
compounds for LCꢀMS/MS were purchased from Cayman Chemical. Calcium ionophore
A23187 was obtained from J&K Chemical. The Symmetry C18 reverseꢀphase column (3.5 ꢁm,
ACS Paragon Plus Environment
20

Page 21 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
2.1 mm × 150 mm) was purchased from Waters Corp. HTꢀ1080 cells were the generous gift of
Professor Chu Wang (Peking University, China). HEK293T cells were received as a gift from
Professor Jincai Luo (Peking University, China). The DualꢀGlo Luciferase Assay System was
from Promega. ChOOH was a kind gift of Prof. Sayuri Myiamoto (Universidade de Sao Paulo,
Brazil).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Allosteric Site Prediction and Virtual Screen. Potential allosteric site in GPX4 was identified
using the CAVITY and CorrSite program16,17,19, and then applied to screen for potential allosteric
modulator. The volume of the predicted allosteric site is smaller than the substrate site (278.4 Å3
versus 407.5 Å3), but their predicted maximal pKds (an index for ligandability assessment of
binding cavities, depending on the volume, lip size, hydrophobic volume, surface area, and
hydrogenꢀbond forming surface area of the cavity) are comparable (7.48 versus 7.82). Rigid
body docking was first performed with default parameters using the molecular docking program
Glide in SP mode28,29, to screen the SPECS compounds library (May 2013 version for 10 mg;
197,276 compounds). Flexible ligands and rigid receptor docking was then performed using the
Glide XP mode with default parameters to further screen the top 10,000 compounds from the SP
results. After this, the binding conformations of the top 2,000 compounds from Glide XP were
exported and manually selected. Finally, 251 compounds were purchased for experimental
testing. Correlation analysis between experimental pEC50 and calculated pKd is given in
Supporting Information, Figure S11 and Table S2.
Protein Expression and Purification. The recombinant mutant plasmid was transformed to
Escherchia coli strain M15. Recombinant cells were cultivated at 37 °C in LB medium
containing kanamycin (25 ꢁg/mL) and ampicillin (100 ꢁg/mL). At an OD600 of 0.6, GPX4
expression was induced by addition of IPTG (1 mM) and the cells were further cultured at 30 °C
ACS Paragon Plus Environment
21

Journal of Medicinal Chemistry
Page 22 of 36
1
2
3
4
5
6
7
8
9
for 5 h. The cells were harvested by centrifugation (6,000 rpm, 15 min), resuspended in lysis
buffer [TrisꢀHCl (100 mM, pH 8.0), NaCl (300 mM), imidazole (20 mM), TCEP (5 mM), and
PMSF (1 mM)] and incubated on ice with 1 mg/mL lysozyme for 30 min. After sonication, the
lysate was centrifuged (17,000 rpm, 40 min) and the supernatant was purified using a
nickel−nitrilotriacetic acid column (HisTrap HP; GE Healthcare) at 4 °C. The hisꢀtagged GPX4
was eluted with an increasing imidazole stepꢀgradient (100ꢀ250 mM). The eluted enzyme was
further subject to dialysis for 12 h [TrisꢀHCl (100 mM, pH 7.4), DTT (1 mM), and glycerol (20
% v/v)] to remove the bulk of imidazole. The final purity of the enzyme was more than 95 % as
confirmed by SDS−PAGE. Protein concentrations were measured using Nanodrop 2000
(Thermo Scientific, USA).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Activation of GPX4 in a CellꢀFree Assay. By coupling the oxidation of NADPH to NADP+ by
oxidized glutathione in the presence of glutathione reductase, GPX4 activity was assessed by
measuring the decrease in NADPH fluorescence emission at 460 nm (excitation at 335 nm). To
evaluate the effects of compounds on GPX4 activity, compounds were first preꢀincubated with
purified GPX4 mutants in the assay buffer [TrisꢀHCl (100 mM, pH 7.4), EDTA (5 mM), Triton
Xꢀ100 (0.1 % v/v), NADPH (0.2 mM), glutathione (3 mM), and glutathione reductase (1 unit)]
for 5 min at 37 °C. Then the reaction was initiated by adding tertꢀbutyl hydroperoxide (25 ꢁM).
Each compound was dissolved in DMSO at a final concentration of 5 %, which did not perturb
the assay. Fluorescence signals were recorded for 6 min on a plate reader (Synergy, BioTek).
The initial reaction rates at different concentrations of compound were used to determine EC50
values, which were calculated with a fourꢀparameter logistical model of log dose against
percentage activation or inhibition and were obtained from at least three sets of experiments.
Assay system without adding GPX4 was used to exclude the influence of the redox properties of
ACS Paragon Plus Environment
22

