Rational design of tetrahydrotetrazinones as potent nonclassical radical-trapping antioxidants for inhibition of ferroptosis

Abstract

Radical-trapping antioxidant (RTA) protection plays a critical role in industrial, biological, and pharmaceutical contexts. The cytoprotective role of RTAs has been increasingly recognized as essential in regulating ferroptosis, a form of regulated cell death driven by lipid peroxidation. However, current ferroptosis inhibitors primarily rely on phenolic or aromatic amine scaffolds, which were found through high-throughput screening or derived from historically known RTAs. We introduce hydrazine-derived tetrahydro-1,2,4,5-tetrazin-3-ones (THTZOs) as a class of nonclassical RTAs. THTZOs operate via a unique mechanism involving hydrazyl and verdazyl radical intermediates that is independent of aromatic stabilization. These compounds demonstrate broad reactivity against autoxidation in both simple hydrocarbons and membrane phospholipids, effectively suppressing ferroptosis induced by RAS-selective lethal 3 (RSL3) or erastin in multiple cell lines. Our findings revitalize the application of hydrazine derivatives in RTA chemistry and establish a foundation for the rational design of RTAs as potent ferroptosis inhibitors, offering previously unidentified antioxidant therapeutic candidates.

Rational design of tetrahydrotetrazinones as potent nonclassical radical-trapping antioxidants for inhibition of ferroptosis.

INTRODUCTION

Autoxidation is a radical chain reaction—comprising initiation, propagation, branching, and termination steps—which spontaneously converts hydrocarbons and oxygen into hydroperoxides: R-H + O₂ → R-OO-H. This process causes deleterious effects in diverse contexts, including the degradation of petroleum-derived products (e.g., lubricants, fuels, and plastics), plant-based compounds, and critical biomolecules such as lipids, proteins, and nucleic acids (1–3). Radical-trapping antioxidants (RTAs) inhibit autoxidation by intercepting propagating peroxyl radicals (ROO•) via hydrogen atom transfer (HAT) or radical adduct formation, thereby diverting chain propagation into termination pathways and mitigating oxidative damage (Fig. 1A) (4, 5).

Fig. 1. Design of THTZO RTAs.

(A) Phenols and aromatic amines represent two main classes of RTAs. Their aromatic moieties play a critical role in facilitating these molecules to serve as effective H-atom donors for radical-trapping activity. (B) Incorporation of an additional heteroatom weakens the X─H bond in hydroperoxide, hydropersulfide, hydroxylamine, and hydrazine derivatives relative to their alcohol, thiol, and amine counterparts. This bond weakening results from spin delocalization and polarized resonance with adjacent X atoms in the perXyl radicals (where X = O for ROO•, X = S for RSS•, X = NO for R¹R²NO•, and X = N for R¹R²NR³N•). (C) THTZOs containing a hydrazine scaffold demonstrate potent RTA through efficient HAT to peroxyl radicals (ROO•), in contrast with the unfavorable reactivity between aliphatic amines and ROO•. This unique reactivity profile establishes THTZOs as effective inhibitors of (phospho)lipid peroxidation and promising suppressors of ferroptosis in this work.

(Phospho)lipid peroxidation (i.e., autoxidation) is a key driver of ferroptosis, a form of regulated cell death implicated under severe pathological conditions, including ischemia-reperfusion injury, degenerative disease, and cancer (6, 7). Beyond iron and lipid metabolism, the antioxidant defense system acts as a primary gatekeeper in regulating ferroptosis (8–10). Both endogenous and synthetic lipophilic RTAs are essential for viability as suppressors of this peroxidative cell death (11–15). The characterization of prototypical antiferroptotic agents—such as α-tocopherol (α-TOH) (7), ferrostatin-1 (Fer-1) (7), and liproxstatin-1 (Lip-1) (16)—has provided fundamental insights into the structure-activity relationship (SAR) of phenolic and aromatic amine inhibitors (17, 18). Mechanistically, to efficiently quench a peroxyl radicals (ROO•) via HAT, an RTA’s X─H bond must have a bond dissociation enthalpy (BDE) lower than that of the O─H bond in ROOH (BDE ~ 88 kcal/mol) (19). Aromatic rings attenuate the O─H or N─H bond strength through π-conjugation and radical stabilization; consequently, their antioxidant efficacy is highly dependent on the electronic structure of the aromatic system. However, the limitations of current RTA candidates—including insufficient potency [often exhibiting only millimolar-range median effective concentration (EC₅₀) values as for phenols], poor metabolic stability, and a propensity to form reactive quinone metabolites that cause hepatotoxicity—highlight the urgent need for innovative scaffold design (17, 20). Extensive research has focused on refining ferrostatin (21–27) and phenoxazine derivatives (28, 29), spurring substantial medicinal chemistry efforts (Fig. 1A). Nevertheless, the expansion into new RTA architectures remains a crucial research priority for developing novel molecule scaffolds that can therapeutically modulate ferroptosis (30). Recent efforts have begun to address this need by identifying noncanonical RTA scaffolds, such as metal bis(thiosemicarbazonates) (31, 32) and naphthalene-1,8-diamide arylboronic esters [ArB(DAN)] (33).

