ZX703: A Small-Molecule Degrader of GPX4 Inducing Ferroptosis in Human Cancer Cells

Abstract

Ferroptosis is a novel form of oxidative cell death triggered by iron-dependent lipid peroxidation. The induction of ferroptosis presents an attractive therapeutic strategy for human diseases, such as prostate cancer and breast cancer. Herein, we describe our design, synthesis, and biological evaluation of endogenous glutathione peroxidase 4 (GPX4) degraders using the proteolysis targeting chimera (PROTAC) approach with the aim of inducing ferroptosis in cancer cells. Our efforts led to the discovery of compound 5i (ZX703), which significantly degraded GPX4 through the ubiquitin–proteasome and the autophagy–lysosome pathways in a dose- and time-dependent manner. Moreover, 5i was found to induce the accumulation of lipid reactive oxygen species (ROS) in HT1080 cells, thereby inducing ferroptosis. This study provides an attractive intervention strategy for ferroptosis-related diseases.

Ferroptosis, proposed by Stockwell in 2012, is an oxidative, non-apoptotic type of regulated cell death characterized by the iron-dependent accumulation of lipid peroxides.¹ Ferroptosis, plays a vital role in various human diseases, including tumorigenesis, nervous system diseases, hematological system diseases, and more.²⁻⁵ It has been established that ferroptosis can be triggered by several mechanisms including labile iron levels, modulation of the intracellular glutathione (GSH) pool, polyunsaturated fatty acids (PUFAs), and the activity of glutathione peroxidase 4 (GPX4).^1,6,7

GPX4, a glutathione-dependent selenoenzyme, is the central regulator of ferroptosis, converting lipid hydroperoxides to lipid alcohols.^2,3 Elevated GPX4 expression level is observed in a wide spectrum of cancer cells, and inhibition of GPX4 activity or GPX4 knockout strategies has shown efficacy in suppressing tumor growth. Therefore, GPX4 represents a promising target for treating drug-resistant cancers by inducing ferroptosis.⁸ However, targeting GPX4 is challenging due to its shallow active site, limiting interactions with small molecules.^9,10 Current GPX4 inhibitors are covalent, primarily targeting the active-site selenocysteine, which poses a risk of off-target effects and lacks cell-type specificity. To date, only a limited number of GPX4 inhibitors have been reported, including RSL3, ML162, and ML210 (Figure 1).^11,12

Figure 1.

Chemical structures of previously reported GPX4 inhibitors and degraders.

In recent years, targeted protein degradation such as proteolysis targeting chimera (PROTAC) technology has emerged as a promising technology. PROTACs are heterobifunctional molecules that consist of two moieties, one binds to the protein of interest (POI) and the other binds to an E3 ubiquitin ligase, connected by an appropriate linker.¹³ Upon binding, PROTACs facilitate the formation of ternary complexes between POI and E3 ligase, leading to polyubiquitination of the POI and subsequent degradation by the ubiquitin–proteasome system (UPS).^14,15 Unlike traditional small molecules, PROTACs work at lower doses due to their catalytic protein degradation mechanism. Furthermore, extended and sustained use of traditional small molecules inevitably leads to a significant decrease in affinity and the development of acquired drug resistance. PROTACs offer a solution to drug resistance problems by completely eliminating the target mechanism.¹⁶ In recent years, GPX4 degraders have been developed utilizing various different classes of GPX4 ligands, including ML162, ML210, and RSL3.¹⁸⁻²² Most reported GPX4 degraders were designed employing cereblon (CRBN) as an E3 ligase, as exemplified by 1 (dGPX4), 2 (PD-4), and 3 (DC-2) (Figure 1). dGPX4 was first reported as a potent degrader of GPX4, albeit with a poor DC₅₀ of 200 nM. Subsequently, the Ge group²¹ repurposed ML162 to discover GPX4 PROTAC PD-4, which efficiently induced GPX4 degradation but with a less favorable IC₅₀ value of 0.86 μM. Compound 4 was an RSL3-based GPX4 degrader utilizing Von Hippel–Lindau (VHL) as an E3 ligase, which showed excellent degradation and antiproliferative activity. However, RSL3 exhibits poor pharmacokinetic properties, which hinders its use in vivo.^10,23 Thus, we hypothesized that using a GPX4 inhibitor ML210 with favorable drug-like properties would enhance the degradation of the resulting PROTAC. While preparing this manuscript, Xu et al. reported their ML210-based GPX4 PROTAC DC-2 by employing the E3 ligase CRBN. While the majority of PROTACs progressing into clinical development are linked to CRBN ligands, it is important to note the advantages of VHL ligands. VHL ligands have demonstrated greater stability, as evidenced by the decomposition observed in thalidomide-based PROTACs even in mild PBS buffer conditions.^24,25 Additionally, VHL E3 ligase exhibits varying expression levels in different cell types and lacks any known neo-substrates.²⁶