Page 23 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
compounds. Control experiments were also conducted to confirm that the compounds were not
glutathione reductase activators or inhibitors.
7
8
9
SPR Experiments. The binding affinities of compounds towards GPX4 were assayed using the
SPRꢀbased Biacore T200 instrument (GE Healthcare, Uppsala, Sweden). GPX4 was
immobilized on a CM5 sensor chip by using standard amineꢀcoupling at 25 °C with running
buffer PBSꢀP [phosphate buffer (20 mM, pH 7.4), NaCl (2.7 mM), KCl (137 mM), surfactant Pꢀ
20 (0.05 %)] as described previously.30 A reference flow cell was activated and blocked in the
absence of GPX4. In the direct binding experiments between GPX4 and compounds, GPX4
immobilization level was fixed at 800 response units (RU), and then different concentrations of
compounds containing 5 % DMSO were serially injected into the channel to evaluate binding
affinity. Regeneration was achieved by extended washing with the running buffer after each
sample injection. The equilibrium dissociation constants (KD) of the compounds were obtained
by fitting binding response units to Hill equation.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Prevention of ErastinꢀInduced Ferroptosis by GPX4 Activator in HTꢀ1080 Cells. HTꢀ1080
cells were cultured in DMEM containing 10 % FBS, 1 % supplemented nonꢀessential amino
acids and 1 % PenꢀStrep mixture (all from Gibco) and maintained in a humidified atmosphere
with 5 % CO2 at 37 °C. The day before the experiment, 2,000 cells/well were seeded in 96ꢀwell
plates (Corning). The day of the experiment, the medium was changed to fresh DMEM
containing different concentrations of GPX4 activator. After 30 min incubation at 37 °C in a
tissue culture incubator, erastin (10 ꢁM) was added to stimulate ferroptosis. After 24 hours, cell
viability was determined by MTS assay according to the instructions of the CellTiter 96®
AQueous One Solution Cell Proliferation Assay System (Promega). In brief, 10 ꢁL CellTiter 96®
AQueous One Solution Reagent was added to each well. Then cells were incubated at 37 °C for 1ꢀ
ACS Paragon Plus Environment
23

Journal of Medicinal Chemistry
Page 24 of 36
1
2
3
4
5
6
7
8
9
4 hours in a humidified, 5 % CO2 atmosphere. After incubation, the absorbance at 490 nm was
recorded with a BioTek synergy 4 MultiꢀMode Microplate Reader. The background absorbance
was subtracted from all reported measures. In addition to MTS assay, cell viability was assessed
by trypan blue exclusion test (Supporting Information, Figure S12).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Influence of the GPX4 Activator in Human Polymorphonuclear Leucocytes. Following the
reported procedure,3 human PMN cells were isolated from the venous blood of healthy
volunteers who had not received nonꢀsteroidal antiꢀinflammatory drugs (NSAIDs) for at least 14
days. Informed consent for participation was obtained from all the volunteers. PMNs were preꢀ
incubated with vehicle (DMSO) and GPX4 activator at 37 °C for 15 min, respectively. To
stimulate the PMNs, 10 ꢁM A23187, 2 mM CaCl2, and 500 ꢁM MgCl2 were added. After
incubation for 120 min at 37°C, reactions were terminated by the addition of cold methanol (700
ꢁL of methanol was added into 300 ꢁL of reaction solution). 15(S)ꢀHETEꢀd8 and PGB2 were
added as internal standards. The upper solvent was evaporated under a stream of nitrogen, and
the residue was dissolved in 150 ꢁL of methanol before LCꢀMS analysis.
LCꢀMS/MS Method. LC separation was performed with a Shimadzu Prominence UFLC XR
(Shimadzu Corporation). The multipleꢀreactionꢀmonitoring (MRM) spectra were obtained with a
QTRAP 5500 mass spectrometer (AB SCIEX) equipped with an ESI source. A Waters
Symmetry reverseꢀphase C18 column (2.1 mm × 150 mm, 3.5 ꢁm) was used for the LC
separation as described.15 Up to 35 eicosanoids and two internal standards [15(S)ꢀHETEꢀd8 and
PGB2] were monitored. Optimized LCꢀMS/MS parameters for the analysis of eicosanoids are in
accordance with previous research.15
ACS Paragon Plus Environment
24