Nonaromatic amines are generally disregarded as scaffolds for effective RTAs. The aliphatic aminyl radicals (R₂N•) exhibit poor stabilization of the unpaired electron, leading to high N─H BDEs (BDE ~ 95 kcal/mol) that render HAT to ROO• thermodynamically unfavorable (34). In stark contrast to these transient R₂N• species, the hydrazyl radical (R¹R²N─N•R³) benefits from significant spin delocalization facilitated by the N─N bond and lone-pair interactions, thereby enhancing its persistence and thermodynamic stability (35, 36). This electronic configuration positions hydrazines as functional analogs to hydroxylamines (37) and hydropersulfides (38), which also have activated X─H bonds capable of donating hydrogen atoms to peroxyl radicals (Fig. 1B) (15, 39–43). Diverging from traditional strategies focused on modulating the electronic properties of aromatic systems, we instead envisioned hydrazines as previously unidentified RTA templates.

To explore this paradigm, we developed tetrahydro-1,2,4,5-tetrazin-3-ones (THTZOs) as a novel class of RTAs, featuring a scaffold that is distinct from classical RTA architectures. Through comprehensive kinetic and mechanistic investigations, we characterized the radical-trapping activity of THTZOs, elucidated the basis of hydrazyl and verdazyl radical stabilization that enables this unique, aromaticity-independent reactivity, and demonstrated their exceptional efficacy in inhibiting (phospho)lipid peroxidation and ferroptosis. In sharp contrast to ferroptosis inhibitors such as Fer-1, Lip-1, or phenoxazine, which were found through high-throughput screening or derived from known RTA scaffolds, our findings provide a framework for developing potent ferroptosis inhibitors through the rational design of novel RTAs (Fig. 1C).

RESULTS

Proof of concept: Identification of THTZO as novel RTA

A series of THTZOs (3a to 3g, 4a, and 5a) was synthesized via condensation of N-substituted carbohydrazides with aldehydes or ketones, following literature procedures (Fig. 2A) (44–46). Notably, this heterocyclic scaffold has an exceptionally weak N─H bond (BDE ~ 76 kcal/mol), comparable to those of hydropersulfides and hydroxylamines, yet it exhibits bench stability—thereby overcoming the rapid degradation typically observed in these analogs. Given that the chemistry and utilization of THTZOs have remained largely unexplored, we investigated their potential as RTAs.

Fig. 2. Identification of THTZO as novel RTAs.

(A) Structures of hydrazine derivatives and tetrahydro-1,2,4-triazine-3-thiones (THTZOs). (B) PBD-BODIPY serves as the signal carrier in its coautoxidation with 1,4-dioxane (λ_max = 588 nm, ε = 123,000 M⁻¹ cm⁻¹). (C) The inhibition rate constant (k_inh) and radical-trapping stoichiometry (n) of an added RTA are determined from the initial rate of PBD-BODIPY consumption and the inhibited period duration (t_inh). (D to G) Representative coautoxidation profiles of PBD-BODIPY (10 μM) and 1,4-dioxane (2.9 M) in PhCl, initiated with AIBN (6 mM) at 37°C, in the presence of THTZOs (4 μM). (H) Inhibition rate constants (k_inh) and stoichiometries (n); for direct comparison with 2,2,5,6,8-pentamethyl-7-hydroxychroman (PMC), k_inh was calculated assuming n = 2. UN, uninhibited.

The RTA activity was assessed by monitoring the inhibition of 2,2′-Azobis(2-methylpropionitrile) (AIBN)–initiated autoxidation of 1,4-dioxane. PBD-BODIPY was used as a spectroscopic probe due to its competitive oxidation with dioxane and distinct spectrophotometric signature (λ_max = 588 nm) (Fig. 2B) (33, 47). Using known rate constants for the reaction of PBD-BODIPY with dioxane-derived peroxyl radicals (k_PBD-BODIPY = 5310 M⁻¹ s⁻¹) (34), the inhibition rate constant (k_inh) and stoichiometric factor (n) of THTZOs were quantitatively determined from the inhibition period (t_inh) using Eqs. 1 and 2 (Fig. 2C). Representative kinetic traces are depicted in Fig. 2 (D to F), with a summary provided in Fig. 2G.

Initial evaluation of 3a demonstrated potent activity, displaying a k_inh of 1.4 × 10⁶ M⁻¹ s⁻¹—nearly four times greater than that of the 2,2,5,6,8-pentamethyl-7-hydroxychroman (PMC) reference (k_inh of 3.2 × 10⁵ M⁻¹ s⁻¹). Compound 3a also exhibited a well-defined, dose-dependent inhibition period (Fig. 2D). Both electron-deficient (3b) and electron-rich (3c) aromatic substitutions at the C(6) position, as well as complete replacement of the aromatic ring(s) with alkyl substituents (3d and 3e), were found to be potent RTA (Fig. 2E). In addition, C(6)-disubstituted THTZOs (3f and 3g) also effectively inhibited coautoxidation, exhibiting activities comparable to or exceeding those of their C(6)-monosubstituted analogs (Fig. 2F). Furthermore, structurally related compounds 4a and 5a—featuring different N(3)/N(6) substitutions—exhibited moderate RTA activities relative to 3a, confirming that the potent RTA activity is inherent to the THTZO scaffold. The slower k_inh observed for 4a and 5a compared to 3a is presumably attributable to steric effects (vide infra in mechanistic studies). Significantly, these THTZO compounds demonstrated radical-trapping activity competitive with benchmark RTAs, e.g., k_inh = 2 × 10⁵ M⁻¹ s⁻¹ for 4,4′-di-tert-butyl-diphenylamine and k_inh = 8 × 10⁶ M⁻¹ s⁻¹ for phenothiazine (48, 49). Overall, these results have identified the RTA reactivities of THTZO in principle.