In this study, we describe the synthesis and characterization of our ML210-based GPX4 PROTACs that utilize the commonly recruited E3 ligase VHL. Through systematic structural optimization of the linker length and composition, we identified 5i (ZX703) as a potent GPX4 degrader with potential for treating drug-resistant cancers by inducing ferroptosis.

To design the ML210-based GPX4 PROTAC, we began by examining the binding model of ML210 in the complex with GPX4 (PDB: 6HKQ). Our analysis revealed that the chlorophenyl moiety in ML210 is solvent-exposed, rendering it a suitable attachment site for connecting with the E3 ligase ligand (Figure 2A). Ligands for VHL have demonstrated successful applications in developing PROTAC degraders for a wide range of protein targets.^27,28 Accordingly, a series of potential PROTAC degraders targeting GPX4 were synthesized by utilizing ML210 and VHL, connected through various linker types (Figure 2B).

Figure 2.

Design of GPX4 PROTACs based on VHL ligands. (A) Docking models of compound ML210 with GPX4 (PDB: 6HKQ). (B) Design of new GPX4 PROTACs based on VHL.

The synthesis of compounds 5a–l is shown in Scheme 1. Commercially available (4-chlorophenyl)(4-hydroxyphenyl)methanone (8) was directly reacted with various alkyl bromides to produce compounds 11a and 11c–k with a terminal tert-butyl ester group. The substitution reaction of 3-bromopropan-1-ol with compound 8 furnished compound 9. Direct oxidation of 9 gave carboxylic acid 10 that was protected by treatment with tert-butanol to obtain 11b. 11a–k were reduced by NaBH₄ to obtain 12a–k, which were successively reacted with oxalyl chloride and piperazine to produce the desired compounds 14a–k. Nitration of 5-methylisoxazole-3-carboxylic acid (15) was accomplished using concentrated sulfuric acid and potassium nitrate. The obtained compound 16 was converted to the corresponding acyl chloride 17. Acylation of compounds 14a–k with 17 afforded compounds 18a–k. Subsequently, 18a–k were deprotected by trifluoroacetic acid (TFA) to obtain 19a–k, followed by amide coupling with VHL ligands to produce the final products 5a–l.

Scheme 1. Synthesis of GPX4 Degraders 5a–l.

Reagents and conditions: (a) 3-bromopropan-1-ol, K₂CO₃, DMF, 80 °C, overnight, 72%; (b) 2,2,6,6-tetramethylpiperidinyloxy (TEMPO), iodobenzene diacetate (BAIB), DCM, H₂O, rt, overnight, 87%; (c) SOCl₂, t-BuOH, 90 °C, overnight, 48%; (d) tert-butyl bromoalkylate of different lengths, K₂CO₃, DMF, 80 °C, overnight, 50–99%; (e) NaBH₄, THF/MeOH, 0 °C, 1.5 h, 94–99%; (f) oxalyl chloride, DCM, DMF, 0 °C, overnight, 94–99%; (g) piperazine, CH₃CN, 90 °C, overnight, 48–76%; (h) KNO₃, H₂SO₄, 50 °C, overnight, 81%; (i) oxalyl chloride, DCM, DMF, rt, overnight, 99%; (j) TEA, DCM, rt, overnight, 53–78%; (k) TFA, DCM, 0 °C, overnight, 100%; (l) (2S,4R)-1-((S)-2-amino-3,3-dimethylbutanoyl)-4-hydroxy-N-(4-(4-methylthiazol-5-yl)benzyl)pyrrolidine-2-carboxamide (20) or (2S,4R)-1-((S)-2-amino-3,3-dimethylbutanoyl)-4-hydroxy-N-((S)-1-(4-(4-methylthiazol-5-yl)phenyl)ethyl)pyrrolidine-2-carboxamide, HATU, DIPEA, DCM, 0 °C, overnight, 13–35%.

