A masked zinger to block GPX4

Abstract

Ferroptosis induced by GPX4 inhibition offers promise for killing drug-resistant cancer cells, yet current GPX4 inhibitors lack selectivity. The discovery of masked nitrile oxide electrophiles as selective prodrug inhibitors of GPX4 points to an attractive path for chemically inducing ferroptosis.

In the popular television series, The Masked Singer, celebrities engage in a singing contest with their identities hidden throughout by adorning head-to-toe costumes. While silly, to say the least, the series premise also provides an opportunity for the audience to engage in creative sleuth work to deduce the identities of disguised singers, which is a key part of the show’s appeal. The mechanistic characterization of bioactive small molecules discovered by phenotypic screening can also present its own mysteries that challenge the intuition of organic and biological chemists, especially when the parent hit compound is found to differ from the actual product causing the cellular phenotype of interest. In this issue, Eaton et al. describe a fascinating set of studies that illuminate the mechanism of action of an unusual class of cellular inhibitors of the lipid hydroperoxide reductase GPX4¹. The manuscript, which reads like a can’t-put-it-down detective novel, culminates in the discovery of masked nitrile-oxide electrophiles as a new class of prodrugs that offer a superior way to covalently inhibit GXP4 and induce ferroptosis in cancer cells.

Ferroptosis is a nonapoptotic, iron-dependent form of cell death² and a feature of cancer cells that show drug-resistance³ and that have undergone an epithelial-to-mesenchymal transition⁴. Ferroptosis can be induced in several ways, but most directly by the inhibition of GPX4⁵, which is a selenoprotein that detoxifies lipid hydroperoxides by reducing them to lipid alcohols (Fig. 1a). Several GPX4 inhibitors have been described, most of which represent alpha-chloroacetamide electrophiles that alkylate the catalytic selenocysteine residue in the GPX4 active site⁶. This category of covalent inhibitors, however, is limited by generally poor proteomic selectivity and pharmacokinetic properties. One exceptional type of GPX4 inhibitor is the compound ML210, which lacks an alpha-chloroacetamide or other obvious electrophilic group⁷ (Fig. 1b).

Figure 1.

(a) Ferroptotic cell death can be induced by covalent inhibition of GPX4, a lipid peroxidase enzyme that reduces lipid peroxides to non-toxic alcohols in cells. (b) The prodrug mechanism of action for ML210, which is converted to a nitrile oxide electrophile JKE-1777 in cells that reacts covalently with GPX4.

ML210 was originally identified by phenotypic screening as an anti-cancer, ferroptosis-inducing compound⁷, but its molecular mechanism of action has remained enigmatic. Despite lacking an apparent electrophilic moiety, ML210 was shown by Eaton et al. to covalently modify the catalytic selenocysteine in GPX4 in cells, producing a protein adduct that was 41 Da lower in molecular weight than that expected for the parent ML210 compound. Furthermore, an alkyne probe analog of ML210 enriched GPX4 from cancer cells and revealed that this compound had fewer off targets than alpha-chloroacetamide inhibitors of GPX4.

Surprisingly, ML210 did not form covalent adducts with purified GPX4, leading Eaton et al. to hypothesis that the compound acts as a cloaked electrophilic prodrug requiring metabolic activation in cells. The chemical transformation that ML210 undergoes in cells was deduced by a remarkably clever series of experiments involving iterative chemical synthesis-driven explorations of the structure-activity relationship of ML210 analogs. A key clue was the requirement of the nitroisoxazole ring for activity, a chemical group that can undergo ring opening to a α-nitroketoxime under basic conditions⁸. JKE-1674, an analog of ML210 in which the nitroisoxazole ring was replaced with an α-nitroketoxime, was synthesized and found to induce ferroptosis and covalently label GPX4 in cells. However, the investigative work of the authors was not done yet, as JKE-1674 also did not react with purified GPX4, indicating that a further chemical transformation occurred in cells to generate the relevant bioactive electrophile.

Subsequent mechanistic studies revealed that the α-nitroketoxime group of JKE-1674 can be converted into a nitrile oxide JKE-1777 (Fig. 1b), which proved capable of reacting with purified GPX4. The molecular weight of the adduct formed by the reaction of JKE-1777 with the selenocysteine of GPX4 was identical to the adduct that was observed for ML210 in cells. Additional analogs of ML210 containing substituents, such as a nitrolic acid, that spontaneously generate nitrile oxides displayed similar activity to ML210, providing further evidence that a nitrile oxide electrophile is responsible for the GPX4-inhibitory and ferroptosis-inducing activity of ML210.

The enhanced selectivity and stability of compounds such as ML210 and JKE-1674 enabled experiments that were not possible with original alpha-chloroacetamide GPX4 inhibitors. For instance, a CRISPR-based suppressor screen using ML210 identified genes essential for ferroptosis that were not previously discovered in screens with other ferroptosis-inducing compounds. And, as a hint of experiments to come, Eaton et al. showed that mice could be orally dosed with JKE-1674 and the compound detected in serum for up to 24 hours.

The precise mechanism by which the nitroisoxazole group of ML210 is converted into a reactive nitrile oxide in a cellular context remains unknown. Identifying the conditions or enzymes, if any, responsible for this conversion may represent a promising avenue of future research. In particular, if enzymatic pathways with greater expression in cancer cells can be leveraged to generate nitrile oxide inhibitors of GPX4 from prodrug precursors, this could allow for more selective induction of ferroptosis, which remains a persistent translational concern, given that GPX4 inactivation has the potential to produce toxic effects in organs such as the kidney⁹.

We also point out that one of the original clues motivating Eaton et al. to investigate ML210 was its activity in a screen of > 800 human cancer cell lines, where the compound produced a pattern of cell killing that closely resembled other GPX4 inhibitors (http://portals.broadinstitute.org/ctrp.v2.1). This finding, and others¹⁰, underscores the power of large-scale, cell-based screening for illuminating unexpected mechanistic relationships across structurally diverse compound classes. While the molecular evidence for these hidden connectivities may only be uncovered by diving into the complex context of the human cell, Eaton et al. demonstrate how the analytical tools available to the modern chemical detective can make such an ambitious mystery tour well worth taking. Finally, we wonder whether further investigation of nitroisoxazoles and α-nitroketoximes as masked electrophiles may reveal the potential to develop selective ligands of other (seleno)cysteine proteins, thereby further expanding the scope of covalent chemistry for probing the human proteome.