Mechanistic studies revealed that hydrazyl and verdazyl radicals act as unprecedented intermediates in the RTA mechanism

THTZOs can be oxidized by reagents such as K₃[Fe(CN)₆], NaIO₄, p-quinone, or PbO₂ to generate verdazyl radicals, which have promising applications in energy storage and molecular electronics (Fig. 3A) (46, 50–52). In this work, we propose a cascade HAT pathway to explain the RTA mechanism of THTZOs. As illustrated in Fig. 3A, the process initiates with HAT from the N─H group of THTZO to a ROO•. The calculated BDE of the N─H bond in compound 3a is 75.3 kcal/mol—substantially lower than the O─H bond of ROOH—making this step thermodynamically favorable (Fig. 3B). This generates hydrazyl radical Int-1, which could subsequently react with ROO• to form dihydrotetrazin-3-one (DHTZO) Int-2. Consequent HAT from Int-2 to another ROO• ultimately produces a stable verdazyl radical. This cascade allows one THTZO to trap up to three ROO•, close to experimental stoichiometry (e.g., n ≈ 2.3 for 3a).

Fig. 3. Mechanistic studies.

(A) Reported synthesis of verdazyl radicals from THTZO derivatives in literature and the proposed mechanism for peroxyl radical trapping by THTZO in this study. (B) Computed (CBS-QB3) minimum energy structures of 3a, exhibiting a N─H BDE of 75.3 kcal/mol. (C) Electron paramagnetic resonance (EPR) spectra of 3a (50 mM) during inhibition of 1,4-dioxane/PhCl [1:3 (v/v)] autoxidation initiated by AIBN (2.0 M) at 37°C after 2 hours, alongside EPR spectra of verdazyl radical 11 for comparison. (D) Coautoxidation kinetics of PBD-BODIPY (10 μM) and 1,4-dioxane (2.9 M) in PhCl, initiated by AIBN (6 mM) at 37°C, in the presence of intermediate Int-2a (4 μM). (E) Kinetic isotope effect (KIE) determined from coautoxidations inhibited by 3a and deuterated analog d-3a (4 μM), conducted in the presence of 1% MeOH or 1% MeOD. (F) CBS-QB3 computational analysis of HAT from 3a to methylperoxyl radical (MeOO•, a model chain-carrying peroxyl radical), including thermodynamic and kinetic evaluation of possible reaction pathways.

First, electron paramagnetic resonance (EPR) spectroscopy confirmed the formation of verdazyl radical (11) during 3a-inhibited autoxidation (Fig. 3C and fig. S7). Furthermore, reduction of 11 yielded Int-2, which itself exhibited potent RTA activity, completely suppressing PBD-BODIPY consumption (Fig. 3D). Complete basis set-quadratic Becke3 model (CBS-QB3) computations further validated this pathway. As shown in Fig. 3F, HAT from 3a to methyl peroxyl radical (MeOO•) has a low barrier (ΔG^‡ = 9.4 kcal/mol) and is highly exergonic (ΔG° = −11.1 kcal/mol) (Fig. 3F). The transition state TS1 adopts a syn geometry, involving a secondary orbital interaction (53) between the lone pair on the N(2) atom and the π* singly occupied molecular orbital of the MeOO•. The hydrazyl radical Int-1a has a weak C(6)─H bond (BDE = 37.1 kcal/mol), enabling a highly exergonic second HAT (ΔG° = −47.7 kcal/mol) to form Int-2a. The final HAT to form verdazyl 11 is also favorable (ΔG^‡ = 12.7 kcal/mol and ΔG° = −21.2 kcal/mol). In addition, results from kinetic isotope effect (KIE) (fig. S3) and kinetic solvent effect (KSE) studies (fig. S4) (54) on 3a and 3f were consistent with the proposed HAT mechanism.

In contrast, direct HAT from the C(6)─H bond of 3a to MeOO• is both kinetically and thermodynamically disfavored compared to N─H HAT from 3a. This pathway exhibits a higher activation barrier (ΔG^‡ = 22.9 kcal/mol via TS2) and is less exergonic (ΔG° = −2.9 kcal/mol) (red profile; Fig. 3F), consistent with its higher C(6)─H BDE (83 kcal/mol) (Fig. 3E). The absence of a KIE [k_inh(H)/k_inh(D) ≈ 1; Fig. 3E] between 3a and C(6)-deuterated (d-3a) provides experimental confirmation that C─H bond cleavage is not the rate-determining step. Mechanistic investigation on C(6)-disubstituted THTZO also demonstrated the formation hydrazyl radical as intermediate in the RTA mechanism (figs. S11 to S14).

THTZOs are potent RTA in inhibiting (phospho)lipid peroxidation

As primary biomolecules, (phospho)lipid is highly vulnerable to oxidative damage, a process that contributes to the pathogenesis of numerous diseases, including cardiovascular diseases, neurological disorders, and cancer. To evaluate the protective activity of THTZOs against membrane lipid autoxidation, we used a liposomal model system composed of egg phosphatidylcholine (egg-PC) to closely mimic biological lipid bilayers. The (phospho)lipid peroxidation was initiated highly lipophilic di-tert-undecylhyponitrite (DTUN), and the RTA activity was assessed using STY-BODIPY as a fluorescent probe (Fig. 4A) (15, 33, 55). Reaction progress was monitored by tracking the fluorescence increase associated with STY-BODIPY oxidation (λ_ex = 488 nm; λ_em = 518 nm). Using the known rate constant for the reaction of STY-BODIPY with lipid peroxyl radicals (k = 894 M⁻¹ s⁻¹), the inhibition rate constants (k_inh) and stoichiometric factor (n) of THTZO can be determined using Eqs. 3 and 4 in Fig. 4B. Representative data are presented in Fig. 4 (D to G).