Compounds 6a–d were synthesized as described in Scheme 2. Compounds 22a–d were obtained by the substitution reaction of 21a–d with tert-butyl bromoacetate. The benzyl group in 22a–d was removed to give 23a–d, followed by reaction with pTsCl to obtain 24a–d. Commercially available 8 was reacted with compounds 24a–d via nucleophilic substitution to furnish 25a–d. Finally, compounds 6a–d were synthesized in a manner similar to that previously described for the synthesis of compound 5a.

Scheme 2. Synthesis of GPX4 Degraders 6a–d.

Reagents and conditions: (a) benzyl-substituted ethylene glycol chains of different lengths, t-BuOK, t-BuOH, rt, overnight, 50–55%; (b) Pd/C, H₂, EtOH, rt, 3 h, 97–99%; (c) TsCl, DMAP, TEA, DCM, rt, overnight, 57–87%; (d) compound 8, K₂CO₃, acetone, 65 °C, overnight, 62–98%; (e) NaBH₄, THF/MeOH, 0 °C, 1.5 h, 93–99%; (f) oxalyl chloride, DCM, DMF, 0 °C, overnight, 92–99%; (g) piperazine, CH₃CN, 90 °C, overnight, 52–67%; (h) compound 17, TEA, DCM, rt, overnight, 69–97%; (i) 1) TFA, DCM, 0 °C, overnight, 100%, 2) compound 20, HATU, DIPEA, 0 °C, overnight, 9–47%.

The syntheses of compounds 7a–c are outlined in Schemes 3 and 4. As shown in Scheme 3, commercially available 2-(4-(tert-butoxycarbonyl)piperazin-1-yl)acetic acid or 1-(tert-butoxycarbonyl)piperidine-4-carboxylic acid was amidated with a VHL ligand (20) to give compounds 30a,b. Under acidic conditions, compounds 30a,b were deprotected to afford 31a,b. 32a,b were synthesized by the reductive amination of 31a,b with tert-butyl 4-formylpiperidine-1-carboxylate. Boc-deprotection and coupling with acid 19a gave 7a,b. As described in Scheme 4, intermediate 33 was synthesized by the reductive amination of compound 20 with tert-butyl 2-oxo-7-azaspiro[3.5]nonane-7-carboxylate and then was deprotected under acidic conditions to afford 34. Amide coupling between compounds 34 and 19a gave the final compound 7c.

Scheme 3. Synthesis of GPX4 Degraders 7a,b.

Reagents and conditions: (a) HATU, DIPEA, DMF, rt, overnight, 61–74%; (b) HCl (2 M), dioxane, 50 °C, 3 h, 81%; (c) tert-butyl 4-formylpiperidine-1-carboxylate, NaBH(OAc)₃, DCE, AcOH, rt, overnight, 68–78%; (d) 1) HCl (2 M), dioxane, 50 °C, 3 h, 100%, 2) compound 19a, HATU, DIPEA, DCM, 0 °C, overnight, 18–57%.

Scheme 4. Synthesis of GPX4 Degrader 7c.

Reagents and conditions: (a) tert-butyl 2-oxo-7-azaspiro[3.5]nonane-7-carboxylate, NaBH(OAc)₃, DCE, AcOH, rt, overnight, 53%; (b) HCl (2 M), dioxane, 50 °C, 3 h, 80%; (c) compound 19a, HATU, DIPEA, DCM, 0 °C, overnight, 20%.