Page 25 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
Inhibition of COXꢀ1 and COXꢀ2 in a CellꢀFree Assay. COXs can convert AA to PGG2, and
further reduce PGG2 to PGH2. The enzyme activities of COXs were determined
7
8
spectrophotometrically by monitoring the oxidation of N,N,N’,N’ꢀtetramethylꢀpꢀ
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
phenylenediamine (TMPD) at 610 nm during the conversion of PGG2 to PGH2. The test
compound was preꢀincubated with purified COXꢀ1 or COXꢀ2 for 15 min. Next, TMPD solution
was added, and the reaction was initiated by adding a solution of AA. Absorbance at 610 nm was
recorded on a plate reader (Synergy, BioTek). Inhibition activities were measured as described.31
Inhibition of mPGESꢀ1 in a CellꢀFree Assay. The enzyme activity of mPGESꢀ1 was measured
by assessment of PGH2 conversion to PGE2.32 The test compound was preꢀincubated with
enzyme sample for 15 min. Next, PGH2 was added to start the reaction. After 1 min, stop
solution was added to terminate the reaction. Production of PGE2 in the reaction mixture was
measured with the PGE2 EIA kit (Cayman Chemical). Inhibition was measured in a similar
manner as that of GPX4.
Inhibition of 5ꢀLOX in a CellꢀFree Assay. The enzyme activity of 5ꢀLOX was determined
spectrophotometrically by monitoring the oxidation of H2DCFDA to the highly fluorescent 2’,7’ꢀ
dichlorofluorescein during the catalytic reaction mediated by 5ꢀLOX as reported.33 The test
compound was preꢀincubated with enzyme for 10 min. Next, the reaction was initiated by the
addition of AA, and fluorescence signals (excitation at 500 nm and emission at 520 nm) were
recorded on a plate reader (Synergy, BioTek). The human 5ꢀLOX protein was prepared as
previous reported.34
Inhibition of LTA4H in a CellꢀFree Assay. LTA4H hydrolase activity was measured by
monitoring the formation of LTB4 with an enzymeꢀlinked immunosorbent assay. The test
ACS Paragon Plus Environment
25

Journal of Medicinal Chemistry
Page 26 of 36
1
2
3
4
5
6
7
8
9
compound was preꢀincubated with enzyme for 15 min. Next, LTA4 was added to initiate the
reaction. After 10 min, the reaction mixture was diluted to stop the reaction, and the production
of LTB4 was measured with the LTB4 ELISA kit (Cayman Chemical). Inhibition activities were
measured as described.35
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Activation and Inhibition of 12ꢀLOX and 15ꢀLOX in a CellꢀFree Assay. The enzyme
activities of 12ꢀLOX and 15ꢀLOX were assessed spectrophotometrically by measuring the
formation of 12ꢀHpETE and 15ꢀHpETE, respectively. The test compound was preꢀincubated
with the enzyme for 1 min. Next, AA was added to initiate the reaction. Absorbance of product
at 235 nm was monitored on a plate reader (Synergy, BioTek). Activation or Inhibition activities
were measured as described.15
Inhibition of the NFꢀκB Pathway by GPX4 Activator in Luciferase Activity Assay. All
plasmid DNAs were prepared using the TIANprep Mini Plasmid Kit (TIANGEN). HEK293T
cells were grown to 70 % confluency in 96ꢀwell plates (Corning) at 37 °C in DMEM
supplemented with 10 % FBS, and treated with EntransterTMꢀH (Engreen) transfection reagent
(0.1 ꢁL) and purified plasmids [0.25 ꢁg pGL4.74(hRluc/TK) and 0.25 ꢁg pGL4.32(luc2P/NFꢀ
κBꢀRE/Hygro plasmid)] in 50 ꢁL DMEM/10 % FBS per well. 24 hours later, GPX4 activator
was incubated with the cells (2×105 cells per well) for 30 min. After incubation, TNFα (5 ng/mL)
or ILꢀ1β (10 ng/mL) was added to stimulate the cells for 6 h or 24 h, respectively. Then the
luciferase assays were carried out using the DualꢀGlo Luciferase Assay System (Promega) with a
BioTek synergy 4 MultiꢀMode Microplate Reader.
Mutagenesis Experiments. All mutagenesis experiments were carried out according to the
instructions of the QuikChange SiteꢀDirected Mutagenesis (SBS Genetech Co., Ltd, Beijing,
ACS Paragon Plus Environment
26