Fig. 4. Assessment of antioxidant activity of THTZO in biomimetic membranes.

(A) STY-BODIPY serves as the signal carrier in the DTUN-initiated coautoxidation of liposomal phosphatidylcholine. (B) The inhibition rate constant (k_inh) and radical-trapping stoichiometry (n) of an added RTA are determined from the initial rate of STY-BODIPY consumption and the inhibited period duration (t_inh). (C) Structures of more THTZOs in addition to Fig. 2A for SAR. (D to F) Representative coautoxidation kinetics of STY-BODIPY (1 μM) with egg-PC liposomes (1 mM) in phosphate-buffered saline (PBS) at 37°C, initiated by DTUN (0.2 mM) and inhibited by THTZOs (4 μM). (G) Inhibition rate constants (k_inh) and stoichiometric factors (n) derived from the kinetic analyses. (H) Coautoxidation kinetics under identical conditions as (D) to (F) but inhibited by varying concentrations of Int-2a (1, 2, 4, and 6 μM). (I) Reduction of verdazyl radical 11 by ascorbic acid regenerates intermediate Int-2a, which itself exhibits RTA activity. (J) Inhibited coautoxidation kinetics with THTZOs supplemented with ascorbic acid (8 μM) as coantioxidants. The k_inh values were calculated with n = 2. UN, uninhibited.

Our initial evaluation revealed that N(2),N(3)-dimethyl–substituted THTZOs (e.g., 3a, 3b, 3d, 3f, and 3g) exhibited minimal RTA activity in liposomal suspensions (Fig. 4D)—a stark contrast to their high reactivity in organic solution. The sole exception was 3c, which functioned as a modest inhibitor of egg-PC peroxidation (k_inh = 0.8 × 10⁴ M⁻¹ s⁻¹; n = 2.7), outperforming α-TOH (k_inh = 0.5 × 10⁴ M⁻¹ s⁻¹; n = 2) (18). Computational analysis revealed a strong correlation between lipophilicity and activity: The active inhibitor 3c displayed a higher cLogP (2.74) than inactive analogs 3a (1.91), 3b (1.77), and 3d (2.13). This prompted further investigation, which confirmed that intrinsic lipophilicity is a critical determinant of RTA efficacy in lipid membranes. A clear trend of increasing activity with lipophilicity was observed in the series 3a [N(2),N(3)-dimethyl; cLogP = 1.91] < 12 [N(2)-methyl,N(3)-n-propyl; cLogP = 2.81] < 4a [N(2),N(3)-di-isopropyl; cLogP = 3.72], with 4a exhibiting a well-defined inhibition period (k_inh = 1.5 × 10⁴ M⁻¹ s⁻¹; Fig. 4E).

We subsequently explored the SAR of THTZO4 derivatives (4a to 4h; Fig. 4E). These compounds exhibited potent lipid peroxidation inhibition, with activity strongly correlated to their lipophilicity. Notably, more lipophilic analogs (e.g., 4c to 4e) exhibited four- to sixfold higher inhibition rate constants compared to 4a, achieving values of k_inh = 8.9 × 10⁴ M⁻¹ s⁻¹ (4c) and 8.2 × 10⁴ M⁻¹ s⁻¹ (4e). These kinetic parameters surpass those of established lipophilic RTAs, including Fer-1 (k_inh = 4.6 × 10⁴ M⁻¹ s⁻¹), Lip-1 (k_inh = 1.2 × 10⁴ M⁻¹ s⁻¹), and α-TOH (k_inh = 0.5 × 10⁴ M⁻¹ s⁻¹) (18). The C(6)-disubstituted derivatives (4f to 4h) displayed reduced stoichiometric factors relative to 4a to 4e, a trend consistent with our previous observations from inhibited autoxidation studies in dioxane. The N(2),N(4)-diphenyl–substituted THTZO5 derivatives (5a to 5h) showed moderately reduced activity compared to their di-isopropyl counterparts. Nevertheless, these compounds maintained similar structure-activity trends, with 5c demonstrating the highest inhibitory potency within the THTZO5 series (Fig. 4F).

C(6)-monosubstituted THTZOs displayed elevated stoichiometric factors (n > 3) during inhibited phospholipid peroxidation compared to homogeneous organic medium (where n ≥ 2). We evaluated two mechanistic hypotheses: (i) Differential stability of Int-2 in biphasic versus homogeneous organic medium: Initial studies confirmed that Int-2a exhibits excellent RTA activity in lipid bilayers (Fig. 4H). Comparative ¹H nuclear magnetic resonance kinetic analysis revealed substantial solvent-dependent stability differences: While Int-2a completely decomposed within 80 min in C₆D₆, the biphasic D₂O/CD₃CN system showed markedly enhanced stability, maintaining >60% of Int-2a after 80 min and requiring 240 min for complete degradation (fig. S9). These results imply that biphasic environments could significantly stabilize reactive intermediates, providing a plausible explanation for the elevated n observed in phospholipid bilayer systems. (ii) The observed n > 3 suggests potential regeneration of RTA species, a phenomenon previously reported for nitroxide antioxidants by Valgimigli and co-workers (41–43, 56). To investigate this possibility, we evaluated potential synergistic effects between THTZOs and the aqueous-phase antioxidant ascorbic acid. At the lipid bilayer-aqueous interface, ascorbic acid can reduce persistent verdazyl radicals to regenerate intermediate Int-2. This reformed intermediate maintains sufficient lipophilicity to diffuse back into the (phospho)lipid phase and intercept lipid peroxyl radicals (LOO•) (Fig. 4I). Notably, while ascorbic acid alone exhibited negligible RTA activity, its combination with THTZOs significantly prolonged the protection duration (Fig. 4J, fig. S6, and table S1). This synergistic effect likely mediated by an antioxidant recycling mechanism that substantially enhances the suppression of (phospho)lipid peroxidation (28). To investigate this further, we used EPR to measure the formation of the verdazyl radical 11′ during 4a-inhibited (phospho)lipid peroxidation. We observed that the detectable concentration of 11′ was much lower in this system than in organic solution. Compared to AIBN-induced autoxidation in dioxane, the kinetics of 11′ accumulation were also apparently slower (fig. S8). This observation presumably implied that verdazyl radical is recycled by hydroxyl radicals (HOO•), which are generated during unsaturated lipid peroxidation (57). Furthermore, other physiologically relevant reductants, including glutathione (GSH) and reduced form of nicotinamide adenine dinucleotide, were also found to extend the inhibitory efficacy of Int-2a (fig. S6).