We designed and synthesized 5a as a potential GPX4 degrader, incorporating ML210 as the GPX4 ligand and VHL as the E3 ligand and employing a linear linker consisting of a single methylene unit. In a cell growth assay, compound 5a exhibited an IC₅₀ value of 2.137 μM in the HT1080 cell line and was approximately 14 times less potent than ML210. Prior studies have demonstrated that the length of PROTAC linkers influences potency and selectivity.²⁹ Therefore, we first aimed to establish the optimal linker length in 5a for cellular potencies by incrementally lengthening the linker by one methylene group, yielding compounds 5b–k (Table 1). Substituting the methylene group in 5a with an ethylene group generated 5b, exhibiting an IC₅₀ value of 2.165 μM in cell growth inhibition and displaying potency almost identical to that of 5a. However, compound 5c, characterized by a three-methylene group, is inactive. Expanding the linker in 5a by three, four, five, or six additional methylene groups yielded 5d–g, without obviously affecting cellular potency. Further investigation revealed that elongating the linker in 5g by one additional methylene group yielded 5h with an IC₅₀ value of 0.767 μM, which was ∼2 times more potent than that of 5g. Encouragingly, extending the linker in 5g by one ethylene group yielded compound 5i with an IC₅₀ value of 0.435 μM in inhibiting HT1080 cell growth, which was approximately 4-fold more potent than 5g. Extending the linker in 5i by one methylene or ethylene group generated 5j or 5k, respectively. Compound 5j exhibited a potency approximately 2-fold lower than that of 5i, while 5k was roughly 5-fold less potent. These data suggested that a nine-methylene unit linker represents the optimal length for GPX4 degraders. Having identified the optimal linker length, we then investigated the modification of the VHL ligand composition in compound 5i (Table 1). Introducing an (S)-methyl group into compound 5i, yielded compound 5l with an IC₅₀ value of 1.210 μM in HT1080 cell growth inhibition, which was ∼3 folds less potent than 5i.

Table 1. Proliferation Inhibition of GPX4 Degraders with Alkyl Chains in HT1080 Cells^a.

Compound	Linker (n)	R	IC50 (μM) in HT1080 cell growth inhibition
ML210	–	–	0.153 ± 0.007
5a	1	H	2.137 ± 0.282
5b	2	H	2.165 ± 0.248
5c	3	H	NA
5d	4	H	1.543 ± 0.124
5e	5	H	2.294 ± 0.162
5f	6	H	1.361 ± 0.142
5g	7	H	1.685 ± 0.054
5h	8	H	0.767 ± 0.067
5i	9	H	0.435 ± 0.021
5j	10	H	0.889 ± 0.045
5k	11	H	2.352 ± 0.067
5l	9	Me	1.210 ± 0.052

^aCells were treated with compounds for 24 h, and viability was assessed with a Cell Counting Kit-8 (CCK-8) assay.

Besides the linker length, the linker type also significantly influence the degradation potency of PROTAC compounds.³⁰⁻³² Due to the hydrophobic nature of the linkers in compounds 5a–k, we incorporated polar groups into the linker to increase solubility. Consequently, we synthesized several potential GPX4 degraders with polyethylene glycol (PEG) linkers. Interestingly, the resulting compounds 6a–d exhibited significantly lower potency compared to compound 5i, with IC₅₀ values ranged from 1.830 to 4.733 μM (Table 2).

Table 2. Proliferation Inhibition of GPX4 Degraders with PEG Chain and Rigid Linkers in HT1080 Cells^a.

^aCells were treated with compounds for 24 h, and viability was assessed with a Cell Counting Kit-8 (CCK-8) assay.

We next introduced conformational restrictions into the linker, yielding compounds 7a and 7b (Table 2). In contrast to degraders containing flexible linkers, compounds 7a and 7b showed potency much weaker than that of 5i, with IC₅₀ values of 4.765 and 4.454 μM, respectively. We also synthesized compound 7c by incorporating a rigid spiro ring system into the linker (Table 2). Compound 7c with a [3.5]azaspiro ring did not improve the antiproliferation effect. These results indicated that 5i exhibited the strongest proliferation inhibitory effect among all of the compounds. Notably, 5i also demonstrated potent inhibition of cell growth in DU145 cells, with an IC₅₀ value of 0.369 μM (Figure S1). Thus, 5i was selected as a representative compound for further investigation.

Encouraged by the promising cellular potency of compound 5i, we next examined its ability to induce GPX4 degradation in the HT1080 cell line. As shown in Figure 3A,B, 5i effectively induced GPX4 degradation in a dose-dependent manner with a D_max value of 86% and a DC₅₀ value of 0.135 μM.

Figure 3.

Compound 5i induced effective GPX4 degradation. (A) Compound 5i degrades the GPX4 protein in a dose-dependent manner. (B) Quantification of GPX4 levels in (A) by using gray-scale analysis, with the calculation of DC₅₀ values from the fitted curve. (C) Degradation kinetics of 5i in the HT1080 cell line. Cells were treated with 0.2 μM 5i at different times. (D) Quantification of GPX4 levels in (C) by using gray-scale analysis. GPX4 protein was examined by Western blotting, and the protein level was quantified by densitometry and normalized to the corresponding density of the GAPDH protein.