Page 27 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
China). The pQEꢀ30 bacterial expression plasmid of the U46C mutant of human cytosolic GPX4
was mutated to obtain the mutants. The DNA sequences of all mutants were verified by DNA
sequencing. The protein expression and activity assays of the mutants were performed as
described for the U46C mutant.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
GPX4ꢀSpecific Activity Assay. GPX4ꢀspecific activity was measured in GPX4 KO MEFs reꢀ
constituted with a mouse WT GPX4 (WT) and a mock transduced control (KO). All cell lines
were cultured in the presence of αꢀToc (10 µM) as this would allow the KO cells to proliferate as
previously described.36 GPX4 specific activity was carried out in a coupled assay, by monitoring
the consumption of NADPH in the presence of glutathione reductase upon addition of
phosphatidyl choline hydroperoxide.37
Viability Assay. Cells were seeded onto 96ꢀwell plates (5,000 cells per well) and allowed to
attach overnight. The next day, the cells were preꢀtreated for one hour with the different GPX4
activators at a fixed concentration (200 µM). Upon that medium was changed and cells were
exposed to 20 µM ChOOH. After 24 hours viability was assessed using the AquaBluer assay as
previously described.7
Testing of Reducing Effects. The stable radical 2,2ꢀdiphenylꢀ1ꢀpicrylhydrazyl (DPPH) assay6
was used for the testing. The antioxidant trolox was used as the positive control.
Iron Chelating Ability Assay. The Prussian blue assay was used for the testing. The iron
chelators deferoxamine (DFO), ciclopirox olamine (CPX), and EDTA were used as the positive
control. 50 µL of each test compound dissolved in DMSO (10 mM) was added to 50 µL of
FeSO4 (3 mM) and incubated for 10 min. After incubation, 50 µL of K3[Fe(CN)6] (2 mM) was
ACS Paragon Plus Environment
27

Journal of Medicinal Chemistry
Page 28 of 36
1
2
3
4
5
6
added and the absorbance at 680 nm was recorded with a BioTek synergy 4 MultiꢀMode
Microplate Reader.
7
8
9
Zinc Chelating Ability Assay. Zinc reacts with thiocyanate and crystal violet to form a ternary
ionꢀassociation complex, with maximum absorption at 522 nm. The zinc chelators EDTA and
thiourea were used as the positive control. 50 µL of each test compound dissolved in DMSO (10
mM) was added to 50 µL of ZnSO4 (2.5 mM) and incubated for 10 min. After incubation, 50 µL
of crystal violet (5 mg/mL) and 100 µL of NH4SCN (0.25g/mL) were added and the absorbance
at 522 nm was recorded with a BioTek synergy 4 MultiꢀMode Microplate Reader.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ASSOCIATED CONTENT
Supporting Information
The following files are available free of charge.
Figures S1ꢀS12, Table S1, and spectra of compounds (PDF)
Docked complexes of GPX4 with compound 1 and 1d4 (PDB)
Molecular formula strings (CSV)
AUTHOR INFORMATION
Corresponding Authors
*Eꢀmail: lhlai@pku.edu.cn.
*Eꢀmail: pedro.angeli@virchow.uniꢀwuerzburg.de.
Author Contributions
C.L. and X.D. have equal contributions of this work.
ACS Paragon Plus Environment
28

Page 29 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
L.L. and C.L. conceived the project. C.L. performed allosteric site prediction and virtual screen,
protein expression and purification, enzymatic assays, SPR experiments, cellꢀbased assays, and
mutagenesis. X.D. performed chemical synthesis and purity analysis of the compounds. X.D. and
Y.L. contributed to the SAR study. J.P.F.A. and M.C. provided reagents, conducted GPX4ꢀ
specific activity assay and ChOOH viability assay. W.Z. performed molecular dynamics
simulations. X.X. participated in cellꢀfree assay of GPX4 and cell experiments. C.L., X.D.,
J.P.F.A., and L.L. wrote the manuscript.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENT
This work was supported in part by the Ministry of Science and Technology (2016YFA0502303,
2015CB910300 to L.L.), the National Natural Science Foundation of China (21633001 to L.L.),
the Rudolf Virchow Jr group leader program from the University of Würzburg (JPFA) and the
Human Frontier Science Program (HFSP) (RGP0013 to M.C.). The authors acknowledge Dr. H.
Kuhn (University Medicine BerlinꢀCharité) for sharing the GPX4 bacterial expression plasmid
and Mrs Irina Ingold (Helmholtz Zentrum München) for the generation of phosphatidyl choline
hydroperoxide.. We thank Dr. C. Wang (Peking University) for helpful discussions and for
providing the HTꢀ1080 cells. We also appreciate Dr. J. Zhao (Peking University) for help with
cell culture and discussions.
ABBREVIATIONS
AA, arachidonic acid; ACSL4, acylꢀCoA synthetase longꢀchain family member 4; ChOOH,
cholesterol hydroperoxide; COXꢀ1, cyclooxygenase 1; COXꢀ2, cyclooxygenase 2; CPX,
ACS Paragon Plus Environment
29