THTZOs are validated as potent ferroptosis inhibitors

Ferroptosis is driven by the overwhelming accumulation of membrane lipid peroxides. Among known inhibitors, lipophilic RTAs represent the main strategy to interfere (phospho)lipid peroxidation and inhibit ferroptosis. In light of the intrinsic RTA activity of THTZOs, we sought to determine whether rationally designed THTZO-based RTAs, which feature a distinct structural scaffold, could serve as effective antiferroptotic agents targeting this mechanism of action.

Here, we explored the antiferroptotic activity of THTZO in HT22 cells (a mouse hippocampal neuron cell line) using RAS-selective lethal 3 [(1S,3R)-RSL3, hereafter RSL3] as the ferroptosis inducer (58). Among the tested compounds, THTZO4 and THTZO5 showed potent ferroptosis inhibition, while THTZO3 exhibited significantly lower efficacy (Fig. 5A). This activity trend correlated with their radical-trapping efficiency in lipid bilayers, reinforcing the critical role of (phospho)lipid peroxidation in ferroptosis. We quantitatively evaluated the inhibitory potency of THTZO4 and THTZO5 by determining their EC₅₀ values against RSL3-induced ferroptosis in HT22 cells (Fig. 5, B to D). All compounds inhibit ferroptotic cell death with nanomolar potencies. The isopropyl- and phenyl-substituted THTZO4 analogs showed similar potencies (EC₅₀ = 92 and 94 nM for 4a and 4f, respectively), while the corresponding THTZO5 counterparts were slightly less effective (EC₅₀ = 126 and 138 nM for 5a and 5f, respectively).

Fig. 5. Evaluation of antiferroptotic activity of THTZO.

(A to C) Evaluation of THTZO3/4/5’s ability to rescue HT22 cells from RSL3-induced ferroptosis. (D) Corresponding EC₅₀ values for THTZO4 and THTZO5 in ferroptosis protection. (E) Assessment of THTZO protective effects against erastin-induced ferroptosis in HT22 cells. (F) Investigation of THTZO capacity to mitigate RSL3-induced ferroptosis in HSC-LX2 cells. (G) Ultraviolet-visible (UV-vis) absorption spectra of deferoxamine (DFO) and compound 4a in the presence and absence of Fe³⁺ or Fe²⁺. a.u., arbitrary units. (H) Confocal microscopy images showing STY-BODIPY–stained HT22 cells treated for 24 hours with (i) vehicle control, (ii) RSL3 (1 μM), (iii) RSL3 plus 4c (30 nM, EC₅₀), and (iv) RSL3 plus Fer-1 (74 nM, EC₅₀).

In general, greater potencies were followed by increasing the lipophilicity. Notably, several THTZO4 compounds surpassed the benchmark ferroptosis inhibitor Fer-1 in efficacy; for instance, 4c and 4e displayed EC₅₀ values of 30 and 31 nM, respectively, compared to 74 nM for Fer-1 under identical conditions. Furthermore, these THTZO derivatives effectively suppressed erastin-induced ferroptosis (7) in HT22 cells (Fig. 5E), as well as RSL3-induced ferroptosis in HSC-LX2 cells (a human hepatic stellate cell line) and MIA PaCa-2 cells (a pancreatic cancer cell line) (Fig. 5F and figs. S18 to S20). Within these assays, THTZO compounds again demonstrated excellent antiferroptotic profiles, e.g., 4c was showed to be more than twofold more potent than Fer-1. Preliminary cytotoxicity evaluation on HT22 cells as a model revealed that these THTZOs displayed favorable safety profiles (figs. S15 to S17).

Considering that hydrazones are known iron chelators (59), we investigated whether the cytoprotective effects of THTZO compounds might involve iron chelation; we therefore assessed the iron-chelating capacity of lead compound 4c using ultraviolet-visible (UV-vis) spectroscopy. While the positive control deferoxamine (DFO) displayed the characteristic ~430-nm absorption band upon Fe²⁺/Fe³⁺ coordination, compound 4c showed no detectable spectral shifts under identical conditions (Fig. 5G). These findings suggested that 4cs inhibits ferroptosis upon interfering the propagation cycle of lipid-derived peroxyl radicals. To further validate this mechanism, we quantitatively assessed lipid peroxidation inhibition using the STY-BODIPY probe in HT22 cells by confocal fluorescence microscopy (32). As shown in Fig. 5H, both 4c and Fer-1 (at EC₅₀) effectively suppressed RSL3-induced lipid peroxidation, as evidenced by the significant reduction in green fluorescence compared to RSL3-treated controls. This direct visualization of lipid peroxidation inhibition, combined with the iron chelation results, provides compelling evidence that THTZOs exert their potent cytoprotective effects primarily through inhibition of (phospho)lipid peroxidation.