Subsequently, we investigated the kinetics of GPX4 degradation induced by 5i in HT1080 cell lines (Figure 3C,D). Our findings revealed that ∼50% of GPX4 was degraded after 6 h treatment with 5i. Remarkably, over 80% of the GPX4 protein was effectively degraded after 12 h. The results also showed that the degradation activity of 5i lasted for at least 24 h. These data suggested that 5i was a potent GPX4 degrader.

The degradation mechanism of GPX4 in HT1080 cells by 5i was investigated through Western blot analysis (Figure 4). Since ML210 has been shown to degrade GPX4 (Figure 3A), the use of the VHL ligand (VHL032) could not completely block 5i-induced GPX4 degradation. Our data showed that degradation of GPX4 was only partially attenuated by pretreatment with the proteasome inhibitor MG132, which indicated that 5i may also degrade GPX4 through other pathways besides the UPS. We pretreated the cells with the lysosome inhibitor chloroquine (CQ) and found that the reduction of GPX4 induced by 5i could be partially inhibited. When the cells were co-treated with MG132 and CQ, the degradation of GPX4 was almost completely blocked. These mechanistic data demonstrated that 5i functions as a GPX4 degrader based on both the UPS and autophagy–lysosome pathways.

Figure 4.

Compound 5i induces GPX4 degradation via both the ubiquitin–proteasome-mediated proteolysis process and the autophagy–lysosome pathway. (A) Western blot analysis of GPX4 in HT1080 cells following a 12 h treatment of 5i with or without MG132 (0.3 μM), CQ (50 μM), or VHL032 (10 μM). (B) GPX4 levels in various groups in (A) were quantified using gray-scale analysis.

To elucidate the effects of 5i on cancer cell ferroptosis, we analyzed the change of intracellular ROS level in HT1080 cells following treatment with 5i treatment. Flow cytometry analysis using the DCFH-DA assay was performed to measure the change of ROS levels in cells treated with increasing concentrations of compound 5i (Figure 5). The results showed that 5i induced ROS accumulation in a dose-dependent manner. Furthermore, pretreatment of HT1080 cells with the ferroptosis inhibitor Fer-1 for 2 h effectively rescued cell death induced by 5i.

Figure 5.

Compound 5i induces ROS accumulation. Flow cytometry analysis of ROS level in HT1080 cells treated with ML210, various concentrations of 5i, or compound 5i in the presence of 5 μM Fer-1 for 12 h, with pretreatment with 10 μM 2′,7′-dichlorodihydrofluorescein diacetate (DCFH-DA).

Upon the basis of a potent GPX4 inhibitor ML210, we developed a series of small-molecule VHL-based PROTAC degraders targeting the GPX4 protein. After extensive optimization of the linker length, composition, and rigidity, we successfully obtained a potent GPX4 protein degrader, compound 5i, with DC₅₀ value of 0.135 μM. Mechanistic investigation suggested that 5i degrades GPX4 protein by both the ubiquitin–proteasome pathway and the autophagy–lysosome pathway. The kinetic study demonstrated that compound 5i rapidly degraded the GPX4 protein within a few hours. Moreover, the DCFH-DA assay suggested that 5i can effectively accumulate ROS, subsequently leading to ferroptosis in HT1080 cells. Taken together, 5i is an efficacious GPX4 degrader and a promising lead compound for developing a novel therapy aimed at inducing ferroptosis in cancer cells.

Glossary

Abbreviations

IC₅₀

half-maximal inhibitory concentration

DC₅₀

half-maximal degradation concentration

GPX4

glutathione peroxidase 4

GAPDH

glyceraldehyde-3-phosphate dehydrogenase

UPS

ubiquitin–proteasome system

ROS

reactive oxygen species

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.3c00571.

• Figure S1, 5i treatment resulted in a significant death of DU145 cells; materials and methods; ¹H and ¹³C NMR, HPLC, and HRMS spectra of compounds 5a–l, 6a–d, and 7a–c (PDF)

Author Contributions

^# M.H. and X.L.contributed equally. Q.X. and D.Z. designed the research; M.H., X.L., L.W., Y.Z., Y.S., H.H., H.L., and T.C. conducted the research; M.H., X.L., Q.X., and D.Z. analyzed the data and wrote the paper. All authors read and approved the final manuscript.

The authors declare no competing financial interest.