Beyond glutathione peroxidase 4 (GPX4), ferroptosis suppressor protein 1 (FSP1) is a GSH-independent regulator that suppresses ferroptosis by synthesizing RTAs from ubiquinone or vitamin K (11, 60, 61). To investigate whether FSP1 could also regenerate THTZO derivatives and thereby potentiating RTA capacity, we used HT1080 cells as a model for studying FSP1 function (11). In the presence of the FSP1 inhibitor (iFSP1), HT1080 cells exhibited increased sensitivity to ferroptosis induced by RSL3 (fig. S21). Phylloquinone and menadione rescued cells from ferroptosis induced by RSL3 alone. Similarly, the antiferroptotic potency of 4a and Fer-1 was significantly enhanced in the absence of iFSP1, compared to cotreatment with RSL3 and iFSP1 (Fig. 6, A and B). Collectively, these results suggest that the observed potency of THTZOs may be attributed not only to their intrinsic RTA activity but also to FSP1-mediated recycling. Our findings reinforce the conclusion that the robust cytoprotective role of FSP1 stems from its dual function in both de novo synthesis of endogenous RTAs (62) and the regeneration of synthetic RTAs, such as Fer-1, THTZO, and seratrodast (Fig. 6C) (63).

Fig. 6. Evaluation of the regeneration of THTZO by FSP1.

(A) Compounds 4a, Fer-1, phylloquinone, and menadione dose-dependently rescue HT1080 cells from RSL3-induced ferroptosis in both the presence and absence of iFSP1. (B) Corresponding EC₅₀ values for 4a and Fer-1. (C) Schematic illustrating the complementary protective roles of GPX4 and FSP1 in suppressing ferroptosis. FSP1 functions to reduce quinones (e.g., ubiquinone or vitamin K) in the synthesis of the corresponding para-hydroquinone RTAs. FSP1 is also capable of regenerating synthetic RTAs such as THTZOs and Fer-1.

DISCUSSION

Research on RTAs is highly valued, with applications spanning petroleum, polymer, food, and biomedical sciences (4, 5, 19). The recent characterization of prototypical antiferroptotic agents—inhibitors that suppress regulated cell death that is linked to lipid peroxidation, such as α-TOH, Fer-1, Lip-1 and phenoxazines—have provided fundamental insights into SAR profiles of lipophilic RTAs. The aromatic ring(s) in these compounds stabilize the resulting radical intermediates, which weakens the O─H or N─H bonds. This thermodynamic profile favors high reactivity toward ROO•, while the aromatic system also kinetically facilitates the reaction through HAT or, more specifically, proton-coupled electron transfer via a transition state stabilized by secondary orbital interactions. In contrast, nonaromatic ring–substituted amines are generally inefficient H-atom donors and exhibit poor antioxidant activity. These findings demonstrate that hydrazine scaffolds can undergo HAT to inhibit autoxidation as effectively as aromatic amine RTAs—yet without requiring aromatic rings! There are challenges in harnessing the HAT chemistry of hydrazines in the development of RTAs, specially associated with their high electron-donating capacity and susceptibility to oxidation. For example, alkyl hydrazines undergo oxidative degradation to form alkyl radicals (64). Similarly, acylhydrazines, such as isoniazid, are known to act as radical precursors activated by KatG (65). As a result, the reverse process—HAT to hydrazyl radicals—has been extensively studied, as evidenced by the widespread use of 2,2-diphenyl-1-(2,4,6-trinitrophenyl)hydrazyl (DPPH•) in assays of antioxidant activity (66). Our mechanistic studies demonstrated the hydrazyl and verdazyl radical as previously unrecognized intermediate in RTA mechanism.

RTA activity is significantly influenced by medium effects, particularly H-bonding interactions, which is especially critical in phospholipid bilayers—a strongly hydrogen-bonding system (${\mathrm{\beta }}_{\mathrm{H}}^2$ = 0.69). H bonding between RTAs and phosphodiester headgroups prevents HAT to ROO• and consequently attenuating RTA kinetics. RTAs with higher H-bond acidity exhibit substantially reduced efficacy in inhibiting phospholipid peroxidation (Fig. 7A). Compared to PMC (a phenolic RTA, ${\mathrm{\alpha }}_{\mathrm{H}}^2$ = 0.37) and phenoxazine (an aromatic amine RTA, ${\mathrm{\alpha }}_{\mathrm{H}}^2$ = 0.44), THTZOs demonstrate lower H-bond acidity (${\mathrm{\alpha }}_{\mathrm{H}}^2$ ~ 0.18). A significantly smaller decline in THTZO activity compared with PMC and phenoxazine was observed when the medium changed from organic solvents ($k_{\text{inh}}^{\text{dioxane}}$) to phospholipid bilayers ($k_{\text{inh}}^{\left(\text{phospho}\right)\text{lipid}}$). Therefore, THTZOs are among the rare examples of weakly H-bond–donating RTAs that maintain efficacy across heterogeneous environments (15, 39).

Fig. 7. Comparison of THTZOs with known synthetic RTA ferroptosis inhibitors.

(A) Hydrogen-bonding interactions modulate RTA reactivity toward LOO•. Representative RTA scaffolds—including PMC, phenoxazine, and THTZO 4e—exhibit varying H-bonding capacities (${\mathrm{\alpha }}_{\mathrm{H}}^2$), leading to medium-dependent RTA activity shifts between organic solutions and (phospho)lipid environments ($k_{\text{inh}}^{\left(\text{phospho}\right)\text{lipid}}$: $k_{\text{inh}}^{\text{dioxane}}$). (B) Structures of representative ferroptosis inhibitors known to act by RTA mechanisms: aromatic ring–dependent RTAs {Fer-1, Lip-1, phenoxazine, and 1,8-diaminonapththalene phenylboronic amide [PhB(DAN)]} versus limited examples of aromatic ring–independent (right) RTAs (CuATSM, JP4-029, and THTZO 4e in this study); the EC₅₀s were obtained from RSL3-induced ferroptosis model from corresponding reference. ^aData were adapted from (28); EC₅₀ values measured in RSL3-induced ferroptosis in HT22 cells^b from this study or in Pfa1 cells^c and HT-1080 cells^d from the references.

Beyond the nonclassical RTA scaffold and its associated novel radical chemistry, the demonstration of THTZO efficacy in physiologically relevant systems was particularly compelling. This was exemplified by their potent ferroptosis-inhibitory activity in cellular models. Not all RTAs function as effective ferroptosis inhibitors. Some, such as 2,2,6,6-tetramethylpiperidine and tetrasulfides (48, 67), operate primarily under elevated temperatures, while others, such as hydropersulfides, lack the stability required for pharmacological application (68). In contrast, we demonstrate that THTZOs inhibit RSL3-induced ferroptosis at nanomolar concentrations, with compounds 4c, 4e, and 4g with EC₅₀s <50 nM, demonstrating superior potency to Fer-1 (EC₅₀ = 74 nM). This antiferroptotic activity was further validated across three distinct cell lines, from mouse neuronal cells to human hepatic stellate cells and pancreatic carcinoma cell (Fig. 5). Therefore, THTZOs, along with metal thiosemicarbazones [e.g., copper(II)-diacetyl-bis(N⁴-methylthiosemicarbazone) (CuATSM)] (31, 32) and nitroxides (e.g., XJB-5-131 and JP4-039) (69), represent a novel class of ferroptosis inhibitors characterized by nonclassical, aromaticity-independent scaffolds (Fig. 7B). Given that aromatic systems with electron-donating substituents often act as structural alerts for drug-induced toxicity, tetrazines have emerged as versatile tools in bioorthogonal chemistry (70). Moreover, 6-oxoverdazyl derivatives have demonstrated cytocompatibility, properties that have enabled their application in medicinal imaging (71). Therefore, the THTZO scaffold represents a promising strategy to circumvent the metabolic liabilities and biocompatibility associated with current ferroptosis inhibitors, potentially enabling more effective antioxidant therapeutics.

Beyond their critical role in industrial materials and food science, RTAs are increasingly recognized as essential regulators of ferroptosis. This study establishes hydrazine-derived THTZOs as a potent class of RTAs that inhibit ferroptosis. In contrast to prototypical ferroptosis inhibitors Fer-1 and Lip-1—found blindly through high-throughput screening—our work presents a rational design strategy for RTAs based on a distinct molecular scaffold to target this (phospho)lipid peroxidation and associated cell death. Mechanistic studies revealed that hydrazyl and verdazyl radicals act as unprecedented intermediates in the THTZO RTA mechanism. Owing to their high potency, synthetic accessibility, and tunable properties, THTZOs represent a highly promising platform for further SAR optimization to develop advanced ferroptosis inhibitors. Investigations toward this goal are currently underway.

MATERIALS AND METHODS

General

For reagents, equipment, and detailed experimental procedures, see Supplementary Text.

General procedure for synthesis of compound THTZOs

Substituted carbohydrazide hydrochloride (0.2 mmol, 1.0 equiv) was dissolved in ethanol (2 ml), followed by the addition of corresponding aldehyde or ketone (0.2 mmol, 1.0 equiv). The mixture was stirred at room temperature and monitored by thin-layer chromatography. Upon completion, dichloromethane and water were added for extraction, and the organic phase was collected, dried with anhydrous sodium sulfate, filtered, and then evaporated under reduced pressure. The residue was purified by recrystallized or flash column chromatography on silica gel to obtain targeted product.

Evaluation of RTA activity in organic solutions

Into a 3-ml cuvette were added 625 μl of freshly distilled dioxane (passed through a 10 cm–by–2.5 cm silica plug) and either 1753 μl of chlorobenzene (PhCl) or 1128 μl of PhCl with 625 μl of dimethyl sulfoxide (DMSO; for KSE studies). The solution was mixed well and then incubated at 37°C in Agilent Cary 3500 UV-Vis. When temperature stabilized, the zero adjustment is corrected, and the samples were blanked. Next, 50 μl of AIBN (0.3 M in PhCl) and 47 μl of PBD-BODIPY (0.54 mM in 1,2,4-trichlorobenzene) were added to all cuvettes in sample cells and then mixed well. The reaction was monitored at 588 nm for 5 min to confirm stabilization and then paused. Next, 25 μl of test RTA in PhCl was added to all sample cuvettes, achieving final concentrations of 4 or 1, 2, 6, and 8 μM. A PMC control (4 μM) was included as a standard reference. Control experiments were run and repeated three times independently.

Assessment of antioxidant activity in biomimetic membranes

Liposomes were prepared by dissolving l–α-PC (egg-PC, Sigma-Aldrich) in phosphate-buffered saline (PBS; 10 mM, pH 7.4). The solution was then extruded through a polycarbonate membrane using an Avanti Polar Lipids extruder to obtain a uniform suspension of 100-nm vesicles, yielding a final lipid concentration of 25 mM. The reaction mixture was prepared by diluting liposomes, STY-BODIPY (1.175 mM in DMSO), and DTUN (12 mM in EtOH) with pH 7.4 PBS buffer to 295 μl per well in a 96-well plate. After thorough mixing, 5 μl of THTZOs (0.24 mM in CH₃CN) was added to each well, yielding final concentrations of 1 mM liposomes, 1 μM STY-BODIPY, and 0.2 mM DTUN. The plate was then incubated in a BioTek Synergy H1 plate reader stabilized at 37°C, with initial mixing performed using a 1-min vibration program. Fluorescence was monitored (λ_ex = 488 nm and λ_em = 518 nm) throughout the experiment. A PMC control (4 μM) was included as a standard reference. Control experiments were run and repeated independently three times.

Quantitative RTA kinetic analysis

Reaction kinetics were analyzed using 2,2,7,8-pentamethyl-6-chromanol (PMC) as reference standard with a defined stoichiometry (n = 2) to determine the radical initiation rate (R_i). The rate constants for the reactions of the BODIPY with peroxyl radicals were described in the literature: PBD-BODIPY reacts with dioxane-derived peroxyl radicals with k_PBD-BODIPY = 5310 M⁻¹ s⁻¹ (34), while STY-BODIPY reacts with lipid-derived peroxyl radicals with k_STY-BODIPY = 894 M⁻¹ s⁻¹ (55). The inhibition rate constant (k_inh) and stoichiometric factor (n) were calculated from the inhibition period duration (t_inh) using Eqs. 1 and 2, providing quantitative measures of antioxidant efficiency in both homogeneous and membrane systems. For dioxane/PBD-BODIPY coautoxidations, R_i was determined to be 2.5 × 10⁻⁹ M s⁻¹ ± 10% from the curve of PMC under the same condition. For egg-PC/STY-BODIPY coautoxidations, R_i was determined to be 9.1 × 10⁻¹⁰ M s⁻¹ ± 10% from the curve of PMC under the same condition

$$\frac{-\mathrm{\delta }\left[\text{BODIPY}\right]}{\mathrm{\delta }t}=\frac{k_{\text{BODIPY}}\left[\text{BODIPY}\right]R_{\mathrm{i}}}{n\times k_{\text{inh}}\left[\text{RTA}\right]}$$ (1)

$$n=\frac{t_{\text{inh}}\times R_{\mathrm{i}}}{\left[\text{RTA}\right]}$$ (2)

Electron spin resonance studies

The EPR spectrometer was calibrated using 2,2,6,6-tetramethylpiperidine-N-oxyl radical in benzene (g = 2.0062) by determining the calibration factor. Instrument parameters were set as follows: microwave frequency of 9.85 GHz, 20.0-dB attenuation (1.84-mW power), 100.00-kHz modulation frequency with 0.1 mT amplitude, and 20.48-s sweep time.

Computational methods

All quantum chemical calculations were conducted using Gaussian 16 (Revision C.01). Geometry optimizations and enthalpy calculations were performed using the CBS-QB3 composite method. All structures were confirmed to be true minima by vibrational frequency analysis.

Evaluation of antiferroptotic activity by cell viability assays

All cell lines (HT22, HSC-LX2, and MIA PaCa-2) were incubated in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal bovine serum and 1% 100× penicillin-streptomycin in incubator with 5% CO₂ at 37°C. Cells were passaged by dissociation with 0.05% trypsin and 0.2% EDTA every other day. Cells were seeded in a 96-well plate at 2000 cells per well and incubated for 24 hours. Cells were then treated with RSL3 (final concentration: 1.0 μM), erastin (final concentration: 0.8 μM), or a concentration gradient of RTA, followed by an additional 24 hours of incubation. After aspiration and removal of the original medium, subsequent steps were performed according to the manufacturer’s instructions using the following kits: AquaBluer assay for HT22 cells and Cell Counting Kit-8 for HSC-LX2 and MIA PaCa-2 cells. All experiments included four technical replicates per RTA concentration and were performed in three independent biological repeats.

Iron chelation properties of 4a

The iron-chelating properties of compound 4a were investigated using a UV-2600 spectrophotometer (Shimadzu, Japan) with 1-cm path length quartz cuvettes. Test solutions were prepared by adding either FeSO₄ or FeCl₃ (100 μM final concentration) to solutions containing DFO or compound 4a (100 μM) in Hepes buffer (20 mM (pH 7.4) and 150 mM NaCl). UV-vis absorption spectra were recorded between 290 and 600 nm.

Confocal microscopy imaging

HT22 cells (10,000 cells per well) were seeded in six-well glass-bottom confocal plates and cultured for 24 hours (37°C and 5% CO₂) before treatment with RSL3 (1.0 μM) alone or in combination with either Fer-1 (74 nM) or 4c (30 nM) for an additional 24 hours. After treatment, cells were washed with PBS, stained with STY-BODIPY (1.0 μM in DMEM) for 30 min at 37°C, washed again with PBS, and imaged in PBS using confocal microscopy with 488- and 561-nm excitation wavelengths detected by a GaAsP detector.

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S21

Tables S1 to S5

Data S1 and S2